Download CONTENTS- The Norwich Economic Papers

Document related concepts

Economic democracy wikipedia , lookup

Economics of fascism wikipedia , lookup

Business cycle wikipedia , lookup

Non-monetary economy wikipedia , lookup

Fiscal multiplier wikipedia , lookup

Transcript
The
Norwich
Economic
PapersVolume 14
SEPTEMBER, 2016
CONTENTSArticles
IN THIS ISSUE
This edition features articles,
exemplary student essays from the
2015/16 academic year and a
“special mention” from our previous
essay competition.
 James Merewood, “To What Extent did Barack Obama Fulfil the Expectations Set Out in the
Famous 'Yes We Can' Speech?
 Andra Tutuianu, “China: Collapse of an Empire or a new fresh start?”
 Alexander Lindsay, “Norwich’s own Launch-pad for Growth”
First Year Exemplary Essays




George Markham, Introductory Economics Assignment
John Kahodi, Introductory Economics Assignment
Wokciech Serwacki, Introductory Economics Assignment
Alex Lovett, Introductory Economics Assignment
Second Year Exemplary Essays
 Alice Hebditch, AEA, “What influences trade union membership? A microeconomic analysis”
 Rebecca Heath, AEA, “The Economics of Housework: Exploring the Determinants and Allocation
of Hours of Household Labour.”
 Winnie Ho, AEA, “Does producing green product create opportunity for companies? Is living area
being important factor for companies to consider?”
 Alexander Lindsay, AEA, “Austerity measures have strained Charities, will income growth relieve
this pressure”
 Natalie Wood, AEA, “What factors determine people’s willingness to donate money to charity
and their generosity when donating?”
1
Third year Exemplary Essays
 Hannah McCartney, Labour Economics, “An investigation into the UK gender pay gap and how it
could be improved through a change in paternity leave policy”
Special Mention for Essay Competition
 Anthony Amoah, “Are we ready for a repeat of financial crisis?”
Foreword
Dear All,
It gives me particular pleasure to welcome you to my final edition of The Norwich Economic Papers, this
edition continues our mission of showcasing the high-caliber work that UEA students have produced
throughout the academic year. I am impressed with the quality of the work that my fellow students but
would especially like to comment on James Merewood’s piece as delivers an insightful and informed
reflection on Obama’s time in office which I hope you all enjoy. I am proud of all the students work and I
have no doubts that this paper will continue to thrive in the future.
I would like to congratulate all students who were nominated by their lectures to have
their work published, the Editorial Board and those who have contributed articles. In
particular my greatest thanks are due to Dr Jibonayan Raychaudhuri our Academic
Advisor who has invested time and energy into ensuring the success of this paper.
Alex Lindsay [Editor]
2
To What Extent did Barack Obama Fulfil the Expectations Set Out in the Famous 'Yes We Can'
Speech?
By James Merewood
Noted as one of the most important political speeches of the 21st century, the famous 'Yes We
Can'1 speech followed the announcement of Barack Obama as the Democratic
presidential nomination. Outlining promises of an approach to necessity over popularity, the
speech gave hope to many Americans searching for change within a disenfranchised system.
With the most important change relating to the affordability of Health Care and the
culmination of the Iraq war, the 'Yes We Can' speech marked a key point in American history.
Due to Obama reaching the end of his second term as President, it is now possible to
examine the extent to which the expectations set out within the 'Yes We Can' speech have
been achieved.
To many, the key policy area in the Presidential race of 2008/09 was that of Health Care and
the insurance system which was in place pre-Obama. Due to the private provision of
American health care, costs within the health care system are at an all time high, with an
average of $8,362 spent per year on Health Care by US citizens compared to the average of
$3,480 spent per year by UK citizens2. As a result of high health care costs it is necessary for
American citizens to purchase insurance, although some are provided insurance through the
work place. Promises of affordable and available health care first arrived within the 'Yes We
Can' speech and were brought to fruition in late 2011 with the introduction of the Patient
Protection and Affordable Care Act (PPACA). Coined Obamacare by media sources,
the PPACA aims to provide affordable health insurance for millions of Americans who "go
without insurance"3 due to the high cost of insurance premiums provided by private firms. In
addition, Obama has extended the Medicaid system, which allows for free health care
provision for those most in need within society, including the elderly who often cannot afford
private health insurance due to the reduced income received from pensions schemes. In
2012 it was estimated that over 30 million previously uninsured American citizens would gain
health care coverage as a direct result of the introduction of Obamacare and the
extensions provided to Medicaid4.
In addition, the introduction of the PPACA was punctuated with further reform on Health
Care Policy; alongside the PPACA, law was introduced mandating that each individual must
obtain health care insurance or pay a tax penalty of up to $2,085 per family5. In economic
theory the provision of health care insurance allows for protection from future health shocks,
consumption smoothing, alongside lower budget costs6. In the absence of health insurance,
and under a private provision of health care, some citizens are forced to wait until hospital
visits are absolutely necessary resulting in a higher number of Emergency Room (ER) visits, which
come at no cost to the patient, and an overall higher cost to the tax payer. Therefore,
imposing a tax on those
Available at https://www.youtube.com/watch?v=HoFqV3qVMGA
http://uk.businessinsider.com/an-american-uses-britain-nhs-2015-1?r=US&IR=T
3 Wilensky, Gail. (2012). The Shortfalls of "Obamacare". The New England Journal of Medicine. 367:1479-1481.
4 Ibid.
5 http://obamacarefacts.com/obamacare-individual-mandate/
6 Lui, Kai. (2014). Insuring Against Health Shocks: Health Insurance, Consumption Smoothing and Household
Choices. Norwegian School of Economics.
1
2
3
without health insurance allows for the funds to be available within the budget to compensate
for the cost of an ER visit by these patients. Whilst economic theory suggests that the imposition
of a tax on those who choose not to purchase health insurance counteracts the higher cost of
ER visits and provides an initial incentive to purchase health insurance, critics have argued that
subsidies provided under Obamacare are unconstitutional. However, a Supreme Court ruling of
June 25, 2015 found that tax credit subsidies for those purchasing health care insurance through
Obamacare exchanges are constitutional7, saving the health insurance of millions of Americans.
In addition to the legal ruling of the Supreme Court regarding tax credit subsidies, further ruling
'effectively made the ACA's Medicaid expansion optional'8 leading to some Republican
governors 'refusing to expand their programs'9. Unable to bring the two main political parties
together, Obama has faced further Republican backlash regarding the introduction of
Obamacare; during the first year of operation, only four out of thirty states with republican
governors chose to put their own exchanges in place, whilst the rest opted to yield power to
Washington10. This intense resistance to Obamacare within Republican states has seriously
undermined the availability and affordability of health care within the USA, due to faltering public
opinions and bureaucratic complications11.
As the second most important policy outline, and one which many believe was 'necessary' for
Obama to gain power12, the popular stance of ending the war in Iraq features heavily during
the 'Yes We Can' speech. However, further infighting of the American government has lead to
the targeted bombing of ISIS following a peaceful negotiation with Iran and the end of the Iraq
war, an outcome that seems contradictory and against the fundamental peaceful principles
laid out in the 'Yes We Can' speech. Whilst initial expectations were met following the end of the
Iraq war in December 201113 , Obama has since followed the doctrine of counterterrorism
begun by Bush. Whilst many believed that the 'Yes We Can' speech marked the start of a more
moral approach to counterterrorism, the Obama presidency has featured bombings in the name
of counterterrorism and with the aim of reducing the threat of terrorism within the US14. Whilst
pressures from the Republican party may be the reasoning behind the move from a peaceful
and moral approach to a further entrenchment of the Bush doctrine, ultimately the expectations
set out within the 'Yes We Can' speech appear to no longer take precedent over the approach
to necessity rather than popularity, another key promise within the speech.
In conclusion, whilst some may consider the election of Obama for a second term to be
evidence for the fulfilment of the promises made within the 'Yes We Can' speech, the steps taken
http://www.supremecourt.gov/opinions/14pdf/14-114_qol1.pdf
Oberlander, Jonathan. (2012). The Future of Obamacare. The New England Journal of Medicine. 367:2165-2167.
9 Ibid.
10 Jones, David and Bradley, Katherine. (2014). Pascal's Wager: Health Insurance Exchanges, Obamacare, and the
Republican Dilemma. Journal of Health Politics, Policy and Law. 39:97-137.
11 Manchikanti, L and Hirsch, JA. (2012). Obamacare 2012: Prognosis unclear for interventional pain management.
Pain Physician. 15(5):629-640.
12 Jacobson, Gary. (2010). George W. Bush, the Iraq War, and the Election of Barak Obama. Presidential Studies
Quarterly. 40(2):207-224.
13 http://www.britannica.com/event/Iraq-War
14 McCrisken, Trevor. (2011). Ten Years On: Obama's War on Terrorism in Rhetoric and Practice. International Affairs.
87(4):781-801.
7
8
4
by Obama to provide available and affordable health care are insufficient in fulfilling the
expectations of many Americans. Whilst Obamacare and Medicaid help those most in need,
the constant backlash of many Republican states has lead to an uneven availability of health
care insurance and in some instances bureaucratic failure in which states do not provide the
coverage required. In addition, the constant battle between Democrat and Republican
representatives regarding both Obamacare and international policy has lead to a more
centralist approach than outlined in the 'Yes We Can' speech in order to maintain support from
both parties; this move towards a centralist position is considered by many as a move away from
progression. As such, whilst Obama has made significant steps towards meeting the expectations
set out in the 'Yes We Can' speech, further reforms are needed in order to realise the full extent
of progressive change that the world expected during the Obama presidency.
Bibliography
Jacobson, Gary. (2010). George W. Bush, the Iraq War, and the Election of Barack Obama. Presidential Studies Quarterly.
40(2):207-224.
Jones, David and Bradley, Katherine. (2014). Pascal's Wager: Health Insurance Exchanges, Obamacare, and the Republican
Dilemma. Journal of Health Politics, Policy and Law. 39:97-137.
Lui, Kai. (2014). Insuring Against Health Shocks: Health Insurance, Consumption Smoothing and Household Choices. Norwegian
School of Economics.
Manchikanti, L and Hirsch, JA. (2012). Obamacare 2012: Prognosis unclear for interventional pain management. Pain Physician.
15(5):629-640.
McCrisken, Trevor. (2011). Ten Years On: Obama's War on Terrorism in Rhetoric and Practice. International Affairs. 87(4):781-801.
Oberlander, Jonathan. (2012). The Future of Obamacare. The New England Journal of Medicine. 367:2165-2167.
Wilensky, Gail. (2012). The Shortfalls of "Obamacare". New England Journal of Medicine. 367:1479-1481.
http://www.britannica.com/event/Iraq-War
http://obamacarefacts.com/obamacare-individual-mandate/
http://www.supremecourt.gov/opinions/14pdf/14-114_qol1.pdf
https://www.youtube.com/watch?v=HoFqV3qVMGA
China: the collapse of the empire or a new fresh start?
For the past decades, the world has seen how the Chinese economy has succeeded to move
from an entirely agrarian society to an industrial force. The outcome of this reforms has
propelled the communist country to become the second largest economy in the world,
experiencing on average a 10% growth almost like a clockwork. Unfortunately, as all good
things must come to an end eventually, it appears that the so called ‘Chinese economic
5
miracle’ hype has reached a turning point in the history.
The growth deceleration experienced by China has happened after the government decided
to stimulate the economy by releasing the restrictions on the domestic stock market which has
created a bubble in the price of stocks and also due to over-investment mainly financed by
leverage and debt into unproductive sectors such as real estate.
The end of 2015 has culminated for the Chinese economy with a contraction in the
manufacturing sector which has resulted in prematurely closed markets due to volatility. The
decrease in growth from a two-digit to an average of 7% has produced a lot of uncertainty all
over the world, especially for the countries that were reliant on China to produce growth.
Unfortunately for China, the country appears to be the bad character in all this story which
seems to be driving the economic crash, but the real forces that drive it are actually the free
market capitalism and the austerity policies that are falling at a global rate. Most of the
content of the recent articles on this issue suggest that the Chinese economic slowdown is a
leading indicator of where the worldwide economies will be heading by the end of 2016, but
no one is taking into consideration the fact that maybe the modern industrial economy is
rather reacting to the slowdown from the US and Europe.
It is well known that the conversion from an agricultural economy to a modern power house
economy has been possible due to the large cheap labour force that China had so far and
also through foreign investments. This has allowed to the Chinese economy to built all this
capacity to produce exports for the rest of the world at a cheaper price and has accounted
for the largest share of its GDP. However, given the fact that the Chinese market has built its
productivity on capitalism, they didn’t consider that capitalism will always have a crash like the
financial crisis that stroked in 2008 and emerging markets such as Europe and America which
had been the main importer of Chinese goods have been seriously affected and therefore
their demand for imports has shrunk considerably, which has lead to the economic slowdown
that China is experiencing nowadays.
After this chain of events that has triggered the slowdown of the Chinese economy and after
an inefficient quantitative easing policy, the government is re-shaping its economy through
four adjustment processes: debt restructuring, economic restructuring – new model for the
economy, capital market challenge which has to be correctly built so that the capital
circulates through the economy and last but not least a balance of payments challenge which
is related to the depreciation in the currency that China is experiencing at this point in time. All
these policies are not something new for the rest of the world, especially for the US which has
been through all this transitions and therefore are manageable given the amount of resources
and buffers that the country has.
6
Through economic restructuring, the Chinese government hopes to stimulate the growth by
influencing the domestic consumption rather than relying on investment and exports to
account as a big share of GDP. It will be a difficult process for China to convert from a capex
economy to an opex economy where the consumer will be the one that is supposed to drive
the economy growth given the fact that the government won’t have any power to influence
the discretionary consumption.
However, at the beginning of 2016, this change made assets price to be revalued and
together with the interest rates raised by the FEDs and the behaviour of many emerging
markets that didn’t performed so well, this has caused a lot of volatility in the markets.
The volatility experienced through every foreign market it is also a problem of communication
that China has regarding the transitions that are happening there and which raised high levels
of uncertainty in the markets. A certain degree of volatility it is alright in an economy because it
is compatible with the market driven principles that the Chinese economy is adhering to and
an intervention would be expected just if the volatility becomes indeed a problem.
The events that are taking place in China should be rather considered some cyclical
adjustments that the economy is facing and it will most probably take between 2 and 3 years
for the country to see the results, but unfortunately for the rest of the world this comes at a time
when the foreign financial markets are still vulnerable after the aggressive financial crisis from
2008.
Many may question the ability of the Chinese government to go through these four adjustment
processes and to get to the desired outcome which has as target the growth of GDP. First of all,
China has managed to move from a poor country to the second largest economy of the world
in few decades and this can be treated as an accomplishment that no other country has done
in such a short time period. Secondly, this manufacturing industrialization has been possible due
to a strong leadership that China has. So, if the Chinese government prove to be as good as it
has been until now and maybe even more determined than the German government and also
communicative in what concerns the transitions that are happening, it will be expected that
the wheels of their economy will start to spin again.
The forecast for the second largest economy or the largest economy in the world if measured
in PPP terms is at two extremes. The first foresee that the economy will continue to slowdown
because of the volatility, whereas the other side foresees that the Chinese economy will
continue to grow because the economy doesn’t afford to let the growth rate to drop too
sharply because it will ignite a lot of financial problems. Even if the Chinese GDP has collapsed
at 6.9% for 2016, when looked at what the growth was for example six years ago with a 12% 13% growth rate, it is the same when compared with this year, so maybe
7
the question is about what a normal rate should be for the Chinese economy after the world
has been used to see a massive growth from this economy.
Norwich’s own Launch-pad for Growth
By Alexander Lindsay
“Jobs, jobs, jobs.” Good government should provide its citizens with both security and aspiration. The
security of a home, a job, protection from violence, security in sickness and old age. Aspiration means the
chance for self-improvement, family formation, material benefits and a home. Without the provision of jobs
government can only disappoint the expectations of the citizens.
What will be the jobs of the future? What skillsets should students develop today to be relevant to an
employer in ten or twenty years? Globalisation and digital technology have de-industrialized the old
western economies and pulled apart long-established business models. AI software (expert systems) will
take bread-and-butter business from many professionals, while straight-through processing has removed
the need for millions of middle office jobs. Entire industries have evaporated or moved offshore. Do
governments have sufficient vision and wisdom to address a tired economic system in which homes are
increasingly unaffordable, good jobs increasingly scarce and the citizens vividly aware of “the squeezed
middle” and intergenerational injustice?
Look no further than The Norwich Research Park! Here on our doorstep is an exceptionally well-founded
Launch-pad for Future Growth. Around 3,000 scientists, researchers and clinicians are driving innovation
for the 12,000 who currently work on site for over 60institutes and businesses. This is one of Europe’s
leading Life Science research centres and world leader in several aspects of Agri BioTech, Food and Health
and Industrial BioTech. The foundations of the excellence in plant science was laid by the John Innes Centre
together with The Sainsbury Laboratory, especially known for researching the interactions between plants
and pathogens. Add the intellectual resources of the UEA, the Norfolk and Norwich University Hospital NHS
Foundation Trust, the Institute of Food Research and BBSRC (Biotechnology and Biological Sciences
Research Council) who have just joined forces to create the new £75m Quadram Institute which will focus
on food-related disease and human health, and you can sense the potential of the site.
But a successful Launch-pad needs five elements to innovate successfully and create jobs for the future. A
critical mass of private and public intellectual capital, here represented by the Institutions and 3,000
researchers on site needs to be stabilised by (second) a solid long-term commitment by government to
support the aims of the enterprises on the park with a secure framework of grants, funding and tax
structures. Third is the crucial matter of shared visions by all participants, especially researchers who may
need to commit years of their lives into difficult projects and the sponsors who need to keep faith with them.
Fourth is the arrival of investment capital to develop the research into a commercial enterprise and to grow
the business until it is self-funding. Finally, the cluster must possess some unique sustainable advantage
strong enough to sustain future growth and to retain talent and young companies even as they evolve.
In the case of the Norwich Research Park, the early access to a gene sequencer on site has helped related
leading positions in bio-imaging and proteonomics. A number of projects are gaining insights into various
metabolic dysfunctions. A future cure for obesity is always the holy grail of research into the behaviour of
gut bacteria.
8
The five structural pre-requisites of a successful Launch-pad are in place. Additionally NRP supplies a
nurturing eco-system with schemes such as TimeBank, a supportive network of business, financial and legal
advisors who donate their time to aid the start-ups. Momentum continues to build with the creation of The
Quadram Institute (QI). This is an £82m investment with completion planned for 2018, with £26m from the
government hoping to seed more successes to follow Inspiralis, the provider of tools for research into
cancer and infectious diseases.
It is noticeable that the talent working on the NRP is very international and especially European. In BrexitBritain we need to make sure that the funding and support for far-sighted initiatives such as the NRP
continue without a pause and that international collaborations have no reason to suspect any softening in
our commitment. Jobs of the future can only come from initiatives such as this cluster of Life Science
research and this is where governments will grow the tax base of the future that will deliver security and
aspiration as we move through our careers.
UEA’s students should utilise this launch-pad by taking advantage of Bite Size Business seminars provided
by Norwich Business School and Innovation New Anglia workshops both of which offer support for people
looking to develop their business ideas / plans. I strongly encourage students to look at the following
programmes or attend the approaching events the soonest of which is October 21st.
Please check out the following links:
http://innovationnewanglia.com/about/
http://www.norwichresearchpark.com/newsandevents/events/bitesizebusinessseminarsleadershipandman
ag.aspx
The following first year written assignments were submitted in response to an assignment set by
Fabio Arico. The task was to review the following article:
“Prospects for the UK Economy”, Journal of the National Institute of Economic and Social
Research, Feb 2016, 235(1).
Describe the current situation of the UK economy and outline the expected consequences for
the near future.
The essay focuses on; expected change in fiscal policy, why the Central Bank might increase
the interest rate in the future and various other economic issues discussed in the article (e.g.
price and earnings, supply side, investment and savings, etc.
9
ECO-4002Y – Introductory Economics (Macroeconomics) Written Assignment
By George Markham
Word Count: 1780
Contents
1.
Introduction .................................................................................................................................. 11
2.
Expected change in Fiscal Policy in the UK ................................................................................... 11
3.
Consequences of contractionary Monetary Policy....................................................................... 14
4.
Fiscal Policy impact on the components of demand .................................................................... 16
5.
Conclusion .................................................................................................................................... 17
6.
Appendix ....................................................................................................................................... 17
7.
Bibliography .................................................................................................................................. 20
10
1.
Introduction
In this essay I will investigate three key points, the expected change in Fiscal Policy in the UK, the consequences of
contractionary Monetary Policy and the impact of Fiscal Policy on the components of demand. The current situation of the UK
economy and the prospects for the future are deliberated. I will present evidence from a variety of sources combined with my
own economic knowledge in order to justify claims made. Much of what is said in this essay is based around on information
from the article “Prospects for the UK Economy”, Journal of the National Institute of Economic and Social Research, Feb 2016.
2.
Expected change in Fiscal Policy in the
UK
The article “Prospects for the UK Economy” discusses in depth the implications of the recent Autumn statement set by
Chancellor George Osbourne. Having read the statement we can see that the Government are promoting expansionary fiscal
policy, with a £4 trillion spending budget being allocated over the next five years (HM Treasury 2016). The substantial increase in
the planned government consumption came as a surprise to many, relative to the summer budget (HM Treasury 2015). Despite
the advancement in fiscal expenditure the Government still expect to meet their primary fiscal target of an absolute surplus in
2019-20 (See figure1).
The reason for the loosening in fiscal policy is an attempt to stimulate economic growth continually as it has been doing since
the recession of 2008 (See figure 2). Increases in government spending lead to an increase in equilibrium income, shown in
figure 3 below.
Figure 3
11
Government spending has increased meaning that equilibrium income has increased from Y* to Y₁ .
The expansionary fiscal policy also effects the interest rate, Figure 4 below shows the effect of this. The IS curve shifts to the right
from IS to IS₁ , income rises from Y* to Y₁ and the interest rate also rises from r* to r₁ , thus showing some crowding out occurs.
Figure 4
12
The rightward shift in the IS curve which causes national income to rise also has an effect on aggregate demand, this shifts from
AD to AD₁ , which in turn increases the price level from P₁ to P₂ and thus there is inflationary consequences from expansionary
fiscal policy.
Judging whether the loosing fiscal policy has been a success or depends on a number of factors. It has improved the economy
as national income has increased (Y* to Y₁ ), however we have also seen that interest rates and the price level increase (r*to r₁
& P* to P₁ ), as well as the sum of money the government have actually spent (ΔG). Washington R (2010) maintained the view
that “in almost all cases, monetary stimulus should be the first option.” But went on to add it can still be useful and necessary to
use expansionary fiscal policy, however he felt it is most suitable for periods where expectations are not high and output is in a
sharp decline.
13
Conversely, Gosselin P (2015), claims that the fiscal austerity of the US is paying them “a big price in growth, jobs and wages”.
The Bloomberg writer claims that if the government were promoting enhanced loosening of fiscal policies then the US
economy would be growing at 3% rather than just over 2% and that an estimated 2.4million more citizens would be in
employment.
In conclusion it seems as though this may not be the best time for expansionary fiscal policies to be implemented by the UK
government, as Caggiano G et al (2014) states; the depth of a recession determines the effectiveness of increases in
government spending. He claims that increasing government spending when the economy is in an expansionary period, as we
are (figure 2), would have mild positive effects for at most a year and then negative effects would be generated upon output.
3.
Consequences of contractionary
Monetary Policy
The Monetary Policy Committee (MPC) of the Bank of England, are responsible for operating and conducting monetary policy
within the UK. The UK HM treasury set the objectives and inflationary targets for UK monetary policy, they also appoint the
members of the MPC. The purpose of this section is to investigate and access the consequences of a proposed increase in
interest rates.
One reason the MPC may want to alter interest rates, is in order to rectify inflation as it may be off target. If the government
want to reduce inflation and aggregate demand they may wish to deploy a tighter monetary policy in order to reach this. As
you can see in figure 5 below, the interest rates have been raised from r* to r₁ and thus Q has fallen from Q* to Q₁ , which
would lead to a fall in the price level.
The current consumer price index level of inflation stands at 0.3% (see figure 6 below), and the inflation target is 2% (Bank of
England 2016). As was forecasted by Monaghan A (2014) the UK economy is still in a recovery stage, and so an increase in
14
interest rates, which would lead to a fall in aggregate demand of a recovering economy, as well as keeping inflation down
when it is way off target, does not seem to be the correct policy the government should currently employ.
Conversely, there are several reasons the Government may want to increase interest rates. The MPC recently voted
unanimously to maintain the Bank Rate at 0.5% (Taborda J 2016), yet with the rate this low, it is not a desirable action for people
to save and so they won’t. The household saving ratio has actually “fallen to its lowest level for more than 50 years.”(The
Telegraph). Chan S.P (2014) claimed that raising interest rates now would help to “support and sustain” Britain’s recovery as well
as ensuring prices rise smoothly in the future. In addition to this, Sentence A (2016) said that the MPC are giving themselves no
room to manoeuvre with relaxing monetary policy if there were to be another global economic shock, he states that if interest
rates are kept low then “the only tool available to the MPC is quantitative easing, which has an uncertain impact on the
economy.”
The combined effect of both expansionary fiscal policy and contractionary monetary policy is shown in figure 7 above. The
fiscal loosing would see the IS curve shift from IS* to IS₁ and the contractionary monetary policy would see the interest rate rise
from r* to r₁ and thus the LM curve shifts from LM* to LM₁ . In the diagram, both the policies have nullified each other in terms of
GDP growth and it stays at Y*, yet the interest rate has still increased. However when put into real life practise we must consider
we do not necessarily know what the size of the shift in each of the curves will be, nor do we know the exact steepness of the
15
curves. The two polices are most effective when applied simultaneously, yet agreeing with Sentence A (2016) the government
want to increase interest rates as not doing so is exposing the UK to potential wider risks, but they also wish to keep the
economy growing throughout its recovery. To do this they have had to apply the two policies in a way that may see them
contradict each other in terms of economic growth, however it can see them reach the goal they desire..
4.
Fiscal Policy impact on the
components of demand
The components of demand in an economy consist of four elements; consumer spending(C), private investment (I),
government expenditure on goods and services (G) and expenditure on exports (X) minus the expenditure on imports (M). Thus
we have the equation AD=C+I+G+(X-M). The aggregate demand curve (shown in figure 4) displays how much national output
(GDP) will be demanded by the economy at any given price level. Within this section, the effects of expansionary fiscal policy
on the components of demand will be investigated.
The article “Prospects for the UK Economy” says that “recent revisions in the components of GDP left the 2014 annual average
unchanged, but led to a reduction in the rate of growth if growth in the second and third quarters of 2015”, this was primarily
due to consumer spending pattern changed through the year. The article goes on to add that domestic demand is expected
the primary driver of economic growth over the next couple of years, with consumption and investment expected to
contribute “2.3 and 1.0 percentage points towards overall growth in 2016, respectively.”
Expansionary fiscal policy as previously stated, leads to an increase in aggregate demand, this in-turn means that domestic
consumers and firms will have more income, consequently leading to consumption and investment increasing. Subsequent to
this larger income firms and consumers will now increase their spending on imports. Figure 8 above shows the effect of this,
increased government spending increases aggregate demand from AD* to AD₁ , however then we must consider the rise in
imports and so demand is at AD₂ , with national output having increased from Q* to Q₁ and the price level rising from P* to P₁ .
Palley T.I (2009) declares that government spending can have an even larger relative impact to tax cuts than was originally
thought. Furthermore, he states that increased government spending can have a smaller impact on the trade deficit than tax
cuts also. Thus showing us the expansionary policy the government has recently employed can be an attempted to reduce the
trade deficit, which stood at £4.14 Billion in October (BBC 2015). From this we can conclude that expansionary fiscal policy has
a positive impact on the components of demand as a whole, yet it can cause an increase in imports which will reduce
aggregate demand.
16
5.
Conclusion
In conclusion, there is sufficient evidence to show the potential outcome from the proposed expansionary fiscal policy, in
addition to this the effects of the possible forthcoming rise in interest rates are discussed, along with both the policies
combined. It seems the government must know what Sentence A (2016) claimed to be true in that interest rates need to rise so
that people can save again and even more essentially so that the MPC can promote expansionary monetary policy to assist
the economy if the is another economic shock. Although the main problem is finding the right balance of fiscal and monetary
policies combined so that the economy can still grow through this recovery.
What will happen in the future? As previously stated, the outcome of the Autumn statement came as a surprise to many, so
predicting the future prospects for the UK economy is very problematic as we do not know what the outcome of the imminent
budget statement will be.
6.
Appendix
Figure 1.
http://www.ifs.org.uk/uploads/gb/gb2016/presentations/gb2016_carl.pdf
Figure 2.
17
http://www.bbc.co.uk/news/10613201
Figure 3.
Figure 4.
18
Figure 5.
Figure 6.
http://www.tradingeconomics.com/united-kingdom/inflation-cpi
19
Figure 7.
Figure 8.
7.
Bibliography
BBC (2015) UK trade deficit widens in October. Available at: http://www.bbc.co.uk/news/business35058982 (Accessed: 27 February
2016).
Bank of England (2016) Monetary policy framework. Available at:
http://www.bankofengland.co.uk/monetarypolicy/pages/framework/framework.aspx (Accessed: 26 February 2016).
Caggiano Et al, G. (2014) Estimating fiscal Multipliers: News from a Nonlinear world !. Available at:
http://economia.unipd.it/sites/decon.unipd.it/files/20140179.pdf (Accessed: 29 February 2016).
20
Chan, S.P. (2014) Ian McCafferty: Four reasons why interest rates must rise now. Available at:
http://www.telegraph.co.uk/finance/personalfinance/interest-rates/11284303/Ian-McCafferty-four-reasonswhy-interest-rates-shouldrise-now.html (Accessed: 2 March 2016).
Gosselin, P. (2015) Here’s how you add 2.4 Million jobs to the economy. Available at:
http://www.bloomberg.com/news/articles/2015-05-28/government-austerity-exacts-toll-on-u-s-jobs-wagesand-growth (Accessed: 29
February 2016).
HM Treasury (2016) Spending review and autumn statement 2015: Key announcements. Available at:
https://www.gov.uk/government/news/spending-review-and-autumn-statement-2015-key-announcements (Accessed: 23 February
2016).
HM Treasury and The Rt Hon George Osborne (2015) Summer Budget 2015: Key announcements.
Available at: https://www.gov.uk/government/news/summer-budget-2015-key-announcements (Accessed: 23 February 2016).
Monaghan, A. (2014) UK economic recovery to continue into 2016, forecasts OECD. Available at:
http://www.theguardian.com/business/2014/nov/25/uk-economic-recovery-continue-2016-oecd (Accessed: 26 February 2016).
Palley, T.I. (2009) ‘Imports and the income-expenditure model: Implications for fiscal policy and recession fighting’, Journal of Post
Keynesian Economics, 32(2), pp. 311–322. doi: 10.2753/pke0160-3477320211.
Sentance, A. (2016) Six reasons why now is the right time to raise interest rates. Available at:
http://www.telegraph.co.uk/finance/economics/12102397/Six-reasons-why-now-is-the-right-time-to-raiseinterest-rates.html
(Accessed: 26 February 2016).
Taborda, J. (2016) United Kingdom interest rate | 1971-2016 | data | chart | calendar. Available at:
http://www.tradingeconomics.com/united-kingdom/interest-rate (Accessed: 26 February 2016).
The Telegraph (2016) Six reasons why now is the right time to raise interest rates. Available at:
http://www.telegraph.co.uk/finance/economics/12102397/Six-reasons-why-now-is-the-right-time-to-raiseinterest-rates.html
(Accessed: 2 March 2016).
Washington, R. (2010) When does fiscal stimulus work?. Available at:
http://www.economist.com/blogs/freeexchange/2010/07/fiscal_policy (Accessed: 29 February 2016).
21
ECO-4002Y – Introductory Economics (Macroeconomics) Written Assignment
By John Kahodi
Introduction
The aim of this essay is to outline expected changes to key economic issues affecting the UK’s economy in the future. This will
be conducted by identifying policy implementations from “Prospects for the UK Economy” a report by Kirby et al (2016),
explaining their potential outcome that will be supported by the use of economic theories and empirical evidence.
1.0 Fiscal policy forecast
According to the prospects for the UK Economy report, there are expectations that there will be discretionary fiscal expansions
within the economy as plans have been made to increase government consumption of goods and services produced in the
economy such as the expenditure on health care and education. There is also prospects that capital spending, also known as
government investment i.e. building new roads/hospitals, is forecasted to increase significantly within the same period. This is
also the case for welfare spending which is total spending on social protection. The plan to conduct expansionary fiscal policy
has been able to occur due to lower public sector net borrowing that will eventually lead into a surplus in 2019-20 as
forecasted by the Office of Budget Responsibility, in aid from increases in tax receipts to boost revenues which was introduced
in the
Public sector net borrowing
(% GDP)
2015-16
2016-17
2017-18
2018-19
2019-20
2020-21
Summer Budget 2015
3.7
2.2
1.2
0.3
-0.4
-0.5
Autumn Statement 2015.
1.1 Effect of expansionary fiscal policy
According to Feldstein, M (2002) discretionary fiscal policy can be effective in stimulating the economy when there has been a
sustained downturn from aggregate demand while interest rates are low and when prices are falling or expected to fall. These
conditions stated by Feldstein outlines the current economic state the UK is as they progressively continue to recover from the
2008 financial crisis while interest rates have been held at a record low of
22
0.5% since March 2009 (Bank of England). Furthermore the price level in the UK has remained weak with uncertainty over future
prices largely due to decreases in global commodity prices, uncertainty around world output and trade growth as well
movements in the sterling as reported in the prospects for the UK economy (Kirby et al, 2016). As the government increases their
expenditure, they buy more goods and services in the economy. Government spending is a component of Aggregate
demand (AD=C+I+G+X-M) therefore as government spending increases, aggregate demand also increases and shifts to the
right. National income increases from Y1 to Y2 with demand pull inflationary pressures from P1 to P2 as shown in figure 2.
Investment in the economy increases as business confidences rises from more consumption of goods and services. The IS-LM
model can be used to demonstrate this effect as the IS shifts to the right resulting in the interest rate to increase from r 1 to r2 as
demonstrated in figure 1.
r
Figure 1
LM
r* 2
r* 1
IS 2
IS 1
Y
Price Level
Figure 2
AS
P *2
P *1
AD 2
Y*1 Y*2 National Income
AD 1
23
As government expenditure is an injection, this will stimulate further spending in the UK because of the multiplier effect when
we consider the income/expenditure equation where
.
W, J
Figure 3
W
J2
J1
Figure 4
Y=E
E2
E *2
E1
∆G
E*1
Y* 1
Y* 2
The decision to implement an expansionary fiscal policy where it involves changing the government expenditure can be
shown to have a greater effect than changing taxes. This is as when we consider the multiplier of changing government
spending to the tax multiplier in changing taxes, the tax multiplier (kt) will only partially multiplies the change in taxes in the
economy as the tax multiple is kt
where 0<c<1. Whereas if we look at the multiplier (k) which is k =
to a
change in government expenditure, the multiplier fully multiplies the increase in government spending. The decision by the
government to implement greater expenditure may however cause a crowding out effect where the change in government
24
spending may lead to a rise in total injections that is smaller than the initial change in government expenditure. This is as
resources may be taken away from the private sector as the government consumes it. Furthermore with higher interest rates, as
shown in figure 1, this makes it more expensive for investors to take out loans.
2.0 Monetary policy
The last increase in the interest rate by the monetary policy committee occurred in July 2007 (Bank of England). It had been
expected that the bank rate would rise at the turn of the year according to the Bank of England governor, Mark Carney (2015).
However due to falling oil prices and growing uncertainty to world trade and output has resulted in the MPC deciding to leave
the interest rate unchanged at 0.5%. There are expectations of a gradual rise in the bank rate to now occur in August 2016
(Kirby et al, 2016).
The report states that the bank rate is expected to rise to 1% by the end of 2016, while gradually increasing to 1.5% by the end
of 2017. Contractionary monetary policy tightens the money supply and slows down economic activity by increasing interest
rates to reduce borrowing from banks, firms and individuals. The Bank of England may conduct this by setting the interest rate
at a higher rate and then manipulating the money supply to meet the new equilibrium, as they have greater control over the
interest rate. They can also contract the money supply in the economy if they have more control over the quantity of money
than the interest rate. Either method results in a shift to the left of the money supply in the money market diagram, with interest
rates increasing as shown in figure 3. The effect on aggregate demand is that as interest rate rises, this encourages more
savings in the economy, reducing the level of consumer spending. This may be a reason why the interest rate is expected to
increase as since the financial crisis, household savings has declined.
Household saving ratio:
Source: ONS
Consumer spending is also reduced from an increase in mortgage interest payments, as households will be left with less
discretionary income. According to P. Bunn et al (2015) despite higher interest rates amounting to more financial pressure to
some households, they are in a better situation than previously to manage an increase in interest rates. The NMG Consulting
survey (2015) suggests if interest rates were raised, borrowers would cut back on their spending than what savers would raise
from their savings for each extra pound of interest. Furthermore investment in the economy would also fall as interest rate rises
as it becomes more expensive for business to take out loans. The overall effect of this change in aggregate demand on
national output and the price level is shown in figure 4 where national output falls from Y1 to Y2 and the price level falls from P1
to P2.
25
r
Figure 5
MS
MS
r* 2
r* 1
L
Price Level
LM
P* 1
Figure 6
P* 2
AD 2
Y*2 Y*1
AD 1
National Income
We can use the IS-LM model to observe the effects of both fiscal and monetary policy simultaneously on the goods and
financial markets. An increase in government expenditure shifts the IS curve to the right increasing the national income and the
interest rate as demonstrated previously in figure
1 (section 1.1). Interest rates will rise from the effects of both fiscal expansions and contractionary monetary policy as shown
below in figure 7 with the short term outcome on national income unclear. The effect on national income depends on the
magnitude of the policy implementations. The slope of both the LM and IS curve also needs to be considered as the fiscal
policy could become more effective if the LM curve is relatively flat while the IS curve is relatively steep. Monetary policy would
be more effective if the slope of the LM curve is relatively steep and the IS curve relatively flat. A further consequence of
increasing the interest rate is that it could also have an effect on the Exchange Rate. This is as UK’s bank rate relatively
becomes more attractive for foreign depositors therefore the demand for the pound increases causing the exchange rate to
appreciate.
26
3.0 Labour Market
The economy is recovering from a labour market slack since the financial crisis which is when there are more workers than
actual jobs. Unemployment since the 2011, which was at its peak of 2.7 million people (BBC), has gradually fallen since and
fallen to its lowest record since 2005 to 5.1% during 2015. Records from the ONS show that employment rate was 73.7% in 2015,
“highest employment rate since comparable records began 1971”. There has been research into whether fiscal expansions
increases employment while reducing unemployment. According to E. Bova et al. (2014) discretionary government
expenditure on goods and services has the largest affect in stimulating employment. This is as the expenditure directly
increases aggregate demand from consumptions of goods and services in the economy therefore more labour are required to
produce more goods and services demanded as labour is a derived demand. In particular if the government increases their
spending on the public sector as more employment may occur in the public sector as in 2013 according to the ONS the
percentage of people in employment who worked in the public sector was the lowest since the records began. According to
Kirby et al (2016) investment has also remained strong in the UK economy increasing by 5.8% which may also derive more
demand for labour. This can be shown in the labour market diagram with a shift to the right of labour demand as shown in
figure 9. This increase in labour demand has reduced the labour market slack significantly (S. Berry et al 2015). The supply of
labour is expected to also increase due to population growth projected to increase by 9.7 million in the next 25 years largely
due to an increase in net migration as well as the birth rate expected to exceed the death rate (Kirby et al. 2016). The effect of
this is that the supply of labour also increases, shifting to the right as shown in figure 8.
27
AS L 1
Average
AS L 2
Figure 8
Wage Rate
W* 1
AD L 2
AD L 1
N*1
N*2
No. of Workers
The diagram above shows the increase in the number of workers from the labour market which has continued to grow since
the financial crisis as employment in the UK continues to rise as shown in the graph below.
Source: ONS – Labour Force Survey
Kirby et al (2016) also state that standard living will improve in the future as real wages are currently at the same level they were
in 2005. Therefore as the price level rises, workers are expected to negotiate with firms for higher wages which would see wages
rise in line with expected inflation so that workers maintain their real income. Wages should therefore rise towards equilibrium in
the long run as shown in figure 9.
28
Average
AS L
N
Figure 9
Wage Rate
Equilibrium Unemployment
W* e
W
AD L
QS L
N* e
QD L
No. of Workers
Disequilibrium Unemployment
Conclusion
Fiscal expansion is expected to have a significant role in increasing the national’s output as the UK continues to recover from
the financial crisis which in effect is positive towards improving the labour market as the labour market grows towards the same
level before the crisis occurred. This may come with some constraint from contractionary monetary policy which would slow
down growth to prevent inflationary pressures from occurring. However this constraint will not deter the UK from increasing their
national output as they are expected to continually increase GDP at a sustainable rate with inflation increasing towards the
inflation target of 2%.
29
Bibliography
S. Berry et al, 2015. Bank of England: Trends in UK labour supply
E. Bova et al, 2014. A Fiscal Job? An Analysis of Fiscal Policy and the Labor Market.
P. Bunn et al, 2015. ‘The potential impact of higher interest rates and further fiscal consolidation on households: evidence from
the 2015 NMG Consulting survey’, Bank of England Quarterly Bulletin, 55, (4) p.357–368
M. Feldstein, 2002. The Role for Discretionary Fiscal Policy in a Low Interest Rate Environment. NBER Working Paper No. 9203.
(Online) Available at:
http://www.nber.org/papers/w9203 (Accessed 9 March 2016)
J. Sloman, (2012). Introductory Economics – Macroeconomics. (8) Pearson (Compiled by F. Aricò), p.32-264.
BBC, 2015. Bank votes 8-1 to keep UK rates at record low. (Online) Available at:
http://www.bankofengland.co.uk/boeapps/iadb/repo.asp. (Accessed 9 March 2016).
BBC, 2015. Economy tracker: Unemployment. (Online) Available at: http://www.bbc.co.uk/news/10604117. (Accessed 9 March
2016).
ONS, 2015. National Accounts articles: The Saving Ratio: How is it affected by Households’ and Non Profit Institutions Serving
Households’ income and expenditure? (Online) Available at:
http://www.ons.gov.uk/economy/nationalaccounts/uksectoraccounts/articles/n ationalaccountsarticles/2015-07-01.
(Accessed 9 March 2016).
ONS, 2016. Statistical Interactive Database - official Bank Rate history.
(Online) Available at: http://www.bankofengland.co.uk/boeapps/iadb/repo.asp. (Accessed 9 March 2016).
ONS, 2015. Participation rates in the UK - 2014. (Online) Available at:
http://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employme
ntandemployeetypes/compendium/participationratesintheuklabourmarket/201 5-0319/participationratesintheuk20141overview. (Accessed 9 March 2016).
30
ECO-4002Y – Introductory Economics (Macroeconomics) Written Assignment
By Wokciech Serwacki
Introduction
The UK and other economies have experienced various changes, such as a fall in economic growth and a rise in
unemployment, following the Great Recession in 2008. Although global growth has remained subdued, many empirical sources
show that the UK economy is amongst the fastest recovering and hence quickest growing nations. The government and the
central bank have been responsible in recovering the volatile and unbalanced economy. The nature of the UK economy has
challenged the government and the central bank to use Fiscal and Monetary Policy to stimulate demand, which has suffered
due to the repercussions of the recession. There has been political and economic debate regarding the future outlook of the
domestic economy and I aim to discuss the impact of policies used for demand management. In this essay, I begin by
identifying the expected changes to fiscal policy used in the forthcoming years and predict the influence this may have on
various macroeconomic indicators such as: interest rates, price level and GDP. I then extend my discussion by looking at a
potentially deliberate decision by Bank of England to increase the interest rate in the economy, establish the links between the
Fiscal and Monetary policies and explore the combined effects of their implementation. I finalise my discussion by exploring the
impact of future fiscal policy in more detail, specifically referring to consequences on the UK labour market.
Expected Fiscal Policy
Current position of the economy in the business cycle and many other economic related factors are likely to influence the
change in fiscal stance in the UK. The unemployment levels have fallen to 5.4% as of 2015, and labour productivity has reached
its highest in 2015 yet still remains 13% below pre-downturn (ONS, 2016) showing positive outlook for the economy. However, as
the nation approaches the EU referendum, the economy is likely to experience an increase in uncertainty, lack of business
confidence and hence delayed investment by firms. In the case of an EU exit, the high uncertainty and lack of confidence is
likely to remain in the short term negotiation period (Kirby et al., 2016). The pessimism regarding investment among economic
agents planning to invest is likely to have an impact on the expenditure and the national income.
45 o Equilibrium line
AE (I 0)
AE0
AE (I 1)
( Such that I
0 > I 1)
AE1
Y1
Y0
The delayed investment effects are modelled on the aggregate expenditure graph above. As expenditure is a function of
investment: E(I), a fall in investment would cause a fall in aggregate expenditure an hence a fall in national income from y 0 to
y1.
Since the Autumn Statement in 2015, the government has announced its spending review. Within the autumn statement, the
government outlined the loosening of its fiscal policy (Kirby et al., 2016). This is evident because as suggested in the Autumn
31
Statement, the government has planned to allocate over £4 trillion total expenditure to be implemented until 2020 (GOV,
2015). The spending changes have been outlined to help the government prioritise its spending through investment on
healthcare, education and housing sectors. Although, based on the changes outlined in the Autumn Statement 2015,
discretionary taxation has significantly increased, which contributes to the rising revenue collected from taxes. Nominal
government plan for consumption, capital and welfare expenditure are predicted to increase by a greater amount (Kirby et
al., 2016). Hence, the fiscal loosening introduced is expected to generate a net expansionary effect on the economy. The
diagram below models the forecasted changes.
r
LM
r2
r1
IS2
IS1
Y1
Y2
Y 1A
Y
The diagram shows the impact of expected expansionary fiscal policy. Originally the economy is at equilibrium interest rate r 1
and income Y1. As the economy experiences an injection of fiscal expenditure, the injections component (J) increase and
hence the national income increases. As a result, the IS curve shifts outwards such that new equilibrium is formed between the
goods and the money market at r2 and Y2. However, the economy may experience a crowding out effect. Rising effects
triggered through expansionary fiscal policy can have effects of absorbing the UK’s scope for lending and therefore
discourage firms from investing in capital projects. As the interest rate rises, borrowing costs increase and investments may fall.
This is demonstrated by a fall in national income from Y1a to Y2.
There are also other opportunity costs associated with expansionary fiscal policy through increases in government (fiscal)
expenditures. As fiscal expenditure is a component of aggregate demand, an expected rise in fiscal expenditure would
increase the aggregate demand. Ceteris Paribus, one could observe a trade-off between a rising aggregate demand and
demand pull inflationary pressures and hence a rise in the price level of the economy. This effect is modelled on the AD-AS
diagram. When fiscal expenditure rises AD1 shifts to AD2 and the price level increases from P1 to P2.
32
P
AS
P2
P1
AD2
AD1
Y1
Y2
Y
Interest rates
There has been speculation over decisions to be taken by the Monetary Policy Committee to alter the official interest rate,
otherwise regarded as the Bank rate. Since the 5th Mach 2009 the official Bank rate was changed from 1% from the previous
month to 0.5% (Bank of England, 2016). The interest rate has remained at 0.5% until this current day. It is suggested, the Bank
rate is expected to remain at 0.5% until August 2016 but expected to rise to 1% by the end of 2016. Further rises are anticipated
in the medium term before 2.75% is reached by 2020. (Kirby et al., 2016). The MPC have taken the decision to delay the rise
upon the complicated global environment such as falling commodity costs, and Chinese stock market turbulence. The
domestic conditions, such as UK experiencing deflation in the last quarter of 2015, have downgraded short term outlook
prospects for the economy. However, if the Bank of England are to raise the interest rate it may be when the UK economy
absorbs its spare capacity. Further reasons for the increase, are likely to stem from “recovering oil price, a move of inflation
closer towards the 2% official target and the right global economic environment” (Mark Carney, January Speech, 2016).
Hence, in order for the MPC to meets its mandated target interest rates, Bank of England would rise the interest rate to avoid
“overshooting its mandated ‘inflation’ target” (Kirby et al., 2016). Increasing the interest rate may arise from a controlled
movement of money supply or a direct control over the bank rate which adheres to the rule of inflation-targeting used by the
central bank. The diagram below shows the consequential effects of a rise in interest rates through keeping the banking system
short of liquidity. Liquidity shortages will force commercial banks to get liquidity from the BofE through gilt repos. As the central
bank has control over gilt repo rate charged, a rise in gilt repo rate may have a domino effect on the other interest rates in the
econ
1
Q of gilts
Gilt
established on the money market.
Source: Economics, Sloman, p.207, 2016
Alternatively, the central bank can sets an interest rate r1 and with time the Ms is moved
to reach an equilibrium on the money market. As a consequence of a rise in the
interest rate, money supply would contract and a new equilibrium would be
repo rate
r2
D2 by central bank
r1
D1 by central bank
Source:
Economics,
Sloman,
p.207, 2016
S of gilts by bank
Q2
Q
33
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
Ms1
r
34
Ms0
r1
r0
L
Ms,L
Combined effect of policy intervention
Rather than stimulating demand by manipulating the fiscal and monetary policies individually, the two policies can be used
simultaneously to obtain the greatest effect. Hence, the simultaneous use of both of the demand management policies will
bring combined effects as modelled below.
r
LM2
LM1
r2
r1
IS2
IS1
Y1
Y
As discussed previously in the essay, expansionary fiscal policy will have an impact on the goods market and hence on the IS
curve such that it would shifts from IS1 to IS2. In contrast, a rise in the rate of interest will have an impact on the money market
such that the LM curve would shift upwards from LM1 to LM2. The overall combined effects would increase the rate of interest,
yet the effects on national income would be ambiguous as it would depend on factors such as: the effectiveness of the two
policies and relative elasticises of the IS and LM curves. For instance, the rise in the interest rate may have a negative impact
on the decisions taken by domestic investors however, a rise in the UK interest rate may attract foreign direct investment (FDI).
Although the loosening of fiscal policy can create powerful effects, the government cannot use it indefinitely due to
unsustainable accumulation of budget deficit. The quantity of effects are also dependent on the expectations of the
economic agents in the economy.
34
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
35
Fiscal Policy effects on labour market
Expansionary fiscal policy conducted through an increase in fiscal expenditure could impact the labour market. Although
unemployment figures have fallen 5.4%, spare capacity still remains in the economy and it is projected that the claimant rate
and ILO rate are forecasted to fall further to 5.0% (Kirby et al., 2016). £23 billion of government spending will be invested into
schools and although unemployment may fall further in the short term, in the long term labour productivity is expected to rise.
With rising labour productivity, the UK economy may experience a reduction of cyclical unemployment due to a rise in
aggregate demand and a rise in labour productivity as shown in the table below. Disequilibrium unemployment would shrink
(ceteris paribus) due to a rise in the aggregate demand for labour. This is likely to contribute to a further rise in national income
and improve long term UK economic growth.
Source: (UK Labour productivity, OECD, 2016)
Conclusion
To conclude, the UK economy is expected to undergo various macroeconomic changes such as growing employment and
rising national income. In this essay, I demonstrated that expansionary fiscal policy would help increase the national income
and the interest rate in the economy, at the expense of a rise in the general price level. Contractionary monetary policy would
affect these macroeconomic variables to a greater extent. Theory dictates, ambiguous effect on the national income as a
result of a combined effect of expansionary fiscal policy and contractionary monetary policy. This may be dependent on the
transition of the UK economy in the future and policies used by other institutions such as the European Central Bank because
the changes in UK’s policies is highly dependent on the conditions of the global economy.
35
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
36
References
Bankofengland.co.uk, (2016). Bank of England Statistical Interactive Database | Interest & Exchange
Rates | Official Bank Rate History. [online] Available at: http://www.bankofengland.co.uk/boeapps/iadb/repo.asp [Accessed
1 Mar. 2016].
Data.oecd.org, (2016). Productivity - Labour productivity forecast - OECD Data. [online] Available at:
https://data.oecd.org/lprdty/labour-productivity-forecast.htm [Accessed 1 Mar. 2016].
Data.worldbank.org, (2016). United Kingdom | Data. [online] Available at:
http://data.worldbank.org/country/united-kingdom [Accessed 1 Mar. 2016].
Gov.uk, (2015). Spending Review and Autumn Statement 2015: key announcements - News stories - GOV.UK. [online] Available at:
https://www.gov.uk/government/news/spending-review-andautumn-statement-2015-key-announcements [Accessed 1
Mar. 2016].
Kirby, S., Carreras, O., Meaning, J., Piggott, R. and Warren, J. (2016). Prospects for the UK
Economy. National Institute Economic Review, 235(1), pp.F47-F76.
Ons.gov.uk, (2016). Labour productivity- Office for National Statistics. [online] Available at:
https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/labourproductivity [Accessed 1 Mar. 2016].
Sloman, J. (2012). Economics with MyEconLab Access Card. Harlow: Pearson Education Ltd., p.207
(compiled version).
36
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
37
ECO-4002Y – Introductory Economics (Macroeconomics) Written Assignment
By Alex Lovett
In November 2015 the government released the spending review of government plans for the next five years that aims to build
on the stable growth the UK has obtained since the 2008 recession. GDP annual growth has stayed between 2% and 3% since
2011 (Source: Data.WorldBank) and the governments plans clearly aim to boost the economy further by allocating £4 trillion to
government spending over the next five years whilst forecasting a £20 billion surplus by 2020 (Source: Gov. HM treasury 2015
spending review). In this essay I aim to explain the impact this fiscal loosening and also the contractionary monetary policy will
have on the economy and also specifically the labour market.
Fiscal Policy
The Autumn statement following the summer budget presented a looser fiscal stance than what was described in the coalition
plans. Primarily by a significant increase in government expenditure. This is used to stimulate growth in the economy if there is
any spare capacity available. The decision to prioritise government expenditure may be because it is more effective at
generating output than a decrease in taxation would. This is due to the difference in their multipliers.
Taxation multiplier
Government expenditure multiplier
c
1
1-c(1-t)
1-c(1-t)
c * kt = k
We know that ‘c’ is the marginal propensity to consume and therefore, it has to be a figure less than one. This means that the
tax multiplier (kt) has to be less than our government expenditure multiplier. This is because reducing taxation only increases a
consumer’s disposable income, it does not guarantee that this additional money will be injected into the economy as they
may choose to save it instead. Where as government expenditure is a direct injection into the economy thus is more efficient
at generating output. Valerie Ramey (2011, 683) estimates the government expenditure multiplier to be between 0.8 and 1.5. It
is also in the governments interest to prioritise one method to another as drastic changes to both may lead to a significant
budget deficit.
37
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
38
Government expenditure is a component of aggregate demand, so any increase in
‘G’ will translate to a proportional increase in ‘AD’. Aggregate demand would shift to
the right resulting in an increase in output from Y1 to Y2 and an increase demand pull
inflation from P1 to P2. As shown in the diagram to the right. This is based on the
assumption that there is spare capacity available in the economy otherwise the
economy would overheat and could generate hyper-inflation.
The effect of this fiscal policy can also be shown on an IS LM diagram. The
expansionary fiscal policy would shift IS to the right. This would result in an increase in output from Y1 to Y2 and a rise in the
interest rate from R1 to R2. However, this may lead to a drawback effect
also called the crowding out effect. This is the case where an increase in government
spending results in a smaller relative increase in GDP, in other words when the
multiplier is less than one. This means that the increase in government expenditure has
crowded out other private spending such as investment or consumer spending
(Burton Zwick, 1974, 563-564). In order to increase government consumption, they will
need to borrow money from the banks. This would drive up money demand and
increase interest rates discouraging investment and actually causing a decrease in
AD. So the actual effect of expansionary fiscal policy is hard to determine however
despite this it is generally believed to result in a net increase in output.
Interest rates
The prospects for the UK economy report states that the bank interest rate is expected to be gradually increased to 1.5% in
2017 and further to 2.75% by the end of 2020. The reason for this projected rise is because the UK economy is forecasted to
have utilised all available spare capacity at some point in 2016. Thus, any further growth would lead to demand-pull inflation.
38
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
39
This diagram shows this demand-pull inflation. At AD1 the economy is at full capacity or at the point of full employment. If the
economy were to expand further to AD2, it would place upward pressure on the price level and raise It from P1 to P2, while
achieving no subsequent growth in Y. The aim of this monetary policy is to
control inflation to help achieve stable growth.
The interest rate increase is an example of contractionary monetary policy.
The effects of this policy can be explained by the Interest rate transmission
theory and the Exchange rate transmission mechanism.

The interest rate mechanism describes the relationship between a change in money supply and changes in
investment. To raise the rate of interest the bank would increase the liquidity ratio and reduce money supply causing
MS1 to shift left to MS2, leading to a shortfall of money supply which increases the interest rate from R1 to R2. This
means it is costlier to borrow and lowers business confidence and investment will decrease from I2 to I1. As investment
is a component of injections it will also lead to a downward shift from J1 to J2, resulting in a fall in output from Y1 to Y2.
This process is shown in the diagrams below.

The exchange rate mechanism shows how the change in money supply impacts an economy’s balance of trade. So
instead when money supply is reduced and interest rates increase we look at the response by financial investors.

Relatively higher interest rates incentivise foreign investors to hold their money or to invest in the UK. To do so they will buy
sterling, increasing demand (D1 to D2). It will also provide a disincentive for domestic investors to invest abroad, so they will sell
less pounds and reduce supply (S1 to S2). A shift to the left of supply and to the right of demand will cause the currency to
39
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
40
appreciate (E1 to E2), increasing the price of exports and cheapening imports. This will increase withdrawals (W1 to W2) and
lower injections (J1 to J2) leading to an overall fall in output. As shown below.
The effectiveness of monetary policy depends on the elasticity of the curves involved in these mechanisms. The elasticity of the
money demand (L) curve describes the responsiveness of money demand to changes in the interest rate. A relatively inelastic
curve would have a stronger impact on investment than a more elastic one. The worst case scenario is a perfectly elastic
curve which describes a liquidity trap. This is where interest rates are believed to be at the lowest possible rate which renders
monetary policy entirely ineffective. Business confidence is also a large factor of the effectiveness of monetary policy. This
represented by the elasticity of the investment curve. An inelastic curve shows low confidence and vice versa. If expectations
are low or unclear the curve may shift left and if they are positive then it may shift right, which makes it difficult to predict as
expectations are sensitive. This may be the case this year with the EU referendum, many investors may hold off investment until
they know what the result is going to be.
The combined effect of both the expansionary fiscal policy and contractionary monetary policy greatly depends on the
elasticity of the IS LM curves, as the elasticity determines how effective the policy actually is. This is where it becomes subjective
40
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
41
and different economists believe in different ideas. The Keynesian approach
is shown on the diagram to the right. Keynesians assume that the LM curve is
strongly elastic and the IS curve is relatively steep. This means that cash and
bonds are very good substitutes and that investment isn’t very responsive to
changes in the interest rate. If investment is less responsive to interest rates,
crowding out will have less of an impact. In this scenario the impact would
be an increase in GDP from Y1 to Y2 and and increase in R from R1 to R2.
On the other hand, monetarists believe LM is actually inelastic and IS is elastic. So that investment is sensitive to a change in the
interest rate. The effect is shown in this diagram. The policies would have an outcome of a fall in GDP from Y1 to Y2 and a rise in
the interest rate from R1 to R2. For Keynesians fiscal policy is more effective and hence the outcome is an increase in Y.
Monetarists believe monetary policy is more effective hence the result is a fall in Y. The monetarists approach would lead to a
fall in AD and a decrease in the price level while Keynesians would forecast
an increase in AD and hence a rise in the price level.
However, David W. Findlay (1999, 380) concluded his journal by saying that
the relationship between the slopes of these curves and the effectiveness of
macro policy is weak. So these outcomes are not certain.
41
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
42
Labour Market
The UK Prospects report also talks about the condition of the labour market. It
describes increasing employment as it reaches a record high of 74.2%
however increasing wages portray a tightening labour market. Labour market
tightness is described as an imbalance between the demand and supply of
labour, a tight market being an excess of labour demand, which increases
labour costs and thus wages (Brigden and Thomas, 2003, 7). This can be
explained using a labour market diagram. The wage rate starts at W1 where
there is an excess of demand of N1 –N2. Firms increase their wage rates so that more people are willing and able to work and
the supply of labour increases until it tends to the equilibrium N*. At this equilibrium we have equilibrium unemployment of N3 N*. This presents the market tightness and rising wage rates described in the report.
The impact of expansionary fiscal policy would follow a similar pattern. An
increase in government expenditure would lead to an increase in national
income, this would increase disposable income and therefore consumer
spending. Firms would react by producing more which requires more labour.
This would shift ADL to the right and create the labour market imbalance
once more. Firms would then raise wage rates to encourage more agents to
work and this would continue until the market reaches equilibrium once more.
The overall result is an increase in wage but also a fall in equilibrium
unemployment, which has fallen to N4-N*2.
The impact of the fiscal expansion may have a more effective impact on the labour market if it actually targeted reducing
unemployment. This way there would be a much more direct effect on the labour market as opposed to the effect of an
increase in aggregate demand and income. This could include policies to improve the efficiency of the labour force. These
could include policies to help trained workers find jobs by investing more in job centres or by training the workforce and
investing in education to enhance the skills of the workforce. These policies would shift the ASL curve to the right. Creating an
excess in the supply of labour. Firms would be able to lower wages as unemployed workers would be willing to work on a lower
42
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
43
wage and this would continue until the market met its new equilibrium. At this equilibrium wage rate would fall from W1 to W2
but the no. of workers would increase from N1 to N2. This shift would
reduce equilibrium unemployment as shown in the diagram.
The aim of this essay was to explain the impact of the fiscal and
monetary policy imposed by the government will have on the
economy. In summary, the expected impact will be stable growth,
efficient government spending should stimulate growth in the
economy and the rise in interest rates should control inflation.
However as previously discussed the EU referendum has clouded
the future of the UK economy, it will be difficult to make accurate
predictions until the UK’s decision is revealed.
Bibliography

Brigden and Thomas. Bank of England. (2003). What does economic theory tell us about labor market tightness? Page
7

Findlay. (1999). The IS-LM Model: Is There a Connection between Slopes and the Effectiveness of Fiscal and Monetary
Policy? Journal of Economic Education. Page 380.

http://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG. Accessed 17/03/16

https://www.gov.uk/government/news/spending-review-and-autumn-statement-2015-key-announcements. Accessed 17/03/16

Kirby, Carreras, Meaning, Piggott and Warren. (2016). Prospects for the UK Economy.

Ramey. (2011). Can government purchases stimulate the economy? Journal of Economic literature. Page 683.

Zwick. (1974). "Snapback" and "Crowding-out" Effects in Monetary and Fiscal Policy: Explanation and Interrelation:
Comment. Journal of Money, Credit and Banking. Pages 563-564.
43
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
44
What influences trade union membership? A microeconomic analysis
By Alice Hebditch
Introduction
I will be investigating what individual variables influence the demand for trade union membership. It has been reported that
25% of all employees are members of a union (Department for business, innovation and skills, 2014), what sort of person still joins
them?
Usually the firm will pursue objectives, such as profit maximisation, that are not always compatible with the objectives of the
employee. Those in the labour market may instead seek real pay increases and better working conditions that may not be
considered by the firms profit maximising agenda. This gives rise to the trade union, a collection of workers who join together to
pursue common goals by negotiating with employers. They create collective benefits that all, even non-unionised employees
can gain from, but also private benefits to members such as legal services and support in the event of an individual facing
grievance procedures.
The trade union decision seems quite topical with recent industrial action from junior doctors being widely publicised. Those
acting within this body are likely to be young, highly skilled and new to the labour market. In addition to this, real pay cuts
following the financial crisis in 2008 have seen inflation outstrip wage growth for many employees. Does this mean people have
lost commitment in joining a union, and if so who do these people tend to be?
Background
Although vast literature exists there is conflicting findings as to how individual’s demographics impact the probability of joining
a union. For example, D Blanchflower (2007) finds the probability follows an “inverted U-shaped pattern in age”. In contrast, a
study by Schnabel et al (2012) rejects this notion and finds no evidence of this non-linear relationship. In addition a psychology
paper written by (Parkes et al, 2004) shows that age has no significance when it comes to union membership. These ambiguous
results lead to questions regarding the functional form that the variable should take.
It is common perception that females are less attached to the workforce, and therefore less likely to be a member of a trade
union (Sutton, 1980). Females are also more likely to work part time, with part time work shown to have a significant negative
impact on the probability of trade union membership (Berg et al, 1992). A Booth (1985) finds that earnings are positively
correlated to the probability of being a union member; suggesting that individuals with high income are more likely to
contribute financially for collective action, and also gain from private benefits such as increased job security. In contrast, it has
been noted in a recent study by (Georke et al, 2012) that when including the log (gross wag) the variable instead has no
correlation with union membership; this was taken from a German study of full time employees.
Whether the individual works for the public or private sector is also likely to influence the probability of membership. According
to the Labour Force Survey 2013 the percentage of unionised workers as a proportion of all employees was 14.2% for the
private sector and 54.3% for the public sector (Department for business, innovation and skills, 2014). This shows far greater
demand for union representation in the public sector, a relationship supported with empirical evidence from (Booth, 1986).
Political preference will also influence the likelihood of becoming a member of a trade union. Trade unions have typically been
supported from the Labour government, when Labour formed as a political party they affiliated with the major trade unions of
the time; this was in a bid to move towards a more socialist society (Cole, 1948). This is in contrast to right wing parties such as
44
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
45
the Conservatives who have been highly critical of the impact that they have. The era of ‘Thatcherism’ saw a demise in the
power of trade unions, which were seen to only disrupt the natural wonders of the market. We therefore expect those with left
wing views to be more likely to join a union.
Job satisfaction also seems to be a contributing factor to the union decision. It has been found by Freeman and Medoff (1984)
that dissatisfied workers are more likely to be a member of a union. NHS junior doctors highlight the impact that industrial
factors, such as employment terms of contract, may have on the satisfaction derived from work. This satisfaction could be
derived from all different factors, the wage received, the conditions you work in, the opportunities for progression etc.
Therefore we suspect that satisfaction should have a direct impact on the need to join an establishment that is designed to
collectively bargain for improvements.
T Degroot (2006) found that individual’s demographics are not always linearly related to union demand. To correct for this, he
suggests that a method of neural network analysis, which treats the data differently to multiple regression analysis, could help
to explain why some variables that we include and expect to be significant, are not. This movement away from traditional
analysis could offer a potential explanation for any variables included that record insignificance, even though they appear to
be relevant factors that contribute to the decision.
What has been noted from many previous studies is the inclusion of industrial factors. Alison Booth (1986) includes variables to
capture the characteristics of the industry. These include the size of the firm, the type of industry for which they operate and the
labour mobility between firms and roles. In addition, she also includes the amount of ‘union-provided goods’ available to the
employee. These variables can all be expected to influence the individual’s decision, however because of data constraints my
analysis will only focus on the impact of individual’s demographics on the union decision.
45
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
46
Methodology and Data
The data that has been used within my analysis is collected in wave 4 of the ‘Understanding Society, The UK Household
Longitudinal Survey’. The original data has been collected from a population of 47,157 people. Available to include within my
analysis was a condensed list of 92 variables. The data is a cross section, given that the information has been obtained at one
point in time. Although previous wave data is available, time trend data extends beyond what has been taught within this
course, and therefore will not be considered.
The dependent variable that I have included is d_tuin1 which is labelled ‘member of a workplace union’. To be part of a union
you must be working and have one available to you in the workplace. There was a precursor question that would gage
whether this was an option available to employees, with the variable d_tujbpl and the label ‘union or staff association at
workplace’. My interest is in the probability of joining a union given that it is an option available to the individual. For this reason
I have focused directly on the variable d_tuin1 (Member of a workplace union). This has resulted in my analysis only including
those within the working population who can opt to say yes (value=1) or no (value=2). The data has been adjusted to remove
all missing and inapplicable results for these purposes, as they are either not in work or do not have the chance to currently join
a union within their organisation.
Given that my dependent variable is qualitative I have recoded this into binary form. This will be called MemberOfUnion, taking
the value of 1 when the individual is a member and 0 when they are not. Having used the existing literature to guide me with
my explanatory variables, descriptive statistics of all the included variables have been provided in figure 1. The original data
was ‘unclean’ therefore all data has been corrected for missing values prior to use.
Figure 1 – Descriptive Statistics
Variable
N
Mean
Max
Min
Std Dev
MemberOfUnion
9829
0.5927
1.00
0.00
0.49135
age2
47157
2588.2165
10816.00
256.00
1856.22413
Age
47157
47.35
104
16
18.595
Lnincome
44836
7.0388
9.62
-6.93
1.24357
Female
47157
0.5380
1.00
0.00
0.49856
PartimeEmp
47157
0.1436
1.00
0.00
0.35074
Public
7541
0.5688
1.00
0.00
0.49528
GrantfundedOrTrust
7541
0.3990
1.00
0.00
0.48973
ManagmentOrProffesional
26555
0.4064
1.00
0.00
0.49117
IntermediateWorker
26555
0.1350
1.00
0.00
0.34169
SupervisorOrTecnicalWorker
26555
0.0742
1.00
0.00
0.26214
WhiteBackground
44866
0.8517
1.00
0.00
0.35539
Single
43097
0.3000
1.00
0.00
0.45828
MarriedOrPartnership
43097
0.5195
1.00
0.00
0.49962
Dissatisfied
24347
0.0506
1.00
0.00
0.21919
Satisfied
24347
0.5473
1.00
0.00
0.49777
46
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
47
LeftWingViews
20984
0.5744
1.00
0.00
0.49444
DontKnowPoliticalView
20984
0.0402
1.00
0.00
0.19637
FemaleMarried
45784
0.2599
1.00
0.00
0.43859
FemaleSingle
45784
0.1482
1.00
0.00
0.35534
Valid N (listwise)
3225
0.5927
1.00
0.00
0.49135
Firstly, the linear probability model will be used as a basis for further analysis. This model is limited as it can provide probabilities
that are greater than 1 and lower than 0, which are not meaningful. The logit and probit models are non-linear and therefore
constrain the dependent variable between 0 and 1, allowing for more meaningful interpretation of my explanatory variables.
The main model that will be used is logit, and comparison between models will form my sensitivity analysis. Below is the
equation for my main OLS, logit and probit regressions;
𝑀𝑒𝑚𝑏𝑒𝑟𝑂𝑓𝑈𝑛𝑖𝑜𝑛 = 𝐵0 + 𝐵1 𝐴𝑔𝑒2 + 𝐵2 Age + 𝐵3 𝐿𝑛(𝐼𝑛𝑐𝑜𝑚𝑒) + 𝐵4 𝐹𝑒𝑚𝑎𝑙𝑒 + 𝐵5 𝑃𝑎𝑟𝑡𝑡𝑖𝑚𝑒𝐸𝑚𝑝 + 𝐵6 𝑃𝑢𝑏𝑙𝑖𝑐𝑆𝑒𝑐𝑡𝑜𝑟 +
𝐵7 𝐺𝑟𝑎𝑛𝑡𝐹𝑢𝑛𝑑𝑒𝑑𝑂𝑟𝑇𝑟𝑢𝑠𝑡 + 𝐵8 𝑀𝑎𝑛𝑎𝑔𝑒𝑚𝑒𝑛𝑡𝑂𝑟𝑃𝑟𝑜𝑓𝑓𝑒𝑠𝑖𝑜𝑛𝑎𝑙 + 𝐵9 𝐼𝑛𝑡𝑒𝑟𝑚𝑒𝑑𝑖𝑎𝑡𝑒𝑊𝑜𝑟𝑘𝑒𝑟 + 𝐵10 𝑆𝑢𝑝𝑒𝑟𝑣𝑖𝑠𝑜𝑟𝑂𝑟𝑇𝑒𝑐ℎ𝑛𝑖𝑐𝑎𝑙𝑊𝑜𝑟𝑘𝑒𝑟 +
𝐵11 𝑊ℎ𝑖𝑡𝑒𝐵𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑 + 𝐵12 𝑆𝑖𝑛𝑔𝑙𝑒 + 𝐵13 𝑀𝑎𝑟𝑟𝑖𝑒𝑑𝑂𝑟𝑃𝑎𝑟𝑡𝑛𝑒𝑟𝑠ℎ𝑖𝑝 + 𝐵14 𝐷𝑖𝑠𝑠𝑎𝑡𝑖𝑠𝑓𝑖𝑒𝑑 + 𝐵15 𝑆𝑎𝑡𝑖𝑠𝑓𝑖𝑒𝑑 + 𝐵16 𝐿𝑒𝑓𝑡𝑊𝑖𝑛𝑔𝑉𝑖𝑒𝑤𝑠 +
𝐵17 𝐷𝑜𝑛𝑡𝐾𝑛𝑜𝑤𝑃𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙𝑉𝑖𝑒𝑤 + 𝐵18 𝐹𝑒𝑚𝑎𝑙𝑒𝑀𝑎𝑟𝑟𝑖𝑒𝑑 + 𝐵19 𝐹𝑒𝑚𝑎𝑙𝑒𝑆𝑖𝑛𝑔𝑙𝑒 + µ
Following on from D Blanchflower (2007) I have chosen to test for an inverted U shaped relationship between age and the
probability of being a member of a trade union. In addition I have also adopted a natural logarithmic function of income. My
preliminary analysis suggested that ln(income) presented more significance compared to the standard functional form, which
was effected by the few individuals that received significantly higher earnings. This amendment was very straight forward given
that my sample is of the working population where individuals have a gross income greater than 0.
As all other variables were nominal, I have recoded these into dummy variables, more information on this can be found in
figure 2. To avoid the dummy variable trap I have removed one category from each original variable to be used as a
basecase for my analysis, which is also clarified in figure 2.
47
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
48
Figure 2–Variables and Coding
Variable
Description
(original variable derived from)
Basecase – for
all dummies
Coding
Expected
sign of
coefficient
Member of Union – dependant
variable
Individual is a member of a
trade union
-
Age2(d_dvage)
Age squared
-
-
-ve
Age (d_dvage)
Age
-
-
+ve
Lnincome
Natural log of total monthly
personal income gross
-
-
+ve
(d_fimngrs_dv)
Female
Individual is female
Male
(d_jbsectpub)
GrantfundedOrTrust
(d_jbsectpub)
ManagmentOrProffesional
(d_jbnssec5_dv)
IntermediateWorker
(d_jbnssec5_dv)
SupervisorOrTechnicalWorker
(d_jbnssec5_dv)
WhiteBackground
Individual works part time
Individual works for public
sector
Individual works for grant
funded
organisation/trust/charity
Individual works in category of
intermediate worker
Individual has white ethnicity
Individual marital status is single
(d_marstat)
Individual marital status is
married/in partnership
1 = works part time
-ve
Some other
public sector
firm
-ve
0 = works full time
1 = works for public sector
+ve
0 = otherwise
1 = works for grant funded
organisation
+ve
0 = otherwise
1 = works as a
manager/professional
Routine
worker, small
employer or
own account
Individual is supervisor or
technical worker
(d_marstat)
MarriedOrPartnership
Full time
employed
Individual works as a manager
or professional
(d_racel_dv)
Single
1 = is female
0 = is male
(d_jbft_dv)
PublicSector
N/A
0 = not a member
(d_sex)
ParttimeEmp
1 = is a member
-ve
0 = otherwise
1 = works as an intermediate
worker
+ve
0 = otherwise
1 = works as a supervisor
+ve
0 = otherwise
Non-white
background
1 = white
Some other
marital status
1 = single
+ve
0 = non white
?
0 = not single
1 = married
?
0 = not married
48
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
Dissatisfied
(d_jbsat)
Satisfied
(d_jbsat)
LeftWingViews
(d_vote4)
DontKnowPoliticalView
(d_vote4)
FemaleMarried
Individual is dissatisfied with
their job
Individuals is satisfied with their
job
Individual has left wing political
views
Individual has no recorded
political view
Individual is female and married
49
1 = dissatisfied with job
Neutral job
satisfaction
+ve
0 = otherwise
1 = satisfied with job
-ve
0 = otherwise
1 = left wing views
Right wing
views
+ve
0 = otherwise
1 = No political view
?
0 = otherwise
-
1 = Female and married
?
0 = is not a married female
FemaleSingle
Individual is female and single
-
1 = female and single
?
0 = is not a single female
Empirical Analysis
I have begun my analysis with some preliminary investigations into the correlation between my explanatory variables (see
appendix 1c). By running an OLS regression with collinearity diagnostics, I can report the correlation between variables; this
gives me awareness to any multicollinearity issues from the very beginning. Predictably age is reported to have a high
correlation with age squared, the correlation coefficient is 0.990. This is expected given that age squared is a function of age.
The same functionality problem arises with the variable SingleFemale, whereby there is a correlation of 0.718 with the variable
Single.
We also observe a large negative correlation between being single and being married (-0.729). This is expected given that an
individual will identify to one category or the other. The variable PublicSector has a high negative correlation with working for
a grant funded establishment, trust or charity. We would expect that those who identify with the category PublicSector are
therefore less likely to work in the grant funded category. Although some correlation exists between variables, these
correlations are predictable and are included with the purpose of providing a better fit to the model.
Within my preliminary analysis I have observed a low variation for ‘small employer and own account’ in the category ‘Current
job: Five Class NS-SEC’. The variable has been excluded along with ‘Routine workers’ to form a base case. As will be evident
from rows 10 – 12 (Figure 3) I have included the three remaining classes of workers for comparison against this group.
Figure 3 - Linear Probability Model (OLS)
49
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
50
Variable
OLS – Model 1
OLS - Model 2
Age2
-0.00036***
-0.00036***
d_dvage
0.035***
0.035***
Lnincome
0.079***
0.086***
Female
0.056***
-0.105**
Parttime
-0.060***
-0.060***
PublicSector
0.060
0.059
GrantfundedOrTrust
-0.010
-0.010
ManagmentOrProffesional
0.075***
0.073***
IntermediateWorker
-0.030
-0.027
SupervisorOrTechnicalWorker
0.093*
0.096**
WhiteBackground
0.101***
0.097***
Single
0.076**
-0.037
MarriedOrPartnership
0.070***
-0.074
Dissatisfied
0.018
0.017
Satisfied
-0.030*
-0.028*
LeftWingViews
0.100***
0.100***
DontKnowPoliticalView
0.060
0.064
FemaleMarried
-
0.197***
FemaleSingle
-
0.147**
Constant
-1.066***
-0.994***
𝑅2
0.070
0.074
Valid N
3225
3225
Significance is shown at the 10% level *, 5% level ** and the 1% level ***
I have undertaken my first OLS regression excluding my 2 dummy interaction terms. The purpose of this was to see if the
interaction terms aid the explanatory power of the model. The 𝑅2 for OLS Model 1 shows that the combined explanatory power
of the variables represent 7% of the changes in the probability of being a union member. Whilst I acknowledge that this
variable is small, it is a common finding when including numerous dummy variables, so overall the model is deemed to have
good explanatory power.
The introduction of the two interaction terms in OLS model 2 results in the 𝑅2 increasing to 7.4%. This model has greater
explanatory power of union membership. Therefore these interaction terms will be included in my Logit and Probit models, and
OLS model 2 will be used for comparison. What this has caused is a change in sign of the female variable from positive to
50
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
51
negative, so that it is now as predicted. The coefficient in model 2 infers that females who are not married or single are 10.5
percentage points less likely to be a member of a union compared to males, ceteris paribus.
The use of these interaction terms makes the Female term less significant, previously this was significant at the 1% level but now
in model 2 it is significant only to the 5% level, based on a two tailed test. The term FemaleMarried has a steeper slope value
than the value for FemaleSingle, therefore the probability of being a union member increases more quickly for females who are
married than those who are single, ceteris paribus.
Overall my model looks to be of a reasonable fit, with 12 significant explanatory variables reported in my OLS, Logit and Probit
models. Shown in figure 4:
Figure 4 - Significance of all three models
Highlighted insignificant results
Variable
Significance
OLS – model
2
Significance
LOGIT
Significance
PROBIT
Age2
0.000
0.000
0.000
d_dvage
0.000
0.000
0.000
Lnincome
0.000
0.000
0.000
Female
0.041
0.051
0.052
Parttime
0.010
0.010
0.010
PublicSector
0.308
0.285
0.293
GrantfundedOrTrust
0.860
0.891
0.884
ManagmentOrProffesional
0.006
0.007
0.006
IntermediateWorker
0.353
0.334
0.363
SupervisorOrTechnicalWorker
0.048
0.063
0.055
WhiteBackground
0.000
0.000
0.000
Single
0.479
0.481
0.519
MarriedOrPartnership
0.121
0.125
0.131
Dissatisfied
0.648
0.653
0.640
Satisfied
0.088
0.084
0.079
LeftWingViews
0.000
0.000
0.000
DontKnowPoliticalView
0.263
0.286
0.290
FemaleMarried
0.000
0.001
0.001
51
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
FemaleSingle
0.016
52
0.020
0.021
Figure 4 demonstrates the P values associated with all three regressions. These significance levels are all reported from a two
tailed test as previous studies often find conflicting results. Focusing on the Logit regression, it is confirmed that 8 variables are
significant to the 1% level, 1 variable to the 5% level and 3 to the 10% level. Before performing diagnostic tests of the model and
going on to interpret my findings, I review a Pearson Pairwise correlation matrix to again check for any signs of multicollinearity.
This confirms the correlations seen in the previous test, as they were justifiable the specification of the model will now be tested.
As the Logit regression provided 7 variables with insignificant results it is appropriate to test for overall joint significance using the
likelihood ratio test. This test will compare a restricted model, which excludes all individually insignificant variables, against one
that includes all variables in the regression, as shown below:
𝐻0 : 𝐵6 𝑃𝑢𝑏𝑙𝑖𝑐𝑆𝑒𝑐𝑡𝑜𝑟 = 𝐵7 𝐺𝑟𝑎𝑛𝑡𝐹𝑢𝑛𝑑𝑒𝑑𝑂𝑟𝑇𝑟𝑢𝑠𝑡 = 𝐵9 𝐼𝑛𝑡𝑒𝑟𝑚𝑒𝑑𝑖𝑎𝑡𝑒𝑊𝑜𝑟𝑘𝑒𝑟 = 𝐵12 𝑆𝑖𝑛𝑔𝑙𝑒 =
𝐵13 𝑀𝑎𝑟𝑟𝑖𝑒𝑑𝑂𝑟𝑃𝑎𝑟𝑡𝑛𝑒𝑟𝑠ℎ𝑖𝑝 = 𝐵14 𝐷𝑖𝑠𝑠𝑎𝑡𝑖𝑠𝑓𝑖𝑒𝑑 = 𝐵17 𝐷𝑜𝑛𝑡𝐾𝑛𝑜𝑤𝑃𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙𝑉𝑖𝑒𝑤 = 0
𝐻1 : 𝐻0 𝑖𝑠 𝑛𝑜𝑡 𝑡𝑟𝑢𝑒
Likelihood Ratio Statistic = (-3834.184) – (-6113.887) = 2279.703
Chi- square 𝒳7,0.01 = 18.48
Given that the LRS >𝒳 2 7,0.01 the test confirms that these individually insignificant results do have joint significance and should be
included within the model. This leads to a rejection of the null hypothesis in favour of the alternative hypothesis. Although this
was performed at the 1% level the null is rejected at all conventional levels of significance (see appendix 2b). We would
expect these results given that these variables have all been included in previous literature, as seen in (Booth, 1985).
I have also performed the Hosmer-lemeshow (HL) test to check the specification of my model. The HL significance was 0.623, as
this p value is large, the null hypothesis is not rejected and the model is deemed a good fit (appendix 2c).These results are
promising, given that the test tends to reject the null hypothesis at large sample sizes. The classification tables show an increase
from 67.3% to 69.2%, again suggesting that my model provides an acceptable fit.
52
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
53
Figure 5 – Coefficients for all three models
Variable
OLS
Logit
Probit
Age2
-0.00036
-0.002
-0.001
d_dvage
0.035
0.160
0.098
Lnincome
0.086
0.415
0.244
Female
-0.105
-0.508
-0.298
Parttime
-0.060
-0.284
-0.172
PublicSector
0.059
0.298
0.177
GrantfundedOrTrust
-0.010
-0.039
-0.025
ManagmentOrProffesional
0.073
0.338
0.210
IntermediateWorker
-0.027
-0.130
-0.075
SupervisorOrTechnicalWorker
0.096
0.439
0.276
WhiteBackground
0.097
0.464
0.278
Single
-0.037
-0.189
-0.102
MarriedOrPartnership
-0.074
-0.377
-0.217
Dissatisfied
0.017
0.086
0.053
Satisfied
-0.028
-0.141
-0.086
LeftWingViews
0.100
0.478
0.291
DontKnowPoliticalView
0.064
0.288
0.174
FemaleMarried
0.197
0.968
0.575
FemaleSingle
0.147
0.711
0.416
Constant
-0.994
-7.027
-4.218
Valid n
3225
3225
3225
Having checked the specification of my model, I can go on to compare the Logit results with those from the OLS and Probit
regressions. As can be seen in figures 3 and 4 the coefficients signs and levels of significance coincide between all three
models. This shows clarity between my methods of estimation, and makes the results more reassuring. The use of predicted
probabilities seen in appendix 4 shows the relative difference in the Logit and Probit regressions. The results are shown for a 30
year old married female of white background, with a monthly income of £800, working part time in the public sector. The
probability that she would be a member of a trade union is 48% in the probit model and 46.5% in the logit model; therefore the
results are statistically similar across both models.
Only three coefficients in figure 5 were opposite to the signs we predicted in figure 2, these were for the dummies
GrantfundedOrTrust, ManagmentOrProffesional and IntermediateWorker. Of these, the most surprising was for managers and
professionals resulting in the estimated logit being 0.338 units more likely to be a member of a union compared to a routine,
53
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
54
small employer or own account worker, ceteris paribus. This is surprising given that these workers would have more powers to
make change on their own, and therefore could be perceived to have less need of union protection (Booth, A, 1985).
Of the variables that did not have an expected sign, Single and Married are found to be negative and DontKnowPoliticalView,
FemaleMarried and FemaleSingle all positive. The results for Single and Married are insignificant; the inclusion of my highly
significant interaction terms may suggest that these variables are not required in the regression, despite the findings of my
likelihood ratio test.
Although the results suggest that worker dissatisfaction leads to a higher probability of union membership, this relationship is
insignificant. We saw that Freeman and Medoff (1984) found statistically significant results for this effect; I suspect that my
insignificance may have arisen through issues of cause and effect. However my estimated logit model suggests a significant
negative relationship for satisfied workers, with workers being 0.141 units less likely to be a member of a union compared to a
neutral worker, ceteris paribus. There is also a significant negative relationship found in the probit model.
What is highly significant in all my models is an inverted U shaped relationship with age, supporting Blanchflower’s (2007)
findings. This suggests that the probability of being a union member increases as age increases but at a decreasing rate,
holding everything else constant. Shown below is the age for which the probability of being a union member is maximised.
𝑀𝑒𝑚𝑏𝑒𝑟𝑂𝑓𝑈𝑛𝑖𝑜𝑛 = −7.027 − 0.00163117𝐴𝑔𝑒 2 + 0.16043809Age
𝜕𝑀𝑒𝑚𝑏𝑒𝑟𝑂𝑓𝑈𝑛𝑖𝑜𝑛
𝜕𝐴𝑔𝑒
=-0.00326234𝐴𝑔𝑒
+ 0.16043809
Age = 49.178 years old
This shows that the probability is rising until you just past 49years old and then the probability begins to decline, ceteris paribus.
Lawler and Hundley (1983) provide one possible explanation for such an effect given the relative decline in the employee’s
ability to amend their own terms and conditions of employment. Workers are less likely to change jobs or employers as age
increases, and therefore the net benefit from joining a union increases, suggesting that the probability also increases.
However, unions take time to implement change, so the benefit of being part of the union as you approach retirement
diminishes, until the costs associated outweigh the benefits.
It was surprising that the variable PublicSector provided insignificant results given that previous literature (Booth, 1985) has
found a significant positive relationship associated with this variable. What I suspect is that any significance has been lost by
grouping too much of the sector together, an issue raised by (McShane, 1986). To investigate this, I have re-specified the
dummy variables associated with this category using a Logit regression shown in appendix 3. The results show highly significant
positive results for NHS workers, employees of the government or council and those working for a nationalised industry, who are
all more likely to be a member of a union compared to some other public sector firm, ceteris paribus. These results now appear
to be in line with our expectations formed in figure 2.
Conclusion
The aim of my paper was to analyse how personal characteristics influence the decision to be a member of a union. My
regressions have provided significant data that compliments the previous literature, whilst the introduction of the interaction
terms FemaleSingle and FemaleMarried have aided explanation for the effects that marital status have on the union decision.
This contradicts findings from (Bain et al, 2009) that marital status is insignificant.
54
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
55
However, my analysis also has its limits. Industrial factors such as the firm’s current level of unionised workers were not available,
and I suspect this would have created encouragement or discouragement effects and therefore influenced the individual’s
decision. (Booth, 1985) follows a social custom model as she argues that peer pressure can have a large influence, therefore
including more industrial and social factors may offer further explanation of the membership decision.
My results confirm that personal attributes do play a significant role in the union decision and this infers that there will be groups
of underrepresented workers within society. My empirical analysis demonstrates who these people tend to be, and it is with
such findings that policy reforms can be implemented. This may be in the form of ensuring reduced pay differentials for workers
who tend to not be part of a union, for example non-white workers, or ensuring regulation of working conditions in industries
where workers are less likely to be members of a union.
Bibliography
Bain, G and Elias, P, 2009, Trade union membership in Great Britain: An individual-level analysis, British journal of industrial
relations, Vol. 23, Issue 1
Blanchflower, D G, 2007, International Patterns of Union Membership, British Journal of Industrial Relations, Vol 45, Issue 1, pages
1-28
Booth, A, 1985, The Free Rider Problem and a Social Custom Model of Trade Union Membership, The Quarterly Journal of
Economics, Vol. 100, No. 1, pages 253-261
Booth, A, 1986, Estimating the Probability of Trade Union Membership: A Study of Men and Women in
Britain, Economica, Vol. 53, No. 209, pages 41-61
Cole, M, 1948, British Trade Unions and the Labour Government, Industrial and Labor Relations Review
Vol. 1, No. 4, pages 573-579
Degroot, T, 2006, Modelling demand for unionisation with non-traditional data analysis methods, Social
Indicators research, Vol. 79, Issue 2, pages 275-289
Department for business, innovation and skills, 2015, Trade union membership 2014, Statistical bulletin, accessed on 8th February
2016, available at:
https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/431564/Trade_Union_Membership_Statistics_2
014.pdf
Freeman, R.B. and J.L. Medoff: 1984, What Do Unions Do? Basic Books, New York
Goerke, L and Pannenberg, M, 2012, Risk Aversion and Trade-Union Membership, The Scandinavian journal of Economics, Vol.
114, Issue 2, pages 275-295
Lawler, J, and Hundley, G. (1983), ‘Determinants of Certification and Decertification Activity’, Industry Relations, pages 335-48.
McShane, S, 1986, The multidimensionality of union participation, Journal of Occupational Psychology, Vol. 59, Issue 3, pages
177-187
55
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
56
Parkes, K R and Razavi, T D B, 2004, Personality and attitudinal variables as predictors of voluntary union
membership, Personality and Individual differences, Vol. 37, No. 2, pages 333- 34
Schnabel, C and Wagner, J, 2012, With or Without U? Testing the Hypothesis of an Inverted U-Shaped Union Membership-Age
Relationship, Contemporary Economics
Sutton, J R, 1980, Some determinants of woman’s trade union membership, The Pacific Sociological Review, Vol. 23, No. 4,
pages 377-391
Appendix
Appendix 1a – OLS model 1 regression output
Dependant variable: MemberOfUnion
Method: OLS – Linear probability model
Total observations: 3225
Model Summary
Model
1
R
R Square
.265a
Adjusted R Square
.070
Std. Error of the Estimate
.066
.45349
Coefficients
56
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
57
Standardized
Unstandardized Coefficients
Model
1
B
(Constant)
Std. Error
-1.067
.193
age2
.000
.000
age
.035
lnIncome
Female
Coefficients
Beta
t
Sig.
-5.536
.000
-.746
-5.944
.000
.006
.813
6.388
.000
.079
.020
.093
4.046
.000
.056
.018
.057
3.106
.002
-.060
.023
-.053
-2.598
.009
.060
.058
.062
1.029
.304
-.010
.059
-.010
-.172
.864
.075
.027
.077
2.804
.005
-.030
.029
-.024
-1.017
.309
SupervisorOrTecnicalWorker
.093
.049
.036
1.917
.055
WhiteBackground
.101
.024
.072
4.112
.000
Single
.076
.030
.069
2.522
.012
MarriedOrPartnership
.070
.025
.072
2.859
.004
Disatisfied
.018
.038
.008
.461
.644
-.030
.017
-.032
-1.791
.073
LeftWingViews
.100
.019
.096
5.366
.000
DontKnowPoliticalView
.060
.057
.019
1.042
.297
PartimeEmp
Public
GrantfundedOrTrust
ManagmentOrProffesional
IntermediateWorker
Satisfied
a. Dependent Variable: MemberOfUnion
57
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
58
Appendix 1b – OLS model 2 regression output
Dependant variable: MemberOfUnion
Method: OLS – Linear probability model
Total observations: 3225
Model Summary
Model
1
R
R Square
.272a
Adjusted R Square
.074
.069
Std. Error of the Estimate
.45273
58
Coefficients
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
59
Standardized
Unstandardized Coefficients
Coefficients
t
B
Std. Error
Beta
Model
1
(Constant)
-.994
.194
age2
.000
.000
age
.035
lnIncome
Sig.
-5.119
.000
-.746
-5.942
.000
.006
.812
6.382
.000
.086
.020
.101
4.374
.000
Female
-.105
.051
-.108
-2.047
.041
PartimeEmp
-.060
.023
-.053
-2.585
.010
.059
.058
.061
1.019
.308
-.010
.058
-.011
-.176
.860
.073
.027
.075
2.735
.006
-.027
.029
-.022
-.928
.353
SupervisorOrTecnicalWorker
.096
.048
.037
1.980
.048
WhiteBackground
.097
.024
.070
3.988
.000
Single
-.037
.053
-.034
-.707
.479
MarriedOrPartnership
-.074
.048
-.076
-1.551
.121
.017
.038
.008
.456
.648
-.028
.017
-.030
-1.707
.088
LeftWingViews
.100
.019
.096
5.381
.000
DontKnowPoliticalView
.064
.057
.020
1.120
.263
FemaleMarried
.197
.056
.206
3.542
.000
FemaleSingle
.147
.061
.108
2.403
.016
Public
GrantfundedOrTrust
ManagmentOrProffesional
IntermediateWorker
Disatisfied
Satisfied
59
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
60
a. Dependent Variable: MemberOfUnion
Appendix 1c – correlation matrix
60
0.154
0.012 -0.101
Public
LeftWingVie
ws
DontKnowP
oliticalView
FemaleMarri
ed
FemaleSingl
e
Satisfied
Disatisfied
MarriedOrPa
rtnership
Grantfunded
OrTrust
Managment
OrProffesion
al
Intermediate
Worker
SupervisorO
rTecnicalWo
rker
WhiteBackgr
ound
Single
age
0.010
0.022
0.177
0.065
0.052
0.085 -0.184
0.247 -0.056
0.299 -0.031 -0.026
0.613
0.046
0.022 -0.037 -0.014
0.024 0.013
0.024 -0.055
0.103 -0.002
0.025 -0.042 -0.084 -0.013
0.025
-0.014 -0.270 -0.295 -0.073
0.022
-0.010
0.030 -0.002
0.017 -0.016 -0.023
0.011 0.009 0.021
0.203
0.094 -0.074 -0.068
0.017
-0.047
0.048
0.004
0.009
0.109 -0.018
0.027
0.067
-0.031 -0.353 -0.387 -0.056 -0.071 -0.108
0.118
0.053
0.065 -0.069 -0.076 -0.027
0.018
0.010
0.069
0.012
0.004 -0.154 -0.006
0.007 -0.194 -0.164
0.111 -0.006 -0.959
0.046
0.008
0.455
0.135 -0.075 -0.066
-0.095
0.107
0.007
0.010
-0.040
0.024
0.069
-0.041
0.015
-0.007
0.058
0.008
-0.016
-0.022
-0.034
-0.070
0.168
1.000
0.007
0.010
0.107
0.111
-0.006
-0.959
0.057
lnIncome Female
0.038 1.000 0.990 0.043 -0.037 0.146 -0.014
0.057 0.990 1.000 0.065 -0.029 0.134 -0.016
lnIncome
0.154 0.043 0.065 1.000 -0.230 -0.497 -0.114
Female
0.012 -0.037 -0.029 -0.230 1.000 0.241 -0.096
PartimeEmp -0.101 0.146 0.134 -0.497
0.241 1.000 0.020
Public
0.041 -0.014 -0.016 -0.114 -0.096 0.020 1.000
0.038
age
0.020
0.023
-0.119
0.081
-0.022
-0.011
0.027
0.018
0.007
-0.252
-0.619
1.000
0.168
-0.075
-0.066
0.455
0.007
-0.194
-0.164
0.135
-0.007
-0.029
0.093
-0.089
0.023
-0.015
-0.028
0.004
0.028
-0.087
1.000
-0.619
-0.070
0.008
0.004
-0.154
-0.006
0.018
0.067
-0.095
0.118
0.109
-0.018
0.004
0.046
0.010
0.027 -0.022 -0.011
0.058
0.029 -0.002
0.066 -0.056 -0.031 0.632 -0.523
0.041 -0.461 0.718
0.045 0.035 -0.032
-0.036 -0.032
0.093 -0.017 -0.331 1.000
0.048 -0.036 -0.066 1.000 -0.331
0.718 -0.523 0.004 -0.037
0.632 -0.038
1.000 -0.066 -0.017
1.000 -0.234 -0.036 0.093
0.041 -0.031 0.045 -0.038 -0.234
0.073 -0.056 0.011 -0.046
-0.058 0.035 -0.461
0.030 0.045
-0.015 -0.178
0.012 -0.002 -0.002 -0.025 1.000 -0.240 0.011 0.045 -0.038 0.004
-0.006 0.025 -0.058 0.066 -0.240 1.000 -0.046 -0.038 0.048 -0.037
1.000 -0.025
0.073
0.025 -0.178
1.000 -0.729 -0.002 -0.058
-0.003 0.029 -0.729
-0.022 -0.036
0.017 1.000 -0.036
0.030 -0.058 -0.036
0.093 -0.029 -0.007
0.081 -0.119 0.023 0.020
0.015 -0.041 0.069 0.024
0.004 -0.028 0.023 -0.015 -0.089
0.018
0.008 -0.007
-0.074 0.022 0.065 -0.270
-0.068 0.025 0.085 -0.295
0.025 -0.042 -0.184 -0.073
0.022 -0.084 0.613 0.299
-0.037 -0.013 0.247 -0.031
-0.014 0.046 -0.056 -0.026
0.094 -0.010 0.052 -0.014
1.000 0.017 -0.022 -0.003 0.012 -0.006 -0.015
-0.087 0.028
-0.252 0.007
0.048 0.017 -0.047
Married
DontKno
OrPartn Disatisfi
LeftWin wPolitica Female Female
ership
ed
Satisfied gViews
lView
Married Single
-0.353 0.177 0.010 0.022
-0.387 0.203 0.017 0.011
-0.056 0.030 -0.016 0.009
-0.071 -0.002 -0.023 0.021
-0.108 0.103 0.024 0.024
0.009 -0.002 0.013 -0.055
-0.034 -0.022 -0.016
0.069
0.065
-0.069
-0.076
-0.027
0.027
0.012 0.053 -0.031
Managm Intermedi Supervisor WhiteB
Grantfund entOrProf ateWorke OrTecnical ackgrou
edOrTrust fesional
r
Worker
nd
Single
-0.040
1.000
age2
Partime
Emp
0.041
MemberOfU
nion
age2
Member
OfUnion
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
61
61
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
62
Appendix 2a – Logit regression output
Dependant variable: MemberOfUnion
Method: Binary Logistic function - drawn from the binomial distribution
Valid N: 3225
B
Step 1a
age2
S.E.
Wald
df
Sig.
Exp(B)
-0.002
0.000
32.134
1
0.000
0.998
d_dvage
0.160
0.026
37.217
1
0.000
1.174
lnIncome
0.415
0.096
18.559
1
0.000
1.514
Female
-0.508
0.260
3.810
1
0.051
0.602
PartimeEmp
-0.284
0.110
6.651
1
0.010
0.753
0.298
0.279
1.144
1
0.285
1.347
-0.039
0.282
0.019
1
0.891
0.962
0.338
0.126
7.150
1
0.007
1.402
-0.130
0.135
0.932
1
0.334
0.878
0.439
0.237
3.445
1
0.063
1.551
0.464
0.116
15.910
1
0.000
1.590
-0.189
0.268
0.496
1
0.481
0.828
-0.377
0.246
2.350
1
0.125
0.686
0.086
0.191
0.202
1
0.653
1.090
-0.141
0.082
2.983
1
0.084
0.869
0.478
0.089
29.002
1
0.000
1.612
0.288
0.270
1.138
1
0.286
1.333
FemaleMarried
0.968
0.280
11.932
1
0.001
2.633
FemaleSingle
0.711
0.305
5.427
1
0.020
2.037
-7.027
0.955
54.147
1
0.000
0.001
Public
GrantfundedOrTrust
ManagmentOrProffesional
IntermediateWorker
SupervisorOrTecnicalWorker
WhiteBackground
Single
MarriedOrPartnership
Dissatisfied
Satisfied
LeftWingViews
DontKnowPoliticalView
Constant
62
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
63
Appendix 2b – Likelihood Ratio Test
I have not reported the output for the Logit regression with all insignificant variables removed, however please note the valid N
increased to 3855.
Number of restrictions: 7
Model Summary
Step
-2 Log likelihood
Cox & Snell R Square
3834.184a
1
.072
Nagelkerke R Square
.101
𝑳𝒖 - Unrestricted model – all variables included
Model Summary
Step
1
-2 Log likelihood
Cox & Snell R
Nagelkerke R
Square
Square
3834.184a
.072
.101
𝑳𝑹 - Restricted model – all insignificant results removed
Model Summary
Step
1
-2 Log
Cox & Snell R
Nagelkerke R
likelihood
Square
Square
6113.887a
.070
.095
LR= 2(log𝑙𝑢 - log𝐿𝑟 ) = (-3834.184)-(-6113.887) =2279.703
Chi square values for common significance levels
63
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
𝒳 2 7,0.01 = 18.48
𝒳 2 7,0.025 = 16.01
𝒳 2 7,0.05 = 14.07
64
𝒳 2 7,0.10 = 12.02
Step
Chisquare
1
df
6.217
Sig.
8
0.623
Appendix 2c–Hosmer-Lemeshow Test
Goodness of fit test for a Logit regression
Contingency Table for Hosmer and Lemeshow Test
MemberOfUnion = .00
Observed
Step 1
Expected
MemberOfUnion = 1.00
Observed
Expected
Total
1
193
190.748
130
132.252
323
2
150
149.782
173
173.218
323
3
127
130.425
196
192.575
323
4
116
115.632
207
207.368
323
5
110
103.674
213
219.326
323
6
81
93.907
242
229.093
323
7
84
83.922
239
239.078
323
8
84
73.679
239
249.321
323
9
57
63.421
266
259.579
323
10
52
48.808
266
269.192
318
𝐻0 : 𝐺𝑜𝑜𝑑 𝑓𝑖𝑡 - Correctly specified
𝐻1 : 𝐵𝑎𝑑 𝑓𝑖𝑡 - Not correctly specified
Hosmer- lemeshow p value is found to be 0.623
0.623> 0.1/0.05/0.01 then we do not reject the null hypothesis. The model is correctly specified.
Appendix 3 – Logit regression model for revised public sector dummies
Dependent variable: MemberOfUnion
Method: Binary Logistic
Valid N: 3225
64
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
65
Appendix 4 - Logit Probability
Method: Binary logit regression
X: included variables within regression
Beta: Coefficient of each variable
X
Beta (b)
Characteristics
Xb
age2
-0.002
900
-1.468057094
d_dvage
0.160
30
4.813142838
65
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
66
lnincome
0.415
6.684611728
2.773826583
Female
-0.508
1
-0.507764478
ParttimeEmp
-0.284
1
-0.284291833
Public
0.298
1
0.29823735
GrantfundedOrTrust
-0.039
0
0
ManagmentOrProffesional
0.338
0
0
IntermediateWorker
-0.130
1
-0.130006576
SupervisorOrTechnicalWorker
0.439
0
0
WhiteBackground
0.464
1
0.463780617
Single
-0.189
0
0
MarriedOrPartnership
-0.377
1
-0.376756347
Disatisfied
0.086
0
0
Satisfied
-0.141
1
-0.140836528
LeftWingViews
0.478
1
0.477645553
DontKnowPoliticalView
0.288
0
0
FemaleMarried
0.968
1
0.968124284
FemaleSingle
0.711
0
0
Constant
-7.027
1
-7.027164003
Sum
-0.140119634
Exp(sum)
0.869254237
Probability
0.465027293
66
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
67
Appendix 4 – Probit probability
Method: Binary probit regression
X: included variables within regression
Beta: Coefficient of each variable
X
Beta (b)
Characteristics
-0.001
900
d_dvage
0.098
30
lnIncome
0.244
6.684611728
age2
Xb
-0.892583357
2.93077475
1.628804803
Female
-0.298
1
-0.297759134
PartimeEmp
-0.172
1
-0.171903023
0.177
1
0.177014577
-0.025
0
0
0.210
0
0
-0.075
1
-0.075250267
SupervisorOrTecnicalWorker
0.276
0
0
WhiteBackground
0.278
1
0.278327485
Single
-0.102
0
0
MarriedOrPartnership
-0.217
1
-0.216880785
0.053
0
0
-0.086
1
-0.086170902
LeftWingViews
0.291
1
0.291027083
DontKnowPoliticalView
0.174
0
0
FemaleMarried
0.575
1
0.574876768
FemaleSingle
0.416
0
0
-4.218
1
-4.218248139
Sum
-0.077970141
Public
GrantfundedOrTrust
ManagmentOrProffesional
IntermediateWorker
Disatisfied
Satisfied
Constant
Exp(sum)
0.924992046
Probability
0.480517334
67
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
68
The Economics of Housework:
Exploring the Determinants and Allocation of Hours of Household Labour.
By Rebecca Heath
1.
Introduction
The job market has always provided the traditional labour that economists have concentrated on as it is easy to analyse this
market due to the ability to monetise workers’ decisions regarding leisure taken, hours worked, and so forth. However, as more
economic schools of thought have developed and traditional labour theory has been re-examined, the economics of the
household has been shown to influence the decisions that consumers make on a day-to-day basis. This project will aim to explore
the determinants of household labour, represented by hours of housework, and how this unpaid labour is allocated. Housework
is defined as cooking, cleaning, doing the laundry and other similar activities. Contemporary issues such as the gender-gap,
gender roles and social norms, and theory such as the domestic division of labour, are key to understanding the amount of
housework undertaken; it makes this topic relevant in modern economics.
2.
Literature Review and Background
Women have “greater household commitments” (Bryan and Sevilla-Sanz, 2010), and by linking in theory that Gary Becker
published on the allocation of time (Becker, 1965) and the economics of marriage (Becker, 1973), there forms the notion that
there is a trade-off between working in the labour market and working in the household. Becker (1973) looks at specialisation
and productivity within the household: the partner who is more productive in the labour market relative to the other, will specialise
in the market whereas the other will specialise in housework which will increase the overall household utility; this is normally
undertaken by the female partner. An out-dated view is that typically women will stay at home, raise any children and perform
any housework that needs doing whilst men work in the labour market. As Bianchi et al (2000) describes, gender perspectives
are important in how a couple decides who should do more of the housework: the more Egalitarian gender ideology a couple
holds, the more evenly split the housework chores are for both partners. This ideology is a recent view that couples have taken
on; attitudes towards women and the labour market have moved from traditional ‘stay at home’ views to a more Egalitarian
view where there’s an equal share of labour market and household labour.
Similar to the study that Layte (1998) conducted, I will be exploring the relationship between satisfaction with leisure time and
hours of housework. With the addition of binary variables, I can look at how individuals classify housework and test if there is a
trade-off between housework and leisure time.
68
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
69
In addition to previous literature, I shall investigate how the hours of housework are affected by income for an individual who is
satisfied with their leisure time, keeping all other variables constant. This can be achieved by creating an interaction term
between the binary variable representing an individual who is satisfied with their leisure time and multiplying it by the natural
logarithm of total monthly personal income. I expect that as the natural logarithm of personal income increases, the hours of
housework will rise less quickly for an individual who is satisfied with their leisure time, than for an individual who is not satisfied with
their leisure time due to the dissatisfaction coming from the individual undertaking more hours of housework.
I shall also consider hours of sleep as a determinant of hours of housework. Looking at this study from a health economics
viewpoint, sleep can be seen as a resource that individuals have to allocate, which will be taken into consideration when an
individual decides how much housework to do. Podor and Halliday (2009) consider the impact of health on the allocation of
time: “better health is associated with more time allocated toward production on the market and at home, but less consumption
of leisure…”, which would support the theory that there is a trade-off between housework and leisure. I predict that there is a
negative relationship between hours of sleep and hours of housework.
A definition of housework activities was always given at the start of the data collection the minimise variation in answers and to
make the results more reliable. The majority of the data was collected through time-use diaries, in which the respondent kept a
record of the time they spent on housework. As Carrasco and Dominguez (2014) highlight, hours of housework is a difficult variable
to measure as you are limited to what the respondent recalls and their ‘perceived time’, which can give room for errors, however,
quantifying the variable therefore allows statistical analysis to be conducted. Bianchi et al (2000) use OLS regression analysis,
along with most of the literature reviewed, to estimate their model, which would seem logical given that hours of housework per
typical week falls into the quantitative category.
Overall, the general conclusion in regards to variation in housework completed between genders was that females did
significantly more than males, in a typical week. This fits in with what we would intuitively think and the study that Marini and
Shelton (1993) produced. Considering pre-disposed gender roles, the female would dominate over the household labour, whilst
the male would primarily work in the labour market. I include a binary variable for females to examine this.
Whilst gender is a main variable to focus on, children is also an important determinant, and plays a key role in the economics of
the household. Studies such as Bond and Sales (2001) look at children as a dependent variable and found that having children
has a positive effect on hours of housework. Bianchi et al (2000) goes one step further and explains that when children create
additional housework, it is normally the woman who has to take on this extra work. I have made an interaction term to look at
the relationship between females and children.
Much of the empirical analysis focuses on wages and income and I acknowledge that a critique is the causality issue that hours
of housework could determine level of income as previous literature such as Bryan and Sevilla (2010) has explored. However, for
my analysis, I am looking at the reverse causality. Furthermore, another potential critique of the literature is the country in which
69
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
70
it is set in. Although many of my references are in the United Kingdom, some like Fernandez and Sevilla-Sanz (2006) are based in
Spain, which can show different results based on different social norms and cultural differences.
3.
Methodology and Data
The data used in this study is sourced from the Understanding Society longitudinal survey and will involve taking a single crosssection (wave 4) of the data. Ordinary Least Squares (OLS) is the most appropriate model due to nature of my dependent
variable, hours per week on housework, which is supported in the literature, such as Brines (1994) who also use this method. OLS
is the best linear unbiased estimator, according to the Gauss-Markov theorem, as my dependent variable is quantitative.
The equation below shows the original model in which all variables show a linear relationship between them all, including the
constant, 𝛽0, and the error term, 𝜇𝑖 .
Original Model: Linear Variables
𝐻𝑂𝑈𝑅𝑆 𝑃𝐸𝑅 𝑊𝐸𝐸𝐾 𝑂𝑁 𝐻𝑂𝑈𝑆𝐸𝑊𝑂𝑅𝐾𝑖 =
𝛽0 + 𝛽1𝐻𝑂𝑈𝑅𝑆𝑆𝐿𝐸𝐸𝑃𝑖 + 𝛽2𝐻𝑂𝑈𝑅𝑆𝑊𝑂𝑅𝐾𝐸𝐷𝑖 + 𝛽3𝐴𝐺𝐸𝑖 + 𝛽4𝐶𝐻𝐼𝐿𝐷𝑅𝐸𝑁𝑖 + 𝛽5𝑆𝐴𝑇𝐼𝑆𝐹𝐼𝐸𝐷𝐿𝐸𝐼𝑆𝑈𝑅𝐸𝑖 + 𝛽6𝑁𝐸𝑈𝑇𝑅𝐴𝐿𝐿𝐸𝐼𝑆𝑈𝑅𝐸𝑖
+ 𝛽7𝐹𝐸𝑀𝐴𝐿𝐸𝑖 + 𝛽8𝑀𝐴𝑅𝑅𝐼𝐸𝐷𝐶𝐼𝑉𝐼𝐿𝑖 + 𝛽9𝐷𝐸𝐺𝑅𝐸𝐸𝑖 + 𝛽10𝐴𝐿𝐸𝑉𝐸𝐿𝑖 + 𝛽11𝑊𝐻𝐼𝑇𝐸𝑖 + 𝛽12𝑃𝐸𝑅𝑆𝑂𝑁𝐴𝐿𝐼𝑁𝐶𝑂𝑀𝐸𝑖 + 𝜇𝑖
Dependent Variable
The final specification for hours per week on housework was a linear and not a natural logarithm format; the reason for this is
studies such as Kan and Laurie (2016) also leave hours of housework as linear. Additionally, when I performed an OLS regression
with the natural logarithm of housework, ln(housework), variables that were individually significant when the dependent variable
was linear, were now individually insignificant.
Explanatory Variables
When deciding the income variable to include in the model, I have used personal income because of the missing variables
associated with labour income. The reduction in sample size was substantial and I wanted to keep the sample size large to ensure
accurate and reliable standard errors. I expect this sign to be negative: as income increases, housework decreases as individuals
will outsource housework.
70
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
71
The level of education an individual possesses has an impact on how much housework they are allocated, according to Hersch
and Stratton (1994). Studies have explored how differences in educational levels between husbands and wives may be how
housework is allocated, in other words, the higher educated you are, the less housework you are expected to do. However, for
this study differences in husbands and wives cannot be looked at as the composition of the household cannot be fully
determined for each individual data piece.
Kan and Laurie (2016) considered the ethnicity of individuals and the impact that it has on housework. Formatted as a binary
variable ‘white’ with the base case being non-white ethnicities, I expect this sign to be negative due to the reduced family
culture that white ethnic groups have and the increased likelihood of them outsourcing housework, for example, hiring a cleaner.
Moreover, those of white ethnicity tend to be in societies where gender equality is practiced every day, such as the United
Kingdom, where an equal split of chores is seen.
Bianchi et al (2000) confirmed my idea for undertaking a Chow test and potentially splitting my data into male and female
categories and regressing separately, or, including interaction terms with gender. I expect this sign to be positive when I create
a female binary variable into the model. In addition, this paper supported my idea to make interaction terms with gender and
children; I expect this sign to be positive when I make a binary variable for female and children.
To add to looking at elements of the household, marital status is a crucial variable as confirmed by Bianchi et al (2000) who found
that being married increased females housework hours. I shall look at the effect of marital status on hours of housework per week,
based on the intuition that once an individual is married, housework will increase due to the higher expectations of household
cleanliness and the consequence of two individuals living together creates double the housework than that of a single person,
hence why I have used marital status and recoded it into a binary variable look at those married and in a legal civil partnership.
I expect this will be positive as individuals who aren’t married or in a legal civil partnership will feel less inclined to do housework
as they don’t have to keep a spouse ‘happy’ through economic payments such as tidying the bedroom, taking a spouse out to
dinner, and so forth (Becker, 1973).
The sample is restricted to those who work in the labour market to allow for further comparison of the differences between males
and females. Presser (1994) also restricted the sample to just those who work. Age is included as intuition would suggest that the
older you become the more housework you do. I have added in age-squared to look at a potential quadratic relationship. The
minimum value of age is 16 years, as this is the minimum age requirement to take part in the Understanding Society survey.
In addition to these studies, I shall include variables regarding satisfaction with leisure time, and hours of sleep as explained in the
previous section. Missing values were removed and a description of my recoded and non-recoded variables are shown in Figure
1.
71
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
72
Figure 1: Variable Descriptions and Expected Signs
Old Variable
New Variable Name
Description
Coding
Expected
Sign of
Coefficient
d_howlng
(dependent)
−
hours per week on
housework
−
N/A
d_sclfsat7
satisfied_leisure
satisfied with
amount of leisure
time
1 = satisfied with
leisure;
0 = not satisfied
-ve
d_sclfsat7
neutral_leisure
neither satisfied nor
dissatisfied with
amount of leisure
time
1 = neither
satisfied nor
dissatisfied;
0 = either satisfied
or dissatisfied
-ve/+ve
d_sex
Female
binary: gender of
individual
1 = female; 0 =
male
+ve
binary: marital
status
1 = married /
legal civil
partnership;
0
= not married / in
legal civil
partnership
+ve
-ve
d_marstat
married_legal_civil
d_hiqual_dv
Degree
binary: degree or
other higher degree
1 = degree or
other higher
degree; 0 = no
degree
d_hiqual_dv
Alevel
binary: if individual
has A-levels
1 = A-levels; 0 =
No A-levels
-ve
d_racel_dv
White
binary: if individual
is white
1 = white; 0 =
non-white
-ve
d_dvage
−
age
−
+ve
d_dvage
Agesquared
age x age
−
-ve
-
total monthly
personal income
gross
-
-ve
d_fimngrs_dv
72
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
73
d_fimngrs_dv
lnpersonal_income
natural logarithm
of personal income
−
-ve
d_hrs_slph
−
hours of actual
sleep
−
-ve
d_jbhrs
−
no. of hours
normally worked
per week
−
-ve
Femalechild
-
interaction term:
Female x number of
children
-
+ve
-
interaction term:
Satisfied_leisure x
lnpersonal_income
-
-ve
Sat_leisure_income
As shown in Figure 2, a table of descriptive statistics can be seen with all of the initial variables. With missing values deleted and
qualitative variables recoded into binary variables, the total valid sample size is 19,089 which is a large sample. By restricting the
sample to individuals who work through including the variable ‘hours worked per week’, the sample size is reduced but is still
large enough to produce reliable results.
Figure 2: Descriptive Statistics
N (sample
size)
Minimum
Maximum
Mean
Std.
Deviation
hours per week on
housework
42913
0
168
10.090
9.157
hours of sleep
43148
0
20
6.670
1.430
hours worked per
week
22937
0.5
97
32.613
11.489
age
47157
16
104
47.350
18.595
agesquared
47157
256
10816
2588.217
1856.224
Number of children
47157
0
7
0.500
0.924
Variable
73
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
74
satisfied_leisure
38895
0
1
0.573
0.495
neutral_leisure
38895
0
1
0.144
0.351
female
47157
0
1
0.538
0.499
married or civil
partner legal
43097
0
1
0.520
0.500
degree or other
higher degree
46937
0
1
0.341
0.474
A-level
46937
0
1
0.213
0.410
white
44866
0
1
0.852
0.355
lnpersonal_income
47078
0
10
6.721
1.860
Valid N (listwise)
19089
4.
Empirical Results
Linear Model
The original model, as described in Figure 3, contained all linear variables. Once an OLS regression was conducted, the R 2 value
revealed that only 25.5% of the variation in hours of housework is accounted for. The White Test for heteroskedasticity was
performed and resulted in rejecting the null hypothesis for homoskedasticity, indicating that the original model has
heteroskedasticity which is most likely due to the income variable. To fix this problem, I re-estimated the model using a bootstrap
tool which adjusts the standard errors of variables to reflect heteroskedasticity. Figure 3 shows the standard errors before and
after adjusting for heteroskedasticity; every test conducted afterwards has been accounted for this also. It is important to note
that none of the variables decreased in significance as reflected in Figure 3.
74
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
75
Figure 3: OLS Regressions - with all linear variables
Coefficient
Coefficient with
bootstrap
6.567***
6.567***
(0.396)
(0.454)
-0.230***
-0.230***
(0.0.38)
(0.044)
-0.085***
-0.085***
(0.005)
(0.005)
0.104***
0.104***
(0.004)
(0.004)
1.521***
1.521***
(0.053)
(0.060)
-0.398***
-0.398***
(0.103)
(0.103)
-0.318**
-0.318**
(0.145)
(0.149)
4.874***
4.874***
(0.101)
(0.091)
married or civil
partner legal
0.688***
0.688***
(0.106)
(0.109)
degree or other
higher degree
-0.472***
-0.472***
(0.114)
(0.120)
-0.590***
-0.590***
(0.128)
(0.128)
-0.561***
-0.561***
Variable
Constant
Hours of actual sleep
Hours worked per
week
Age
Number of children
satisfied_leisure
neutral_leisure
female
A-level
white
75
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
monthly personal
income gross
76
(0.139)
(0.153)
0.000***
0.000***
(0.000)
(0.000)
Dependent Variable: hours per week on housework.
Significance levels: ***/**/* correspond to 1%/5%/10%, respectively.
Note: standard errors are in parenthesis.
Non-Linear Model
Figure 4 shows the non-linear model. Age-squared was added in to look at whether a quadratic relationship could be explored,
based on pre-disposed logic: as age increases, individuals will engage in more hours of housework. However, after a certain age
hours of housework will begin to increase at a decreasing rate and will eventually start falling. This can be due to health problems
associated with humans aged 45 and older. Moreover, it could be partly due to levels of income and wealth, as older individuals
are more likely to be in a financial situation where they can outsource activities like housework in order to gain more leisure time.
To consider the different specifications this model could take, I tried various different functional forms, for example, the natural
logarithm of dependent variable housework, but when performing a Ramsey RESET test, the f-statistic had a p-value of 0.000
which resulted in a rejection of the null hypothesis, indicating that there is a specification problem. As an attempted solution, I
wanted to explore the effect of people who had reported zero or very small number of hours of housework. Initially, I deleted the
cases of those who weren’t in paid employment or self-employed to restrict my sample, then I added a filter to include only those
with positive hours of housework. However, the RESET specification test still implied I had a specification problem. This will be
further explained in the conclusion with suggestions of extended research.
Based on the linear model’s OLS results, I decided to take the natural logarithm of personal income. This is because the coefficient
for total monthly personal income gross was very small, 0.000351, and by taking the natural logarithm it makes the data easier to
interpret.
The equation below shows the modifications made on the functional form of the model; the addition of age-squared and the
natural logarithm of personal income.
Non Linear Model: Linear and Non-Linear Variables
𝐻𝑂𝑈𝑅𝑆 𝑃𝐸𝑅 𝑊𝐸𝐸𝐾 𝑂𝑁 𝐻𝑂𝑈𝑆𝐸𝑊𝑂𝑅𝐾𝑖 =
76
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
77
𝛽0 + 𝛽1𝐻𝑂𝑈𝑅𝑆𝑆𝐿𝐸𝐸𝑃𝑖 + 𝛽2𝐻𝑂𝑈𝑅𝑆𝑊𝑂𝑅𝐾𝐸𝐷𝑖 + 𝛽3𝐴𝐺𝐸𝑖 + 𝛽4𝐴𝐺𝐸 2 𝑖 + 𝛽5𝐶𝐻𝐼𝐿𝐷𝑅𝐸𝑁𝑖 + 𝛽6𝑆𝐴𝑇𝐼𝑆𝐹𝐼𝐸𝐷𝐿𝐸𝐼𝑆𝑈𝑅𝐸𝑖
+ 𝛽7𝑁𝐸𝑈𝑇𝑅𝐴𝐿𝐿𝐸𝐼𝑆𝑈𝑅𝐸𝑖 + 𝛽8𝐹𝐸𝑀𝐴𝐿𝐸𝑖 + 𝛽9𝑀𝐴𝑅𝑅𝐼𝐸𝐷𝐶𝐼𝑉𝐼𝐿𝑖 + 𝛽10𝐷𝐸𝐺𝑅𝐸𝐸𝑖 + 𝛽11𝐴𝐿𝐸𝑉𝐸𝐿𝑖 + 𝛽12𝑊𝐻𝐼𝑇𝐸𝑖
+ 𝛽13𝑳𝑵𝑃𝐸𝑅𝑆𝑂𝑁𝐴𝐿𝐼𝑁𝐶𝑂𝑀𝐸𝒊 + 𝜇𝑖
As shown in model 1 within Figure 4, every variable, with the exception of neutral leisure, is significant to the 1% significance level.
Therefore, multicollinearity is not a problem in my model as the variables are all individually significant. Furthermore, by having a
large sample size, this should reduce any unwanted effects of multicollinearity.
A Chow test was conducted to look at the differences between males and females. The f-statistic displayed was 101.551 which
is a large figure, however, the coefficient for the binary variable ‘female’ was also large which is reflective of the significant
difference between males and females. The critical value shown in the F-tables was 1.67 resulting in the rejection of the null
hypothesis of keeping males and females pooled. This confirms that males and females are significantly different in the hours of
housework they complete, therefore they should be regressed separately, or, like I have done in my study, should be computed
into a binary variable and relevant interaction terms which is essentially splitting the sample up. This is reflected in model 3 shown
in Figure 4.
Figure 4: OLS Regressions – ln(personalincome), age2 and interaction terms included
Coefficients.
Coefficients.
Model 1
Model 2
Coefficients.
Model 3
3.529***
3.529***
5.723***
(0.692)
(0.738)
(1.041)
Hours of actual
sleep
-0.211***
-0.211***
-0.213***
(0.038)
(0.044)
(0.044)
Hours worked per
week
-0.104***
-0.104***
-0.087***
(0.005)
(0.006)
(0.006)
0.569***
0.569***
0.541***
(0.025)
(0.026)
(0.026)
-0.006***
-0.006***
-0.005***
(0.000)
(0.000)
(0.000)
Variable
Constant
Age
Age-squared
77
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
78
1.164***
1.164***
0.170**
(0.057)
(0.065)
(0.068)
-0.341***
-0.341***
-2.460**
(0.102)
(0.104)
(1.075)
-0.331**
-0.331**
-0.336**
(0.144)
(0.146)
(0.147)
4.668***
4.668***
3.505***
(0.101)
(0.099)
(0.111)
married or civil
partner legal
0.447***
0.447***
0.526***
(0.106)
(0.111)
(0.110)
degree or other
higher degree
-0.638***
-0.638***
-0.663***
(0.115)
(0.128)
(0.127)
-0.508***
-0.508***
-0.540***
(0.127)
(0.135)
(0.133)
-0.525***
-0.525***
-0.577***
(0.138)
(0.151)
(0.148)
-0.748***
-0.748***
-0.975***
(0.089)
(0.100)
(0.134)
femalechild
-
-
Sat_leisure_income
-
-
Number of children
satisfied_leisure
neutral_leisure
female
A-level
white
ln(personalincome)
2.006***
(0.108)
0.284**
(0.140)
Dependent Variable: hours per week on housework
Significance levels: ***/**/* correspond to 1%/5%/10%, respectively.
Note: standard errors are in parenthesis.
R2 was improved when heteroskedasticity was taken into account, meaning model 2 accounts for 26.5% of the variation in hours
of housework. However, I ran a third and final model, which included my interaction terms.
78
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
79
Final Revised Model
𝐻𝑂𝑈𝑅𝑆 𝑃𝐸𝑅 𝑊𝐸𝐸𝐾 𝑂𝑁 𝐻𝑂𝑈𝑆𝐸𝑊𝑂𝑅𝐾
= 5.723 − 0.213𝐻𝑂𝑈𝑅𝑆𝑆𝐿𝐸𝐸𝑃 − 0.087𝐻𝑂𝑈𝑅𝑆𝑊𝑂𝑅𝐾𝐸𝐷 + 0.541𝐴𝐺𝐸 − 0.005𝐴𝐺𝐸2 + 0.17𝐶𝐻𝐼𝐿𝐷𝑅𝐸𝑁
− 2.46𝑆𝐴𝑇𝐼𝑆𝐹𝐼𝐸𝐷𝐿𝐸𝐼𝑆𝑈𝑅𝐸 − 0.336𝑁𝐸𝑈𝑇𝑅𝐴𝐿𝐿𝐸𝐼𝑆𝑈𝑅𝐸 + 3.505𝐹𝐸𝑀𝐴𝐿𝐸 + 0.526𝑀𝐴𝑅𝑅𝐼𝐸𝐷𝐶𝐼𝑉𝐼𝐿
− 0.663𝐷𝐸𝐺𝑅𝐸𝐸 − 0.54𝐴𝐿𝐸𝑉𝐸𝐿 − 0.577𝑊𝐻𝐼𝑇𝐸 − 0.975𝐿𝑁𝑃𝐸𝑅𝑆𝑂𝑁𝐴𝐿𝐼𝑁𝐶𝑂𝑀𝐸
+ 2.006𝐹𝐸𝑀𝐴𝐿𝐸𝐶𝐻𝐼𝐿𝐷 + 0.284𝑆𝐴𝑇𝐿𝐸𝐼𝑆𝑈𝑅𝐸_𝐼𝑁𝐶𝑂𝑀𝐸
R2 was improved, meaning model 3 with interaction terms accounts for 28.2% of the variation in hours of housework. With
interaction terms, the variables ‘number of children’ and ‘satisfied leisure’ decreased in significance from 1% to 5% level, but this
is to be expected as they are being used within the interaction terms. T-tests and p-values show that all of my variables are
individually significant, therefore there is no reason to test for joint significance.
Interpretations
For simple, clear and concise interpretations I have converted the coefficients into hours and minutes. The constant, 5.723,
indicates that when all other explanatory variables are zero, and individual will complete 5 hours 43 minutes of housework per
week. However, in reality it doesn’t make much logical sense due to the age variable included in the model: at zero years of
age undertaking housework isn’t feasible.
Household determinants
Looking at the components of the household, a key determinant explored was gender, and the information revealed in the final
revised model is what we would expect: if the individual is female, hours of housework is estimated to be 3 hours and 30 minutes
higher than if the individual is male, ceteris paribus. Being female has a significant impact on housework and is reflected through
having the largest magnitude of the explanatory variables. Alongside this, I found that if an individual is female and has children,
the number of hours of housework per week increases by 2 hours, ceteris paribus. An explanation for this is that typically mothers
will take time away from the labour force and stay at home, specialising in childcare and housework whist the male will specialise
in the labour market. Both of these variables are significant to the 1% level and are consistent with previous literature: there is
proof that females take the burden of this unpaid labour more so than males which confirms Bianchi et al (2000) findings.
Marital status and number of children are crucial components of a household and both variables reported a positive relationship
with housework. If an individual is married or in a legal civil partnership, hours of housework per week is estimated to be 32 minutes
higher than if the individual is not married or in a legal civil partnership, ceteris paribus, which supports Bianchi et al (2000) findings.
79
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
80
Similarly, with number of children, holding everything else constant, for each additional child in the household, individuals are
predicted to see a rise of 10 minutes of housework per week.
Age, health and leisure
Age, sleep and satisfaction with leisure time are variables that contribute to an individual’s general wellbeing and each one has
an individual effect on housework. The impact of age on hours of housework per week, each additional year increases the hours
of housework by 33 minutes, ceteris paribus. The OLS regression in Figure 4 confirmed an ‘inverted-u’ shaped quadratic
relationship between age and hours of housework per week; the turning point is 54 years, found by differentiating the household
equation with respect to age.
The two leisure time binary variables reflected a negative relationship with housework: if the individual is satisfied with their leisure
time, hours of housework per week is estimated to be 2 hours and 28 minutes lower than if the individual is dissatisfied with their
leisure time, ceteris paribus. What is also interesting is that if an individual is neither satisfied nor dissatisfied with their leisure time,
in other words they are indifferent to their leisure time, hours of housework per week is estimated to be 20 minutes lower than if
the individual is dissatisfied with their leisure time, ceteris paribus. A theory for this could be that individuals who are dissatisfied
with their leisure time are completing more housework which would indicate that leisure time is traded-off with housework: the
opportunity cost of doing housework is reduced leisure time.
In regards to sleep, holding everything else constant, for each additional hour of sleep, individuals are predicted to lose 13
minutes of housework per week, which confirms the negative relationship previously thought.
Income and education
Exploring the aspect of human capital, variables like hours worked per week, education levels and income have proven to have
a negative effect on household labour. By logic of time allocation, intuition will say that as hours worked per week increases,
hours of housework per week decreases, the regression results confirmed this.
Educational levels showed relatively large negative coefficients, supporting initial thoughts that the more educated an individual
is, the less housework they complete in a typical week.
If the individual has a degree or other higher degree, hours of housework per week is estimated to be 40 minutes lower than if
the individual has GCSEs or below, ceteris paribus. A similar result is revealed for an individual with A-levels: hours of housework
per week is estimated to be 32 minutes lower than if the individual has GCSEs or below, ceteris paribus.
80
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
81
Unexpectedly, as income increases, the hours of housework will rise more quickly for an individual who is satisfied with their leisure
time, than for an individual who is not satisfied with their leisure time, ceteris paribus, which goes against the theory that people
with more income will outsource housework and therefore be happier with their leisure time.
5.
Conclusion
The aim of this investigation was to explore the determinants of household labour and to test the pre-disposed assumptions of
gender roles. The final revised model has exposed the variables that impact on housework and has been successful in finding
variables that show strong statistical significance. All of the variables were significant to at least the 5% level, with female,
satisfaction with leisure and being a female with children has the largest impact on hours of housework per week, which is
consistent with previous literature.
Further research could be conducted on income as a determinant of housework, linking in satisfaction with leisure time and
looking at the trends. This study has formed a basis for which research could delve into how individuals allocate their leisure time
and housework hours.
A likely reason for the specification problem stated earlier could be due to there being two dividing groups within hours of
housework: those who do very little to zero hours of housework a week, which would need to be estimated using a probit model,
and OLS would be used for those who do significantly more hours. To fully analyse the impact of these two groups, a statistical
model such as the Tobit model would have to be undertaken which will combine a probit regression and an OLS regression.
However, Tobit is an advanced statistical technique not covered in this study, but could further encourage more research in this
area.
81
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
82
Bibliography
Becker, G. (1965) “A Theory of the Allocation of Time”, The Economic Journal, Royal Economic Society,
Vol 75, No. 29, Pages 493-517.
Becker, G. (1973) “A Theory of Marriage: Part I”, Journal of Political Economy, University of Chicago
Press, Vol 81, No. 4 , Pages 813–846.
Bianchi, S. Milkie, M. Sayer, L. and Robinson, J. (2000) “Is Anyone Doing the Housework? Trends in the
Gender Division of Household Labor”, Social Forces, University of North Carolina Press, Vol 79, No. 1,
Pages 191-228.
Bond, S. and Sales, J. (2001) “Household Work in the UK: An Analysis of the British Household Panel Survey
1994”, Work, Employment and Society, Vol 15, No. 2, Pages 233–250.
Brines, J. (1994) “Economic Dependency, Gender, and the Division of Labor at Home”, American
Journal of Sociology, University of Chicago Press, Vol 100, No. 3, Pages 652-688.
Bryan, M. and Sevilla-Sanz, A. (2010) “Does housework lower wages? Evidence for Britain”, Oxford
Economic Papers, Oxford University Press, Pages 1-24.
Carrasco, C. and Domınguez, M. (2014) “Measured time, perceived time: A gender bias”, Time and
Society, Sage Publishing, Pages 1-22.
Fernandez, C. and Sevilla-Sanz, A. (2006) “Social Norms and Household Time Allocation”, IESE Business
School Working Paper, Social Science Electronic Publishing, No. 648, Pages 1-28.
Hersch, J. and Stratton, L. (1994) “Housework, Wages, and the Division of Housework Time for Employed
Spouses”, The American Economic Review, American Economic Association, Vol 84, Pages 120–125.
82
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
83
Kan, M.Y. and Laurie, H. (2016) “Gender, ethnicity and household labour in married and cohabiting
couples in the UK, ISER Working Paper Series, University of Essex, Pages 1-23.
Marini, M. and Shelton, B. (1993) “Measuring Household Work: Recent Experience in the United States”,
Social Science Research, Vol 22, Pages 361-382.
Podor, M. and Halliday, T.J. (2009) “Health status and the allocation of time”, IZA Discussion Paper,
Health Economics, No. 4368, Pages 1-19.
Presser, H. (1994) “Employment Schedules Among Dual-Earner Spouses and the Division of Household
Labor by Gender”, American Sociological Review, Sage Publications, Vol 59, No. 3, Pages 348–364.
Layte, R. (1998) “Gendered Equity? Comparing Explanations of Women's Satisfaction with the Domestic
Division of Labour”, Work, Employment and Society, BSA publications, Vol 12, No. 3, Pages 511-532.
Understanding Society (2016) “About the study”. Last Accessed 08/03/2016. Available at:
https://www.understandingsociety.ac.uk/about
83
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
84
Appendix.
Appendix 1a: OLS regression 1 – total monthly personal income gross.
84
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
85
Appendix 1b: Descriptive statistics on total monthly personal income gross
Statistics
total monthly personal income gross
N
Valid
Missing
47078
79
Mean
1655.6528
Median
1300.0001
Mode
Std. Deviation
Range
.00
1615.62235
15000.00
Minimum
.00
Maximum
15000.00
Appendix 1c: OLS regression 2 – ln(total monthly personal income gross)
85
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
86
Appendix 2a: RESET test - testing satisfied leisure and neutral leisure binary variables
Appendix 2b: RESET test - testing the natural logarithm of dependent variable housework
Appendix 3: OLS using natural logarithm of hours of housework per week
86
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
87
Appendix 4: Chow Test
Pooled Data
Female Data
87
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
88
Male Data
88
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
89
Does producing green product create opportunity for companies? Is living area being important factor for
companies to consider?
By Winnie Ho
1. Introduction and Background
In last several decades the number of environmentally conscious individuals is increasing. Nowadays, almost 76%
of UK adults mentioned that they would pay attention to ethical and green products according to the UK customer
trend 2015 (2014).This concern might raise the level of environmental consciousness and green product purchasing
behaviour (Gan et al.,2008; Miller and Layton,2001; Roberts,1996). The possible reason is that they have realized
their behaviour would have direct impact on the environment.
The Research from Pure Earth/Blacksmith Institute indicated that pollution kills 8.4 million people per year. (UK
Customer trend 2015, 2014) Moreover, pollution has driven up the economic cost of deaths. For instance, a study
by World Health Organisation (WHO) in 2005 shows that financial cost of air pollution increases dramatically in
Europe. It is more than $1.6 trillion each year. The bar chart in Figure 1 shows the ranking of suffering Economic
cost of deaths among the 10 courtiers.UK is in top six. (Harvey, 2015)This cost might keep rising by development
of science and technology in the world since these activities worsen pollution. It is clearly enough to show that
environment issue would bring the problem of death and the increase the burden of economic cost in economy.
Since the environment continues to worsen, it has become a public concern in what products they buy and what is
impact on environment from the products. Purchasing green products is good for economy because it can reduce
the burden of government spending on pollution and health. It can also improve the efficiency in the economy
because of the better health.
At the same time, there is a good opportunity for business to product green products. Several companies have
already tried to produce green products to the market and they are successful. Some famous examples are IKEA,
NIKE and LUSH. Selling green products could create a positive image and build the customer loyalty to firms due
to the good company’s reputation by ethical behaviour.
Moreover, green consumers become a large group of potential customers to the companies. Since 2000 the sales of
green products have soared 451% from £1.5 billion to £8.04 billion in 2010. The average household currently
spends over £305 per year on green retail goods and it predicts will rise to £648 in 2015. Although UK customers
need to pay 44% more for green products than standard alternative nowadays, the sales of green products keeps
increase. (The Telegraph, 2010) It shows that even customers should pay premium on the green products, there are
89
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
90
more customers are willing to pay for it. It also implies that they are prepared to show environmentally friendly
attitudes. Therefore, there is a trend that customers are willing to buy more and pay more for green products.
In the following paragraphs, this paper would conduct some models to identify what factors affect people to pay
more for the green products.
2. Literature Review
This report considers several literatures. Many literatures are about the relationship between environmentally
friendly purchase behaviour and different demographic characteristics. Few articles use the extent of willingness to
90
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
91
pay more for green products as dependent variable. However, it is not the big problem. Both of the articles assess a
similar idea and both suggest a tendency to display environmentally friendly attitudes. Therefore, these articles can
be the supporting documents to show whether some variables affect customers’ environmentally friendly attitudes
to result in paying more for green products.
This paper would also mainly focus on living area explanatory variable because companies are interested to know
whether it is important factor and what residential area should the stores mainly located. This report would analyse
other possible demographic variables as well, using the data collected in Understanding Society survey.
The first research is done by the Barbaro-Forleo, et al.(2001). The survey aimed at finding out the target customers
who are willing to pay more for environmentally friendly products. It used age, sex, number of children, level of
education and marital status as explanatory variables. The result identified that female will be willing to pay more
for green products as being more environmentally concerned than males. (Barbaro-Forleo, et al.,2001; Berkowitz
and Lutterman, 1968; Webster, 1975; McIntyre et al. 1993; Banerjee and McKeage, 1994). Moreover, household
who are married and have at least one number of children would pay more for green products. The reason behind is
they concern health status of their family member including their future generations.
The second survey is conducted by Gan et al. (2008). This research investigates environmentally friendly purchase
behaviour. Logistic regression is used in this analysis and the model is estimated by maximum likelihood method.
Moreover, it creates dummy variables for different qualitative explanatory factors. It covers the similar explanatory
variables as pervious study. In addition, it includes other variables which are level of education and monthly
income.
Age difference according to the report that done by Chhay, et al.(2015) provides that the relationship between pay
more for green products and age is U-shaped. (Gan, et al. ,2008) Firstly, young customers might want to pay more
on green products because new generation have more environmental knowledge and they are less affected by
conflicts between environmental concern and economic interests.(Grendstad et al., 2001; Jones & Dunlap, 1992;
Malkis & Grasmick, 1977) These reasons arouses their environmentally friendly attitude. For the middle age, they
would concern their economic interests more than environment since they need to save money for their children’s
education and purchasing properties. For the elderly, they might be less conflict between environmental concern
and economic interests because they have already bought properties and their children go to work. Additionally,
elderly might concern their health status because they might be easier to have health problems. Thus, they are
willing to pay more for green products which are healthier than conventional alternatives and without harmful
chemicals.
Since the relationship between dependent variable and age is non-linear and U-shaped, it will include age square in
91
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
92
the model. The research also reveals that consumers who are better educated, higher income, and married more
likely to purchase green products.
The third survey is conducted by Roozen(1997). In this survey, it focuses on the environmentally friendly
detergents. This survey is based on the binomial logit model. It mentioned living area as another explanatory
variable that can be considered in model. Urban residents are more likely to purchase green products because they
are frequently facing the pollution including air, water, noise and waste disposal problems. Therefore, they are
more likely to realize that the seriousness of environmental issue. In addition, because the pollution affects their
daily life directly, it increases their environmentally friendly attitudes to buy green products. On the other hand,
referring to Grendstad et al (2001), rural residents might be more likely to purchase green products. The reason
behind is that they have more utilitarian approach to nature than urban residents since their survival depends on it.
It indicates that they would be easier to exhibit environmentally friendly behaviour. In this paper, it would use
urban as the dummy variables for analysis.
Overall those researches, some of the explanatory variables should be included in the model. They are age, sex,
income, educational level, marital status, number of children and living area.
3. Data
The data used in this study is wave 4 from Understanding Society (The UK Household Longitudinal Study) which
is building upon the British Household Panel Survey data. These survey data pay attention on the changing UK life. The data
indicates information about attitudes, social and economic circumstances, behaviour and health. There are 92
variables with 47157 observations in condensed dataset that would be used in this survey. Besides, this study
would deal with missing values and recode some of the selected dependent and explanatory variables from the
dataset. Dependent variable would be mentioned first, and then explanatory variables.
For the dependent variable, it is qualitative variable called “pay more for the environmentally friendly products”
(d_scenv_pmep). The answers are given by 1. Strongly agree; 2. Tend to agree; 3. Neither agree nor disagree; 4.
92
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
93
Tend to disagree; 5. Strongly disagree. The inapplicable and missing values have been removed. The frequencies
table is shown in the Figure 2. The percentage of missing value is 17.7%. After deal with the missing values,
remaining samples are 38809 samples which are larger than 500 samples. Thus, sample size of the dependent
variable is large enough.
Next, dependent variable need to be recoded into binary form. (1= Agree 0= Disagree) 1, 2 are classified as agree
and 3, 4, 5 classified as disagree. There is a new variable called Agree_paymore. It classifies 3 as disagree because
observations will be more even distributed to 4 and 5 to have a more accurate analysis. Crosstab table in Appendix
1 shows that the new variable has been coded correctly.
For explanatory variables, most of them have been outlined in the literature review. There are two variables which
are supports a particular political party and whether belong to a religion that have not been mentioned. Referring to
Lee and Norries(2000),environmentalism suggests that environmental concern might be linked to political attitude.
Individual who is interested in politics is expected to protect environment and might be willing to buy and pay
more for green products. Moreover, religion can affect consumer attitudes and behaviour such as food purchasing
decisions. (Suki, 2015; Pettinger et al., 2004). Therefore, these two variables would be included. All of the
selected variables should be deal with the missing values.
After considering the dependent and explanatory variables, dummy variables for qualitative variables should be
created. Sex(d_sex) , Urban or rural area, derived (d_urban_dv),highest qualification(d_hiqual_dv), legal marital
status(d_marstat), supports a particular political party(d_vote1), and whether belong to a religion(d_oprlg) turn into
dummies. Expected signs of those variables are determined by the pervious researches. This model would take natural
93
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
94
logarithm of monthly labour income in order to normalize the data. Figure 3 shows the summary of expected
values and recoding.
94
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
95
95
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
96
Then, Figure 4 presents the descriptive statistics of dependent and explanatory variables. It shows minimum,
maximum, mean, standard deviation and skewness of the variables. There are no negative numbers in minimum
column as missing values are removed. The lnincome is negative in the minimum column because it is natural
logarithm of very small amount of monthly labour income. Age2 has large standard deviation because it squares
the age.
After that, this paper would consider which models would be used for analysis.
96
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
97
4. Model
This report would include three different models which are Ordinary least squares model (OLS), logit and probit
model.
OSL model would be used as starting point. It gives results based on the Linear Probability Model (LPM) which is
the simplest binary choice model. It figures out the data that minimises sum of the residuals as the best fits.
However, there is a problem that OLS is no guarantee predicted values of dependent variable will lie between
range of 0 and 1. Since the dependent variable is binary, some of the predicted values might lie over 1 or below 0.
It shows in Figure 5. As there is a drawback of OLS model, it would mainly focus on logit and probit models as the
next models for empirical analysis. Logit and probit model could guarantee the predicted probabilities lie between
0 and 1.Moreover, both logit and probit models allow determining the predicted probability for selected
explanatory variables.
Here is the empirical relationship:
Agree_paymore = β0 + β1Age + β2Age2+β3Female+ β4 lnincome+ β5Urban+β6Degree+ β7Married+ β8Number of
own children in household +β9 Support_party+ β10 Religion + u
After testing OLS model, its results would be compared with the results from logit model by looking at their
significance levels using sensitivity analysis. It would also conduct the Hosmer-Lemeshow test to test the goodness
of fit to the model. Then, probit model would be performed and compare the output with logit model.
97
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
98
5. Empirical Analysis
Before running the regression models, it should run the correlation matrix between explanatory variables shown in
Appendix 2a and collinearity diagnostics table shown in Appendix 2b to identify any problem of multicollinearity.
The reason is that multicollinearity usually occurs if two or more variables are highly correlated.
According to the correlation matrix, if the value between two variables is closed to one, it is likely to have perfect
multicollinearity problem. The matrix shows that Age and age2 have strong correlation with 0.980. In addition,
referring to collinearity diagnostics table, VIF is 45.587 and 45.612 for age and age2 respectively. VIF is a
measure of extent to which the variance of the OLS estimator is inflated because of multicollinearity. Moreover,
the tolerance(TOL) is 0.022 and 0.022 for age and age2. If there is perfect collinearity, TOL is zero. In this case,
age and age2 have large VIF and TOL is closed to 0. This is the expected result because age2 is the function of
age. Other variables show small correlation with each other. It means that there is no multicollinearity problem in
the model.
Then, it can run the first regression which is OLS. Appendix 3a shows the output of OLS. Firstly, R2 would be
considered to understand the overall explanatory power of the model. R2 is 0.032. It means explanatory variables
explain 3.2% of variations in agree to pay more for green products. It is not unusual that R2 would be small when
dependent variable is binary.
Next, signs and significance levels would be considered. Most of the associated signs with the coefficients are
same as expected, except age,age2, female and number of children variables. Most of them are negative sign,
except age2 is positive. Furthermore, most of the variables are statistically significant at 1% level and income is
significant at 5% level. Female, married and religion are shown as insignificant. After excluding these 3
insignificant variables, all the variables become significant at least to the 5% level.
After that, Appendix 3b shows that predicted values of maximum and minimum are 0.1241 and 0.6787
respectively in the OLS model. These values are inside the 0 and 1. It means that OLS could be a proper regression
for estimating the model. However, this research will also focus the result of logic and probit models because they
are more suitable models for binary dependent variable.
Referring to Gan(2008), logit model could be better explain consumers’ purchasing behaviour towards green
products since the random term is assumed to have a logistic distribution. Therefore, next step is to produce logit
models using same variables.
98
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
99
Appendix 4a shows the logit model. The model summary table in Appendix 4b indicates how much the variance of
dependent variable is explained by predicted variables. Nagelkerke R square shows that 4.3% of variance is
affected by predicted variables. The value of R2 is far from one. Nevertheless, R2 is not the suitable estimator to
show whether the model is good fit or not. Hosmer and Lemeshow goodness-fit test would be performed to show
whether the data fit to the model later.
Then, logit model would compare the significance level with OLS model. In the logit regression, there are 5
variables are significant at 1% level, age and income variables are significant to 5% level and female, married and
religion variables are insignificant. Comparing the significance levels with OLS model, the outputs in Figure 6
shows that Female, married and religion are individually insignificant in both models.
In order to know whether these 3 variables are jointly significant, likelihood ratio test will be carried out. This test
will compare the difference between restricted model and unrestricted model with the chi-square critical value to
find out whether it should reject the null hypothesis or not. Restricted model will exclude the insignificant
variables from the model. The figures of Log likelihood could be found in chi-square table in the Appendix 5 and it
also includes the calculation procedure.
99
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
100
H0: β3=0, β7=0, β10 = 0
H1: β3≠0, β7≠0, β10 ≠0
LR is larger than Chi-square statistic at even 1% significance level so reject the null hypothesis and conclude that
female, married and religion are collectively effect on the dependent variable. These 3 variables would remain in
the model.
Next, Hosmer and Lemeshow test will be used in order to identify the goodness of fit to the regression model.
Appendix 6 shows the output for Hosmer and Lemeshow Test.
H0: No specification error (Good Fit)
H1: Specification error (Bad Fit)
There is large p value which is 0.170 that null hypothesis cannot be rejected even by 10% confident interval. Thus,
the model is good fit with the data and no specification error.
In addition to this test, classification tables in Appendix 7 would be considered in order to look at in conjunction
with other information. This table provides percentage of correct predictions which is 65.3% if it assumes all the
predicted values are coded as 0. Moreover, the next table reports how well our model does using cut value of
0.500. The figure is 65.5%. This result shows that our model is acceptable.
After that, this research would consider coefficient and odd ratio of explanatory variables in the logit model. Figure
7 shows the outputs. Odd ratio is the exponentiation of the coefficient. It can explain the odds are in favour or not
against toward dependent variable. If the odd ratio is closed to 1, it shows that the effect of coefficient to the
dependent variable is small, such as married and religion. If the coefficient is positive, the odd ratio would be
greater than 1. It indicated that there is more favour in paying more for green products. For instance, individual
who lives in urban area is willing to pay more for green products than individual who live in rural area by 0.144
times. From this result, we can understand that individual who has degree, support particular party and lives in
urban area are more favour to pay more for green products.
On the other hand, if the coefficient is negative, the odd ratio would be smaller than 1. It indicated that there is less
favour in paying more for green products. Age, female and number of own children have the negative coefficient
and the odd ratio less than 1. It means that individual is female, aged, more children in the household are less
favour on paying more for green products. These results are different as expected from the literatures. Here are the
possible reasons. For female variable, it might due to more males concern their environmentally friendly purchase
behaviour nowadays because of the exacerbated environmental issues. For the number of children, if the number of
100
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
101
own children in the household increases, the individuals might not be affordable to pay more on green products
due to high burden on living expenses.
The main reason of different expected sign might be most of the literatures are about the relationship between
environmentally friendly purchase behaviour and different demographic characteristics. Thus, the expected sign of
the coefficients might be different if there is about relationship between pay more for green products and
demographic characteristics in my model.
In the logistic regression, it also shows that the coefficients of age is negative (-0.017096) and coefficient of age2
is positive (0.000308).It indicated that age has U-shaped relationship with dependent variable. (Gan, et al.
,2008)The probability of agree to pay more for green products reduces at increasing rate as age increases.
Moreover, their p-values are 0.020644 and 0.000317 for age and age2 respectively. Thus, there is strong evidence
of a U-shaped relationship as the P-value is small.
The literatures of Gan et al. (2008) and Chhay et al. (2015) also show that there is U-shaped relationship in age. In
order to prove these findings , the following would calculate the turning point of Age so as to find the age at which
probability of agree to pay more for green products is minimised. This calculation would use the coefficient of age
and age2 from the OLS model.
F = -0.004Age + 0.000073Age2
101
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
102
dF
 0.004  2 * 0.000073 Age
dAge
Set equal to 0 and solve for Age,
0= -0.004 + 2*0.000073Age
Age=0.004/0.000146
Age= 27.397
Age = 27.40 years old
The second derivative is positive (0.000146). Therefore, turning point is a minimum. The probability of agree to
pay more is minimised at age = 27.40 years old.
Additionally, predicted probabilities table could be performed to shows the predicted probability of agree to pay
more for green products across age in different sex using the coefficient from logit model. The result is shown in
Appendix 8a and the graph is shown in figure 8.Two curves of male and female both shows the U-shaped
relationship and the minimum point is at 27.4 years old.
This result proves that younger and elderly would be willing to pay more for green products than the middle-aged
individuals.(Gan et al.,2008)
102
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
103
Then, predicted probabilities table could be used to find the difference of predicted probability between urban and
rural, assume that everything remains the same (ceteris paribus) except the characteristic of living area variable.
Calculation is shown in Appendix 8b. Predicted probability is 0.381242354 for urban area and predicted
probability is 0.34788104 for rural area. The difference is 0.0334.Thus, result proves that urban residents has
marginal effect on agree to pay more for green products than rural residents.
Further estimation method would be probit model. It would compare the coefficients and significance level with
logit model. Appendix 9 shows results of probit regression. According to Figure 9, signs of coefficients are same
and significance levels are similar between two models. The magnitudes are different because logit results are
based on a logistic distribution and the probit results are based on standard normal distribution. Overall, there is
small statistical difference between two models. Predicted probabilities table in appendix 10 could prove they have
small difference in predicted probabilities. The probability of is 0.347881044 and 0.404330413 for logit and probit
models respectively. Thus, the difference is 0.05645 between two models.
103
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
104
Then, it would split the sample up into different group to understand whether the results are different in different
ethnic group based on OLS model. It would split the samples into White and Nonwhite group using ethnic group
variable (d_racel_dv). The new variable is named White. (d_racel_dv <= 4) Output is shown in the appendix 11.
The result shows that individual who is non-white might not be willing to pay more for green products even their
have higher income. The reason behind could be they tend to save their income rather than spend it if they have
more income. Urban resident is still willing to pay more to buy green products no matter they are white or nonwhite. The relationship between age and dependent variable is inverted U-shaped in the non-white group. It means
that middle-age group would be willing to pay more for green products than younger and elderly.
To sum up, the results between white and non-white are different. The possible reason could be the culture
difference that affects their environmentally friendly purchase behaviour.
6. Conclusion
104
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
105
In conclusion, the purpose of this research is identifying whether there is an opportunity for companies to produce
green products and whether living area is important factor to affect them to pay more.
Firstly, there is a trend that consumers are more concern what impact of their behaviour on the environment as
mentioned in Introduction part. Producing green products might have higher cost than standard alternative. Thus,
the price of green products might go up. Even consumers need to pay premium on the green products, many of
them are still willing to purchase for it. Selling green products does not only improve the overall public health in
the economy, but create the huge potential market for firms as well. It can also create more jobs because of the
investigation of green products. Therefore, producing green products is beneficial to companies and economy.
Secondly, if companies decide to produce green products, it should consider who the target customers are. Living
area is important factor as firms need to know where should the stores mainly located to sell the final products.
According to my survey, the results conclude that urban area is important variable for company to consider as it is
significant to the model and its coefficient is positive in all of the regressions. It means that urban residents are
willing to pay more to buy green products. Thus, firms should mainly focus on selling their green products in
urban area.
Moreover, the model shows that consumers who are younger or elderly, male, higher income, degree, married, few
numbers of children, supporting a particular party and having religion would be willing to pay more for green
products. However, these variables might be changed based on different ethnic groups. Firms could also consider
these variables.
In fact, this model might not be perfect to reflect the potential customers, since few literatures use the extent of
willingness to pay more for green products as dependent variable. Moreover, it does not include some relevant
variables as some of them cannot find from the dataset, such as price, quality, brand, and convenience which are
traditional product attributes. (Gan,2008; Anderson and Hansen, 2004; Ottman, 2000).
However, this paper can be a good reference for further research on the relationship between the green purchasing
behaviour and the demographic variables.
Reference List
Journals:
105
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
106
Barbaro-Forleo ,G. , Bergeron ,J. and Laroche ,M. (2001) 'Targeting consumers who are willing to
pay more for environmentally friendly products', Journal of Consumer Marketing, 18(6), pp.
503 - 520 [Online]. Available
at:http://www.emeraldinsight.com/doi/pdfplus/10.1108/EUM0000000006155 (Accessed: 26
February 2016).
Chhay, L. , Mian, M. and Suy, R. (2015) 'Consumer Responses to Green Marketing in Cambodia', Open Journal of
Social Sciences, 3(10), pp. 86-94 [Online]. Available
at:http://www.scirp.org/journal/PaperInformation.aspx?PaperID=60503 (Accessed: 1 March 2016).
Gan ,C., Kao ,T., Ozanne ,L. and Wee ,H. Y. (2008) 'Consumers’ purchasing behavior towards green products in
New Zealand ', Innovative marketing : IM.- Sumy : Business Perspectives, 4(1), pp. 93-102 [Online]. Available
at:http://businessperspectives.org/journals_free/im/2008/im_en_2008_1_Gan.pdf(Accessed: 23 February 2016).
Grendstad ,G., Olli,E. and Wollebaek ,D. (2001) 'Correlates of Environmental Behaviors Bringing
Back Social Context', Environment and Behavior, 33(2), pp. 181-208 [Online]. Available
at: http://eab.sagepub.com/content/33/2/181.full.pdf (Accessed: 26 February 2016).
Lee ,A. and Norries ,F.A. (2000) 'Attitude toward environmental issues in East Europe',International Journal of
Public Opinion Research, 12(4), pp. 372-397 [Online]. Available
at: https://ijpor.oxfordjournals.org/content/12/4/372.full.pdf (Accessed: 1 March 2016).
Suki ,N.M. (2015) 'Does religion influence consumers’ green food consumption? Some insights from
Malaysia', Journal of Consumer Marketing, 32(7), pp. 551 - 563 [Online]. Available
at: http://www.emeraldinsight.com/doi/pdfplus/10.1108/JCM-02-2014-0877(Accessed: 3 March 2016).
Websites:
Harvey ,F. (2015) Air pollution costs Europe $1.6tn a year in early deaths and disease, say WHO
, Available at: http://www.theguardian.com/environment/2015/apr/28/air-pollution-costseurope-16tn-a-year-in-early-deaths-and-disease-say-who (Accessed: 24 February 2016).
MINTEL (2014) UK Consumer Trends for 2015, Available at:http://www.mintel.com/press-centre/social-andlifestyle/mintel-identifies-four-key-uk-consumer-trends-for-2015 (Accessed: 23 February 2016).
106
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
107
The Telegraph (2010) Green goods cost nearly 50% more, Available
at:http://www.telegraph.co.uk/news/earth/earthnews/7785705/Green-goods-cost-nearly-50more.html (Accessed: 25 February 2016).
Appendix
Appendix 1
Crosstab table
Appendix2
2a: Correlation Matrix:
Age
Age2
Female
lnincome
Urban
Degree
Married
Number
of own
children
Support
_party
Religion
Age
1
0.980
0.007
0.120
-0.109
-0.051
0.269
-0.251
0.229
0.236
Age2
0.980
1
0.007
0.055
-0.109
-0.088
0.200
-0.300
0.228
0.237
Female
0.007
0.007
1
-0.197
0.001
0.015
-0.053
0.038
-0.063
0.102
107
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
108
lnincom
e
0.120
0.055
-0.197
1
0.019
0.251
0.139
0.061
0.059
-0.001
Urban
-0.109
0.109
0.001
0.019
1
0.005
-0.068
0.043
-0.031
-0.006
Degree
-0.051
0.088
0.015
0.251
0.005
1
0.108
0.092
0.057
0.012
Married
0.269
0.200
-0.053
0.139
-0.068
0.108
1
0.210
0.097
0.157
Number
of own
children
-0.251
0.300
0.038
0.061
0.043
0.092
0.210
1
-0.068
0.007
Support
_party
0.229
0.228
-0.063
0.059
-0.031
0.057
0.097
-0.068
1
0.162
Religion
0.236
0.237
0.102
-0.001
-0.006
0.012
0.157
0.007
0.162
1
2b: Collinearity diagnostics from linear regression:
Explanatory variable:
Tolerance
VIF
Age
0.022
45.587
Age2
0.022
45.612
Female
0.918
1.089
lnincome
0.757
1.320
Urban
0.986
1.014
Degree
0.908
1.101
Married
0.735
1.360
Number of own children in
household
0.749
1.336
Support_party
0.946
1.057
Religion
0.923
1.083
108
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
109
Appendix 3
3a: Output from OLS regression
3b: Unstandardized Predicted value using OLS
Appendix 4
4a: Regression model with Binary logistic
109
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
110
4b: Model Summary
Appendix 5
Likelihood ratio test in the Logistic regression
-2 Log likelihood
Unrestricted model (LogLU)
Restricted model (LogLR)
28309.038
28414.742
LR=2(LogLU-LogLR)
= -28309.038-(-28414.742) =105.704
Degree of freedom is 3 and Chi-Square critical value at 10%, 5% and 1% significant level.
10 % level: χ 2 (3, 0.10) = 6.25
5% level: χ 2 (3, 0.05) = 7.82
1% level: χ 2 (3, 0.01) = 11.34
LR > χ 2 (105.704>6.25; 105.704 >7.82; 105.704 >11.34) so reject null hypothesis.
Appendix 6
Hosmer and Lemeshow Test
110
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
111
Appendix 7
Classification table of logit model
Block 0:
This table assume all predicted values are coded as 0.
111
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
112
Bock 1: ( include all variables)
This table uses 0.500 as the cut value. Predicted values greater than 0.5 are classified as 1 and predicted values less
than or equal to 0.5 are classified as 0.
Appendix 8
8a: Predicted Probability based on logit model
Female
Coefficient(A) Characteristic(B)
AB
d_dvage
-0.0171
20
-0.34192
Age2
0.000308
400
0.1232
Female
-0.04731
1
-0.047311
lnincome
0.040451
8.29404964
0.335502602
Urban
0.144088
0
0
Degree
0.605702
1
0.605702
Married
0.027975
1
0.027975
d_nchild_dv
-0.04946
2
-0.098922
Support_party 0.239593
0
0
Religion
0.024774
0
0
Constant
-1.232593
1
-1.232593
SUM (AB): -0.628366398
EXP: 0.533462555
Predicted Probability : 0.347881044
The calculation is based on the independent variable from database with age: 20, female, lnincome is
8.29404964( ln(4000) ) , living in rural, degree, married, 2 children in households, not support a particular political
party and no religion.
112
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
113
Male
Coefficient(A) Characteristic(B)
AB
d_dvage
-0.0171
20
-0.34192
Age2
0.000308
400
0.1232
Female
-0.04731
0
0
lnincome
0.040451
8.29404964
0.335502602
Urban
0.144088
0
0
Degree
0.605702
1
0.605702
Married
0.027975
1
0.027975
d_nchild_dv
-0.04946
2
-0.098922
Support_party 0.239593
0
0
Religion
0.024774
0
0
Constant
-1.232593
1
-1.232593
SUM (AB): -0.581055398
EXP: 0.559307763
Predicted Probability: 0.358689783
The calculation is based on the independent variable from database with age: 20, male, lnincome is
8.29404964( ln(4000) ) , living in rural, degree, married, 2 children in households, not support a particular political
party and no religion.
Predicted Probability for male and female in different age range(20 to 80):
Age
Female
Male
20
0.347881044
0.358689783
25
0.344219544
0.354976446
30
0.344043494
0.354797871
35
0.347350381
0.358151687
40
0.354186672
0.365081646
45
0.36464343
0.37567295
50
0.378846671
0.390042066
55
0.396941235
0.408319848
113
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
114
60
0.419066567
0.430626527
65
0.445322927
0.457037355
70
0.475727495
0.487538731
75
0.510162001
0.521976981
80
0.548317154
0.560005555
The predicted probability is calculated based on the independent variable from database with age(20 to 80), female
or male, lnincome is 8.29404964( ln(4000) ) , living in rural, degree, married, 2 children in households, not support
a particular political party and no religion.
8b: Predicted Probability between urban and rural based on logit model
Urban:
Coefficient(A) Characteristic(B)
AB
d_dvage
-0.0171
20
-0.34192
Age2
0.000308
400
0.1232
Female
-0.04731
0
0
lnincome
0.040451
8.29404964
0.335502602
Urban
0.144088
1
0.144088
Degree
0.605702
1
0.605702
Married
0.027975
1
0.027975
d_nchild_dv
-0.04946
2
-0.098922
Support_party 0.239593
0
0
Religion
0.024774
0
0
Constant
-1.232593
1
-1.232593
SUM (AB): -0.484278398
EXP: 0.616141645
Predicted Probability: 0.381242354
114
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
115
The predicted probability is calculated based on the independent variable from database with age:20, female ,
lnincome is 8.29404964( ln(4000) ) , living in urban, degree, married, 2 children in households, not support a
particular political party and no religion.
Rural:
Coefficient(A) Characteristic(B)
AB
d_dvage
-0.0171
20
-0.34192
Age2
0.000308
400
0.1232
Female
-0.04731
1
-0.047311
lnincome
0.040451
8.29404964
0.335502602
Urban
0.144088
0
0
Degree
0.605702
1
0.605702
Married
0.027975
1
0.027975
d_nchild_dv
-0.04946
2
-0.098922
Support_party 0.239593
0
0
Religion
0.024774
0
0
Constant
-1.232593
1
-1.232593
SUM (AB): -0.628366398
EXP: 0.533462555
Predicted Probability : 0.347881044
The predicted probability is calculated based on the independent variable from database with age : 20, female,
lnincome is 8.29404964( ln(4000) ) , living in rural, degree, married, 2 children in households, not support a
particular political party and no religion.
Appendix 9
Probit model
115
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
116
Appendix 10: Predicted Probability between logit and probit models
Based on Logit:
Coefficient(A)
Characteristic(B)
AB
d_dvage
-0.0171
20
-0.34192
Age2
0.000308
400
0.1232
Female
-0.04731
1
-0.047311
lnincome
0.040451
8.29404964
0.335502602
Urban
0.144088
0
0
Degree
0.605702
1
0.605702
Married
0.027975
1
0.027975
d_nchild_dv
-0.04946
2
-0.098922
Support_party 0.239593
0
0
Religion
0.024774
0
0
Constant
-1.232593
1
-1.232593
SUM (AB): -0.628366398
EXP 0.533462555
116
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
117
Predicted Probability 0.347881044
The calculation is based on the independent variable from database with age: 20, female, lnincome is
8.29404964( ln(4000) ) , living in rural, degree, married, 2 children in households, not support a particular political
party and no religion.
Based on Probit:
Coefficient(A)
Characteristic(B)
AB
d_dvage
-0.011051
20
-0.22102
Age2
0.000195
400
0.078
Female
-0.030035
1
-0.030035
lnincome
0.023752
8.29404964
0.197000267
Urban
0.088492
0
0
Degree
0.371447
1
0.371447
Married
0.017148
1
0.017148
d_nchild_dv
-0.0302
2
-0.0604
Support_party 0.147544
0
0
Religion
0.017298
0
0
Constant
-0.739594
1
-0.739594
SUM (AB): -0.387453733
EXP 0.678783039
Predicted Probability 0.404330413
The calculation is based on the independent variable from database with age: 20, female, lnincome is
8.29404964( ln(4000) ) , living in rural, degree, married, 2 children in households, not support a particular political
party and no religion.
Difference between probit and Logit probabilities: 0.404330413-0.347881044=0.056449373
Appendix 11
Splitting Up the Sample(White and Non-white)
117
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
118
Create new variable: White (d_racel_dv <= 4)
White=1:
:
118
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
119
White=0 (Non-white):
119
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
OLS
White
Nonwhite
0.226666
0.210341
0.143948
Age(computed) -0.004206
-0.005370
0.012628
Age2
0.000073
0.000094
-0.000144
Female
-0.010386
-0.004472
-0.010464
lnincome
0.008809
0.013386
-0.003346
urban
0.031354
0.015864
0.026599
Degree
0.136094
0.146760
0.019630
Married
0.005967
-0.004640
0.078338
Number of
own children
-0.010311
-0.012186
-0.016576
Support_party
0.054526
0.054347
0.043249
Religion
0.004873
-0.015717
-0.009571
Constant
120
120
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
121
Austerity measures have strained Charities, will income growth relieve this pressure?
Introduction and Background
usterity measures have been introduced at every level of government, and we have seen the scope of
public services cut. This has increased the necessity for charities to step in and provide for those who are
no longer supported by social services. This has resulted in headlines such as "Charities in crisis as austerity
bites" in the FT, previously charities have been dependent on public sector outsourcing contracts, and these have
also been widely condensed (Wigglesworth, 2016). Corporate donations are also shying away,“giving by companies
in the FTSE 100 declined by 17 per cent in 2014” than 2013; this has only made the situation tougher for charities
(Gordon). Oxfam was widely criticised by the government for being political when openly stating its opposed
austerity. This shows the desperation that our third sector is facing, and now non-profits are even more reliant on
individual donations. It is essential that for charities to continue functioning that they understand their donator
and impose cost efficient marketing strategies to raise funds. However, this is not easy, not just the government
are tightening their wallets, households have seen stagnant income since 2008 making it ever more difficult for
charities. Figure 1 displays how real incomes had fallen during the interview period 2012-2014. The UK have seen
a slight improvement coming into 2016, but it is still nothing significant enough to reverse the implications of a fall
in real income.
Figure 1: Contributions to the growth of real regular pay: Consumer Price Index (CPI) inflation and the growth of average regular weekly earnings, 2007 to 2015
Source: Office for National Statistics
In principle, everyone sees the advantage of donating for one cause or another, but not all people donate in
practice. Charitable donations are particularly intriguing to study when put in an economic context, a majority of
theoretical frameworks are based on humans having self-interest (Homo economicus). Andreoni (2004) applies
rational choice theory but in the context of charitable giving, he displays its limitations in explaining individual
contributions to collective goods. This OLS model aspires to specify what fundamentally determines the amount
121
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
122
one donates. These forces must be strong as it is predicted UK adults contributed to the £10.6bn in 2014
(Charities Aid Foundation). Evidence suggests our less rational side and our susceptibility to more human traits, a
number of research articles identify donation occur for three primary reasons demonstrable, emotional and
familial. Demonstrable is the most selfish due to personal benefit in past or future, emotional due to our utility
gained from helping others and lastly familial, this is the concept it may help someone we know, this has been
widely researched (Sargeant, Ford and West, 2006). This paper will identify what influences individuals to donate
and what are the common economic, social and demographic factors have am implication and is income a
significant enough factor so that when incomes rise charity donations will rise enough to compensate
government cuts. To do this the paper will highlight variables that are determined as significant in explaining
levels of charity donation in previous literature. The model will be outlined, explained and the results of the
regression analysis interpreted to draw a conclusion on variables that impact donations, hopefully, the findings
can be used to predict if the financial strain on charities is only temporary.
Background
Research articles universally identify religion as a major component that determines charity donation, the
introduction of a Wiley article begins with“Charity and religion go hand in hand” (Ranganathan and Henley). This
correlation is so strong it justified its own variable “rlcharity” which was composed of answers to “My religious
beliefs affect my decisions for charitable giving and helping others” in the questionnaire of the UK Household
Longitudinal Study. For example, a mandatory religious duty for Muslims is Zakat al-Fitr15 (Opoku, 2012). Religion
is undoubtedly an essential explanatory variable. Its significance is enhanced as the dataset fails to include any
indicator of “Network extension” as in Wiepking (1995), we could assume that attending church integrates you
into a network that has positive tendencies towards donation. Due to this assumption church attendance is more
representative of this then self-identified religion.
The empirical analysis on charitable donation from the U.S. and Ireland cite income as a key explanatory variable.
This is very logical as if you have a higher income one will therefore be able to allocate some/more towards a
chosen cause (Ostrower). This model will help predict that when incomes rise will individuals increase donations.
The third most widely supported factor is age, and it is thought that as life experiences increase sympathy grows
increasing the likelihood to donate. These factors are all widely accepted. There is basis to claim once 65+
expenditure falls, resulting in a higher disposable income, in turn translating to higher donations.
The most cite can be quoted saying “Empirical evidence points to self-identified political conservatives’ greater
financial generosity when compared with liberals” (Vaidyanathan, Hill and Smith). This refers to U.S. data and may
also hold true for the UK. However, empirical evidence is not one sided or specific to the UK. Level of political
interest will display a sense of greater awareness outside one’s life, resulting in increased donation behaviour.
15
an annual charitable donation
122
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
123
Being environmental should display empathy and a feeling of responsibility outside one's own actions, the variable
also shows a willingness to part with money for moral reasons. This should be highly correlated to donating money
to charity.
Another interesting factor is education, it is not immediately obvious that this would have an effect. Carroll 1999
(an Irish study) identified individuals with “higher than secondary [school] education” are more likely to donate,
for this reason I will include degree. However, this study is from 1999 and with easier access to news and other
information sources through digitalization it may no longer have as significant effect as previously predicted.
These may be that having more children in the household will decrease the amount donated per child. However,
Carroll et al. stated the opposite effect (Carroll, Mccarthy and Newman, 2005). With added costs of raising a child, disposable
income will be and, therefore, donation. It is likely once kids move out of the house may then increase donation. I
intend to build upon current literature and to do this I will accumulate many models to include the most significant
variables. In addition, to what literature supports the inclusion of leisure satisfaction (satisfaction with the amount
of leisure time) would be interesting. Currently, there is no literature to support this. Being satisfied displays a
certain level of satisficing and a level of content, An individual who is not satisfied is likely to feel their resources
are too scarce to donate. For this reason, Leisure satisfaction will be included as an explanatory variable in order
to allow me to develop on current literature.
Model
The regression will use the amount of money donated to charity in 12m as the dependent variable. I will run 3 different
models, amount donated to charity will be logged as other works of literature have also done (in 2 of the 3 models),
this helps reduce skewness as without logging looking at descriptives it is 7.57 after logging it falls to .091, this skewness
distribution in order is more suited to an ordinary least squares regression analysis. In the case of model 3 it has the
added benefit allowing elasticities to be interpreted. This will be regressed against the explanatory variables identified
above. The dependent variable is both continuous and quantitative making Ordinary Least Squares (OLS) the most
appropriate method for the regression. OLS is a very simple model and could cause problems of misspecification, as it
assumes a linear relationship between the explanatory variables. As age is often non-linear, age has been divided into
categories to avoid this issue as there was mixed literature about if it was linear or non-linear, this should be a successful
coping strategy as done in previous literature.
Model 1
CharityPer12m=
C+𝛽1churchgoer+𝛽2degree+𝛽3thru25to29+𝛽4 thru30to34+𝛽5thru35to39+𝛽6thru40to44+𝛽7 thru45to49+𝛽8thru50t
o54+𝛽9thru55to69+𝛽10thru70plus+𝛽11 𝐼𝑛𝑐𝑜𝑚𝑒𝑃𝑒𝑟+𝛽12leisuresat+𝛽13political+𝛽14 noch*+𝛽15married+𝛽16
eco_peeps+𝛽17 𝑣olunteer
Model 2
LnCharityPer12m=
C+𝛽1churchgoer+𝛽2degree+𝛽3thru25to29+𝛽4 thru30to34+𝛽5thru35to39+𝛽6thru40to44+𝛽7 thru45to49+𝛽8thru50t
o54+𝛽9thru55to69+𝛽10thru70plus+𝛽11 𝐼𝑛𝑐𝑜𝑚𝑒𝑃𝑒𝑟+𝛽12leisuresat+𝛽13political+𝛽14 noch*+𝛽15married+𝛽16
eco_peeps+𝛽17 𝑣olunteer
Model 3
123
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
124
LnCharityPer12m=
C+𝛽1churchgoer+𝛽2degree+𝛽3thru25to29+𝛽4 thru30to34+𝛽5thru35to39+𝛽6thru40to44+𝛽7 thru45to49+𝛽8thru50t
o54+𝛽9thru55to69+𝛽10thru70plus+𝛽11 𝐿𝑛𝐼𝑛𝑐𝑜𝑚𝑒𝑃𝑒𝑟+𝛽12leisuresat+𝛽13political+𝛽14noch*+𝛽15married+𝛽16
eco_peeps+𝛽17 𝑣olunteer
*number of own children in household
Data
Data used in this study is compiled from a single cross-section (wave 4) from the UK Household Longitudinal
Study between 2012 and 2014. Focusing on Employment and socioeconomic characteristics. The original survey
interviewed 47,157 individuals, and all missing values have been removed.
Dependent variable:
The dataset possessed three charity related variables, however both “frequency donated” (d_charfreq) and
“donated money to charity” (d_chargv) failed to display the amount that the individual donated. The third option
was “amount given to charity last 12 months” this was the same recorded variable as in both Steven T. Yen
(2002) and Wiepking (1995), making it ideal. Zero values were excluded as done in Feldstein did in his early
studies on charity donation (Feldstein and Clotfelter, 1976), (Feldstein and Taylor, 1976). The variables were
recoded into binary variables this can be seen in figure 2 in the appendix.
The result was a large sample size of 26,917 observations. Figure 3 is a table of the descriptives, we can analyse
these figures in order to identify any issues. In the recoded binary variables we can see for the majority the mean
is not close to either one or zero. This indicates that there is a suitable amount of variation in the variables,
potentially the “thru25to29” and “thru30to34” variables are too close to zero,
124
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
125
Figure 3: Summary Statistics of Variables
Variable
Mean
Maximum
lnCharity12m
4.262
9.210
0.000
1.424
0.910
203.27
9997
0.000
491.023
7.577
7.039
9.620
-9.630
1.244
-3.580
1651.742
15000
-8750.000
1617.950
3.255
churchgoer
0.221
1.000
0.000
0.415
1.348
degree
0.344
1.000
0.000
0.475
0.657
Eco_peeps
0.463
1.000
0.000
0.499
0.149
thru25to29
0.066
1.000
0.000
0.249
3.476
thru30to34
0.077
1.000
0.000
0.267
3.173
thru35to39
0.081
1.000
0.000
0.273
3.076
thru40to44
0.096
1.000
0.000
0.294
2.748
thru45to49
0.094
1.000
0.000
0.292
2.787
thru50to54
0.088
1.000
0.000
0.283
2.917
thru55to69
0.223
1.000
0.000
0.416
1.334
thru70plus
0.1368
1.000
0.000
0.343
2.114
LeisureSat
0.3311
1.000
0.000
0.471
0.718
Political
0.418
1.000
0.000
0.493
0.331
0.500
7.000
0.000
0.924
1.987
married
0.517
1.000
0.000
0.500
-0.067
Volunteer
0.191
1.000
0.000
0.393
1.575
Charity12m
LnIncome Per
Income Per
Number of
own children in
household
Minimum
Std. Dev.
Skewness
therefore, meaning the model unlikely to capture its effect due to the small variation (bellow .10). This can be
monitored and if any issues occur later, these could be removed. We can also observe the minimum’s being at 0
as all missing values have been removed.
Figure 4: Original regression
125
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
Variable
β
Standa
rd Error
LnIncomePer
.163
***
.007
churchgoer
.805
***
.018
degree
.413
***
.017
Eco_peeps
.218
***
.015
thru25to29
.297
***
.043
thru30to34
.398
***
.042
thru35to39
.507
***
.042
thru40to44
.553
***
.040
thru45to49
.679
***
.039
thru50to54
.669
***
.040
thru55to69
.710
***
.34
thru70plus
.736
***
.036
LeisureSat
.129
***
.016
Political
.229
***
.016
.038***
.011
.175
***
.017
Number of own
children in
household
married
126
126
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
volunteer
.413
***
127
.020
***=significant at the 1% level (N=26917)
Empirical Analysis
As a preliminary check for multicolinearity a correlation matrix between the explanatory variables. This can be
seen in the Appendix figure 5. The highest correlation occurred between “Number of own children in household”
and “thru35to39” at 0.318. This is a medium value (between 0.3 and .49) and is well below 0.8, therefore, being
an acceptable figure and should be of no concern.
The initial regression of model 3 proved to be fairly inaccurate with an 𝑟 2 value of .236, meaning the explanatory
variables explained just 23.% of the variation in “LnCharity12m” This is relatively low, however for cross-sectional
data (rather than time series) it is expected. All of the variables can be accepted at the 1% significance level. This
means there is no need for a redundant variable test. All of the signs are as the literature stated and as the
predictions show in figure 2. We can see the large explanatory power that church goer has, this is in line with the
studies of Ranganathan and Henley (2008). We can also interpret the coefficient of “LnIncomeper” as an
elasticity. This means a 10% increase in income leads to a 1.6% rise in donation to charity. This should mean as
UK productivity increases and wages rise we should see an increase in the amount donated to charity, although
this will be very small. Model 1 and Model 2 regressions were also carried out; the results are shown in figure 6.
Model 2 was recorded the highest 𝑟 2 value at .26, being higher when compared to Model 3. Model 2
interestingly shows a coefficient on thru70plus of .925, meaning that individuals over 70 donate 93% more than
individuals under 25. This figure is slightly lower in Model 3 but the trend of donations is increasing as age
increases are consistent throughout the models. A probable cause is the result of lower spending from individuals
over 70+ as younger individuals have rent/mortgages to pay.
Diagnostic Tests
In order to verify if there was a no specification, a Ramsey RESET tests must be carried out. We test the null
hypothesis (the correct specification is linear) against the alternative (the correct specification is non-linear) that
does not indicate what the correct specification should be. This regression was found the p-value of the F-statistic
was highly significant at 0.000 for all three models (shown in figure 7 a-c in the appendix), which strongly
suggests there is an issue. The Null hypothesis is therefore rejected, this indicates that there is a specification
issue. The initial concern this test caused was that I had included an irrelevant variable (LeisureSat) which had
resulted in multicollinearity due to the unnecessary increase in the standard error of the coefficients,
consequentially making the usual t tests unreliable. I then attempted to exclude explanatory variables, as there
was a potential that some of them could be irrelevant. Even without “LeisureSat”, “Eco_peeps” and “Political” the
RESET test results still rejected the null hypothesis (***shown in appendix--excluded reset file--***). To improve
the specification I looked at another possibility being the exclusion of a relevant explanatory variable. However,
this model is based on literature and the most relevant variables within the limited available had been included.
After adjusting my models variables and their functional form while repeatedly re-doing Ramsey’s RESET test
where I continually rejected the null hypothesis. The variables that I added in order to ensure I had all of the
127
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
128
relevant variables were (NotSmoking16, Widowed, GCSE, Subjective financial situation- future). All of these have
some level of basis in current literature however, they did not add value to the model or rectify the misspecification issue. Potentially variables that are not in the dataset could offer some solution. An example of this
may be personality traits or amount inherited17 Both of these have some literature and would contribute to a
higher 𝑟 2 value. The last step is to change the functional form of some explanatory variables; An age2 variable
was then created in order to investigate a non-linear specification. I continually found the results of the Ramsey
RESET test to be unchanged, this has made me question the source of the mis-specification. Having tried all of the
main causes of specification issues I looked at the variability in my explanatory variables to ensure their
significance. As identified previously “thru25to29” and “thru30to34” variables had a low variation. I excluded
these from the model and still found the p-value of the F-statistic was highly significant at 0.000. The large
sample size of 26917 could be a source for this issue. Despite my efforts to source the problem seems apparent
the model has some issues in its OLS form. The evidence suggests that as the model’s explanatory variables have
a large number of zeros making it possible that a more advanced technique is needed (Tobit).
Estimating VIF:
VIF=
1
1−𝑅𝑗2
1
= 1−0.24= 1.32
1.32 is relatively low showing us that the model is not inflated by multi-collinearity. This is best interpreted in
tolerance (the inverse of VIF), if TOL is 1 that indicates perfect collinearity.
𝑇𝑂𝐿 =
1
= 0.76
𝑉𝐼𝐹
This is less than one meaning we do not have perfect collinearity but the model does not have multi-collinearity.
The White Test can be used in order to identify any heteroskedasticity in the model.
Ho= homoskedasticity
H1= heteroskedasticity
The test statistic is calculated by 𝑛𝑟 2 with n being the sample size. The 𝑟 2 is calculated via running a regression
where the quantitive explanatory variables (Number of own children in household and LnIncome) are squared and
the dependent variable is the residual value (from the original regression) squared. If the Null cannot be accepted
it indicates heteroskedasticity in the model (as shown in appendix figure 8).
Model 1
Carroll suggested that smoking had an implication with donation, it provided limited explanation for this variable, and I wish to include
smoker due to its surprising nature.
16
Avery found the average person gives $4.56 to charity each year for every $1,000 of (non)inherited wealth, but only $0.76 out of
inherited wealth (Avery 1994:29)
17
128
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
129
The 𝑟 2 value from this regression was 0.025 with the sample size being 26,917 observations.
Test statistic: 0.025 ∗ 26917 = 697.926
Model 2
The 𝑟 2 value from this regression was 0.008 with the sample size being 26,917 observations.
Test statistic: 0.008 ∗ 26917 = 223.336
Model 3
The 𝑟 2 value from this regression was 0.006 with the sample size being 26,917 observations.
Test statistic: 0.006 ∗ 26917 = 161.502
The critical value comes from the chi-squared table, the degrees of freedom are determined on the number of
explanatory variables present in the auxiliary model.
2
𝑥(17),0.05
=25
As even the lowest test statistic (161.2>25) is greater than the critical there is sufficient evidence that there is
heteroskedasticity is present in all of the models. This is an issue as observations will have a “fan” shape meaning
that observations made on the region of low variance are more predictable then that in the high variance region.
In order to improve the models reliability observations with lower variance must be more highly weighted in order
to increase the reliability of the model. The result of this heteroskedasticity is an issue with the OLS standard errors;
this has the knock-on effect with the t-values and the confidence intervals. This must be dealt with as OLS is no
longer the BLUE. Bootstrapping can be a useful tool to recalculate more accurate standard errors due to the
inflation from heteroskedasticity. By using estimates from bootstrapping the standard deviation of regression
coefficients can be calculated and used as the standard error of the regression coefficients. The largest changes
occurred in two of the age categories with small changes in degree. The model has not been massively improved
but its adjustments to the standard errors show some impact from homoskedasticity. All of the β are still Significant
at a 1% level which shows the power of the explanatory variables.
Figure 9: Bootstrap results
Variable
β
Standar
d Error
LnIncome
Per
.163***
.008
churchgoer
.805***
.020
degree
.413***
.018
129
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
130
Eco_peeps
.218***
.015
Conclusion
thru25to29
.297***
.045
thru30to34
.398***
.043
thru35to39
.507***
.041
thru40to44
.553***
.041
thru45to49
.679***
.040
thru50to54
.669***
.040
The aim of this investigation was to identify determinates of
charitable donation and see if they could provide an insight
into future charity contributions. What can be taken away with
confidence from model 3 is the elasticity of .163 between
income and charity donation. This is inelastic, as a 1% increase
in income will result in just a .163% increase in donation
(ceterius parabus). This could be of concern for charities as
when the economy improves; productivity and wages recover
the inflows of donations to charities will not significantly rise,
but we could expect an improvement in the current situation.
thru55to69
.710***
.33
thru70plus
.736***
.035
LeisureSat
.129***
.016
Political
.229***
.061
-.038***
.011
Number of
own children in
household
Charities may feel both a greater burden but also receive more
support through the result of an aging demographic. As a larger
portion of UK adults will fall into the age categories with the
largest coefficients we are likely to see a rise in donations.
There is potential that this is a generational preference rather
than an inevitable transition that occurs in tangent with aging.
Charities still have reason for concern,
This model has also found 12 variables that are all statistically
significant in predicting the amount of charitable donation. This
married
.175***
.017
model could instead be used to help reduce costs of marketing
for charities and be more meticulous with their spending. We
volunteer
.413***
.019
can conclude those who are married, political, satisfied with
their leisure time, hold a degree and attend church are most
***=significant at the 1% level (N=26917)
likely to donate; this information can be used to advertise in
local church newsletters, political websites and potentially leisure centers. All of these would implement
elements of findings that have been uncovered
One issue with the data coming from individuals rather than the charities is the potential for social desirability
bias when surveyed individuals may over-report their "good behavior" such as the amount donated. In a survey
one relies on honesty and the lack of an incentive for people over-report. In government papers and research
done by charity UK they use data from charities, this is more reliable and could provide more accurate results.
Further work
The issues that this regression has come across are a combination of 2 factors, first the missing variables and
second the regression method used. The inclusion of more variables such as personality traits, inheritance,
parents donation habits, what religion individuals were (e.g. Muslim, Jewish, Christian etc.), social networks. Also
a distinction in the type o charity, whether religious or not can be an essential factor. All of these would improve
the 𝑟 2 of the regression. Secondly this study is using OLS, previous studies have used tobit and double hurdle
model as proposed by Cragg (1971), these more advanced techniques may have been more appropriate and
130
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
131
would have prevented mis-specification. This is particularly essential due to the unique characteristic of
charitable donation, which makes it even more problematic to analyse accurately. 0 and missing values are
difficult to deal with, this issue originates on the factor “that the missing values probably are not Missing at
Random, an assumption that needs to be satisfied when imputing missing data” The basic OLS regression
struggles in this area.
Another area that may require further consideration is the relationship between charity donation and
volunteering, there is potential for an endogeneity problem. Literature is unclear if charity donation leads to
volunteering or volunteering leads to charity donation.
The last area I would like to look into is having a disposable income variable, this may help to see if older age
categories simply donate more due to lower expenditures and therefore a higher disposable income or if it is the
result of life experiences. This would give a further understanding into the predictors of charitable giving, there is
currently still no definitive argument for this trend.
131
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
132
Figure 2: Variable Descriptions
Old
variable
New variable
d_fimngrs_dv
LnIncomePer12
m
d_scenv_pm
ep
Eco_Peeps
Description/Values
Expected
Sign of
Coefficient
Log of total monthly personal income
+
Individuals who pay more for Eco
products
+
gross
Dummy. 1=purchase 0=does not
d_oprlg2
churchgoer
Individuals who attend church more than
once a month
+
Dummy. 1=more than once 0=less than
once
d_hiqual_dv
degree
Whether individual has academic
degree
+
Dummy. 1= holds degree 0=no degree
d_dvage
thru25to29
Individual ages 25 to 29
Dummy. 1=between 25-29 0=not 25-29
d_dvage
thru30to34
Individual ages 30 to 34
Dummy. 1=between 30-34 0=not 30-34
d_dvage
thru35to39
Individual ages 35 to 39
Dummy. 1=between 35-39 0=not 35-39
d_dvage
thru40to44
Individual ages 40 to 44
Dummy. 1=between 40-44 0=not 40-44
d_dvage
thru45to49
Individual ages 45 to 49
Dummy. 1=between 45-49 0=not 45-49
d_dvage
thru50to54
Individual ages 50 to 54
Dummy. 1=between 50=54 0=not 50-54
d_dvage
thru55to69
Individual ages 55 to 69
Dummy. 1=between 55-69 0=not 55-69
d_dvage
thru70plus
Individual aged 70 or older
Dummy. 1=between 70+ 0=not 70+
d_sclfsat7
LeisureSat
Whether individual is satisfied with the
amount of leisure
+
+
+
+
+
+
+
+
+
132
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
133
Dummy. 1= satisfied 0=not satisfied
d_vote6
Political
Whether individual has an interest in
politics
+
Dummy 1=Interest 0=no interest
d_nchild_dv
Number of own
children in
household
Number of own children in household
-
d_marstat
married
Whether individual is married
+
Dummy. 1=married 0=not married
d_volun
Volunteer
Whether individual volunteers
Dummy. 1=volunteer 0=not a volunteer
+
133
134
Figure 5: Correlation matrix-
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
134
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
135
Figure 6- Full regression table
Model 1
Variable
β
Model 2
Stand
ard Error
β
Model 3
Stand
ard Error
β
Stand
ard Error
IncomePer
00.053*
**
.002
.000
***
.000
-
-
LnIncome
Per
-
-
-
-
.163
***
.007
churchgo
er
240.250
***
6.598
.809
***
.018
.805
***
.018
degree
61.161*
**
6.144
.365
***
.017
.413
***
.017
Eco_peep
s
41.831*
**
5.627
.261
***
.015
.218
***
.015
thru25to29
15.472
15.01
7
.452
***
.040
.297
***
.043
thru30to34
03.038
14.63
6
.527
***
.039
.398
***
.042
thru35to39
-02.717
14.80
9
.612
***
.040
.507
***
.042
thru40to44
18.743
14.03
8
.668
***
.038
.553
***
.040
thru45to49
54.084*
**
13.65
4
.776
***
.037
.679
***
.039
thru50to54
71.769*
**
13.70
1
.773
***
.037
.669
***
.040
thru55to69
72.435*
**
11.66
4
.856
***
.031
.710
***
.34
thru70plus
82.814*
**
12.18
6
.925
***
.033
.736
***
.036
LeisureSat
31.394*
**
5.909
.122
***
.016
.129
***
.016
Political
29.921*
**
5.788
.189
***
.015
.229
***
.016
135
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
Number of
own children
in household
06.369*
3.869
136
0.37***
.010
.038***
.011
married
32.090*
**
6.191
.156
***
.017
.175
***
.017
volunteer
141.388
***
6.788
.387
***
.018
.413
***
.020
𝒓𝟐
.138
***=significant at the 1% level
.260
Figure 7a:
.236
**= significant at the 5% level *= significant at the 10%
level (N=26917)
Figure 7b:
136
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
137
Figure 7c
Figure 8- White Test
137
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
138
BibliographyBennett, R. (2006), “Predicting the Lifetime Durations of Donors to Charities”, Journal of Non-profit and Public Sector Marketing, Vol. 15,
No. 1/2, pp. 45-67.
Carroll, J. Mccarthy, S. and Newman, C. (2005). The Determinants of Charitable Donations In The Republic Of Ireland. The Economic and
Social Review, [online] 2. Available at: http://www.esr.ie/ESR_papers/vol36_3/03_Carroll_Article.pdf [Accessed 1 Mar. 2016].
Charities Aid Foundation, (2014). UK GIVING 2014. London.
Feldstein, M. and Clotfelter, C. (1976). Tax incentives and charitable contributions in the United States. Journal of Public Economics, 5(1-2),
pp.1-26.
Feldstein, M. and Taylor, A. (1976). The Income Tax and Charitable Contributions. Econometrica, 44(6), p.1201.
Gittell, R. Tebaldi, E.,(2006). Charitable Giving: Factors Influencing Giving in US States. Nonprofit and Voluntary Sector Quarterly 35, 4: 72136.Giving Japan 2012.
Gordon, Sarah. "FTSE 100 Companies Give Less To Charity - FT.Com". Financial Times. N.p., 2016. Web. 10 Mar. 2016.
Opoku, R. (2012). Examining the motivational factors behind charitable giving among young people in a prominent Islamic country.
International Journal of Nonprofit and Voluntary Sector Marketing, 18(3), pp.172-186.
Ranganathan, S. and Henley, W. (2008). Determinants of charitable donation intentions: a structural equation model. International Journal of
Nonprofit and Voluntary Sector Marketing, 13(1), pp.1-11.
Sargeant, A., Ford, J. and West, D. (2006). Perceptual determinants of nonprofit giving behavior. Journal of Business Research, 59(2), pp.155165.
138
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
139
Schokkaert, E. (2006). Empirical analysis of transfer motives. In: Kolm, S.C., Ythier, J.M. (Eds.), The Handbook of the Economics of Giving,
Altruism and Reciprocity, vol. 1. Elsevier, North Holland (Chapter 2).
Steven T. Yen (2002) An econometric analysis of household donations in the
USA, Applied Economics Letters, 9:13, 837-841
Wigglesworth, R. (2016). Charities in crisis as austerity bites - FT.com. [online] Financial Times. Available at:
http://www.ft.com/cms/s/0/be15e318-7c26-11e1-9100-00144feab49a.html#axzz42KLMuGwo [Accessed 8 Mar. 2016].
139
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
140
What factors determine people’s willingness to donate money to charity and their generosity when donating?
By Natalie Wood
1.
Introduction
As economists, we assume that everyone is rational, meaning that we are all self-interested and only gain utility from what we
consume. However, I believe that this is untrue and that we are able to gain utility from other methods besides consumption,
such as good will. Donating to charity could be classed as being altruistic and might therefore lead to an increased utility for
many people, which could be a reason that so many people are willing to donate money to charities. Having volunteered and
donated money to charities myself before, I am interested in asking the question: what factors determine people’s willingness
to donate to charity and their generosity when donating?
In order to answer this question, my report will include 8 sections: the introduction (1), background (2), model and data (3),
empirical analysis first model (4), empirical analysis second model (5) my conclusion (6), the bibliography (7) and the appendix
(8).
2.
Background
In order to see which variables I could include in my model, I firstly resorted to looking at previous literature. After completing
my initial research, I looked through the given data set and picked out the most relevant variables, checking for any missing
data to ensure that I didn’t reduce my sample size significantly.
140
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
141
From other studies such as Leslie et al (2013) and Willer et al (2014), I have seen the conclusion that women donate more than
men, found repeatedly. I have also seen that this can vary depending on the type of charitable cause such as education and
health related charitable organisations (Einolf, 2011). However, the dataset that I am using is restricted and so I can only
investigate this to the point of seeing if the determinants of donating to charity are different for men and women. Other
research may suggest that women donate due to empathetic and compassionate reasons whereas men’s decisions to
donate tend to be more calculated. To look at this further, I could also add an interaction term into my equation.
Clotfelter (2002) reports that contributions to charity rises with age until middle age (40s and 50s) and then decreases, which
would follow the pattern of age squared, which I will test in my investigation. This paper also suggests that more people give
money to charity as their income increases, which is something that I can test for within my first model, and then in my second
model I can test for how much their donations increase as income rises. Other characteristics that have been related to
charitable giving that are mentioned within this paper include: education, marriage, number of children and if an individual
lives in a city; all of which I can include using a selection of variables given in the dataset. Clotfelter (2002) also concludes that
church-goers and other religious organisations contribute more donations than any other person.
It has also been found that other ethnicities are less likely to donate than those of white ethnicity (Rooney et al, 2005). However
this is still only a more recent topic for many investigations and so many academic’s results are different, so it will be interesting
to see the results that I obtain from this investigation. This paper also looks at interaction terms between marital status and sex,
finding that married men are 5% more likely and married women are 11.6% more likely to donate compared to single men. For
my investigation I can put an interaction term into my second model to see if my results are similar to Rooney et al (2005).
The level of education has also been found to have a positive correlation with charitable giving in many previous empirical
investigations. Wiepking and Bekkers (2011) list numerous references to other investigations which also share this conclusion.
There has been lots of research into the success of methods of charitable giving (for example, door to door collections or online
donating), and to which charities, people of different education levels, donate money (Schlegelmilch et al, 1997). But since the
data is limited in this investigation, this analysis isn’t possible.
However in Cheung and Chan’s (2000) paper, the education variable did not display much significance or impact when
having the intention to donate to IRO charities (international relief organisations), alongside income, age and sex. This is highly
surprising due to the number of papers who have found these variables to be important and significant. These conflicting ideas
make this investigation into these variables even more intriguing.
141
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
3.
142
Model and data
The data used in this investigation has been collected from the fourth wave of the Understanding Society survey – the UK
household longitudinal study. This is a UK based survey, previously known as the BHPS survey, which is designed to capture life in
the UK and how it is changing over time. The original number of observations collected from the survey was 47157, collected
from interviews between 2012 and 2014 (Brown and Taylor, 2015). From this survey, there were a limited number of possible
variables that I could include in my models.
By conducting research and previous literature allowed me to then select the variables that I would need to construct wellrounded models and the dataset was adjusted removing inapplicable answers and missing values. I was left with 24446
observations, which is a substantial amount to be able to draw reliable conclusions from. The qualitative variables were then
recoded into binary variables, shown in my descriptive statistics table (Figure 1 below).
142
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
143
Figure 1 – a table of descriptive statistics including all initial variables.
Descriptive Statistics
N
Minimum
Maximum
Mean
Std.
Deviation
number of children aged under 18 responsible for
43217
0
8
0.34
0.807
amount given to charity last 12 months
27695
1
9997
203.74
491.492
age
47157
16
104
47.35
18.595
total monthly personal income gross
44836
0
15000
1738.443
1611.469
attends a religious service often
43193
0
1
0.2206
0.41465
financially stable
43123
0
1
0.6367
0.48095
financially unstable
43123
0
1
0.1086
0.31111
dissatisfied with life overall
38908
0
1
0.1844
0.3878
satisfied with life overall
38908
0
1
0.7165
0.45072
male
47157
0
1
0.462
0.49856
married
43097
0
1
0.5195
0.49962
urban
47128
0
1
0.7531
0.43124
white ethnicity
44866
0
1
0.8517
0.35539
mixed ethnicity
44866
0
1
0.0157
0.12428
asian ethnicity
44866
0
1
0.0875
0.28251
143
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
My
144
black ethnicity
44866
0
1
0.0378
0.19083
arab ethnicity
44866
0
1
0.0033
0.05773
degree or higher
46937
0
1
0.3409
0.47402
a levels
46937
0
1
0.2132
0.40957
gcses
46937
0
1
0.2119
0.40866
gives money to charity
43155
0
1
0.6813
0.46598
other ethnicity
44866
0
1
0.0039
0.06269
Valid N (listwise)
24446
first
model is a Logit model, which poses the question ‘what factors determine people’s willingness to donate to charity’. This has
been added in for completeness and I will only briefly investigate this model, so I can examine my second model in more
depth.
Figure 2 - Logit equation 1:
GIVES MONEY TO CHARITY = β₀ + β₁ ATTENDS A RELIGIOUS SERVICE OFTEN + β₂ FINANCIALLY STABLE + β₃ FINANCIALLY UNSTABLE
+ β₄ SATISFIED WITH LIFE OVERALL + β₅ MALE + β₆ AGE + β₇ MARRIED + β₈ DEGREE OR HIGHER + β₉ ALEVELS + β₁ ₀ GCSES +
β₁ ₁ URBAN + β₁ ₂ WHITE ETHNICITY + β₁ ₃ MIXED ETHNICITY + β₁ ₄ ASIAN ETHNICITY + β₁ ₅ BLACK ETHNICITY + β₁ ₆ ARAB ETHNICITY
+ β₁ ₇ INCOME + μᵢ .
My second model uses the dependent variable ‘the amount donated to charity in the last 12 months’ which is quantitative,
meaning that I am using Ordinary Least Squares (OLS) to do the second part of my statistical investigation. This is looking at
those who donate money to charity (who gave money to charity in the first model), in order to see what factors determine
people’s generosity when donating money.
Figure 3 - OLS equation 1:
AMOUNT DONATED TO CHARITY IN THE LAST 12 MONTHS = β₀ + β₁ ATTENDS A RELIGIOUS SERVICE OFTEN + β₂ FINANCIALLY
STABLE + β₃ FINANCIALLY UNSTABLE +β4SATISFIED WITH LIFE OVERALL + β₅ MALE + β₆ AGE + β₇ MARRIED + β₈ DEGREE OR HIGHER
+ β₉ ALEVELS + β₁ ₀ GCSES + β₁ ₁ URBAN + β₁ ₂ WHITE ETHNICITY + β₁ ₃ MIXED ETHNICITY + β₁ ₄ ASIAN ETHNICITY + β₁ ₅ BLACK
ETHNICITY + β₁ ₆ ARAB ETHNICITY + β₁ ₇ INCOME + β₁ ₈ NUMBER OF CHILDREN AGED UNDER 18 + μᵢ .
Below is a table of my variables (figure 4), including my recoded binary variables that I am using in my investigation, their
recoding and their expected signs for when I run the regressions.
144
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
145
Figure 4 – table of all variables
Variable
1)
2)
3)
4)
5)
6)
7)
8)
Donated money
to charity
(d_chargv)
Amount given to
charity last 12
months
(d_charam)
Natural log of
‘amount given to
charity in the last
12 months’
(lnd_charam)
Attendance at
religious services
(d_oprlg2)
Recoded binary
variable
Recoding
Expected sign
Gives money to
charity
(d_chargvd1)
1= yes
DEPENDENT VARIABLE
FOR MODEL ONE
0=no
-
-
INITIAL DEPENDENT
VARIABLE FOR MODEL
TWO
-
-
MODIFIED DEPENDENT
VARIABLE FOR MODEL
TWO
Attends a religious
service often
(d_oprlg2d1)
1= attends a religious
service often
Subjective
financial situation
– current
(d_finnow)
Subjective
financial situation
– current
(d_finnow)
Satisfaction with
life overall
(d_sclfsato)
Financially stable
(d_finnowd1)
1= financially stable
Financially
unstable
(d_finnowd2)
1= financially unstable
Satisfied with life
overall
(d_sclfsatod2)
1= satisfied with life
overall
Sex (d_sex)
Male (d_sexd1)
1= male
+ve
0= otherwise
+ve
0= otherwise
-ve
0= otherwise
+ve
0= otherwise
-ve
0=female
9)
Age (d_dvage)
10) Age squared
(d_dvage2)
-
-
+ve
-
-
-ve (upside down ‘U’
shape)
11) Legal marital
status (d_marstat)
Married
(d_marstatd1)
1= married
12) Highest
qualification
(d_hiqual_dv)
Degree or higher
(d_hiqual_dvd1)
1= degree or higher
13) Highest
qualification
(d_hiqual_dv)
A-levels
(d_hiqual_dvd2)
1= alevels
+ve
0= not married
+ve
0= otherwise
+ve
0= otherwise
145
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
14) Highest
qualification
(d_hiqual_dv)
GCSEs
(d_hiqual_dv3)
1= gcses
15) Urban or rural
area derived
(d_urban_dv)
Urban
(d_urban_dvd1)
1= urban
16) Ethnic group
(d_racel_dv)
White ethnicity
(d_racel_dvd1)
1= white
17) Ethnic group
(d_racel_dv)
Mixed ethnicity
(d_racel_dvd2)
1= mixed
18) Ethnic group
(d_racel_dv)
Asian ethnicity
(d_racel_dvd3)
1= Asian
19) Ethnic group
(d_racel_dv)
Black ethnicity
(d_racel_dvd4)
1= black
20) Ethnic group
(d_racel_dv)
Arab ethnicity
(d_racel_dvd5)
1= arab
21) Ethnic group
(d_racel_dv)
Other ethnicity
(d_racel_dvd6)
1= other ethnic group
22) Total monthly
personal income
gross
(d_fimngrs_dv)
23) Natural log of
‘total monthly
personal income
gross’
(lnd_fimngrs_dv)
24) Number of
children aged
under 18 that the
respondent is
responsible for
(d_nchund18resp)
25) Interaction term of
‘marriedmale’
146
+ve
0= otherwise
?
0= rural
+ve
0= otherwise
-ve
0= otherwise
-ve
0= otherwise
-ve
0= otherwise
-ve
0= otherwise
-ve
0= otherwise
-
-
+ve
-
-
?
-
-
-ve
-
-
?
From looking at previous literature, I have included a religious variable (base case - do not attend a religious service often)
along with sex (base case - female), age, marital status (base case - not married), education (base case - no or other
qualifications), location (base case - live in a rural area), ethnicity (initial base case - other ethnicity) and income variables.
I have also included the ‘subjective financial situation – current’ (base case - neither financially stable nor unstable), because I
believe that your personal belief in your financial situation would be important when determining if you are able to donate
some of your income to charity and how much. I have also included the variable ‘satisfied with life overall’ because I believe
that if you are generally satisfied or happy with life, you are more likely to be willing to donate money to charity than those who
are not (which is the base case).
146
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
4.
147
Empirical analysis – first model
Initially I ran my first Logit regression, with the dependent variable ‘gives money to charity’ including explanatory variables as
suggested in Figure 2. From this, I discovered that the variables Mixed, Asian, Black and Arab ethnicities were not significant to
any level. In order to get the ethnicity variables significant, I decided to change the base case from ‘other ethnicity’ to ‘white
ethnicity’, and this then made all of my variables significant to at least a 10% level (see Figure A in the appendix).
Looking at Figure 4, the signs of the coefficients matched my educated predictions, and outcomes from the results of similar
studies, as I continue on to mention.
As predicted, those who are of different ethnicity to white, are less willing to donate money than those whom are of white
ethnicity, ceteris paribus, which also corresponds to the findings of Rooney et al (2005). The male coefficient is also negative,
suggesting that males are less willing to donate money to charity, ceteris paribus. Before running the regression, I was unsure
whether an individual’s location (urban or rural) would affect their willingness to donate money to charity, but from these
results, it seems that those who live in a more urban area are less willing to donate money to charity than those who live in a
rural area, ceteris paribus.
Income has a positive coefficient, meaning that as income rose, more people would donate to charity, ceteris paribus, which
Clotfelter (2002) suggested. All of the education variables (‘degree or higher’, ‘a-levels’ and ‘gcses’) also have a positive
coefficient, so those with qualifications are more willing to donate to charity that those who have no or other qualifications,
ceteris paribus, which corresponds with the findings of Wiepking and Bekkers (2011).
I ran this Logit model just for completeness when conducting this investigation, it was needed for further understanding in the
next section of empirical analysis.
5.
Empirical analysis – second model
The second model that I investigated used an OLS estimation method, initially using the dependent variable of ‘amount given
to charity in the last 12 months’. However when checking that my model was correctly specified, by running a Ramsey reset
test, I found that my model was incorrectly specified. In order to solve this problem I created the natural logarithm of the
147
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
148
dependent variable and then completed the reset test once again to find that my model was now correctly specified (see
Figure B in the appendix).
Originally, I also used ‘white ethnicity’ as an explanatory variable (and ‘other ethnicity’ as the base case), but this was causing
all of my other ethnicity variables to be insignificant. In order to deal with this, I decided to use ‘white ethnicity’ as the base
case and include ‘other ethnicity’ within the regression, which resulted in ‘Asian ethnicity’ and ‘black ethnicity’ becoming
significant. I have performed an F-test to test for joint significance between these variables which concluded that they are not
jointly significant in the model, however I have decided to keep the variables in the model because in my opinion they are
relevant.
Figure 5 – re-specified OLS (equation 2):
LN(AMOUNT GIVEN TO CHARITY IN THE LAST 12 MONTHS) = β₀ + β₁ ATTENDS A RELIGIOUS SERVICE OFTEN + β₂ FINANCIALLY
STABLE + β3FINANCIALLY UNSTABLE + β4SATISFIED WITH LIFE OVERALL + β₅ MALE + β₆ AGE + β₇ MARRIED + β₈ DEGREE OR HIGHER
+ β₉ ALEVELS + β₁ ₀ GCSES + β₁ ₁ URBAN + β₁ ₂ OTHER ETHNICITY + β₁ ₃ MIXED ETHNICITY + β₁ ₄ ASIAN ETHNICITY + β₁ ₅ BLACK
ETHNICITY + β₁ ₆ ARAB ETHNICITY + β₁ ₇ INCOME + β₁ ₈ NUMBER OF CHILDREN AGED UNDER 18 + μᵢ .
In order to get reliable results from any regressions that are run, I then tested my re-specified model (Figure 5) for
heteroskedasticity using the White test and found that there was a problem (Figure C in the appendix). To combat this, I reestimated my model using a bootstrap tool, which should now mean that there is not a heteroskedasticity problem within my
models.
The first regression run, used my initial, modified variables (from Figure 5), in order to see the regression results and analyse them
in terms of magnitude, sign and significance (can be seen in Figure D in the appendix and Figure 6 below). The highlighted
numbers in the table are those that I mention in my interpretation below.
Figure 6 – OLS 1
Coefficient, (standard error)
Constant
P-value
2.134 (0.04559)
0.001
Number of children aged under 18 responsible for
0.007 (0.011)
0.540
Age
0.018 (0.001)
0.001
0.910 (0.02)
0.001
0.273 (0.019)
0.001
-0.127 (0.033)
0.001
0.030 (0.018)
0.0999
-0.043 (0.018)
0.031
Attends a religious service often
Financially stable
Financially unstable
Satisfied with life overall
Male
148
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
149
Married
0.154 (0.017)
0.001
-0.029 (0.017)
0.093
Mixed ethnicity
0.009 (0.077)
0.906
Asian ethnicity
0.218 (0.037)
0.001
Black ethnicity
-0.276 (0.059)
0.001
Arab ethnicity
0.021 (0.161)
0.909
Degree or higher
0.731 (0.025)
0.001
A-levels
0.452 (0.027)
0.001
GCSEs
0.311 (0.028)
0.001
-0.015 (0.169)
0.941
0.0002 (0.0000056)
0.001
Urban
Other ethnicity
Total monthly personal income gross
Dependent variable: ln(amount given to charity in the last 12 months)
Those that are highlighted in the p-values column are not significant in my model, but I have kept them in my model for reasons
previously stated. Some of the variables may not be significant due to the limited number of observations within that category,
for example, those of ‘Arab ethnicity’ and ‘other ethnicity’ have very few observations.
Most of my coefficients have the same expected sign, as previously predicted in Figure 3. The one exception being the
variable ‘number of children aged under 18 responsible for’. In the model (Figure 6), it has a positive sign, meaning that the
more children there are, the more money the individual donated to charity in the last 12 months, specifically, for each extra
child within the household the amount donated to charity increases by 0.7%, ceteris paribus. However, I had predicted this
variable to have a negative sign. But, this variable isn’t significant, so this interpretation may be incorrect.
The constant can be interpreted as: when all explanatory variables are 0, an individual would give £8.45 to charity in a year,
ceteris paribus. However this interpretation seems unrealistic with respect to age in particular as well as many of the other
explanatory variables, as when you are 0 years old, you are unlikely to be donating money to charity.
One variable that is quite striking is that if you attend a religious service often you are estimated to donate around 91% more
than those who do not attend a religious service often, ceteris paribus, which corresponds with previous literature. Other
variables that have a significantly large coefficient include the education variables (degree or higher, a levels and GCSEs), in
which people donate around 73.1%, 45.2% and 31.1% respectively more over the last 12 months, than those who have no
qualifications, ceteris paribus. This is expected, as many academic papers have also reached the same conclusion (Wiepking
and Bekkers, 2011).
Additionally, from this model I have found that if you are of Asian ethnicity, you are predicted to donate 21.8% more than if you
are of white ethnicity, ceteris paribus, which contradicts papers such as Rooney et al (2005) and my prediction from Figure 4.
On the other hand, if you are of black ethnicity, you are predicted to donate 27.6% less than if you are of white ethnicity,
ceteris paribus, which follows the findings from Rooney et al (2005) and my previous prediction.
149
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
150
Furthermore, as age increases by one year, you are likely to donate 1.8% more, ceteris paribus. I would have expected this
coefficient to be of greater magnitude, however the magnitude may be low because age may follow a quadratic relationship
with the amount donated to charity, rather than a linear relationship. To investigate this, I have included age squared into my
next regression (OLS 2, Figure 7).
Also, the R² of 25% suggests that the model is a good fit, when considering the number of variables included. This means 25% of
the variation in ‘ln(amount given to charity in the last 12 months)’ is explained by the model.
For the next regression, I have included an age squared variable (OLS 2, figure 7 below). This new variable turned out to be
significant, meaning that the relationship between the natural logarithm of the dependent variable and age is not linear,
implying that a quadratic relationship exists, which was also suggested by Clotfelter (2002). Through differentiation, I found out
that the turning point was a maximum, suggesting an inverted ‘U’ shape curve with the maximum amount of money being
donated at age 78 (see figure E in the appendix), which is older than other studies have suggested. Other studies have
suggested that the maximum amount of money is donated by people who are in their 40s and 50s.
Adding in the age squared variable (seen in OLS 2, Figure 7) also changed the sign of the coefficient of the variables ‘number
of children aged under 18 responsible for’ (which is still not significant) and ‘urban’; as well as the magnitudes of the
coefficients in the model, although most are very similar to OLS 1. By adding in the age squared variable, if you live in an urban
area, you are now likely to donate 9% more to charity than those who are living in a rural area.
Additionally, by adding in ‘age squared’ to the model also increases the goodness of fit, as R² has increased from 25% to 25.4%.
The next regression, includes the variable ‘ln(income)’, so I will have a double logarithm model. From this I obtained an
elasticity of 0.139 between the variables ‘amount given to charity in the last 12 months’ and ‘total monthly personal income
gross’.
Figure 7, below, shows the latest regression (OLS 3) along with the previous regressions.
Figure 7 – a table of three OLS regressions
150
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
(Constant)
number of children aged under 18
responsible for
OLS 1
2.13400***
(0.04558)
0.00682
(0.011)
age
0.01794***
(0.001)
attends a religious service often
0.90993***
(0.020)
financially stable
0.27315***
(0.019)
financially unstable
satisfied with life overall
-0.12720***
(0.033)
0.03033*
(0.018)
male
-0.04288**
(0.018)
married
0.15361***
(0.017)
urban
-0.02948*
(0.017)
mixed ethnicity
0.00949
(0.077)
asian ethnicity
0.21805***
(0.037)
black ethnicity
-0.27639***
(0.059)
arab ethnicity
0.02104
(0.161)
degree or higher
0.73126***
(0.025)
a levels
0.45219***
(0.027)
gcses
0.31128***
(0.028)
other ethnicity
-0.01514
(0.169)
OLS 2
151
OLS 3
1.47548***
(0.072)
0.60078***
(0.081)
-0.00439
(0.011)
-0.01319
(0.011)
0.04989***
(0.003)
0.05509***
(0.003)
0.92900***
(0.020)
0.92479***
(0.021)
0.29700***
(0.020)
0.34833***
(0.019)
-0.14600***
(0.033)
-0.14959***
(0.034)
0.04600**
(0.018)
0.05549***
(0.019)
-0.03066*
(0.018)
-0.03249*
(0.018)
0.09048***
(0.018)
0.10177***
(0.018)
0.09048*
(0.017)
-0.03519*
(0.018)
0.01541
(0.077)
0.03402
(0.074)
0.22849***
(0.037)
0.22203***
(0.036)
-0.29147***
(0.058)
-0.27925***
(0.060)
0.03406
(0.159)
0.06187
(0.161)
0.69635***
(0.025)
0.77177***
(0.025)
0.43699***
(0.027)
0.44869***
(0.027)
0.27730***
(0.028)
0.28116***
(0.027)
-0.02303
(0.170)
-0.06380
(0.177)
total monthly personal income gross
0.00017***
(0.000006)
Age squared
Ln(income)
0.00015***
(0.000006)
-0.00032***
(0.000027)
-0.00038***
(0.000028)
0.13869***
(0.008)
151
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
152
In Figure 7, the coefficients are reported along with their standard errors in brackets and the stars indicate whether a variable is
significant (***= 1% level, ** = 5% level and * = 10% level).
To further test how the difference between men and women in their generosity when donating to charity, I have performed a
chow test using my initial re-specified OLS model (Figure 5). I have pooled the data and split it up for men and women.
Workings out for this test can be seen in Figure F in the appendix. From this test, I concluded that the determinants of the
‘amount given to charity in the last 12 months’ are different for men and women.
To expand this investigation further, an interaction term comprising of ‘male’ and ‘total monthly personal income gross’ could
have been added to further what was discussed in section 2 (background), however this explanatory variable was
insignificant. After gaining the results of the Chow test, I wanted to find out if men made more calculated decisions when
donating to charity and would have based this on their income.
Due to the previous idea not being viable in my investigation, I could follow Rooney et al’s (2005) idea of the interaction term
of ‘married male’ (Figure 8 below). This new variable is significant (10% level) within my model and has a positive coefficient married men are predicted to donate 5.9% more, ceteris paribus. This could be due to the female influence within the married
relationship.
Figure 8 – table of OLS regression including ‘marriedmale’
Coefficient, (standard error)
P-value
Constant
2.170 (0.046)
0.001
number of children aged under 18 responsible for
0.038 (0.011)
0.003
age
0.019 (0.001)
0.001
attends a religious service often
0.861 (0.02)
0.001
financially stable
0.321 (0.02)
0.001
-0.137 (0.033)
0.001
satisfied with life overall
0.034 (0.019)
0.076
male
0.071 (0.026)
0.005
married
0.179 (0.022)
0.001
urban
-0.022 (0.019)
0.240
mixed ethnicity
-0.007 (0.076)
0.938
asian ethnicity
0.171 (0.038)
0.001
black ethnicity
-0.286 (0.055)
0.001
arab ethnicity
0.046 (0.166)
0.808
degree or higher
0.953 (0.023)
0.001
a levels
0.516 (0.026)
0.001
gcses
0.337 (0.026)
0.001
-0.066 (0.181)
0.719
financially unstable
other ethnicity
152
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
marriedmale
153
0.059 (0.032)
0.060
Those highlighted in the p-value column are insignificant within this model. Through adding in this interaction term into the
model, I have also found that the ‘number of children aged under 18’ is now significant, whereas it has not been significant in
all previous models. If you are responsible for one extra child, the amount you donate to charity increases by 3.8% ceteris
paribus. Likewise, ‘urban’ isn’t significant within this model, but has been in previous models.
6.
Conclusion
Overall, within my Logit model, which all variables are significant, and so age, income, attendance at religious services,
financial situation, satisfaction with life, sex, marital status, location, education and ethnicity are all key determinants to if
someone is willing to donate money to charity or not.
I can also conclude that the greatest determinants for the generosity of people when donating to charity is their attendance
to religious services, their age squared, their level of education and their sex. From the investigation, I believe that those who
attend church regularly are the most generous when donating money to charities, closely followed by those who are welleducated (gcses, a-levels or degree) and those whom feel that they are financially stable.
There have been many limitations within this investigation, mostly because the dataset used has not got everything within it
that would help form a well-rounded answer, such as looking into what types of charities people donate money to and the
difference between how men and women make the decision when donating to charity, as mentioned previously. I believe
that there should also be more research into different ethnicities when donating to charities, as I couldn’t gain reliable results
from some of my ethnicity variables due to the limited number of responses within that category. So, in the future, a wider
range of people with different ethnicities should be included in empirical investigations.
153
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
7.
154
Bibliography
Brown, S and Taylor, K. 2015. Charitable behaviour and the big five personality traits: Evidence from UK panel data. Sheffield
economic research paper series. [Online] Available: https://www.sheffield.ac.uk/polopoly_fs/1.480008!/file/paper_2015017.pdf
[Accessed: 24/02/2016]
Cheung, C and Chan, C. 2000. Social-cognitive factors of donating money to charity, with special attention to an international
relief organization. Evolution and programme planning. Vol. 23 No. 2 p241-253. [Online] Available:
http://www.sciencedirect.com/science/article/pii/S0149718900000033 [Accessed: 28/02/2016]
Clotfelter, C. 2002. The economics of giving. Duke University. [Online] Available:
http://philvol.sanford.duke.edu/documents/giving.pdf [Accessed: 16/02/2016]
Einolf, C. 2011. Gender differences in the correlates of volunteering and charitable giving. Nonprofit and voluntary sector
quarterly. Vol. 40 No. 6 p1092-1112.
Leslie, L., Synder, M and Glomb, T. 2013. Who gives? Multilevel effects of gender and ethnicity on workplace charitable giving.
Journal of applied psychology. Vol. 98 No. 1 p49-62 [Online] Available: http://ssir.org/pdf/Leslie_et_al_2013_who_gives.pdf
[Accessed: 16/02/2016]
Rooney, P. et al. 2005. The effects of race, gender, and survey methodologies on giving in the US. Economics letters. 86 p173180 [Online] Available:
https://philanthropy.iupui.edu/files/research/the_effects_of_race_gender_and_survey_methodologies_on_giving_in_the_us.pdf
[Accessed: 24/02/2016]
Schlegelmilch, B, Love, A and Diamantopoulos, A. 1997. Responses to different charity appeals: the impact of donor
characteristics on the amount of donations. European journal of marketing. Vol. 33 No. 8 p548-560. [Online] Available:
http://www.emeraldinsight.com/doi/pdfplus/10.1108/03090569710176574 [Accessed: 27/02/2016]
Wiepking, P and Bekkers, R. 2011. Who gives? A literature review of predictors of charitable giving. I – religion, education, age
and socialization. Voluntary sector review. [Online] Available:
http://scholar.google.co.uk/scholar_url?url=http%3A%2F%2Frepub.eur.nl%2Fpub%2F32755%2Fmetis_172194.pdf&hl=en&sa=T&oi
=gga&ct=gga&cd=0&ei=C9bRVu6kMpaO2Aaoi4CICA&scisig=AAGBfm0m9njO6l619UGBAoSOXMsq_ei97g&nossl=1&ws=1440x8
05 [Accessed: 27/02/2016]
Willer, R., Wimer, C and Owens, L. 2014. What drives the gender gap in charitable giving? Lower empathy leads men to give
less to poverty relief. Social science research. Vol. 52 p83-98 [Online] Available:
http://www.sciencedirect.com.ueaezproxy.uea.ac.uk:2048/science/article/pii/S0049089X15000058? [Accessed: 16/02/2016]
8.
Appendix
Figure A – logit regression
Variables in the Equation
B
Step 1a
d_dvage
d_fimngrs_dv
d_oprlg2d1
d_finnowd1
S.E.
Wald
df
Sig.
Exp(B)
0.021
0.001
647.533
1
0.000
1.021
0.0003
0.000
249.051
1
0.000
1.000
0.720
0.036
398.997
1
0.000
2.054
0.387
0.028
187.827
1
0.000
1.472
154
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
d_finnowd2
d_sclfsatod2
d_sexd1
d_marstatd1
d_urban_dvd1
d_racel_dvd2
d_racel_dvd3
d_racel_dvd4
d_racel_dvd5
d_hiqual_dvd1
d_hiqual_dvd2
d_hiqual_dv3
d_racel_dvd6
Constant
155
-0.242
0.042
33.645
1
0.000
0.785
0.160
0.027
34.127
1
0.000
1.173
-0.425
0.025
288.017
1
0.000
0.654
0.198
0.026
59.124
1
0.000
1.220
-0.099
0.029
11.463
1
0.001
0.906
-0.286
0.096
8.905
1
0.003
0.751
-0.224
0.051
19.111
1
0.000
0.799
-0.671
0.068
98.569
1
0.000
0.511
-0.606
0.224
7.327
1
0.007
0.546
0.987
0.037
717.150
1
0.000
2.683
0.738
0.039
358.565
1
0.000
2.091
0.480
0.037
165.741
1
0.000
1.616
-0.381
0.197
3.744
1
0.053
0.683
-1.217
0.064
367.201
1
0.000
0.296
a. Variable(s) entered on step 1: d_dvage, d_fimngrs_dv, d_oprlg2d1, d_finnowd1, d_finnowd2,
d_sclfsatod2, d_sexd1, d_marstatd1, d_urban_dvd1, d_racel_dvd2, d_racel_dvd3, d_racel_dvd4,
d_racel_dvd5, d_hiqual_dvd1, d_hiqual_dvd2, d_hiqual_dv3, d_racel_dvd6.
Figure B – reset test indicating correct model specification
Model Summary
R
Square
Adjusted
R
Square
Std.
Error of
the
Estimate
Change Statistics
R
Square
Change
F
Change
df1
df2
Sig. F
Change
Model
1
R
.468a
0.219
0.219
1.26059
0.219
413.091
17
25011
0.000
2
.468b
0.219
0.219
1.26059
0.000
0.878
2
25009
0.416
a. Predictors: (Constant), number of children aged under 18 responsible for, other ethnicity, mixed ethnicity, satisfied
with life overall, arab ethnicity, a levels, black ethnicity, asian ethnicity, married, urban, financially unstable, gcses,
attends a religious service often, male, age , financially stable, degree or higher
155
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
156
b. Predictors: (Constant), number of children aged under 18 responsible for, other ethnicity, mixed ethnicity, satisfied
with life overall, arab ethnicity, a levels, black ethnicity, asian ethnicity, married, urban, financially unstable, gcses,
attends a religious service often, male, age , financially stable, degree or higher, PRE_4a, PRE_2a
Figure C – White test for heteroskedasticity
H₀ = homoscedasticity and H₁ = heteroskedasticity
W = 0.01 x 24446 = 244.46
Df = 21 and critical value (5% significance level) X² = 32.671
W > X², therefore we reject H₀ and conclude that there is evidence of heteroskedasticity.
Figure D – before and after bootstrap
Coefficientsa
Unstandardized
Coefficients
Model
1
(Constant)
number of children aged under 18 responsible
for
B
Std. Error
2.134
0.046
0.007
0.011
0.018
Standardized
Coefficients
Beta
t
Sig.
46.565
0.000
0.004
0.605
0.545
0.001
0.219
33.002
0.000
0.910
0.020
0.272
46.592
0.000
0.273
0.020
0.089
13.847
0.000
-0.127
0.033
-0.024
-3.847
0.000
0.030
0.019
0.009
1.611
0.107
-0.043
0.018
-0.015
-2.404
0.016
0.154
0.017
0.054
9.107
0.000
-0.029
0.018
-0.009
-1.626
0.104
0.009
0.069
0.001
0.138
0.890
0.218
0.035
0.037
6.274
0.000
-0.276
0.051
-0.031
-5.408
0.000
age
attends a religious service often
financially stable
financially unstable
satisfied with life overall
male
married
urban
mixed ethnicity
asian ethnicity
black ethnicity
156
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
arab ethnicity
degree or higher
a levels
gcses
other ethnicity
157
0.021
0.176
0.001
0.120
0.905
0.731
0.024
0.254
30.178
0.000
0.452
0.027
0.129
16.692
0.000
0.311
0.027
0.086
11.676
0.000
-0.015
0.147
-0.001
-0.103
0.918
0.000
0.000
0.209
34.478
0.000
total monthly personal income gross
a. Dependent Variable: log of amount given to charity in the last 12 months
Bootstrap for Coefficients
Bootstrapa
Model
1 (Constant)
number of children aged under
18 responsible for
B
Bias
Sig. (2tailed)
Std. Error
95% Confidence
Interval
Lower
Upper
2.134
-0.001
0.0455867
0.001
2.045
2.225
0.007
0.000
0.011
0.540
-0.015
0.029
0.018
1.168E05
0.001
0.001
0.017
0.019
0.910
0.000
0.020
0.001
0.869
0.949
0.273
0.000
0.019
0.001
0.233
0.312
-0.127
-0.001
0.033
0.001
-0.192
-0.064
0.030
0.001
0.018
0.100
-0.006
0.068
-0.043
0.000
0.018
0.031
-0.080
-0.008
0.154
0.000
0.017
0.001
0.121
0.189
-0.029
0.000
0.017
0.093
-0.066
0.004
0.009
-0.003
0.077
0.906
-0.143
0.150
0.218
0.001
0.037
0.001
0.150
0.293
-0.276
0.001
0.059
0.001
-0.391
-0.159
age
attends a religious service often
financially stable
financially unstable
satisfied with life overall
male
married
urban
mixed ethnicity
asian ethnicity
black ethnicity
157
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
arab ethnicity
degree or higher
a levels
gcses
other ethnicity
total monthly personal income
gross
158
0.021
0.001
0.161
0.909
-0.297
0.335
0.731
-0.001
0.025
0.001
0.681
0.782
0.452
0.000
0.027
0.001
0.398
0.504
0.311
0.000
0.028
0.001
0.257
0.365
-0.015
-0.001
0.169
0.941
-0.340
0.338
0.000
2.502E07
5.552E-06
0.001
0.000
0.000
a. Unless otherwise noted, bootstrap results are based on 1000 bootstrap samples
Figure E – turning point workings out (5dp)
As taken from the results of OLS 2
Let F = ln(amount given to charity in the last 12 months)
F = 0.04989age – 0.00032age²
dF/d(age) = 0.04989 – 0.00064age = 0
0.04989 = 0.00064age
Age = 78
To check that it is a maximum: d²F/d(age)² = -0.00064 < 0
Therefore we have a maximum turning point.
Figure F – Chow test
F = [SSRP – (SSRM + SSRF) / (k+1)
(SSRM + SSRF) / [nM + nF – 2(k+1)]
SSRP = 36832.407
SSRM = 15844.041
SSRF = 20836.244
(k+1) = 18
NM = 10269
NF = 14177
F statistic = 5.62412027
F18, 24410 = 1.67
F stat > F18, 24410
158
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
159
An investigation into the UK gender pay gap and how it could be improved through a change in paternity leave policy
By Hannah McCartney
Introduction
Discrimination occurs in the labour market when individuals receive differing rewards on the basis of a characteristic, or set of
characteristics, which have no bearing on their productivity. There is evidence to show there is a pay gap between men and
women in the labour market, leading people to conclude that women are discriminated against, despite legislation which is
meant to outlaw this. A common explanation for this pay gap is the fact that many women choose to have a career break in
order to have children, and it is this during this break that their male colleagues progress up the career ladder and therefore
earn more money. This is shown by the fact that, in the OECD, women account for less than a third of senior managers and
only one woman for every 10 men gets to the boardroom (OECD, 2013). Consequently, one option for narrowing the gender
pay gap might be to encourage women to return to work sooner after having a child, thus reducing the length of their career
break and allowing less disparity between the career progressions (and therefore wages) of men and women. This essay will
focus on paternity leave policy as a means of achieving this, although other policies (such as addressing the rising costs of
childcare) are also incredibly significant. Therese Murphy, Head of Operations at the European Institute for Gender Equality,
said of paternity leave: “men are generally being denied their human right to be a parent because they are expected to work
12 hours-a-day. There is no flexibility, they are being penalised if they want parental leave. That needs to change because it
means women are then staying at home as a result – which then impacts on the gender pay gap.” (Harris, 2015)
This essay will begin by outlining the theory of statistical discrimination, which gives an explanation of why the gender pay gap
exists from a theoretical point of view. The current situation in the UK will then be explained, including generating an estimation
of the gender pay gap using data from the Labour Force Survey 2014 (April-June). Paternity leave policies, participation rates
and gender pay gaps from different countries will then be compared to the UK and the implications of these findings will be
discussed.
Theory of Statistical Discrimination
Although there are four types of discrimination theories (worker, customer, employer and statistical), this essay will focus only on
the theory of statistical discrimination as it gives the most likely explanation of why there is a gender pay gap. Where all of the
other three theories focus on the intentional discrimination of a group of workers - who are equally as productive as other
workers - by other agents (ie. other workers, customers, or employers respectively), statistical discrimination is concerned with
asymmetric information about a worker’s productivity when employing them, and employers using average productivities to
determine wages.
159
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
160
When making hiring decisions, employers do not know an individual’s productivity as they have not been able to observe them
working, so they make generalisations about potential employees based on certain characteristics; in this case, employers may
discriminate against women as they are more likely to leave (at least temporarily) in order to have a family, thus reducing their
productivity in the long run. As Booth (2009) states, “employers base their behaviour on averages”: if an employer thinks that
men are on average more productive than women (as they are less likely to have a career break), they will pay members of
each gender a different wage, equal to the average productivity of that gender. However, since individuals of either gender
may have productivities differing from the average, workers of equal productivity but of different genders will be paid different
wages, hence discrimination against individuals occurs. The extent of the discrimination (in this case measured by the gender
pay gap) will depend on how close the average productivities of each group are to each other: if the averages are similar, the
pay gap between workers with equal productivity will be smaller. The longer the average career break a woman takes, the
more it will negatively affect the average productivity of women, and there will therefore be a greater potential for statistical
discrimination as the gap between average productivities widens.
Diagram 1 shows how employers may make generalisations about workers and choose to set wages on the basis of gender. An
employer sees males as more productive than females so chooses to pay males a higher wage, Wmale (equal to the average
productivity of a male in that role), and women a lower wage, Wfemale (equal to the average productivity of a female in that
same role). Individuals A-C have different levels of productivity. Individual A is a low productivity female; in a perfectly
competitive market she would be paid a wage equal to her marginal productivity, but in this model she would be paid a
higher wage, Wfemale. Individual C is a high productivity male; in a perfectly competitive market he would be paid a wage
equal to his marginal productivity, but in this model he would be paid a lower wage, Wmale. Individual B is either a high
productivity female, in which case she will be underpaid according to her productivity and receive Wfemale, or a low
productivity male, in which case he will be overpaid and receive Wmale, despite the fact that their productivity is equal. This
form of discrimination, although illegal, is still apparent: legislation may have changed over the years, but the fact that women
take a leave of absence from work to have children has not, which is why the gap remains.
Diagram 1
Wmale = MPmale
Wfemale = MPfemale
A
B
C
MPL
160
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
161
This theory illustrates that workers with the same productivity can earn significantly different wages if they are of different
genders due to the generalisation made by employers about how a career break impacts women’s long run productivity. This
assumption about employers’ attitudes seems to hold true in reality: a survey of 500 hiring managers in the UK found “a third
would rather employ a man in his 20s or 30s than a woman of the same age for fear of maternity leave” (Slater & Gordon,
2014) and Misra & Strader (2013) found that, “parenthood, and specifically perceived caregiver responsibilities that are
entrenched in employers' perceptions and reinforced through legislation and policies, are central factors in explaining the
persisting gender wage gap.”
The UK Gender Pay Gap
Most reported gender pay gap statistics are unadjusted - they do not take into account factors that affect worker productivity
(for example a person’s level of education or years of work experience). Instead, the figures purely show the difference
between the median male wage and the median female wage. According to the OECD (2015), the unadjusted gender pay
gap in the UK was 17.5% in 2013. Change this if different on SPSS. Using the data from the Labour Force Survey 2014 (AprilJune), the gender pay gap can be decomposed into explained discrimination (that is to say differences in pay that can be
explained by productivity factors) and pure discrimination.
Initially, a regression was run to show which factors impacted the sample population’s wages. The sample was first filtered so
only working respondents aged between 16 and 64 were used. Using the natural log of hourly wage as the dependent
variable, the independent variables can be shown in Table 1, along with how they have been coded.
Table 1
Variable
Information
Sex
1 = female, 0 = male
Age
Respondent’s age in years
Location
LondonSE
1 = lives in London or South East England, 0 = lives elsewhere
RestEng*
1 = lives in England, but not London or South East, 0 = lives elsewhere
ScotIreWal
1 = lives in Scotland, Ireland or Wales, 0= lives elsewere
Ethnicity
1= white, 0 = other ethnicity
Marital Status
1 = married/civil partnership/cohabitating, 0= otherwise
Education
None*
1 = no education, 0 = otherwise
GCSE
1 = GCSE highest qualification, 0 = otherwise
ALevel
1 = A Level highest qualification, 0 = otherwise
161
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
FurtherEd
Work Experience
162
1 = more advanced highest qualification, 0 = otherwise
Length of time the respondent has been in continuous employment
* indicates that the variable has been omitted from the regressions to avoid the dummy variable trap
Both the age variable and the work experience variable have been included as they are not highly correlated enough to
create a problem with multi-collinearity (proof of which is shown in Appendix A). In addition, as Table 1 shows, certain variables
have been excluded from the regression in order to avoid the dummy variable trap. The results of the regression (which can be
found in Appendix B) show that being female has a highly significant negative effect on a person’s predicted hourly wages: a
female worker is predicted to earn 19.8% less than a male worker, ceteris paribus. All of the other independent variables used
are also shown to be highly significant in their impact on wage, although some have a very small effect.
The sample was then divided into males and females, and separate regressions were run on each group, using the same
independent variables used in Regression 1 (excluding the sex variable). The results of these regressions can be found in
Appendix C and D respectively. The male only regression shows all variables to be highly significant, while the female only
regression finds the white variable to be significant at the 5% level and the ScotIreWal variable to be statistically insignificant.
Using the regression results for each group, the hourly wage of a hypothetical individual was calculated; this individual was a
married, white, 30 year old university graduate, who had 7 years of work experience and lived in London or South East England.
If this individual was male, he would be predicted to earn £18.43 an hour according to the male only regression results; if they
were female, the same individual would be predicted to earn £13.94 an hour (24.4% less). This shows that, even when taking
into account factors that affect salary, women are predicted to earn less than men.
Comparing the UK with Other Countries
Despite the fact that the gender pay gap is a problem that is apparent worldwide, gaps vary significantly between countries.
Slovenia, for example, has the smallest (unadjusted) pay gap in Europe at 3.2% (Eurostat, 2015). They also have a high
participation rate of mothers in the labour market: 75.5 % of mothers of children under six are employed in Slovenia, compared
to 59% in the UK and a 59.1% average across the EU (OECD, 2015). In addition, almost all parents in Slovenia are employed fulltime; part-time work - which is usually paid at a lower hourly rate than full time work - is rare among Slovenian women (13.1 %),
but very popular among British women, with 43.3% in part-time roles (compared to an EU average of 32.5%). Although there are
many other factors that affect the gender pay gap, the fact that the country with the highest participation rate of mothers in
the EU also has the lowest gender pay gap seems more than circumstantial.
The high participation rate is likely due to the fact that Slovenia has progressive paternity leave and childcare policies (it is one
of only five countries in the world to offer more than two weeks paternity leave (ILO, 2014)), both of which encourage women
to return to work soon after having a child. Fathers in Slovenia have the right to 90 working days of paternity leave, at 100% pay
for 15 days and for the remaining 75 days the government pays social security contributions. Each parent also has the right to
transferable parental leave, which lasts 130 days: the mother can transfer 100 of her 130 days of parental leave to the father,
162
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
163
while the father can transfer all 130 days of his parental leave to the mother if he wishes. In addition, one of the parents has a
right to leave work in order to look after or care for a child for a period of 260 days immediately upon expiry of maternity leave.
This means that a father could take up to 480 days paternity leave in total (Republic of Slovenia Ministry of Labour and Equal
Opportunities, no date). In the UK, fathers can choose to take either one week or two consecutive weeks’ leave at either
£139.58 a week or 90% of their average weekly earnings, whichever is lower. From 5th April 2015, new regulations have also
brought Shared Parental Leave and Statutory Shared Parental Pay into force in the UK. This can be used to take leave in blocks
separated by periods of work instead of taking it all in one go, and the parents can choose how to split this leave between
them.
In contrast, Estonia’s gender pay gap stands at 29.9%, the highest in the EU (Eurostat, 2015). Although this is an improvement
from its 2010 figure of 31.5%, the gender pay gap had been steadily increasing in Estonia prior to this, despite recent equality
legislation being put in place. Due to the country’s predominantly traditional views about gender roles, limited and expensive
childcare options and a lack of part-time jobs (only 14.2% of women work part time (Statistics Estonia, 2015)), many Estonian
mothers remain economically inactive after having a child. This is confirmed by the fact that the employment gap between
women and men with children under 6 years old was 38.6% in 2008, while women and men in the same age group without
children had an employment gap of 0.1% (United Nations Economic Commission for Europe, 2010).
In January 2008 an amendment to the Holidays Act significantly increased the amount paid to Estonian fathers for their ten
days paternity leave from a low flat-rate of €4 per day to the sum of 100% of his pay (with a maximum of the average pay in
Estonia). The amount of fathers using this benefit rose from 10% to 40%. Unfortunately, due to the economic crisis, this change
was in force only for 2008. It was revoked from 2009 and fathers are now still entitled to take the ten days leave but will not
receive any money for the period. A mother or father is also granted an additional child care leave of three working days per
year at his or her request if they have one or two children under fourteen years of age, and six working days if they have three
or more children under fourteen years of age or at least one child under three years of age. In 2007, only 15% of those taking
advantage of this benefit were fathers, which was an increase from 11.6% in 2006.
Not only could paternity leave policy improve the gender pay gap, it also has the potential to grow economies; according to
research by the OECD (2013), if female labour force participation rates converged to that of men by 2030, we could see and
overall increase of 12% in GDP. In Japan, 70% of Japanese women drop out of the workforce after having their first child.
According to “Womenomics” (Matsui et al, 2010), getting these mothers back to work could boost Japan’s GDP by as much as
15%. In the long run, countries that make sure parents, regardless of their gender, can exploit the greatest possible value from
their time and skills will be those whose economies and families thrive.
163
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
164
Policy implications & Things to consider
The fact that policies such as those in place in Estonia are available to fathers but are rarely taken advantage of shows that
there is only so much that policy can do; if fathers choose not to take advantage of such schemes then it is unlikely to make a
difference to the wider issue of the gender pay gap. One reason for fathers choosing not to use their leave is that, particularly
in certain countries, there is a widely-held belief that men and women should stick to their traditional gender roles of provider
and caregiver respectively. If these attitudes are strong enough, then a policy challenging these traditional roles is unlikely to
be successful. Kristen Sobeck, an economist at the ILO said a gender pay gap “will persist insofar as societies shift care
responsibilities disproportionately to women” (Topping, 2015). Although some countries are much more progressive than others
in terms of embracing shared parental responsibility and relying less on the traditional male and female roles, women still spend
26 hours per week on care and home activities, compared with nine hours for men in the EU (Grimshaw & Rubery, 2015).
However, it may be the case that paternity leave policies in countries with traditional gender stereotypes could be successful in
the long run if other initiatives were also introduced to encourage more men to take advantage of the benefits. Looking at
Scandanavian countries, where employer mandated paternity leave policies have been in place for many years, it took some
time for men in these countries to actually take it, and in many cases they did so only after the government added incentives.
Bengt Westerberg, a Swedish politician, said of the problem: “Society is a mirror of the family. The only way to achieve equality
in society is to achieve equality in the home. Getting fathers to share the parental leave is an essential part of that” (Bennhold,
2010). In 1974, when Sweden became the first country to replace maternity leave with parental leave, only 562 fathers took
advantage of it (Guilford, 2014) and they were mockingly nicknamed “velvet dads.” Despite government campaigns —
including one featuring a champion weightlifter with a baby perched on his bare biceps — the share of fathers on leave was
stalled at 6% in 1991 (Bennhold, 2010). Sweden had already gone further than many countries have now in relieving working
mothers: the parent on leave got almost a full salary for a year before returning to a guaranteed job, and both parents could
work six-hour days until children entered school. Female employment rates and birth rates had surged to be among the highest
in the developed world, yet very few men were taking up the responsibility of caregiving. In 1995, “daddy leave” was
introduced and had an immediate impact. This new rule meant that no father was forced to stay home, but the family lost one
month of subsidies if he did not and parents were paid 90% of their wages, with the result that soon more than 80% of fathers
took parental leave. Peer influence can also have a substantial effect on male attitudes towards paternity leave; research has
found that when a man’s co-workers took paternity leave, it increased the chance that he would take it by 11 percentage
points — and that if his brother took it, by 15 points (Dahl et al., 2014). This shows that, although it may take time to change
people’s attitudes towards men having a larger share of responsibility for looking after a child, it can be achieved.
As well as increasing the participation rate of mothers, successful policies such as Sweden’s have also been shown to have had
a significant positive effect on women’s wages; according to a study published by the Swedish Institute of Labor Market Policy
Evaluation (Johansson, 2010), for each additional month that a father stays on parental leave, the mother’s future earnings
increase by an average 6.7%. However, the length of parental leave has been found to have a curvilinear relationship with
mother’s earnings (Pettit & Hook, 2005). Leave length impacts employers' perceptions of mothers' employability and,
subsequently, mothers' earnings; moderate leaves reduce the pay gap by ensuring that women remain attached to their
workplaces while children are infants, whereas long leaves are linked to decreased employment continuity and earnings.
164
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
165
Policy decisions should therefore take into account how the parental leave is structured to encourage men to take advantage
of it, as well as the length of leave available.
However, it is not only paternity leave policy that impacts on the gender pay gap. Affordable childcare policies are also
extremely important and would help to return mothers to work more quickly if they were available in the UK. Looking again at
Slovenia, parents only have to contribute one third of the cost of childcare on average, as pre-schools are heavily subsidised.
In Sweden, children have access to highly subsidized preschools from 12 months old and grandparents are offered statesponsored elderly care. This is in contrast to recent figures showing that it costs £6003 a year on average to have a child in fulltime childcare in the UK (Rutter, 2015), meaning that for many mothers it is not worthwhile for them to work, resulting in a lower
participation rate and therefore a wider gender pay gap.
Conclusion
The gender pay gap exists worldwide, though varies significantly between countries. Using the theory of statistical
discrimination, one way to tackle the gender pay gap would be to encourage women to return to work more quickly after
having children, thus increasing the long run average productivity of women and therefore raising their wages. Increasing the
participation rate (and therefore wages) of women would also be beneficial for the economy in terms of productivity and
growth. One option for achieving this would be to offer a more generous paternity leave package and to encourage men to
take a more active role in caring for children, although the policy may also need to be reinforced by other incentives to
encourage men to actually take advantage of it, as there is still a stigma attached to men who take paternity leave in some
countries.
The key to success for a policy like this would be to publically show that men and women are equally valued. At present,
governments promote the expectation that businesses should compensate women’s time equally with men’s, while their own
parental leave policies pay fathers less than mothers for providing childcare; until governments and firms value women’s and
men’s work the same (both in the office and at home) this handicap against women will still exist.
References
Bennhold, K. (2010), “In Sweden, Men Can Have It All” New York Times
Booth, A. (2009), “Gender and competition”, Labour Economics, Vol 16, pp. 599-606
Dahl, G.B et al. (2014), “Peer Effects in Program Participation”, American Economic Review 2014, 104(7), pp. 2049–2074
165
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
166
Davies, R. & Pierre, G. (2005), “The family gap in pay in Europe: a cross-country study”, Labour Economics, Vol 12 (4), pp. 469486
Eurostat, (2015), “Women earned on average 16% less than men in 2013 in the EU”, Available at
http://ec.europa.eu/eurostat/documents/2995521/6729998/3-05032015-AP-EN.pdf/f064bb11-e239-4a8c-a40b-72cf34f1ac6f
Accessed on October 26, 2015
Grimshaw, D. & Rubery, G. (2015), “The motherhood pay gap: a review of the issues, theory and international evidence”, ILO,
Geneva
Guilford, G. (2014), “The economic case for paternity leave”, Available at http://qz.com/266841/economic-case-for-paternityleave/ Accessed January 13, 2016
Harris, C. (2015), “Which EU country has the biggest gender pay gap?”, Euronews, Available at
http://www.euronews.com/2015/03/06/which-eu-country-has-the-biggest-gender-pay-gap/ Accessed December 31, 2015
ILO, (2014), “Maternity and Paternity at Work: Law and Practice Across the World”, Available at
http://www.ilo.org/wcmsp5/groups/public/---dgreports/---dcomm/---publ/documents/publication/wcms_242615.pdf Accessed
December 31, 2015
Johansson, E. (2010), “The effect of own and spousal parental leave on earnings”, Institute for Labour Market Policy Evaluation
(Sweden)
Matsui, K. et al (2010), “Womenomics 3.0: The time is now”, Goldman Sachs, Japan Portfolio Strategy
Misra, J. & Strader, E. (2013), “Gender Pay Equity in Advanced Countries: The Role of Parenthood and Policies”, Journal of
International Affairs
OECD, (2013), “Gender Dynamics: How Can Countries Close the Economic Gender Gap?” Available at
http://www.oecd.org/about/secretary-general/genderdynamicshowcancountriesclosetheeconomicgendergap.htm
Accessed January 12 2016
OECD, (2015), Family and Children Database, Data available at http://www.oecd.org/els/family/database.htm Accessed
October 26, 2015
Pettit, B. & Hook, J.L. (2005), “The Structure of Women's Employment in Comparative Perspective” Social Forces, Vol 84(2), pp
779-801
Republic of Slovenia Ministry of Labour and Equal Opportunities (no date), Translations of legislature available at
http://www.mddsz.gov.si/en/legislation Accessed December 31, 2015
Rutter, J. (2015), “Childcare Costs Survey 2015”, Family and Childcare Trust
Slater & Gordon, (2014), “Slater and Gordon highlight maternity discrimination”, Press release available at
http://www.slatergordon.co.uk/media-centre/news/2014/08/slater-gordon-highlights-maternity-discrimination Accessed
October 26, 2015
Statistics Estonia (2015) Statistical Database, Available at http://pub.stat.ee/px-web.2001/dialog/statfile1.asp Accessed
January 2, 2016
Topping, A. (2015), “
Gender pay gap will not close for 70 years at current rate, says UN”, The Guardian, Available at
http://www.theguardian.com/money/2015/mar/05/gender-pay-gap-remain-70-years-un Accessed January 11, 2016
United Nations Economic Commission for Europe, (2010), “Estonia: Overview of achievements and challenges in promoting
gender equality and women’s empowerment”, Available at
http://www.unece.org/fileadmin/DAM/Gender/documents/Beijing+15/Estonia.pdf Accessed January 2, 2016
166
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
167
Appendix A
Correlation Matrix between Age and Work Experience
Correlations
Length of time
Age of respondent
Age of
contin employ
respondent
(inc self)
Pearson Correlation
.480**
1
Sig. (2-tailed)
.000
N
62480
44744
.480**
1
Length of time contin employ Pearson Correlation
(inc self)
Sig. (2-tailed)
.000
N
44744
44744
**. Correlation is significant at the 0.01 level (2-tailed).
Appendix B
Results of Regression 1
Model Summary
Model
R
.570a
1
R Square
Adjusted R
Std. Error of the
Square
Estimate
.325
.324
.49544
a. Predictors: (Constant), Work experience, ALevel, ScotIreWal, Sex of
respondent, White, Married, LondonSE, GCSE, Age of respondent,
FurtherEd
ANOVAa
Model
1
Sum of Squares
df
Mean Square
Regression
1133.686
10
113.369
Residual
2355.192
9595
.245
Total
3488.878
9605
F
461.861
Sig.
.000b
a. Dependent Variable: Natural log of hourly pay
167
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
168
b. Predictors: (Constant), Work experience, ALevel, ScotIreWal, Sex of respondent, White,
Married, LondonSE, GCSE, Age of respondent, FurtherEd
Coefficientsa
Standardized
Unstandardized Coefficients
Model
1
B
Coefficients
Std. Error
Beta
(Constant)
1.460
.034
Sex of respondent
-.198
.010
Age of respondent
.007
LondonSE
ScotIreWal
t
Sig.
42.586
.000
-.164
-19.446
.000
.001
.145
14.469
.000
.173
.012
.123
14.036
.000
-.037
.014
-.023
-2.593
.010
White
.147
.018
.069
7.981
.000
Married
.124
.011
.103
11.348
.000
GCSE
.231
.025
.162
9.345
.000
ALevel
.337
.025
.241
13.699
.000
FurtherEd
.743
.024
.615
31.539
.000
.001
.000
.151
15.894
.000
Length of time contin employ
(inc self)
a. Dependent Variable: Natural log of hourly pay
Appendix C
Results of Regression for Male Respondents only
Model Summary
R
Sex of
respondent =
Model
1
male (Selected)
.575a
R Square
.330
Adjusted R
Std. Error of the
Square
Estimate
.329
.51169
a. Predictors: (Constant), Length of time contin employ (inc self), GCSE,
LondonSE, White, Married, ScotIreWal, ALevel, Age of respondent,
FurtherEd
168
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
169
ANOVAa,b
Model
1
Sum of Squares
Regression
df
Mean Square
579.385
9
64.376
Residual
1174.567
4486
.262
Total
1753.952
4495
F
Sig.
.000c
245.870
a. Dependent Variable: Natural log of hourly pay
b. Selecting only cases for which Sex of respondent = male
c. Predictors: (Constant), Length of time contin employ (inc self), GCSE, LondonSE, White,
Married, ScotIreWal, ALevel, Age of respondent, FurtherEd
Coefficientsa,b
Standardized
Unstandardized Coefficients
Model
1
B
(Constant)
Std. Error
1.258
.049
Age of respondent
.008
.001
LondonSE
.176
ScotIreWal
Coefficients
Beta
t
Sig.
25.438
.000
.164
11.158
.000
.019
.122
9.525
.000
-.072
.022
-.042
-3.286
.001
White
.250
.028
.114
9.077
.000
Married
.195
.017
.155
11.591
.000
GCSE
.269
.037
.177
7.249
.000
ALevel
.380
.036
.268
10.476
.000
FurtherEd
.788
.035
.629
22.559
.000
.001
.000
.124
9.012
.000
Length of time contin employ
(inc self)
a. Dependent Variable: Natural log of hourly pay
b. Selecting only cases for which Sex of respondent = male
Appendix D
Results of Regression for Female Respondents Only
Model Summary
Model
R
R Square
169
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
170
Sex of
respondent =
female
Adjusted R
Std. Error of the
(Selected)
Square
Estimate
.537a
1
.289
.288
.47629
a. Predictors: (Constant), Length of time contin employ (inc self), FurtherEd,
ScotIreWal, White, Married, LondonSE, ALevel, Age of respondent, GCSE
ANOVAa,b
Model
1
Sum of Squares
Regression
df
Mean Square
470.027
9
52.225
Residual
1156.931
5100
.227
Total
1626.958
5109
F
Sig.
.000c
230.220
a. Dependent Variable: Natural log of hourly pay
b. Selecting only cases for which Sex of respondent = female
c. Predictors: (Constant), Length of time contin employ (inc self), FurtherEd, ScotIreWal, White,
Married, LondonSE, ALevel, Age of respondent, GCSE
Coefficientsa,b
Standardized
Unstandardized Coefficients
Model
1
B
(Constant)
Std. Error
1.476
.046
Age of respondent
.006
.001
LondonSE
.168
ScotIreWal
Coefficients
Beta
t
Sig.
31.881
.000
.119
8.393
.000
.016
.127
10.244
.000
-.011
.019
-.007
-.572
.568
White
.054
.024
.027
2.195
.028
Married
.065
.014
.057
4.559
.000
GCSE
.186
.033
.143
5.685
.000
ALevel
.275
.033
.204
8.241
.000
FurtherEd
.685
.032
.606
21.637
.000
.001
.000
.189
13.934
.000
Length of time contin employ
(inc self)
a. Dependent Variable: Natural log of hourly pay
b. Selecting only cases for which Sex of respondent = female
170
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
171
Special Mention for the NEP Essay Competition-
Are we ready for a repeat of financial crisis?
Author
Anthony Amoah
School of Economics
University of East Anglia, UK
1.
Introduction
Financial crises are nothing new. They have been in existence since the development of money and financial markets
(Reinhart & Rogoff, 2009), and will continue to be associated with developments in these markets. Financial crises have been
very complex, damaging and, of course, contagious.
Historically, prior to 1929 (specifically 1907 & 1921), the world18 experienced some economic downturn, however brief. From
1929 to 1930, four events which significantly hit the worldwide economy include stock-market crash in New York, the SmootHawley tariff of 1930, the first banking [financial] crisis, and the global collapse of commodity prices (Temin, 1994).
Generally, Economists refer to the global economic downturn characterising this period as the Great Depression as described
by the Federal Reserve in 1931. Mankiw (2000) reports that some economists believe that a large contractionary shock to
private spending caused the Great Depression, while other economists believe that the large fall in money supply caused the
Great Depression. Milton Friedman and Anna Jacobson Schwartz have indicated the banking or financial crisis as the main
causal factor of the Great Depression (Temin, 1994). In fact they argue that the banking failures in November and December
of 1930 reduced the supply of money by increasing banks’ demand for reserves, and the public’s demand for currency led to
the depressed spending. This was a reflection of the falling aggregate demand that resulted from the preceding credit
stringency. The period 1982-1983 also witnessed recession, but this was followed by rapid recovery which extended to a boom
in 1987-1989. These periods of recession/financial crisis saw financial reforms in most countries. For instance deregulation of
credit and greater competition in its supply were particularly noticeable in the US and UK. However, in early 2000s; the US,
Japan and Germany were all plunged in simultaneous recession (See Begg et al., 2005). This is commonly described as ‘Early
2000s Recession’.
The implications associated with these global economic downturn include sales falling more quickly than firms had
anticipated because of fall in consumption, high unemployment, lower wages, fall in taxes, lower government revenue and
high budget deficits among others.
Indeed, the periods of recession and depression are observed to have a strong underlying relationship with financial crisis.
Financial crisis is therefore regarded as a virulent cause of recession and or depression (see Labonte and Makinen, 2002).
2.
Definition and Types of Financial
Crisis
Financial crisis is used when financial assets suddenly lose a large part of their nominal value (Kindleberger and Aliber, 2005).
Financial crises are often preceded by asset and credit booms which turns into busts. A financial crisis is often driven by a
18
Three major players in the world economy: the US, Japan and EU.
171
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
172
blend of several events, including substantial changes in credit volume and asset prices, severe disruptions in financial
intermediation, the need for large scale government support, and most conspicuously, the supply of external financing and
large scale balance sheet problems. Therefore this is regarded as a multidimensional event which does not appear plausible if
being described by a single event (Claessens and Kose, 2013). Obviously, the multidimensional nature of financial crisis
suggests a feature with various shapes and forms. In the literature, four major types of crisis are outlined which include:
currency crisis; balance of payments (or capital account or sudden stop) crisis; debt crisis; and banking crisis. We now do a
brief discussion of each as follows.
2.1 Currency Crisis
Frankel and Rose (1996, p.2) define a currency crisis otherwise ‘currency crash’ as a nominal depreciation of the currency of at
least 25% that is also at least a 10% increase in the rate of depreciation relative to the previous year. As a build-up on this
definition, Laeven and Valencia (2008) define currency crisis as a nominal depreciation of the currency of at least 30% that is
also at least a 10% increase in the rate of depreciation compared to the year before. Following this definition, these authors
have identified 208 currency crisis from 1970 to the current 2007 crisis.
2.2 Balance of Payments or Capital Account or Sudden Stop Crisis
“Fisher's style debt-deflation mechanisms can then cause sudden stops through a spiralling decline in asset prices and holdings
of collateral assets” (Fisher, 1933 cited in Claessens and Kose, 2013, p.15). They add that countries that are characterised by
relatively small tradeable sectors and large foreign exchange liabilities are mostly prone to sudden stops. In addition, countries
with widely disparate per capita GDPs, levels of financial development, and exchange rate regimes, as well as countries with
different levels of reserve coverage have been affected.
Empirical literature has shown an association between sudden stops and global shocks. For instance, some emerging markets
in Asia, Latin America, Central and Eastern Europe have shown that a period of large capital inflows, a sharp retrenchment or
reversal of capital flows occurred, mainly triggered by global shocks19. Milesi-Ferretti and Tille (2011) provides evidence that
rapid changes in capital flows are key triggers of local crisis during the recent crisis. On the contrary, Rose and Spiegel (2011),
find little role for global factors, including capital flows, in the spread of the recent crisis.
2.3 Debt Crisis
Countries that experience unfavourable economic conditions mostly rely on short-term loans foreign currency denominated
debt as their main source of capital (Eichengreen and Hausmann, 1999). This has an associated effect of increasing
vulnerabilities, especially when the domestic financial system is underdeveloped and poorly managed. Indeed, this is not new,
however economic theory has not put enough weight on this. According to Claessens and Kose (2013), when models assumes
Ricardian equivalence they tend to make government debt less important. Reinhart and Rogoff (2009a) have demonstrated
that few countries are able to escape default on domestic debt, but with adverse economic repercussions.
2.4 Banking Crisis
By definition, a systemic banking crisis occurs when “a country’s corporate and financial sectors experience a large number of
defaults and financial institutions and corporations face great difficulties repaying contracts on time” (Laeven and Valencia,
2008, p.5). They argue that this leads to a rise in non-performing loans, depressed asset prices, increases in real interest rates,
and a slowdown or reversal in capital flows. In addition, they identify 124 systemic banking crises over the period 1970 to 2007
in their study.
19
The global shocks include increases in interest rates or changes in commodity prices.
172
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
173
Repercussions of Financial Crisis
As presented by Reinhart & Rogoff (2009), we outline three main aftermath features of financial crisis.
1.
First, asset market collapses are deep and prolonged. Real housing price declines average 35 percent stretched
out over six years, while equity price collapses average 55 percent over a downturn of about three and a half years.
2.
Second, the aftermath of banking crises is associated with profound declines in output and employment. The
unemployment rate rises an average of 7 percentage points over the down phase of the cycle, which lasts on
average over four years. Output falls (from peak to trough) an average of over 9 percent, although the duration
of the downturn, averaging roughly two years, is considerably shorter than for unemployment.
3.
Third, the real value of government debt tends to explode, rising an average of 86 percent in the major post–World
War II episodes. Interestingly, the main cause of debt explosions is not the widely cited costs of bailing out and
recapitalizing the banking system. Admittedly, bailout costs are difficult to measure, and there is considerable
divergence among estimates from competing studies. But even upperbound estimates pale next to actual
measured rises in public debt. In fact, the big drivers of debt increases are the inevitable collapse in tax revenues
that governments suffer in the wake of deep and prolonged output contractions, as well as often ambitious
countercyclical fiscal policies aimed at mitigating the downturn.
These show how important it is to plan against financial crisis as its repercussions weigh hugely on country’s prospects. It is
important to acknowledge that even planning against financial crisis can be very tricky and very uncertain. Claessens and
Kose (2013) have observed that banking crisis have common patterns, yet their timing remains empirically hard to pin down.
3.
Stylised Facts of Financial Crisis on
Key Macroeconomic Variables
Available data from the World Bank database is used to produce the facts and figures that provides evidence of the impact
of the most recent financial crisis on some key macroeconomic variables such as Gross Domestic Product (GDP), Inflation,
Unemployment and Oil prices.
173
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
174
Impact of Financial Crisis on Gross
Domestic Product (GDP)
Fig. 1 shows that most of the economies in the world were growing at an average of approximately 3% from 1990 to 2006.
However, the impact of the financial crisis between 20072009 slummed these growth rates into negatives with the exception of
Sub Saharan Africa (SSA) which was performing incredibly well between 2003-2006.
Fig. 1: Impact of recent financial crisis on GDP growth.
Author’s Construct, World Development Indicators (WDI, 2016 Dataset)
Impact of Financial Crisis on Inflation
Economic theory suggests that, given demand constant and a corresponding fall in output, inflation is inevitable. This is
evidenced in Fig.2. Thus, following the imapct of the crisis on GDP, inflation rose abnormally during the crisis era due to
increase in factor prices.
Fig 2: Impact of recent financial crisis on Inflation.
Author’s Construct, WDI (2016 Dataset)
174
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
175
Impact of Financial Crisis on
Resources (Oil Prices)
A rise in inflation is generally associated with high factor/input prices as earlier mentioned. One key production input with high
demand globally is oil. The prices of oil sored astronomically during the 2007-2009 crisis peaking above $130 a barrel. This is
shown in Fig. 3.
Fig 3: Impact of recent financial crisis on Oil Prices.
Adapted from Research.Stlouisfed.org
Impact of Financial Crisis on
Unemployment
The fall in GDP, rise in inflation and cost of production has a direct link with lay-offs. The unemployment rates in most
economies rose in a phenomenal fashion with its associated social consequences.
Fig 4: Impact of recent financial crisis on Unemployment.
175
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
176
Author’s Construct, WDI (2016 Dataset)
It is therefore evident that the financial crisis severely affected macroeconomic variables in most economies as shown in
figures 1-4.
4.
Are we ready for a repeat?
Evidence from the United Kingdom
(UK)
The macroeconomic indicators shown in figures 1-4, show the extent to which economies are struggling to recover fully from
the crisis and totally get back on their feet. To better show the readiness of economies, we picked one developed country
which has performed well over the years yet was hardly hit by the financial crisis of 2007, to justify whether or not countries are
ready for a repeat. We evaluate some key macroeconomics fundamentals (GDP growth, inflation, unemployment, exchange
rate), to ascertain the robustness of the economy to withstand a repeat.
Fig 5a. 5-Year GDP Growth Rates Before and After Crisis
Fig 5b. 5-Year Inflation Rates Before and After Crisis
176
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
2002
2006
GDPG Before Crisis(%)
2002
2006
Unemp Before Crisis(%)
2010
2014
GDPG After Crisis(%)
2010
Unemp After Crisis(%)
2014
2002
177
2006
Inf Before Crisis (%)
2002
2006
Forex Before Crisis
2010
2014
Inf After Crisis(%)
2010
2014
Forex After Crisis
Fig 5c. 5-Year Unemployment Rates Before and After Crisis. Fig 5d. 5-Year Exchange Rates Before and After Crisis
Source (5a,b, c and d): Author’s construct with WDI (2016) Dataset
Fig 5a shows that the average GDP growth was about 2.80% before the crisis. Although UK’s recovery has been strong showing
a sharp rise between 2012 and 2014, however the average GDP growth within the last five years after the crisis is about 1.96%.
Showing that, in average terms, the economy after the crisis has not grown enough relative to the period before the crisis.
Fig 5b illustrates an average inflation rate of about 1.67% before the crisis. Again, the trend shows an impressive marked fall in
the rate of inflation from 4.5% to 1.5% in 2011 and 2014 respectively, yet the average rate of inflation after the crisis is 2.92%. This
shows that the average inflation rate after the crisis is higher relative to the case before the crisis.
Fig 5c demonstrates that the unemployment rate in 2004 was rising steadily until the crisis era which further got worse. The
period after the crisis has shown an impressive fall in unemployment. That notwithstanding, the average unemployment rate
after the crisis is 7.50% compared to 5.02% before the crisis.
In Fig 5d, the period before the crisis shows a very sharp decline in exchange rate from 2002 to 2005 and gradually rose before
the crisis. The period after the crisis has shown an impressive performance in the exchange rate. The average exchange rate
from 2010 to 2015 was 0.63 per US dollar ($) showing a stronger exchange rate relative to the 0.62 per US$ from 2001 to 2006.
However, one can argue that the average change is only 0.01 per US$ suggesting a very small change in the two periods
under consideration.
Following our discussion, one can argue that the macroeconomic fundamentals have seen significant improvement after the
financial crisis. However, the fundamentals do not show a robust economy with enough absorbers to withstand any immediate
shock. According to a report from the New Economics Foundation as cited by Kikialdy (2015, p.1), “The UK’s financial system is
the least resilient of any economy in the G7”. “In a new analysis the UK comes last in five out of seven indicators of financial
resilience, leaving it more susceptible to financial shocks than any other G7 state”. The study found the UK economy's
resilience to have declined in the early 2000s. Although it recovered a bit, the recent crisis makes it still lags significantly behind
other leading economies. Tony Greenham, Head of Economy and Finance at the New Economics Foundation (NEF) is
reported by Kikialdy (2015) to have said: “[w]ithout real structural reform, we remain extremely vulnerable to future financial
storms. Yet even the limited progress made since 2008 now seems at risk of being unpicked by lobbying from the big banks.”
“Far from being done and dusted, banking reform is serious unfinished business….there can be absolutely no room for
complacency”. Decisions regarding banking policies vis a vis financial crisis must be measured against the yardstick of
financial system resilience if we intend to protect the economy from financial shocks.
177
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
5.
178
Conclusion and Policy
Recommendations
In sum, this essay argues that the UK economy is still recovering as shown in the post-crisis era illustrated in figures 5a, b, c and
d. However, this is observed not to be robust relative to precrisis era. In addition, with reference to the NEF Financial Resilience
Index scores (2012 data) the UK’s resilience is at 0.27, with the US on 0.56, Canada 0.62, Italy 0.63, France 0.66, Japan 0.71 and
Germany 0.73. These conclusions are guided by the analysis of some key macroeconomic fundamentals and the seven 20 key
financial system resilience factors.
For policy purposes, first we propose strong policies to speed up the performance of the macroeconomic fundamentals.
Again, we perfectly agree with the policy recommendations as reported by NEF that: “It is vital that policymakers develop a
more sophisticated understanding of resilience and use it to help reshape our financial system. Unless this is achieved, our
economy and society remain at risk of a future financial crisis.” NEF further gave the following recommendations: separating
retail from investment banking, promoting bank diversity, making efforts to increase peer-to-peer lending to increase the
resilience of the financial system.
All in all, we argue that that countries experiencing financial crisis have suffered from deep recessions and sharp current
account reversals. The repercussions of the most recent 2007-09 global financial crisis have been described by Claessens and
Kose, (2013, p.3) as a painful reminder of the multifaceted nature of crises. They hit small and large countries as well as poor and rich ones.
Similarly, Reinhart and Rogoff (2009a) argue that, financial crises are an equal opportunity menace. We point out that its occurrence could
stem from a policy mistake or a shock. Against this background, this study assesses whether current economic structures in UK
can withstand a repeat of this menace. This study finds that in spite of the fact that the UK’s economy is recovering, a lot still
needs to be done to have a robust economy that can withstand a repeat. Based on the case of the UK, we argue that if
proper economic forecasting and robust management systems are not put in place, withstanding a repeat of financial crisis
would pose an unsurmountable challenge to developed and developing economies alike. The example of UK suggests that
globally, more needs to be done to withstand a repeat of any unforeseen financial crisis.
References
1.
Begg, D., Fischer, S and Dormbusch, R. (2005, p.545). Economics 8th Edition, McGraw-Hill Education, Berkshire.
2.
Claessens, S., Kose, M. A., Laeven, L., & Valencia, F. (2013). Financial Crises:
Explanations, Types, and Implications, IMF Working Paper WP/13/28, IMF.
3.
Eichengreen, B., & Hausmann, R. (1999). Exchange rates and financial fragility (No. W7418). National Bureau of Economic Research.
These include diversity, interconnectedness and network structure, financial system size, asset composition, liability composition, complexity
and transparency; and leverage.
20
178
THE NORWICH ECONOMIC PAPERS- VOLUME 14 | Issue #
4.
179
Frankel, J. A., & Rose, A. K. (1996). Currency crashes in emerging markets: An empirical treatment. Journal of international Economics,
41(3), 351-366.
5.
Kindleberger, C. P., & Aliber, R. Z. (2005). Manias, Panics, and Crashes, Fifth Edition John Wiley & Sons, Inc. Hoboken, New
Jersey.
6.
Kirkaldy, L. (2015) UK financial system found to be least resilient in G7, https://www.holyrood.com/articles/news/ukfinancial-system-found-be-least-resilientg7, [Accessed on 04/02/2016].
7.
Labonte, M., & Makinen, G. E. (2002, January). The Current Economic Recession: How Long, How Deep, and How Different
From the Past?. Congressional Research Service, Library of Congress.
8.
Laeven, L., & Valencia, F. (2008). Systemic banking crises: a new database. IMF Working Papers, 1-78.
9.
Mankiw, N. G. (2000, p. 390). Macro-economics, 4th Edition, Worth Publishers, New York.
10.
Milesi-Ferretti, G. M., & Tille, C. (2011). The great retrenchment: international capital flows during the global financial crisis.
Economic Policy, 26(66), 285-342.
11.
Reinhart, C. M., & Rogoff, K. (2009). This time is different: eight centuries of financial folly. Princeton University Press.
12.
Rose, A. K., & Spiegel, M. M. (2011). Cross-country causes and consequences of the crisis: An update. European Economic Review,
55(3), 309-324.
13.
Temin, P (1994). The Great Depression, National Bureau of Economic Research, Working Paper Series, Historical Paper
No.62.
179