Download Scarth 221527 en

A joint Initiative of Ludwig-Maximilians-Universität and Ifo Institute for Economic Research
LABOUR MARKET INSTITUTIONS
AND PUBLIC REGULATION
A CESifo and ISPE Conference
Villa La Collina, Cadenabbia (Como)
2-4 June 2002
Some Macroeconomic Consequences of
Basic Income and Employment Subsidies
Thomas Moutos & William Scarth
CESifo
Poschingerstr. 5, 81679 Munich, Germany
Phone: +49 (89) 9224-1410 - Fax: +49 (89) 9224-1409
E-mail: [email protected]
Internet: http://www.cesifo.de
May, 2002
Some Macroeconomic Consequences
of Basic Income and Employment Subsidies
Thomas Moutos
(AUEB and CESifo)
and
William Scarth
(McMaster University)
Abstract
A small open-economy macro model is used to examine the scope for income
redistribution and employment creation. In particular, the introduction of a guaranteed
annual income is evaluated, and it is compared to the introduction of an employment
subsidy. While several models are considered, in the central one, all factors of production
except unskilled labour are mobile internationally, so the financing of these initiatives
can involve undesirable indirect effects. The models are used to assess the relative
importance of these competing effects – first in a setting where saving and changes in
foreign debt are not involved, and then in a setting where these considerations are
highlighted.
Paper presented at the CESifo-ISPE conference on “Labour Market Institutions and
Public Regulation,” Munich, October 26-27, 2001. We wish to thank our discussant,
Martin Werding, the organizers (Jonas Agell, Michael Keen and Alfons Weichenreider),
Manos Matsaganis, Apostolis Philippopoulos, as well as the conference participants for
many helpful comments and suggestions.
1
1. Introduction
Recent years have witnessed growing income inequality and persistently high
unemployment among the less skilled. In the United States and in the United Kingdom
the worsening prospects have taken the form of decreases in the real earnings of lowerskilled workers – the real hourly wages of young males with 12 or fewer years of
schooling has dropped by more than 20 percent in the last two decades. In continental
Europe real wages at the bottom of the skill distribution have risen, but at the cost of
significant increases in unemployment – especially for this group (Freeman (1995) and
Machin and Van Reenen (1998)). Although there have been competing explanations for
the primary reasons behind the increased inequality (international trade versus
technological change), there is widespread agreement among economists that the major
cause of increasing inequality has been a shift in demand away from unskilled labour in
favour of skilled workers.1 It is for this reason that subsidies for the employment of lowskill (or unskilled) workers have been proposed (see, for example, Phelps (1997) and
Solow (1998)) as a solution to the “plight of the less-skilled.”
At the same time, these developments have stimulated renewed interest in policies that
may redistribute income in ways that minimize undesirable indirect effects. This is
particularly so since one of the usual charges made against the social welfare system is
that in many countries it nourishes a collectively sub-optimal incentive structure, ranging
from excessive early retirement to “poverty traps” for unemployed workers (especially
single mothers) who return to low-wage employment. In many countries, the implicit tax
rate at the low end of the earnings distribution is often very large because of the phasing
out of transfer programs as income rises. For example, Atkinson and Sutherland (1990)
report that in Britain in 1989 almost half a million families faced marginal tax rates of 70
per cent or higher, as a result of means-tested social assistance benefits. Blundell and
MaCurdy (1999) also provide an extensive analysis of marginal tax rates faced by lowincome households in the US and in the UK, and show that the implicit tax rate may
sometimes exceed 100% when two or more transfer programs are phased-out
simultaneously. Despite these costs, the welfare state can still have a net positive
(efficiency enlarging) effect in modern industrialized economies. Indeed, as Sinn (1995)
and Atkinson (1999) have persuasively argued, the welfare state should not be viewed as
something that only disturbs the market process. Especially in a second-best setting, it is
something that can encourage risk-taking, foster efficiency and facilitate the growth
process.
Many individuals have expressed support for an unconditional payment of a guaranteed
income to all citizens. The features that distinguish the Basic Income (BI) proposal –
variously called “guaranteed annual income”, “universal basic income”, or “demogrant”
(see, Meade (1948, 1972), Tobin et al. (1967), Atkinson (1995), Van Parijs (1995, 2000))
– from other social security proposals, are that it is paid irrespective of any other income,
it does not require any present or past work performance, it is not conditional on the
willingness to accept a job, and it is paid to individuals rather than households (Van
Parijs, 1992). Some proponents argue that a BI system should be accompanied by
widespread social security reform, including deregulation of the labour market. For
2
Atkinson (1995), the BI proposal, in its pure form, would replace all social security
benefits and it would be accompanied by a flat comprehensive income tax rate that would
replace the existing income taxes and social security contributions. In this way,
proponents of basic income hope to provide a solution to the “impossible trinity” of
welfare reform objectives: to raise the living standards of low-income families, to
encourage employment, and to keep budget costs low.
In this paper, we examine some of the macroeconomic effects of BI, and we compare
these outcomes to what follows from employment subsidies. Despite the fact that some
proponents of BI expect this initiative to be financed by cuts in existing social programs,
we also consider alternative sources of financing in this paper. We do so not only because
we want to examine the possibility of redistribution, but also for reasons of political
feasibility. Even within the European Union (EU) there are significant differences across
countries in the relative importance attached to redistributive social policy goals, in the
instruments used, and in the extent to which social policy achieves its intended effects.2
The evolution of the social welfare system in each country has created constituencies
which will strongly resist any reductions in the benefits to which they have become
"entitled." Thus, while we examine BI and employment subsidies financed by cuts in the
generosity of unemployment insurance, we also consider financing these initiatives by
taxing the "rich."
The remainder of the paper is organized as follows. In section 2, we outline a baseline
model – a closed economy in which agents do not save. This is the standard setting for
analyses that focus on labour-market outcomes (see, for example, Pissarides (1998)). The
baseline model assumes the existence of only one factor of production – homogeneous
labour. We use this structure to gain intuition and some confidence that our results are not
significantly tied to the particular models used in the rest of the paper.
In section 3, we introduce the main model of this paper. It involves a fairly standard
open-economy macro model with three factors of production: skilled labour, unskilled
labour, and physical capital. Following the recent literature on skill-biased technological
change (see, for example, Goldin and Katz (1998) and Krusell et al (2000)) we assume
that physical capital and skilled labour are complements in the production process
whereas unskilled labour and physical capital are substitutes. It is assumed that BI and
employment subsidies are financed by taxing the "rich" (either skilled workers or the
owners of capital). The small open-economy setting is chosen so that the analysis can
address the concern that globalization may limit the effectiveness of redistribution policy
– since internationally mobile factors of production can escape taxation.
In section 4, we compare two methods of financing redistribution and labour market
policies – taxing the earnings of either skilled labour or the owners of physical capital.
This is (partly) motivated by Lucas’s (1990) assertion that societies can enjoy a “free
lunch” if the tax on physical capital is cut. Then, in section 5, we derive the effects of the
introduction of BI and employment subsidies on unemployment and on the incomes of
both the working poor and the non-working poor. Some extensions are considered in the
remainder of the paper. The sensitivity of our results to different assumptions concerning
3
the nature of the production function, the reason for unemployment, and the international
mobility of skilled labour are examined at the end of Section 5. In section 6, the model is
broadened to consider saving and wealth accumulation among the rich. In this setting it is
possible to define a progressive expenditure tax – a levy that is very often proposed as the
favoured method of financing government policies that are designed to foster equity. The
efficacy of BI (relative to unskilled employment subsidies) is re-assessed in this longerterm setting. Section 7 offers concluding remarks and suggestions for further analysis.
Finally, in an Appendix, a micro model of the family's labour-supply decision is used to
assess one feature of our macro specifications – that BI has no effect on labour-force
participation.
2. The Baseline Model
In this section, we follow convention (for example, Pissarides (1988)) and examine a
closed economy model with homogeneous labour as the only variable factor of
production. Initially, we focus on an efficiency wage theory of unemployment. Then, as a
sensitivity test, we consider unemployment that results from bargaining between unions
and firms.
In efficiency wage models (see, for example, Akerlof and Yellen (1986)) the output of a
firm’s workforce depends on the wage the firm pays. The efficiency wage theories are
based on the premise that employers can not acquire full information about the
productivity of their workers (and this proposition is reflected in most employment
contracts, since they do not involve precise specifications of productivity). A higher wage
offer by the firm may increase the average productivity of its workforce for several
reasons. First, a high-wage firm may (on average) attract workers of higher quality.
Second, a higher wage increases the magnitude of punishment incurred by a worker who
is fired after being found offering a sub-standard amount of effort. Third, high wages may
lead workers to believe that they are treated “fairly”, and they may reciprocate to this
“gift” by offering higher effort. The implication of this dependence of worker
productivity on wages is that the firm will want to choose a wage rate such that the
marginal benefit from a wage increase is equated with the associated increase in costs.
Thus, the profit-maximizing wage rate chosen by firms may be compatible with
involuntary unemployment. The unemployed may be willing to work at lower wages, yet
if firms employ them, marginal revenue would decline more than marginal cost.
In this section we adapt a particularly compact version of efficiency wages (due to
Summers (1988) and highlighted by Romer (1996)), which is compatible with any of the
motivations mentioned above (and which is more readily calibrated than the efficiencywage model used by Pissarides). In this setting, output, Y, depends on effective labour,
bN, where b is the index of labour efficiency and N is the quantity of labour. The
production function is Cobb Douglas:
Y = (bN ) β .
(1)
Following Summers, we specify b as:
4
b = [( w(1 − t ) + pw) − x ]α .
(2)
The term in round brackets defines what individuals receive if they are working; x
denotes their alternative option that is available should they leave their current job. t is
the income tax rate applied to labour earnings, and p is a parameter that defines BI. This
parameter measures the generosity of an unconditional (and tax-free) transfer of income
from the government to all individuals – independent of employment status. We assume
that this BI (which is the same for all individuals) is proportional to the wage rate. α is a
positive fraction; as a result, a higher wage in the current job raises each worker’s return
relative to her alternative, and thereby induces higher productivity. The worker’s
alternative option is defined as
x = (1 − u )(1 − t) w + ufw + pw,
(3)
where f is a parameter measuring the generosity of the unemployment-insurance system.
That is, benefits paid to the unemployed are a proportion (f) of the wage, and these
benefits are untaxed. The worker’s alternative option is a weighted average of the wage
offered at other firms (which equals w in full equilibrium) and what is received if the
individual cannot find work. The weights are the employment rate, (1-u), and the
unemployment rate, u, respectively. When firms optimize, they regard x as independent
of their individual wage and employment decisions.
Firms choose employment and the wage rate to maximize profits (π) – respecting the two
constraints specified by equations (1) and (2). Profits are defined as
π = Y − wN + QN .
Q is a per-employee subsidy paid to the firm. We assume that Q is proportional to the
economy-wide wage rate, that is: Q = qw . Nevertheless, when individual firms optimize,
they do not think of this equation holding at the individual level. That is, when choosing
w, firms do not think that the subsidy rate that they will receive depends on their
individual wage policy.
Setting the derivatives of the profit function with respect to w and N equal to zero,
manipulating the first-order conditions, and using the definition of x, we derive the
following relationships:
u = α(1 − q )(1 − t ) /(1 − t − f )
(4)
βY / N = (1 − q )w .
(5)
Equation (4) states that the unemployment rate depends positively on the generosity of
unemployment insurance and the income tax rate paid by workers, and negatively on the
5
subsidy rate paid to firms for employing workers. Equation (5) states that the marginal
product of labour should be equal to its (net of subsidy) rental rate (the pre-tax wage).
Initially, we assume a fixed labour force (population, P, equals unity) so employment, N,
equals (P – U). Since the unemployment rate, u, equals U/P, we have
N = 1 − u.
(6)
Finally, we note the government's budget constraint. Since this basic version our model
ignores wealth accumulation issues (by implicitly assuming that individuals consume all
their income) we must assume that the government balances its budget. This is stipulated
in equation (7):
fwu + pw + qw(1 − u ) = ξw(1 − u ).
(7)
Equation (7) states that spending on unemployment insurance, unconditional income
transfers (the BI), and employment subsidies must be covered by part of the income tax.
It is worth noting at this point that in any real economy, the overall income tax rate,
t = g + ξ, is larger than component ξ since taxes also have to cover government spending
on other goods (or programs). We assume that the ratio of per-capita spending on these
other things to per-capita labour income is held constant (at g), so this additional
component of the income tax rate is constant.
Having completed the description of the baseline model, we note that equations (1) to (7)
solve for Y, N, u, w, b, x and one government policy variable (which we take to be either
p or q – depending on whether we are considering BI or the employment subsidy).
To derive policy effects, we take the total differential of equations (1) to (7), and then
simplify the coefficients of the resulting system that relates the changes in all variables in
two ways. First, we evaluate the coefficients subject to the restrictions implied by the
initial steady state. Second, we set the initial values of BI (p) and the employment
subsidy (q) equal to zero. We focus on two effects that follow from a reduction in the
generosity of employment insurance – the effect on the unemployment rate and the
average income of each individual (variable x). We examine these outcomes in two cases;
first, when the revenue saved from providing unemployment insurance is used to finance
the introduction of BI, and second, when this revenue is used to introduce an employment
subsidy.
The unemployment-rate effects follow from equations (4) and (7). It is left for the reader
to verify that the unemployment rate must fall more when the wage subsidy is introduced.
That is, du/df is positive in both cases, but du/df must be bigger when the revenue saved
by reducing the generosity of unemployment insurance is used to introduce the
employment subsidy – as opposed to BI. Figure 1 illustrates the effect of each policy and
gives intuition behind this difference in unemployment-rate outcomes.
6
The vertical line indicates the fixed (for now) number of individuals in the economy. The
solid downward sloping line is the initial labour demand function, and the solid upward
sloping line is the initial wage claims line. The height of the wage claim line depends on
workers' outside option – this option gets larger as employment increases since in this
case the probability of unemployment decreases. Initially, the economy's observation
point is the intersection of the demand and the wage claim line – point A. Unemployment
is indicated along the horizontal axis. A cut in the generosity of unemployment insurance
reduces workers' outside option, and this lowers the wage claim line. When the saved
revenue is used to provide BI, there are no further shifts in any of the curves – since this
transfer payment is independent of employment status. As Figure 1 illustrates, the
outcome is point B; unemployment falls, but the wage received by workers also falls.
Workers can still be better off however – despite the fall in wages – since they now
receive the BI. The reader can verify that (dx/x)/df is of ambiguous sign, but this
uncertainty is readily resolved when representative parameter values are inserted into the
multiplier expression. We consider the following illustrative values: initial
unemployment rate: 0.06; initial BI and employment subsidy parameters (p and q): zero;
generosity of employment insurance (f) and the overall tax rate (t): 0.33; production
function parameter (β): 0.67. Equations (4) and (7) then determine α and ξ residually.
With these parameters, it is easily seen that the average income of each worker rises
when BI is introduced.
w
Figure 1
Labour
Market
supply of
people
wage
claim
A
C
B
demand
N
unemployment
The employment subsidy also involves the downward shift in the wage claims line –
since unemployment generosity is reduced by the same amount. But in this case, the
saved revenue is used for subsidizing employment, and as a result, the labour demand
curve shifts to the right, as shown by the dashed demand curve in Figure 1. The outcome
is point C – a bigger reduction in unemployment and less downward pressure on the
wage rate. Indeed, the wage can rise in this case, but even if it does, average labour
income, x, may still not rise as much as with BI, since there is no transfer payment at the
personal level in this case. For the representative parameter values noted above, average
income rises more with BI, than it does with the employment subsidy.
7
The intuition behind this set of results is straightforward. Imperfect information
concerning employee effort creates a second-best problem. There is only one price, the
wage rate, to clear two markets – that for the individuals wanting work and that for
worker effort. Private firms choose to have the wage perform the latter function, with the
result that unemployment is higher than its optimal value.3 It is the imperfect information
(or the market power in the model of unions below) that raises the private cost for the
firm to hire labour (so that it exceeds the social cost). The employment subsidy attacks
this market failure at source, so it is an efficient way to lower unemployment. This is an
application of a standard principle in applied welfare economics (Bhagwati and
Ramaswami (1963)) that it is best to correct a distortion in the sector that it occurs. BI
just ameliorates the symptoms of this market failure, so it is not as effective at lowering
unemployment. Nevertheless, since low average income is the symptom, it should not be
surprising that BI - which directly addresses the level of income – can (and does for all
reasonable parameter values) raise incomes more.
In summary, this basic model suggests that there is a trade-off regarding the two
proposals for dealing with economic hardship. An employment subsidy is preferred if we
wish to lower unemployment (and thereby help some of those which were initially out of
work), while BI is preferred if we wish to raise the incomes of those who have jobs and
of those still out of work after the enactment of either policy. By focusing on the x
variable, we have calculated the effect on the "average" person – who is sometimes
unemployed.
The remainder of the paper investigates whether these conclusions need to be modified
when sensitivity tests are considered. Two straightforward variations of the model are
considered in this section, and more extensive changes are pursued in later sections of the
paper.
Since there is controversy concerning how best to model unemployment, we consider an
alternative to efficiency-wage theory. In particular, in Europe, the role of unions in
pushing the wage above market clearing levels is often stressed. Pissarides (1998, p. 162)
outlines a compact specification of the interaction between unions and firms. It involves
firms choosing employment after the wage is set as a result of a Nash bargaining process.
If individuals are risk neutral, and the production process is Cobb-Douglas, Pissarides
shows that the closed-form solution for the unemployment rate is precisely our equation
(4) above. In this case, parameter α is defined differently: α = (ε(1 − β )) /(β (1 − ε)),
where β is labour's exponent in the production function (as before), and ε is labour's
bargaining power parameter in the Nash-product involved in the theory of wage setting
(this parameter determines the share of the surplus resulting from the employment
relationship which goes to the workers). The only other change in the model is that, with
unions instead of efficiency wages, parameter b is unity. With the structure of the model
almost identical, it is not surprising that the results (and the intuition provided by Figure
1) are the same as above. We conclude that, by focusing on the efficiency-wage
specification in later sections of the paper, we do not expect our results to be significantly
different from those which would have arisen had we focused on unions.4
8
A second model-specification issue concerns labour mobility. In Europe especially, it
may be expected that individuals could migrate into any country that provides more
income and employment support. Such a possibility was ignored above. To check on how
this possibility can affect the results, we now consider the opposite polar case. We now
assume perfect labour mobility – which ensures that individuals enter or leave this
economy so that variable x always equals what is available elsewhere ( x = x ).
Technically, with x now given and fixed, one other variable that was previously
exogenous must become endogenous. That variable is the size of the population, P. But
since this change in model set-up has no effect on either equations (4) or (7), the
unemployment-rate effects are exactly as before. Also, while the "average" person's
income, x, is unaffected by either policy, it is again the case that workers are better off
with the employment subsidy, and the unemployed fare better with BI. Given that all
these outcomes are the same as those that emerged with no labour mobility, we feel
comfortable reverting to the baseline specification (concerning this issue) in later
sections.
We are particularly interested in the scope for redistribution in an increasingly
"globalized" world. Protesters fear that any one country's ability to redistribute may be
frustrated by the fact that those who must be taxed to finance redistribution – capital
owners and skilled workers – are mobile and can escape taxation. To pursue this concern,
we add physical capital and skilled labour to the model and assume international mobility
for both these factors – but not for unskilled labour. Since much modern analysis (for
example, endogenous growth theory) emphasizes the importance of human capital, we
specify the production process so that skilled labour has a special role. We assume that
the cost of moving precludes any mobility for the unskilled (who are the only individuals
who suffer unemployment in the extended models examined below). Also, in keeping
with the concern for redistribution, we make different assumptions concerning how BI
and the employment subsidy are financed in the remaining sections of the paper. Instead
of cutting an existing income-support program (unemployment insurance), we tax the
"rich" – with either taxes on the earnings of capital or skilled labour, or an expenditure
tax on the owners of these factors. We now turn to the first of these more elaborate
changes in specification.
3. A Three-Factor Model
In this section of the paper, we model a small open economy with three factors of
production: skilled labour (S), unskilled labour (L) and physical capital (K). Both skilled
labour and physical capital can move in and out of the country without cost. In contrast,
and in some accordance with the evidence, we assume that unskilled labor is
internationally immobile, and in inelastic supply. We set L = 1. In addition, we assume
that firms (correctly) believe that wages exert an influence on the productivity (effort) of
their unskilled workforce. No such effect is involved with skilled workers; since they
have “good jobs,” these individuals are motivated and they exert a level of effort that is
independent of wages received. Firms have an incentive to manipulate the wage offered
to the unskilled so that costs per efficiency unit of unskilled labour are minimized. As a
9
result, the (unskilled) wage rate exceeds the level that would clear the market, and this
determines the equilibrium unemployment rate for unskilled labour. We continue relying
on Summers’ (1988) specification to model unemployment among unskilled workers.
Given the lack of incentive among firms to manipulate the wage of skilled workers, it is
assumed to adjust freely to make skilled workers indifferent between working in this
country and in the rest of the world (so no skilled workers are unemployed).
The fact that we assume that only unskilled workers are motivated by higher wages
seems to be in stark contrast with the overboarding literature on manager pay. Yet, wages
seem to be the main mechanism through which a firm could motivate (unskilled) workers
who are locked into a particular position and are virtually certain to remain there without
a promotion for the rest of their careers. In contrast, skilled workers (managers and other
professional staff) are subject (to a much greater degree) to a score of other incentive
mechanisms in addition to wages. For example, firms give preference to internal
candidates for promotion, sometimes promoting an internal person (who is deemed to
have performed satisfactorily in the past) when an outside candidate is superior. Also,
there is widespread use of tournament-like promotion procedures, the selective granting
of human capital investment opportunities, and the granting of stock options (see, Lazear
(1995)). Thus, our assumption that efficiency wages are paid only to unskilled workers
should be interpreted as an attempt to capture the fact that firms have more powerful
incentive mechanisms at their disposal for eliciting effort among skilled workers.
Motivated by the literature on skill-biased technological change, we assume that skilled
workers are critical in the production process. Neither unskilled labour nor physical
capital can be substituted for skilled individuals. Indeed, as Wood (1994, p.34) has
persuasively argued, the possession of particular types or amounts of physical capital
cannot give a firm (or a country) a technological advantage, any more than a competitive
advantage in the production of aircraft can be gained by a country simply by purchasing
in the international market a lot of machinery which is used in their production. The
following production function is the simplest that can capture this special role for skilled
labour (along with factor substitution being possible among the other two factors):
Y = min[ K 1−γ (bL) γ , S / θ] .
γ and θ are positive coefficients (fractions); L is the level of employment of unskilled
labour, and b is the index of work effort supplied by unskilled individuals. Parameter θ
defines the Leontief coefficient that stipulates the fixed proportion that skilled labour and
the other composite (K-L) input must be combined. Parameter γ pins down the factor
shares in the Cobb-Douglas function that defines that remaining value-added process.
Before proceeding with the analysis, we note that – at the end of this section – we report
how our results are affected by dropping both this special role for skilled labour and the
efficiency-wage rationale for unemployment. In that sensitivity test, we consider a CobbDouglas production function for all three factors, and unions as the reason for
unemployment among the unskilled.
10
Proceeding with the core model, we note that the firm’s profit function is:
π = Y − w(1 − q) L − vS − (r + δ ) K .
r is the interest rate; δ is capital’s depreciation rate; and w and v are the wage rates paid to
unskilled and skilled workers, respectively. The rest of the variables are as specified in
the previous section. Equations (2) and (3) of the previous section continue to define the
effort function and the alternative option for unskilled workers.
Setting the derivatives of the profit function with respect to w, L and K equal to zero,
manipulating the first-order conditions, and using the definition of x, we derive the
following relationships:
u = α(1 − q )(1 − t ) /(1 − t − f )
(8)
(1 − θv )(1 − γ )Y / K = ( r + δ )
(9)
(1 − θv)γY / L = (1 − q) w .
(10)
Equation (8) states that the unemployment rate depends positively on the generosity of
employment insurance and the income tax rate paid by unskilled workers, and negatively
on a subsidy rate paid to firms for employing unskilled individuals. Equation (9) states
that the net (after the payments made to skilled labour) marginal product of physical
capital should be set equal to the rental cost of capital. In similar fashion, equation (10)
states that the net (after payments to skilled labour) marginal product of unskilled labour
should be equal to its (net of subsidy) rental rate (the pre-tax wage). Profit maximization
also implies that the firm does not waste resources by hiring factors in any other
proportions than those dictated by the Leontief part of the production function. Thus, we
specify:
Y = K 1−γ (bL) γ
(11)
S = θY
(12)
The assumptions that skilled labour and physical capital are perfectly mobile
internationally imply that their after-tax rewards are equal to those prevailing in the rest
of the world. This implies that
v(1 − φ) + pw = v
(13)
r (1 − τ ) = r
(14)
where φ,τ , v and r are: the tax rates applied to the wages of skilled workers and to the
(net of depreciation expenses) returns of capital, and the after-tax rewards that can be had
by skilled workers and the owners of capital in the rest of the world. Equations (13) and
11
(14) constrain the government’s ability to redistribute income, since they imply that the
“rich” (the skilled individuals and the owners of capital) stand ready to withdraw their
services to whichever degree is required to insulate their net returns from any taxes
imposed.
As noted in the previous section, there is one additional limitation on the government’s
use of fiscal incentives and transfers – the fact that it must respect its budget constraint. A
balanced budget is stipulated in equation (15):
G + fwu + pw (1 + S ) + qw (1 − u ) = tw(1 − u) + φvS + τrK.
(15)
Equation (15) states that spending on goods (G), unemployment insurance, unconditional
income transfers (the BI), and employment subsidies must equal the sum of the three
forms of income tax revenue. Recall that, with the supply of unskilled labour equal to
unity, employment of the unskilled, L, equals (1-u).
Having completed the description of the model, we note that equations (2), (3) and (8) to
(15) solve for Y, S, K, u, w, v, r, b, x and one government policy variable (which we take
to be τ). Proceeding with the solution of the model, we first divide both sides of equation
(15) by Y, define g = G/Y, k = K/Y and use equation (10) to substitute out w/Y. Equation
(15) becomes:
γ ( fu + p (1 + θY ))(1 − θv ) = (1 − u )(γ (t − q)(1 − θv ) + (1 − q )(τrk + φvθ − g )). (15a)
To derive policy effects, we take the total differential of equations (2), (3) and (8) to
(15a), and then simplify the coefficients of the resulting system that relates the changes in
all variables in two ways. First, we evaluate the coefficients subject to the restrictions
implied by the initial steady state. Second, we set the initial values of BI (p), the
employment subsidy (q), and the tax on capital (τ) equal to zero, and the initial values of
Y and v equal to unity. After some tedious algebra, we have derived some straightforward
formulae that are reported and explained below. We focus on four effects that follow
from the introduction of BI (an increase in p) that is financed by the introduction of a tax
on the earnings of capital (an increase in τ). We examine the effects on: the
unemployment rate (for unskilled labour, u), the income of each employed unskilled
individual (z), the income of each unemployed unskilled individual (n), and the expected
income of each unskilled individual (who undergoes periods of employment and
unemployment, x). Recall that the incomes of skilled individuals and the owners of
capital are unaffected. The percentage change in each of the indicators of unskilled labour
income are calculated as follows:
(dz / z ) = (dw / w ) − (1 /(1 − t ))dt + (1 /(1 − t ))dp
(dn / n ) = (dw / w ) + (1 / f )df + (1 / f )dp
(dx / x ) = (dw / w) − ((1 − u ) /ψ )dt − ((1 − t − f ) /ψ)du + (u /ψ )df + (1 /ψ) dp
12
where ψ = (1 − t )(1 − u ) + uf + p.
As was the case with the baseline model, in what is reported below, a few results require
illustrative values for some parameters to resolve sign ambiguities. In particular, we refer
to values for: f, t, φ, θ and γ. For representative values, we assume that the “replacement
rate” in the employment-insurance system, f, is 0.33, that the initial tax rates levied on
both unskilled and skilled labour income, t and φ, are both 0.33, and that the factor shares
for skilled and unskilled labour, θ and γ (1-θ), are both 0.33. The only other parameter
value that is needed for illustration is the initial unemployment rate, which we take as
0.06. Note that – taken together – the assumptions concerning t, f and u (along with
equation (3) and the q = 0 initial condition) pin down a value for α (equal to 0.03). We
are now in a position to examine the model’s policy implications.
4. Financing Redistribution
Before considering BI and the employment subsidy, we focus on alternative financing
options. Should redistribution be financed by taxing the earnings of skilled labour or by
taxing the income derived from owning physical capital? This question can be addressed
by considering a separate revenue-neutral tax substitution. Hence, in this section of the
paper, we examine the effects of offering a tax cut for skilled labour that is financed by a
tax on capital.
Before addressing this question directly, we focus on a special case of this perfectcapital-mobility model. Suppose that there is only one skill level for labour, that
unemployment is removed from the system, and that the supply of (unskilled) labour is
an exogenous constant. In this setting, a cut in the tax on wage income that is financed by
imposing a tax on physical capital must lower living standards for labour. The fall in the
tax rate applied to wages is dominated by the fall in the pre-tax wage. Labour is less
productive, since the higher tax on capital’s earnings drives capital out of the country.
This tax is distortionary (since capital is not fixed in supply), while the wage tax is not, so
labour is worse off when the government relies less on the non-distortionary tax. The
reason we note this standard result is so readers can appreciate that – in its simplest form
– our perfect-capital-mobility model supports Lucas’ (1990) conclusion that capital is a
bad thing to tax. In more policy-oriented language, this special case of our model
supports “trickle down” – the way for the government to help labour is to limit direct tax
relief to the “rich.”
But let us now move beyond this standard setting in two ways. First, when there is an
asymmetric information problem and market failure in the labour market (and only one
other factor – capital), the optimal tax on capital is no longer zero. Even though this tax
distorts, it permits a lower tax to be paid by labour. This decreases the difference between
the net wages of the employed and the income received by those out of work, and this
reduces unemployment. This conclusion – that it may make sense to tax internationally
mobile capital – is similar to the one derived by Koskela and Schob (2000) in the context
of optimal factor income taxation. They note that, in the presence of involuntary
13
unemployment, labour supply is locally infinitely elastic. Thus, the inverse elasticity rule
suggests that labour should not be taxed at a higher rate than capital (whose supply is also
infinitely elastic at the world rate of interest). Moreover, the presence of unemployment
due to the wage rate being higher than the competitive one implies that the private
marginal cost of labour is higher than its social marginal cost. Thus, welfare can be
increased by subsidizing the labour input relative to the capital input (whose social
marginal cost equals the world interest rate).
Now we return to our full perfect-capital-mobility model which has three factors –
unskilled labour (fixed supply), physical capital (perfectly elastic supply) and skilled
labour (perfectly elastic supply). We consider a similar revenue-neutral introduction of a
tax on capital in this broader setting – but in this case with the revenue used to cut the tax
on skilled labour – not the tax that is paid by the unskilled. Since both skilled labour and
capital are mobile internationally, the taxes on both these factors distort. But the
distortion concerning skilled labour is more important, since we have specified that this
factor is absolutely required for production. Firms have an option regarding capital; they
can use unskilled labour to substitute for this input. So in this setting, skilled labour is the
“worst thing to tax,” and this tax substitution (which reduces the government’s reliance
on this worst tax) raises the wages received by unskilled labour. This is still an example
of “trickle down” – the provision of tax relief for the skilled generates an indirect benefit
down the economic ladder for the unskilled. But in this setting, this benefit is created by
imposing a tax on capital – not by removing such a tax. As long as the unskilled workers
“need” skilled workers more than they need physical capital to generate a demand for
their own services, this version of trickle down is the applicable one.5
This result should be encouraging for those who are concerned that it may be difficult for
a small country to perform income redistribution – on its own – in a world that is
becoming increasingly “global.” The standard challenge is – how do we acquire the tax
revenue that is necessary for redistribution if everyone that we might tax is perfectly
mobile and can escape taxation by relocating their factor services to the rest of the world?
Some, for example, Helliwell (2000), have argued the world is still a long way from
being this global. Nevertheless, since it must be admitted that we are moving ever closer
to a world in which it may be appropriate to assume that both capital and skilled labour
are in perfectly elastic supply, it is useful to have models which allow us to explore the
fundamental challenge for redistribution that results. Our analysis shows that this
challenge can be met. As just reported, the unskilled can be helped without reducing the
incomes of either skilled workers or capital owners.
It must be admitted that our model does not address the concern that fiscal competition
among countries in a world of mobile capital might eliminate the tax on capital as an
option (see, for example, Edwards and Keen (1996) and Sinn (1994)). A similar point –
applied to mobile skilled labour – is made by Wildasin (1991). But Kessler, Lulfesmann
and Myers (2000) have considered mobile capital and labour together. In this setting,
fiscal competition is lessened. When redistribution is pursued in one country, the
immigration of labour raises the tax base and decreases the incentive to attract capital.
Kessler, Lulfesmann and Myers identify circumstances in which increased redistribution
14
in one country makes the majority of the population in both countries (in their two-county
model) strictly better off. We conclude that fiscal competition may not undermine the
applicability of analyses such as ours after all.
We are now in a position to evaluate the provision of either BI or an employment
subsidy, since we now appreciate that – as long as other social programs such as the
unemployment insurance system are left in place and the government relies on income
tax revenue – the appropriate instrument for financing either initiative in this setting is a
tax on capital. To our knowledge, none of the existing studies of a social wage or
employment subsidies have focused on the challenge facing a small open economy – that
the tax base can shrink via international migration making the “affordability” of these
initiatives a serious concern.
5. The Provision of Basic Income
For the proponents of BI, one of its appealing features is that – since it is not conditional
– it involves no direct incentive effects (unlike an unemployment insurance system). In
addition, since the tax on capital is being used to finance the social wage in this analysis,
and since this tax rate does not enter the reduced form for the unemployment rate
(equation (8)), this initiative has no effect on the unemployment rate. But the wage paid
to the unskilled is affected. Since the tax on capital forces some capital to leave the
country, each unskilled worker has less to work with. The resulting drop in the marginal
product of the unskilled leads to a reduction in their wage (by an amount equal to
Ω /((1 − u )(1 − α)) times each percentage point that BI is raised), where parameter Ω
equals 1 + θ + [θ /(1 − φ)][1 + ((γfu 0 /(1 − u )) + φ − γt]. Nevertheless, despite the fall in their
wages, the unskilled can still be helped by the introduction of BI. After all, it provides all
individuals with an additional source of income. The conditions that must be satisfied for
overall income (including the BI) to increase are:
for the employed: ((1 − t )Ω ) /((1 − u )(1 − α)) < 1,
for the unemployed: ( fΩ ) /((1 − u )(1 − α)) < 1,
for the average unskilled individual: (ψΩ) /((1 − u )(1 − α)) < 1.
The reader can readily verify that for our baseline parameter values, only the second of
these conditions can be satisfied.6 This fact means that the BI lowers the income of the
working poor, but it raises the income of the unemployed poor – just as we found in the
baseline model of section 2. Because, it is much more likely that an unskilled individual
is a member of the working poor, it is perhaps not surprising that this initiative lowers the
expected income of the average unskilled individual. Nevertheless, this outcome is
different from what we found in section 2 – that BI increases the expected income of the
average unskilled individual. Thus, BI receives less support in this perfect-capitalmobility setting, and when it is financed by taxing the "rich." In this limited sense, then,
our analysis provides some support for those who fear that globalization makes
redistribution more difficult.
15
Several reactions to this set of results are possible. Some might argue that we should
focus on the implications for the least well off. Those with such a Rawlsian perspective
will conclude that the BI proposal is supported by the analysis. A second reaction is to
focus on the expected income of an unskilled individual – that is, on variable x in the
model. As just noted, no reasonable calibration can support the conclusion that this
measure increases with the provision of BI. As a result, we conclude that – in this
expected sense (that is, given the probabilities that each individual faces concerning
whether she will be employed or unemployed) – the unskilled are hurt by this initiative.
Yet another way to interpret the effect on x is to consider the standard criterion used in
applied economic policy – the hypothetical compensation principle. The fact that this
policy lowers x means that the “winners” (the unemployed) cannot compensate the
“losers” (the employed) and still have something leftover (if a non-distortionary
mechanism to effect this transfer were available). A final way to interpret this outcome is
to assume that all unskilled individuals flow in and out of the unemployment pool (so that
u and (1-u) represent the proportions of each year that each individual spends
unemployed and on the job). In this interpretation, all the unskilled are hurt by this
policy.
To put this rather limited support for BI in perspective, it is useful to consider at least one
other policy (that governments could adopt instead) as a base for comparison. As noted
above, to provide this perspective, we consider the introduction of employment subsidies
to firms for hiring unskilled labour.
It is clear from equation (8) that employment subsidies decrease the natural
unemployment rate. For a subsidy that is calibrated as a percentage of the wage, the
unemployment rate is reduced by an amount equal to u times the number of percentage
points of subsidy. (As an example, if the initial unemployment rate is 6% and a 10%
subsidy is introduced, the unemployment rate falls by six-tenths of one percentage point.)
While the employment effect of this policy is definite, the implications for the wage rate
are less clear. While the direct effect of the subsidy is to stimulate the demand for
unskilled labour (as in Figure 1), there is a competing indirect effect here. The subsidy
must be financed, and in this case, the resulting increase in the tax on capital makes
capital more expensive and this limits firms’ ability to pay higher wages. We have
derived that the wage rate must change by a percentage amount equal to
[1 /(1 − α)(1 − u )][(1 − u )(t − α) − (u 2 f /(1 − u ))] times each percentage point increase in the
subsidy that is offered for hiring unskilled labour. Given the representative parameter
values noted above, this expression is clearly positive, so there is no serious chance that
the lower capital stock effect can dominate. Indeed, this expression – when combined
with the representative parameter values – implies that unskilled wages rise by 3 percent
when a 10% employment subsidy is introduced. Since both the working and non-working
poor are helped by this policy, and since the proportion of the poor that are unemployed
falls, we conclude that the analysis supports this initiative. Further, since there exists this
policy alternative to the BI that receives more clear-cut support, the provision of a
guaranteed annual income is called into question in this perfect-capital-mobility setting
16
(although, as pointed out above, a Rawlsian social welfare function would still view BI as
a superior option).
As explained above, in constructing this three-sector model, we have been heavily
influenced by the challenge posed by globalization – does a small open economy have
sufficient degrees of freedom to perform meaningful income redistribution? We are
drawn to this focus because, over time, the assumption of perfect mobility for both capital
and skilled labour may become ever more relevant. Since no studies of these initiatives
have focused on the small open-economy constraints, we feel it is important to fill at least
part of this gap. But some readers may argue that – for more immediate relevance – the
analysis should not yet assume perfect mobility for skilled labour. To provide balance,
therefore, we now briefly consider the case of a completely inelastic supply of skilled
labour. This assumption implies that we need no longer impose the constraint that the
after-tax income of skilled labour must be equal to that prevailing in the rest of the world.
As far as the model is concerned we note an important implication of relaxing this
constraint. Output does not respond to any of the policies considered above, since (given
equation (12)) the fixed supply of skilled workers pegs output. This fact implies that any
policy that results in a decrease in the amount of physical capital used in the domestic
economy results in an increase in the amount of unskilled labour (in efficiency units)
employed. Firms simply shift along a given isoquant.
We can see this by considering the introduction of BI (financed by a rise in the tax on
capital). Since we are still assuming perfect capital mobility, capital owners react to the
higher tax by withdrawing capital from the domestic economy. With the fixed output
constraint, this development forces firms to hire more efficiency units of labour. With a
fixed quantity of unskilled individuals employed as well, the only option firms have is to
induce a higher level of efficiency – by offering higher wages. The outcome is that
incomes for both employed and unemployed unskilled workers rise, so they are all better
off – even without adding in the BI – since the unemployment rate is unchanged.
However, there are political economy concerns (as before), since the incomes of skilled
workers fall. These individuals also receive the BI, but the share of national income going
to others (unskilled workers) is higher, and this development lowers the market earnings
of skilled individuals. We conclude that there is no stronger case for a BI after all. The
working poor are not adversely affected in this case, but skilled workers are. In each case,
therefore, we could expect limited support for the initiative.
The final sensitivity test we consider in this section of the paper concerns the nature of
the production process and the rationale for unemployment. We have replaced the
specification of production that assigns a special role for skilled labour with a standard
Cobb-Douglas production function:
Y = Lλ S λ K λ .
1
2
3
Also, we have replaced the efficiency-wage model of the unskilled labour market with
the model of unions described in section 2. Three comments are warranted regarding the
introduction of this model of union-firm interaction into the three-factor setting. First, for
17
unions to have any power, they must be able to control the available supply of labour to
the firm. As a result, the assumption that skilled labour is perfectly mobile internationally
means that we must restrict unions to the unskilled labour market. Second, for there to be
rents for unions and firms to bargain over in a perfectly competitive product market, we
must assume decreasing returns to scale (λ1 + λ2 + λ3 < 1) . Third, the unemployment
rate is still given by equation (8), but now α = [ε(1 − λ1 − λ2 − λ3 )] /[λ1 (1 − ε)], where (as
above) ε is unskilled labour's bargaining power.
It is left for the reader to verify four things. First, as long as there is no depreciation of
capital and there are equal tax rates initially, the fact that skilled labour no longer plays a
unique role in the production process means that there is no longer any reason to prefer
taxing capital – as opposed to taxing skilled labour – when revenue is needed to finance
either BI or an employment subsidy. Second, the effect of both policy initiatives on the
unemployment rate (du/dp = 0 and du/dq < 0) are exactly as reported above. Third, for
similar calibrations (for example, equal factor shares – now λ1 = λ2 = λ3 = 0.30 ) the sign
and size of the effect of the employment subsidy on the wages of the unskilled,
(dw/w)/dq, is unchanged. Finally, the signs of, (dz/z)/dp, (dn/n)/dp and (dx/x)/dp are all
positive. This means that the introduction of BI raises both the income of the working
poor, and the income of the unemployed poor, so the average unskilled individual is
helped, not hurt, by BI in this specification. Only the unemployed individuals were
helped in the core specification.
The wages of the unskilled still fall with the introduction of BI. In our core specification,
this fall in w dominated the rise in p for all unskilled individuals except the unemployed
(for whom BI represents a much bigger portion of income). Recall that the wage falls
since the higher tax on capital (introduced to finance BI) forces both some capital, and
therefore some skilled labour, to leave the country, and this makes the unskilled less
productive. In this new specification, firms have a production function that allows them
to cope more easily with the loss of K and S. Since it is so much easier for firms to
substitute all factors for each other in this new setting, firms are better able to make do
with unskilled labour as a replacement (for the lost skilled labour in particular). The
result is that, while w still falls, it falls much less than in our core specification. This is
why, in this case, the introduction of BI can more than make up for this drop in market
earnings for all groups of unskilled (unemployed and employed). Despite this difference
in results, it seems that our overall conclusion remains intact. With the core specification,
we had concluded that (unless one adopts the Rawlsian view) the employment subsidy
dominates BI. This conclusion is still warranted, since the support for the employment
subsidy is independent of the uncertainty that we have concerning model specification,
while the support for BI hinges centrally on the role that skilled labour plays in the
production process. In short, our analysis has raised doubts about BI that do not emerge
for employment subsidies, and to be "safe" on this score, it would seem that a prudent
policy maker should favour the implementation of employment subsidies.
18
6. Saving and Foreign Indebtedness
Thus far, we have assumed that the workers in the perfect-capital-mobility model differ
in just two ways – one group is less skilled and these individuals dislike their work. The
latter assumption leads to the possibility of low productivity and unemployment among
this group. In this section of the paper, we introduce a third difference between the two
groups. For the skilled, we assume a rate of time preference that is equal to the fixed
world rate of interest, while for the unskilled we assume a rate of impatience that is so
high that these individuals never save. As a result, the unskilled own no physical capital;
all the returns to this factor go either to skilled individuals residing in the domestic
economy or to foreigners. The revised model is an open-economy version of what
Mankiw (2000) has recommended for fiscal policy analysis – a system in which one
group of individuals is “Ricardian” and the other lives “hand-to-mouth.”
With the consumption-savings choice for the “rich” now modeled, we can levy an
expenditure tax – as opposed to an income tax – on this group. Since we restrict this levy
to the rich, it amounts to a progressive expenditure tax – the method of raising revenue
that is often regarded as the one which least sacrifices economic efficiency when
increased equity is being pursued (see, for example, Frank and Cook, 1995). In this
section of the paper, therefore, we introduce this progressive expenditure tax to finance
the BI, and no tax on the earnings of capital is introduced.
The following equations are added to the model:
(1 + σ)C = r (K − A + H )
(16)
H = [v (1 − φ) + pw]S / r
(17)
rA = Y − δK − G − C − E
(18)
E = w((1 − t )(1 − u ) + uf + p)
(19)
and the new notation is defined as: C – total consumption spending by skilled individuals,
E – total consumption expenditures by the unskilled, H – human wealth of the skilled
workers, A – that part of the domestically employed capital stock that is foreign owned,
and σ – the expenditure tax paid by the rich. The rationale for each new relationship is
now briefly explained.
Equation (16) is the standard (aggregate) consumption function that follows from
intertemporal optimization if the instantaneous utility function is the log of consumption
and the rate of time preference is r (see Ramsey (1928) and Blanchard and Fischer
(1989)). Consumption expenditures are proportional to wealth; (K – A) is non-human
wealth; and human wealth is defined as the present value of all future after-tax wage and
basic income (in equation (17)). Since some of the capital stock can be foreign-owned,
full equilibrium requires that the interest payments on that foreign debt be covered by net
export earnings. The steady-state version of this financing requirement is stipulated by
19
equation (18). Finally, expenditures on the part of the unskilled simply equal the total
income of these individuals, and this is defined in equation (19).
There are two additional changes. First, with the progressive expenditure tax, there is
now a σC term on the right-hand side of the government budget constraint (equation
(15)). Second, we now define indifference on the part of skilled workers – between
working in the domestic economy and living in the rest of the world – as existing when
per-capita steady-state consumption is the same in both locations. For simplicity, we have
ignored the transitional effects of domestic government policy on the consumption of
skilled individuals. This would seem to be an acceptable simplification – given that these
individuals make decisions on the basis of their being part of an infinitely lived dynasty.
In any event, this assumption is imposed by making (C/S) an exogenous constant, and
this stipulation now replaces equation (13). The revised model consists of 14
relationships (equations (2), (3), (8)-(12), (14)-(15) and (16)-(19)) with the new term in
(15), and the d(C/S) = 0 restriction). The endogenous variables are the same 10 as defined
in section 2 (except that σ now replaces τ) plus C, H, A and E. For simplicity, we assume
that both σ and A are zero initially.
Some stark results follow from this specification. The introduction of BI has no incentive
effects, and since there is no tax on capital, there is no effect on the capital stock, or on
the level of pre-tax wages for the unskilled. The introduction of the progressive
expenditure tax drives some of the skilled individuals out of the country. However, this
levy also leads to increased saving during the transition to the new steady state on the part
of those that remain living in the domestic economy, and so to a higher level of nonhuman wealth. By itself, this outcome would result in per-capita consumption among the
skilled being higher. Since it is assumed that international mobility precludes this, the
domestic wage paid to skilled workers must (and does) fall. There are competing effects
on the demand for the unskilled. The fact that there are fewer skilled workers in the new
steady state means that demand for the unskilled falls. The fact that firms pay less for
each skilled worker means that the ability to pay wages to the unskilled rises. The reader
can verify that these two effects must exactly cancel off. Thus, the percentage change in
per-capita consumption among the unskilled equals (1/ψ) times the level of BI
introduced. With our illustrative parameter values, if a BI equal to 5% of the unskilled
wage is introduced (and financed by the introduction of the progressive expenditure tax),
the standard of living of the average unskilled individual rises by 7.5%. Further, since
there is no effect on the unemployment rate, we need have no concern about any trade-off
between the implications of the initiative for the working poor, as opposed to the nonworking poor. We conclude that much stronger support for a policy of Basic Income is
found when the analysis is extended to allow for endogenous saving on the part of the
non-hand-to-mouth portion of the population.
There is mixed support for the employment subsidy in this setting. This initiative must
reduce the unemployment rate (as in earlier models) so on this account, this policy is
preferred to BI. Nevertheless, the subsidy for hiring unskilled labour results in firms
choosing to hire less capital. Relative to the outcome with BI, this reduces the
productivity of unskilled labour and so lessens firm's willingness to pay higher wages to
20
these individuals. Indeed, for plausible parameter assumptions, the standard of living for
an average unskilled individual (variable E) falls with the introduction of the employment
subsidy. For this reason, this initiative is inferior to BI in this long-run setting.
Most analyses of labour market institutions and policies (for example, Pissarides, 1998)
abstract from household saving and wealth accumulation. The analysis of this section –
which has combined some of the insights of labour economists and those specializing in
public finance – suggests that it is important that the integration of these different
perspectives proceed. Of all the sensitivity tests pursued in this paper, this one appears to
be one of the more important.
7. Conclusions
Our analysis does not allay all the fears of those who think that globalization may weaken
our ability to effect redistribution in a small economy. Nor does it offer a clear-cut verdict
concerning the provision of basic income. In some cases, our analysis identifies a version
of "trickle-down" economics that helps the unskilled – despite the constraints implied by
the international mobility of factors other than unskilled labour. Nevertheless, the support
for this initiative varies across specifications. We find similar ambiguity concerning the
introduction of BI. If saving and the associated changes in foreign indebtedness are
ignored, and skilled labour is assigned a crucial role in the production process, we find
that this policy harms the employed unskilled while benefiting the unemployed (when
both physical capital and skilled labour are mobile). Yet, when only physical capital is
internationally mobile, the introduction of a guaranteed annual income results in higher
incomes for both employed and unemployed unskilled workers. Nevertheless, in this
case, it is the skilled workers that lose. When saving on the part of the "rich" is made
endogenous, however, there is solid support for the introduction of BI. The results are
opposite for employment subsidies. When saving is ignored, the analysis supports
employment subsidies, but when long-run wealth accumulation is highlighted,
employment subsidies lower the average incomes of unskilled individuals.
Of course, what we have just summarized are only the results of some simple but
conventional models. It has not been possible to incorporate within the analysis all
aspects of the debate concerning BI and other policies that analysts have advocated as a
substitute for universal basic income, such as employment subsidies. For example,
proponents of the BI stress issues such as its beneficial effects on the level of real
freedom for disadvantaged groups, while some opponents stress self-sufficiency and the
value people derive from making a contribution through employment. These wider
philosophical issues are well summarized in the debate between Phelps (2000) and Van
Parijs (2000) in the Boston Review. Some of the wider issues can be included in an
extended version of our analysis. One relates to the encouragement of work sharing, and
a second to reductions in administrative costs. (The BI would do away with complicated
means-tested benefits.) Another issue stressed by proponents of the BI is that – when it
replaces unemployment insurance – it offers incentives for skill acquisition. We have
verified that – in our perfect-capital-mobility model – a reduction in unskilled labour
supply results in higher incomes for the unskilled. Also in the appendix of this paper, we
21
have included a preliminary analysis of the effect of a BI on the aggregate participation
rate, and we have concluded that it is likely to be quite small. Nevertheless, we plan to
pursue this question more fully, and to include work-sharing effects, in future work. Also,
in the version of our model which highlights wealth accumulation, we can consider
Blanchard's (1985) and Nielsen's (1994) extensions to the infinitely lived representative
agent analysis (that involves overlapping generations, retirement and a public pension).
In that setting, we can examine BI financed by a reduction in the generosity of either (or
both of) the unemployment-insurance or the public-pension system (as envisioned by
some proponents of BI). In the meantime, it is hoped that the present paper has clarified
some of the macroeconomic implications of basic income, and the major alternative that
is under debate – employment subsidies.
Appendix: Basic Income and Labour-Force Participation
In the macro models analyzed in the text, it is assumed that labour force participation is
independent of the provision of BI. To assess this simplification, we now briefly review
the empirical literature concerning the impact of income taxation on work incentives. In
addition, we present a stylized model of family labour supply, and use it to examine the
implications of introducing a guaranteed annual income.
The empirical literature on labour supply has identified two margins in which labour
supply can respond. First, there is the response along the intensive margin. That is,
individuals can vary their hours or effort intensity on the job. Second, individuals may
respond along the extensive margin; that is, they decide whether or not to enter the labour
force. In the case of response along the intensive margin, if leisure is a normal good, the
income and substitution effects of tax changes work in opposite directions. The empirical
literature has been at pains to establish the size of the net effect. Burtless (1986) in his
summary of twenty-six estimates of male labour supply, surveyed by Killingsworth
(1983), finds an average total elasticity close to –0.10. Taking into account the labour
supply decisions of married women, (see Mroz (1987), Blundell (1992)) the picture does
not change by much; the evidence for both the United Kingdom and the United States
suggests a labour supply elasticity closer to 0 than to 1.
With respect to the extensive margin, the Negative Income Tax (NIT) experiments and
Earned Income Tax Credit (EITC) policies in the United States have provided most of the
evidence. In the case of NIT, the income and substitution effects operate in the same
direction, both tending to reduce labour supply. Based on data from the New Jersey
experiments (in which eight NIT plans were examined), Burtless (1986) concluded that
the reductions in labour supply occurred mostly in the form of withdrawals from
employment or active labour force participation rather than marginal reductions in
weekly hours of work, with the effect on women’s participation being twice as large as
that of men. This has led economists and politicians to advocate programs that would
22
make work sufficiently attractive to reduce the need for income support. The EITC does
not provide any income support for families with no earnings, but all earnings below a
given threshold are partially matched by the government. It is usually thought that,
relative to a NIT program, incentives to work are enhanced with an EITC because the
implicit tax rate inherent in the latter program is smaller.
It should be noted that this view is not universal, since it is based on models of individual
(rather than family) labour supply considerations. Some authors (for example, Browning
(1995) and Eissa and Hoynes (1998)) have questioned the incentive structure of the
EITC. A major concern in this respect is that the EITC is effectively subsidizing married
mothers to stay at home (while it has only a small positive effect on married men’s labour
supply). Eissa and Hoynes (1998) suggest that a possible reform of the EITC so that it is
based on individual earnings (as opposed to family earnings) would offset the incentives
for secondary earners to leave the labour force. Since the BI proposal provides support to
working age individuals (rather than families), it may be less prone to causing reductions
in participation rates. We now consider a simple model of family labour supply to assess
this possibility. The notation used here is defined independently of that in the text.
The model is based on Mincer (1962) and Newark and Postelwaite (1998). We assume
that there is a continuum of couples, each consisting of a man and a woman, that may
possess different abilities, and which are randomly matched. Each woman has an ability
level denoted by a, a ∈ [0, A] and each man has an ability level denoted by b, b ∈ [0, B ].
We assume that there is a wage rate, w, per unit of ability, and that all men work. In order
to capture the idea that there is an opportunity cost if both household members work, we
assume that there is lost household production if the woman works outside the home. The
value of this lost production, v, is assumed to depend only on the husband’s income:
v = v( wb ), v ' > 0.
A couple whose abilities are represented by ( a, b ) will have utility (a + b) ⋅ w if the
woman works and bw + v(bw ) if she does not. A consequence of this form of opportunity
cost is that, among the women that are matched with men of the same ability level b , it
will be only women with higher ability that will work, since women work only if
a > v( b.w ) / w. If we fix the ability of a man at level b , a woman with ability a * who is
married to this man, will be indifferent between employment and non-employment if
a* = v (bw ) / w . Women of higher ability which are married to a man of the same ability
will strictly prefer employment, whereas women with a < a * , will prefer to abstain from
market work. In other words, the marginal woman attached with a man of ability b has
ability a* = v (bw) / w.
Differentiating this relationship with respect to the wage rate, we find
da * / dw = (v ′b − a*) / w. This result implies that, if v '.b < a * , increases in the wage rate
decrease the critical skill level of women, thereby ensuring that women’s participation in
the labour market increases. In what follows we assume that this condition holds.
23
Consider now the impact of introducing BI and wage-income taxes on women’s decision
to participate in the labour force. If both household members work, then household utility
is (a + b) ⋅ w(1 − t ) + 2 pw , where t stands for the income tax rate and pw is the (tax-free)
value of basic income. If only the husband works, then household utility is equal to
bw (1 − t ) + 2 pw + v((1 − t ) ⋅ w.b ) . Thus, among women who are married to men with
ability b , there is one with ability a* such that a * ⋅w ⋅ (1 − t ) = v((1 − t ) ⋅ w ⋅ b ) who will
be indifferent between market sector employment and household production. We note
that basic income does not directly influence the decision to participate in paid
employment. However, it may do so indirectly, since the payment of BI may be financed
by an increase in t. Differentiating the last equation with respect to the tax rate we find
that da * / dt = (a * − v′b ) /(1 − t ) > 0 , assuming da * / dw < 0. We conclude that, as long as
BI is financed by some other levy – such as a tax on the earnings of capital (as in the text
of this paper) – labour force participation is unaffected by BI. However, if it is financed
by a tax on labour, there can be a decrease in participation.
Even when financed by a labour tax, the negative effect of BI on the participation rate
can be partly remedied if, as Atkinson (1995) has proposed, the payment is conditional on
participation. Atkinson defines participation not only as paid work, but also as any form
of social contribution (such as caring for the elderly or disabled dependants, and
undertaking approved forms of voluntary work). Leaving aside problems of
administering such a system, we proceed to analyze its implications for the household’s
choice. If the woman participates in paid employment, household utility will be equal to
(a + b) ⋅ w ⋅ (1 − t ) + 2 pw , whereas if the woman does not participate utility will be equal
to bw(1 − t ) + pw + v((1 − t )wb ) . Indifference exists if a * w(1 − t ) + pw = v((1 − t ) wb ) .
This equation implies da * / dp = (1 /(1 − t ))[(a * −v ′b)(dt / dp) − 1], which is negative if
dt / dp = 0, which is true if BI is not financed by the imposition of a tax on labour. In this
case, then, BI raises labour force participation. Of course, if a labour tax is involved in
financing the BI, dt / dp > 0, and this makes the sign of da * / dp ambiguous. Overall,
these ambiguities suggest that a zero effect (as assumed in the text) may be reasonable.
24
References
Akerlof, G. and J. Yellen, (eds.), 1986, Efficiency Wage Models of the Labor Market,
(Cambridge: Cambridge University Press).
Atkinson, A., 1995, Public Economics in Action: The Basic Income/Flat Tax Proposal
(Oxford: Oxford University Press).
Atkinson, A., 1999, The Economic Consequences of Rolling Back the Welfare State,
(Cambridge, Mass: MIT Press).
Atkinson, A., and H. Sutherland, 1990, “Scaling the ‘Poverty Mountain’: Methods to
Extend Incentives to All Workers,” in A. Bowen and K. Mayhew (eds.), Improving
Incentives for the Low Paid, (London: Macmillan).
Bhagwati. J. and V. Ramaswami, 1963, “Domestic Distortions, Tariffs, and the Theory of
Optimal Subsidy,” Journal of Political Economy 71.
Blanchard, O., 1985, “Debts, Deficits and Finite Horizons,” Journal of Political Economy
93, 223-247.
Blanchard, O. and S. Fischer, 1989, Lectures on Macroeconomics, (Cambridge, Mass:
MIT Press).
Blundell, R. W., 1992, “ Labour Supply and Taxation: A Survey,” Fiscal Studies 13, 1540.
Blundell, R. W., and T. MaCurdy, 1999, “Labour Supply: A Review of Alternative
Approaches,” in O. Ashenfelter and D. Card (eds.), Handbook of Labour Economics,
Volume 3A, (Amsterdam: North-Holland).
Browning, E., 1995, “ Effects of the Earned Income Tax Credit on Income and Welfare,”
National Tax Journal 48, 23-43.
Burtless, G., 1986, “The Work Response to a Guaranteed Income: A Survey of
Experimental Evidence,” in A. H. Munnell (ed.), Lessons from the Income Maintenance
Experiments, (Boston: Federal Reserve Bank of Boston, Conference Series No. 30).
Edwards, J. and M. Keen, 1996, “Tax Competition and Leviathan,” European Economic
Review 40, 113-134.
Eissa, N., and H.W. Hoynes, 1998, “ The Earned Income Tax Credit and the Labour
Supply of Married Couples,” NBER Working Paper No. 6856.
Esping-Andersen, G., 1990, The Three Worlds of Welfare Capitalism, (Cambridge, UK:
Polity Press).
25
Frank, R. and P. Cook, 1995, The Winner-Take-All Society (New York: Free Press).
Freeman, R., 1995, “Are your Wages Set in Beijing?” Journal of Economic Perspectives
9, 15-32.
Goldin, C., and L. F. Katz, 1998, “The Origins of Technology-Skill Complementarity,”
Quarterly Journal of Economics, 118, 693-732.
Gordon, D. N., 1996, Fat and Mean: The Corporate Squeeze of Working Americans and
the Myth of Managerial “Downsizing” (New York: The Free Press).
Heady, C., T. Mitrakos, and P. Tsakloglou, 2001, “The Distributional Impact of Social
Transfers in the European Union: Evidence from the ECHP,” Fiscal Studies 68, 547-565.
Helliwell, J., 2000, Globalization: Myths, Facts, and Consequences (Toronto: C.D. Howe
Institute).
Kessler, , A., C. Lulfesmann and G. Myers, 2000, “Redistribution, Fiscal Competition
and the Politics of Economic Integration,” Simon Fraser University, Department of
Economics, Discussion Paper 00-11.
Killingsworth, M. R., 1983, Labour Supply, (Cambridge: Cambridge University Press).
Koskela, E., and R. Schob, 2000, “Optimal Factor Income Taxation in the Presence of
Unemployment,” CESifo Working Paper No. 279.
Krusell, P., L. Ohanian, J.V. Rios-Rull, and G. Violante, 2000, “Capital-Skill
Complementarity and Inequality: A Macroeconomic Analysis,” Econometrica 68, 10291045.
Lazear, E., 1995, Personnel Economics, (Cambridge, Mass: MIT Press).
Lucas, R., 1990, “Supply-Side Economics: An Analytical Review,” Oxford Economic
Papers 42, 293-316.
Machin, S. and J. Van Reenen, 1998, “Technology and Changes in Skill Structure:
Evidence from Seven OECD Countries,” Quarterly Journal of Economics 113, 12151244.
Mankiw, G., 2000, “The Savers-Spenders Theory of Fiscal Policy”, American Economic
Review Papers and Proceedings 90, 120-125.
Manning, A., 1995, “How Do We Know that Real Wages are Too High?” Quarterly
Journal of Economics 110, 1111-1126.
26
Meade, J., 1948, Planning and the Price Mechanism (London: Allen and Unwin).
Meade, J., 1972, “Poverty in the Welfare State,” Oxford Economic Papers 24, 289-326.
Mincer, J., 1962, “Labour Force Participation of Married Women: A Study of Labour
Supply,” in Gregg, H. and Lewis, H.G. (eds.), Aspects of Labour Economics, (Princeton,
Princeton University Press).
Mroz, T., 1987, “The Sensitivity of an Empirical Model of Married Women’s Hours of
Work to Economic and Statistical Assumptions,” Econometrica 55, 765-799.
Nielsen, S., 1994, "Social Security and Foreign Indebtedness in a Small Open Economy,"
Open Economies Review 5, 47-63.
Newark, D., and A. Postlewaite, 1998, “Relative Income Concerns and the Rise in
Married Women’s Employment,” Journal of Public Economics 70, 157-183.
Phelps, E., 1997, Rewarding Work (Cambridge, Mass.: Harvard University Press).
Phelps, E., 2000, “Subsidize Wages,” Boston Review 25, 5.
Pissarides, C., 1998, “The Impact of Employment Tax Cuts on Unemployment and
Wages: The Role of Unemployment Benefits and Tax Structure,” European Economic
Review 42, 155-183.
Ramsey, F., 1928, “A Mathematical Theory of Saving,” Economic Journal 38, 543-559.
Rebitzer, J. and L. Taylor, 1995, “The Consequences of Minimum Wage Laws: Some
New Theoretical Ideas,” Journal of Public Economics 56, 245-256.
Romer, D., 1996, Advanced Macroeconomics (New York: McGraw-Hill).
Sinn, H.-W., 1994, “How Much Europe? Subsidiarity, Centralization, and Fiscal
Competition,” Scottish Journal of Political Economy 41, 85-107.
Sinn, H.-W., 1995, “A Theory of the Welfare State,” Scandinavian Journal of Economics
97, 495-526.
Solow, R., 1998, Work and Welfare (Princeton, N.J.: Princeton University Press).
Summers, L., 1988, “Relative Wages, Efficiency Wages and Keynesian Unemployment,”
American Economic Review, Papers and Proceedings 78, 383-388.
Tobin, J., J. Pechman and P. Mieszkowski, 1967, “Is a Negative Income Tax Practical?”
Yale Law Journal 77, 1-27.
27
Van Parijs, P., 1992, Arguing for Basic Income: Ethical Foundations for a Radical
Reform, (London : Verso).
Van Parijs, P., 1995, Real Freedom for All: What (If Anything) Can Justify Capitalism?
(New York: Oxford University Press).
Van Parijs, P., 2000, “A Basic Income for All,” Boston Review 25, 5.
Wildasin, D., 1991, “Income Redistribution in a Common Labor Market,” American
Economic Review 81, 757-774.
Wood, A., 1994, North-South Trade, Employment and Inequality (Oxford, UK:
Clarendon Press).
28
Endnotes
1
Some people have pointed to alternative explanations which do not involve a shift in relative demand in
favour of skilled labour. For example, Gordon (1996) has argued that the weakening of labour market
institutions and the erosion of the real value of the minimum wage are responsible for the increased
inequality in the United States.
2
Following Esping-Andersen (1990) we can identify four “models of welfare capitalism” in the EU: the
Scandinavian model of universal social protection as a right of citizenship; the ‘Bismarckian’ employmentbased model of Germany, Austria, France and the Benelux countries; the Anglo-Saxon model of the United
Kingdom and Ireland; and the fragmented and highly idiosyncratic arrangements of the remaining southern
EU members. Also, Heady, Mitrakos and Tsakloglou (2001) document the very large variance of social
transfers in the EU. Social transfers as a percentage of GDP vary between 16.9% (Portugal) to 37.6%
(Sweden). Wide differences also exist in the allocation of such transfers by type of benefit ( for example,
family related benefits account for 0.7% of total social transfers in Spain and for 15.2% in Ireland), by their
impact on inequality ( the proportional decline in the Gini index of inequality due to social tranfers in cash
varies between 46% for Denmark and 22.7% for Portugal), and in poverty (the existence of unemployment
benefits reduces poverty by 66.4% in Denmark and by 1.7% in Greece).
3
Others (such as Manning (1995) and Rebitzer and Taylor (1995)) have shown that – in such a second-best
setting – non-standard results can emerge. For example, in their efficiency-wage models, minimum wage
laws can raise employment. We have verified that this is not possible in Summers’ version of efficiency
wage theory that we rely on in this paper. Indeed, our results are quite standard since the policy that
addresses the market failure at source is the one that is more effective for raising employment.
4
Indeed, we verify this again in section 5 below.
5
There are two senses in which skilled labour can be considered the most important factor – in a technical
sense, and in a cost-share sense. To emphasize intuition and policy relevance, we ignored this distinction in
the previous paragraph. In fact, it is the cost-share that is important, and this tax substitution helps unkilled
labour only if γθ[t − ( fu /(1 − u ))] > φ(θ − φ). The reader can readily verify that – for the
representative parameter values given above –this condition is certainly satisfied (so the intuitive reasoning
given in this paragraph is applicable).
6
We have considered other parameter values, and found that it is possible for all three conditions to be
satisfied, but only for quite extreme factor shares.
29