Download Expectations of future income and real exchange rate

Faculty of Business and Law
School of Accounting, Economics and Finance
Financial Econometrics Series
SWP 2012/05
Expectations of future income and real
exchange rate movements
A. Hayat, B. Ganiev, and X. Tang
The working papers are a series of manuscripts in their draft form. Please do not
quote without obtaining the author’s consent as these works are in their draft form.
The views expressed in this paper are those of the author and not necessarily
endorsed by the School or IBISWorld Pty Ltd.
Expectations of future income and real exchange rate movements
Aziz Hayata,∗,
a Financial
Bahodir Ganievb,
Xueli Tangc
Econometrics Group, School of Accounting, Economics, and Finance, 70 Elgar Road,
Burwood Highway, VIC 3125, Deakin University, Melbourne, Australia.
b Westminster
International University in Tashkent, 12 Istikbal Street, Tashkent, Uzbekistan.
c School
of Accounting, Economics, and Finance, 70 Elgar Road,
Burwood Highway, VIC 3125, Deakin University, Melbourne, Australia.
This version: November 29, 2012
Abstract
We show that the changes in expectations of future income driven by exogenous factors
(such as the discovery of oil, an increase in global demand for natural resources, etc.) can
cause movements in the real exchange rate (RER) in excess of, and sometimes even in
the opposite direction to, what one would expect given the changes in current income.
We provide both a theoretical model and empirical evidence of this. In particular, we
show that the signing of numerous production sharing agreements (PSAs) between the
government of Azerbaijan and foreign oil companies in 1994–99 fuelled expectations of
higher future incomes, resulting in a considerable appreciation of the RER. Some of these
PSAs subsequently failed or ran into difficulties, which led to a downward revision of
expected future income and a depreciation of the RER in 1999–2003, even though the
current income started to rise, due to an increase in the current oil revenue.
JEL classification: C22; P28; E61; F31
Keywords: Expectations of future income; Real exchange rate; Booming sector; Azerbaijan
economy
Corresponding author. Tel.: +61 3 924 46358; fax: +61 3 924 46283.
E-mail addresses:
[email protected]
(A. Hayat), [email protected]
[email protected] (X. Tang)
∗
1
(B. Ganiev),
1. Introduction
Azerbaijan is one of the fifteen countries of the former Soviet Union (FSU) that became
an independent state after the disintegration of the FSU in 1991. It has been undergoing a
transition from a centrally planned to a market-based economy since 1992 and experiencing
an oil boom since 1994, and its economic fundamentals have generally been strengthening
since 1996. All of these factors tended to put upward (appreciation) pressure on the real
exchange rate (RER) of the country’s national currency, the Manat, in the second half of
the 1990s and the early 2000s, yet the RER of the Manat depreciated significantly over
the period 1999–2003.
Some of the researchers working on Azerbaijan’s economy (including those at the IMF)
view the depreciation of the Manat in real terms over the period 1999–2003 as being a
result of the foreign exchange interventions of the National Bank of Azerbaijan (NBA),
the country’s central bank, and its loose monetary policy (IMF, 2004). Accordingly, the
depreciation of the RER led to its deviation from the equilibrium level (that is, to exchange
rate misalignment). The reason for this is that only real variables (known as RER fundamentals) can affect the equilibrium RER (ERER). Monetary policy cannot impact ERER.
Net purchases of foreign exchange by a central bank can cause a depreciation of RER, but
put appreciation pressure on ERER through the accumulation of official reserves.
We argue that the depreciation of the Manat in real terms over the period 1999–2003
was, to a considerable degree, a result of the oil euphoria that occurred in Azerbaijan in
the mid-1990s. The signing of numerous production-sharing agreements (PSAs) between
the government of Azerbaijan and foreign oil companies between 1994 and 1998 led to the
expectation of a higher future oil income and an excessive (after controlling for the effects
of ERER fundamentals on RER) appreciation of RER at that time, even though actual oil
income remained small. Subsequently, some of the PSAs failed or ran into difficulties, which
led to a downward revision of expected future oil income and, thus, to the depreciation
of RER over the period 1999–2003, even though the actual oil income started to increase.
The excessive appreciation from 1994–1998 seems to have been undone by the depreciation
of 1999–2003, as RER started appreciating again in 2004. Similar phenomena appear to
have happened or be happening during boom and doom times in other economies, such as
Australia, the UK and the USA.
2
A more general point we make is that expectations of future income, driven by exogenous factors such as oil euphoria, an increase in the global demand for natural resources,
high levels of external financial assistance during economic crises, etc., can have a strong
influence on RER in a booming/dooming economy. To the best of our knowledge, this
point has been completely overlooked in the literature to date. We provide both a theoretical model and empirical evidence showing that changes in expectations about future
income levels cause movements in RER in excess of, and sometimes even in the opposite
direction to, what one would expect given the changes in current income.
The balance of the paper is organized as follows. In the next section, we provide a
plausible list of the fundamentals of RER for Azerbaijan, taken from other studies which
have been conducted on transition and natural resource boom economies. In Section 3, we
derive and construct Azerbaijan’s RER. We set up the theoretical model for the problem
in Section 4, and discuss the construction of measurements of future income in Azerbaijan
in the following section. We then provide empirical results in Section 6. In Section 7
we discuss the implications of RER depreciation in Azerbaijan’s economy and provide
a plausible explanation of the influence of expectations of future income on the RER of
other major economies during their boom and doom times. Finally, in Section 8 we provide
concluding remarks.
2. Fundamentals of RER
In this section, we summarize the findings of other studies as to the ways in which
being in a transition state, having a resource boom, and the strengthening of economic
fundamentals would affect RER.
In a centrally planned economy, RER is generally set above the equilibrium level. In
other words, RER is overvalued. Therefore, the transition from a centrally planned to a
market-based economy brings about a sharp depreciation of RER (Halpern and Wyplosz,
1997). This sharp initial depreciation of RER is then followed by its gradual appreciation.
One reason for this is that the transition process involves the dismantling of old production
structures and structural reforms, which lead to increases in productivity, especially in
the production of tradable goods. Consequently, real wages and the relative prices of
non-tradable goods rise, inducing an appreciation of RER (Halpern and Wyplosz, 1997;
3
Obstfeld and Rogoff, 1996).1 In many transition economies, another reason for the gradual
appreciation of RER (after its sharp initial depreciation) is high inflation and a less-thancommensurate depreciation of the nominal exchange rate (Orlowski, 1998). Furthermore,
the liberalization of capital accounts, and subsequent capital inflows, can also lead to an
appreciation of RER at later stages of the transition process (Orlowski, 1998; Coricelli and
Jazbec, 2004).
A number of studies demonstrate the way in which economic fundamentals drive movements in RER. For example, the relationship between economic fundamentals and movements in RER is investigated by Paiva (2001) for Costa Rica, Kuralbayeva et al. (2001)
for Kazakhstan, and Beguna (2002) for Latvia. The economic fundamentals used in these
studies that affect movements in RER rate are fiscal stance, net capital flows, the degree of
openness, the terms of trade (TsOT), productivity, and the interest rate differential. Égert
et al. (2006) investigate whether productivity and net foreign assets explain the appreciation of RER in 11 transition economies in the period of the 1990s and the early 2000s.
They find that productivity causes RER to appreciate in five Central and Eastern European economies (CEE-5), and has no effect in Baltic transition economies. Contrary to our
expectations, an increase in net foreign liabilities led to a real appreciation in the Baltic
countries, but to the expected depreciation in the CEE-5 economies. Lim (1992) provides
empirical evidence of the rejection of hypotheses of purchasing power parity (PPP) (as a
theory of the long-run behavior of the real exchange rate), and the uncovered interest parity (as a theory of short-run behavior) for RER between the US and other G-10 countries.
Instead, he finds support for the idea of RER fundamentals (such as productivity, ToT, and
interest rate differentials) explaining the variation in RER. In connection with the PPP
theory for RER (i.e., whether RER can be a stationary process), there is a considerable
body of literature that provides empirical evidence either for or against the PPP theory.
See for instance Zhou and Kutan (2011); Chortareas and Kapetanios (2009); Sarno and
Valente (2006); Nikolaou (2008); and Parsley and Wei (2007), among many others.
There are also studies that focus solely on the relationship between RER and the
variable of interest in a controlled environment. For example, the relationship between
1
This phenomenon is often referred to as the Balassa-Samuelson effect.
4
productivity and the movement of RER is explored by Égert (2002) for the Czech Republic, Hungary, Poland, Slovakia and Slovenia; Égert (2005) for three south-eastern European countries (Bulgaria, Croatia and Romania), two CIS economies (Russia and Ukraine)
and Turkey; and De Broeck and Sløk (2006) for 26 different countries.2 The relationship between RER and oil price movements is examined by Kutan and Wyzan (2005) for
Kazakhstan; Koranchelian (2005) for Algeria; Zalduendo (2006) for Venezuela; and Korhonen and Juurikkala (2009) for the OPEC countries, among others. Likewise, Bagella
et al. (2006) analyze the relationship between the volatility of real effective exchange rates
(REER) and growth in per capita income. Their approach to testing such a hypothesis
is unique, in the sense that they test the effects of flexible exchange rates and volatile
exchange rates jointly, which had not been possible previously, due to the high correlation
between exchange rate regimes and bilateral exchange rate volatility measures.
There is an expanding body of literature on commodity currencies, which relates commodity prices to RERs (Chen and Rogoff, 2003; Cashin et al., 2004, and followers). These
studies, among others, find a link between RER and the commodity prices, although the
results are somewhat weak and mixed. For instance, this link only exists for about onethird of the 44 commodity exporting countries studied by Cashin et al. (2004), whereas
Chen and Rogoff (2003) find RER being affected by commodity prices in two of the three
commodity exporting countries they consider. However, we believe that the missing link
in this literature is the effect of a substantial change in commodity prices on expectations
of a country’s future income, which depend not only on current commodity prices, but
also on expected future commodity prices. In some countries and in some sub-periods,
the changes in the commodity prices are not sufficient to change the expectations of the
country’s future income, and hence to cause significant RER movements, as appears to be
the case when RER does not respond to changes in commodity prices as per this literature.
To the best of our knowledge, there has been no study to date which has discussed the
role of oil companies in creating oil production hype by providing estimates of oil reserves
2
Ten EU accession countries (the three Baltics, Bulgaria, the Czech Republic, Hungary, Poland, Romania, the Slovak Republic, and Slovenia) and 16 other transition economies (Albania, Croatia, the former
Yugoslav Republic of Macedonia, Mongolia, Russia, and the other countries of the former Soviet: Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, the Kyrgyz Republic, Moldova, Tajikistan, Turkmenistan,
Ukraine and Uzbekistan).
5
which are in excess of actual reserves, which in turn raises expectations of future income
levels, thereby affecting the real exchange rate movements in a resource booming economy.
This study is an attempt to fill this gap.
3. What can drive RER movements?
There are many different definitions of the RER.3 Following Dornbusch (1974), Krueger
(1982), and Frenkel and Mussa (1985), we define the RER as follows:4
RER =
P rice of non-tradable goods
.
P rice of tradable goods
(1)
In our empirical analysis, we use the following two measures of the RER:
RER1 =
CPIAZ
Services
CPIAZ
F ood & non-f ood products
wi
18
Y
CPIAZ
Services
,
RER2 =
ei ×
WPIi
i=1
(2)
(3)
where CPI is the consumer price index in Azerbaijan (AZ), WPI is the wholesale price index
in Azerbaijan’s main trading partners,5 ei is the nominal exchange rate of Azerbaijan’s
national currency, the Manat, against the currency of its trading partner i (expressed in
the foreign currency per Manat), wi is the trade share of trading partner i in Azerbaijan’s
18
X
non-oil trade turnover in 2001, and
wi = 1. In both RER1 and RER2, we use the CPI
i=1
for services as a proxy for the aggregate price of non-tradable goods in Azerbaijan. In
RER1, we use the CPI for food and non-food products (excluding services) as a proxy
for the price of tradable goods in Azerbaijan. In RER2, we use the weighted average
WPI of Azerbaijan’s main trading partners as a proxy for the price of tradable goods in
Azerbaijan. This is because we believe that prices of tradable goods in Azerbaijan are more
closely related to the WPI than to the CPI of the country’s trading partners, as services
(that is, non-tradables) have a relatively large weight in the CPI.
3
See Edwards (1989) for a brief review of some of these definitions.
According to this definition, an increase in the RER corresponds to its appreciation. In this study, we
use the terms ‘goods’ and ‘goods and services’ synonymously.
5
These trading partners are China, France, Germany, Greece, Iceland, Iran, Israel, Italy, Japan, Kazakhstan, the Netherlands, Russia, Spain, Turkey, UAE, the UK, Ukraine and the USA.
4
6
Following Edwards (1989), we define ERER as the relative prices of non-tradable to
tradable goods that result in the simultaneous attainment of both an internal and an
external equilibrium, for given long-run equilibrium values of other relevant variables (fundamentals). An internal equilibrium occurs when the non-tradable goods market clears in
the current period and is expected to be in equilibrium in the future. An external equilibrium occurs when the current account balances both in the present period and in the
future satisfy the inter-temporal budget constraint which states that the discounted value
of the current account balances has to be equal to zero. This leads to the concept of the
behavioral equilibrium exchange rate. Thus, we expect the actual real exchange rate to
be in equilibrium in a behavioral sense; i.e., its movements should reflect changes in the
fundamentals of the economy (Clark and MacDonald, 1999).
Below, we discuss the ideas involved in the estimation of the equilibrium relationship
between RER and its fundamentals. Assume that the prices in an economy are determined
by the market. Then, the best guess for the actual RER (with no other information) would
be
E[RERt ] = ERERt .
(4)
Now suppose that we do not know ERER, but we do have a theory of how it is determined,
i.e.,
ERERt = f (tradf undamentalst , zt ),
(5)
where tradf undamentalst is a vector of traditional fundamentals of RER and zt is a candidate new fundamental representing the expected future (oil) income. We do not observe
f ; however, we can estimate it. Equations (4) and (5) imply
RERt = f (tradf undamentalst , zt ) + εt .
(6)
Equation (6) is a valid regression equation, as long as the following conditions hold:
E[εt |tradf undamentalst , zt ] = 0
E[ε2t |tradf undamentalst , zt ] = constant
E[εt εt−k |tradf undamentalst , zt ] = constant ∀t, k and k 6= 0.
7
(7)
Equation (6) implies that ERER could be estimated meaningfully by simply regressing
the actual RER on traditional RER fundamentals and zt (a candidate new fundamental), provided that equation (7) holds. The traditional RER fundamentals discussed in
the literature are productivity, foreign direct investment (FDI), government expenditures,
openness, interest rate differentials, and TsOT, stock of net foreign assets, among others.
We obtained the data used in this study from the Asian Development Bank, Economist
Intelligence Unit (EIU), IMF and NBA. The frequency is quarterly and the sample period
is from the first quarter of 1996 to the fourth quarter of 2003.
The traditional RER fundamentals that we use as control variables are: productivity
(as measured by the ratio of real GDP to employment), TsOT (the ratio of export prices
to import prices), country risk6 (an index of economic and political stability computed by
EIU), openness (the ratio of the sum of exports and imports to GDP), the interest rate
differential (vis-á-vis the US), government expenditures (as a percentage of GDP), and
FDI inflows (as a percentage of GDP). The data for all of these variables, except for the
country risk, are converted to an index, with 2001 as the base year. Productivity, openness
and government expenditures are adjusted for seasonal factors.
A priori, we would expect RER to have a positive relationship with productivity, TsOT,
openness and government expenditure and FDI inflows, and a negative relationship with
country risk. Productivity has a positive effect on the exchange rate through the BalassaSamuelson effect. An improvement in TsOT leads to appreciation of RER through the
wealth effect. Greater openness generally causes RER to appreciate by enhancing competition, raising productivity and increasing capital inflows. An increase in government
expenditure often leads to the appreciation of RER, because governments usually spend
more on non-tradable goods than on tradable goods. We would expect stronger inflows
of FDI and other capital to cause an appreciation of RER, as they increase the supply of
foreign exchange. We would also expect a higher country risk to lead to a depreciation of
RER, since it reduces capital inflows and the supply of foreign exchange.
We plot actual values of these fundamentals (solid lines), along with their predicted
6
We include the country risk to adjust for endogeneity, which may be present due to its relationship
with FDI. This is because a country with a lower risk will attract a greater FDI relative to a similar
country with a higher country risk.
8
2004
1996
1998
2000
2002
2004
80
60
40
2004
1996
1998
2000
2002
2004
1996
1998
2000
2002
2004
TsOT Index
0.4
0.8
1.2
100 150 200
2002
8
2002
2000
7
2000
1998
6
1998
1996
5
1996
Govt. Expend. Index
2004
4
Openness Index
120
100
80
2002
70
Productivity Index
60
2000
60
FDI Index
1998
50
Country Risk Rating
1996
Notes: A locally-weighted polynomial regression is used to smooth the curves in the plots. The solid lines
represent actual values of the variables, while the dashed lines represent their smoothed (predicted)
values.
Fig. 1: Evolution of the fundamentals of RER in Azerbaijan.
9
values from a locally weighted polynomial regression (dashed lines), in Fig. 1. Clearly,
the smooth curves (dashed lines) for productivity, openness, FDI and TsOT point to the
appreciation of RER. The government expenditures and the country risk suggest that the
appreciation of RER began some time around 2001.
In light of the evolution of the fundamentals, the features of transition economies and
the findings of previous studies, we would expect the RER in Azerbaijan to appreciate
more or less steadily over the period 1996–2003. However, the RER in Azerbaijan actually
appreciated in 1996–1998 but depreciated in 1999–2003. We plot RER, as defined in
equations (2) and (3), in Fig. 2. Each of the RERs shows a depreciation between 1999 and
2004, which is contrary to the features of transition economies, coupled with the booming
3.8
4.0
4.2
RER
4.4
4.6
4.8
oil sector and the movements in the fundamentals of RER in Azerbaijan.
RER1
1996
1998
2000
2002
RER2
2004
Note: We constructed RER1 and RER2 as per equations (2) and (3) respectively.
Fig. 2: RER movements in Azerbaijan from March 1996 to December 2003.
4. Basic model
Following Van der Ploeg (2011), we assume a small open economy populated by a large
number of identical households. The number of households remains unchanged over time
10
and is normalized to unity. The economy possesses considerable reserves of oil. However,
the exact amount of its oil reserves cannot be determined. Moreover, the economy does not
have the technology for oil production, and therefore oil can only be produced by foreign
firms, using foreign labor and foreign capital. Since there is no domestic consumption of
oil, the entire output of oil needs to be exported. The economy signs PSAs with several
leading foreign oil firms. The PSAs specify the total amount of oil to be produced under
each agreement. They also stipulate that the proceeds from oil exports are to be divided
into two parts. One part is to be retained by the foreign oil firms to cover production costs.
The other part is to be equally divided among the (domestic) households as households’
oil income Wt .
The households are informed about the terms of the PSA. However, they cannot predict
the exact amount of their oil income in future periods, due to uncertainties relating to the
volume of domestic oil production and the world price of oil in those periods. Instead,
the households form expectations about their future oil income, using all of the available
information, such as the total amount of oil to be produced under the existing PSAs. The
larger the total amount of oil to be produced under the existing PSAs, the more oil income
the households expect to receive in the future. This means that their future realized oil
income can change if the PSAs or the volume of oil production change (see Pindyck, 1980,
for example), new alternative technology is invented, or other sources of resources are
discovered, as Dasgupta et al. (1978) discuss.
Following Cashin et al. (2004), we assume that the households consume a (composite)
non-oil tradable good, CtT , imported from foreign countries, and a (composite) non-tradable
good, CtN . They do not value leisure, and supply labor inelastically in domestic firms
producing non-tradable goods. The price of the tradable good is in the local currency
of one unit of the tradable good, is taken as exogenous, and is determined by the world
consumption or demand, Cwt, i.e., PT = g (Cwt ). The price of the non-tradable good, PN ,
is determined by the domestic consumption demand, i.e., PN = f (Ct ).7
7
In the study by Cashin et al. (2004), the price of tradable and non-tradable goods depends on a
constant labour supply. In this paper, we instead normalize the labour supply to one. However, following
Obstfeld and Rogoff (2001), we set the price as a function of consumption, and consider the expected
present value of future income as an important factor for price, as was implicitly shown by Carroll (2001).
11
Each agent consumes the non-tradable and tradable goods to maximize their utility.
Following Carroll (2001), an agent’s optimal level of consumption, Ct = CtT + CtN , is equal
to the sum of their oil income, Wt , and their labour income, Ht :
Ct = kt (Wt + Ht ) ,
(8)
where kt includes the preference rate and other factors, and income, Wt and Ht , includes
both current and expected present values of future income,
#
" T̄
X YW +YH
i
i
Wt + Ht = Et
i−t ,
(1 + R)
i=t
(9)
where R and T̄ denote the interest rate and life expectancy, respectively, and YiW and YiH
denote current income. These factors, along with others, affect
W
W
W
Yi+1
= ΨW AW
i+1 , Yi , εi
and
H
H
H
Yi+1
= ΨH AH
,
i+1 , Yi , εi
(10)
(11)
H
where AW
i+1 represents factors such as the productivity of foreign firms, Ai+1 represents
government expenditures or FDI, and εW
εH
denotes shocks to income such as resources
i
i
booms or PSA shocks (financial crises).
RER is a function of the price of non-tradable, PN , to tradable, PT , goods. Consequently, RER can be expressed as a function of expected future income and world consumption or demand as follows
RERt =
PN
,
PT
(12)
" T̄
#
!
X YW +YH
i
i
= R Et
i−t , Cwt .
(1
+
R)
i=t
(13)
The above equation shows that RER is determined not only by the consumption of imported tradable goods or world consumption/demand, Cwt , but also by the current and
expected future oil and labour income, Et (.), which can be influenced by government expenditures, FDI, productivity, or PSAs. This is in contrast to the finding of Korhonen and
Juurikkala (2009) that the price of oil has a significant effect on real exchange rates in a
group of oil producing countries.
12
Interestingly, the average growth of yearly consumption in Azerbaijan for the period
1995–1998 was around 17.4%, whereas the average consumption growth for the period
1999–2003 was around 7.34%—a drop in consumption growth of around 57.9%. Below,
we argue that the appreciation of the RER over the period 1996–1998 was due to the oil
euphoria, which caused expectations of ever higher future incomes from the revenues of oil
production; as a result, consumption growth increased and the RER appreciated. Later
on, over the period 1999–2003, when some of the PSAs began to fail or ran into difficulties,
the expectations of ever higher incomes slipped, the consumption growth dropped and the
RER depreciated.
5. Expectations of future income
We construct two measures for determining the evolving expectations of future incomes
in Azerbaijan over the period 1992–2003. As per the model in equation (13), the measure
uses the cumulative (over time) estimated oil reserves (in billions of barrels), which we
refer to as CEOR,8 to proxy for the expected cumulative future income, based on PSAs
signed between the Azerbaijan government and oil extracting companies. This measure is
based on the estimated oil reserves from the 25 PSAs that were signed between 1992 and
1999. However, of these 25 PSAs, seven subsequently failed, and we adjust the CEOR9
accordingly between 1999 and 2003. We refer to this measure of evolving expectations
of future income in Azerbaijan as CEOR1 hereafter. Another four PSAs,10 while not
completely failing, ran into difficulties in the period 2000–2003, and we likewise remove
their estimated oil reserves from the CEOR1 measure in the same period. We refer to this
measure of the expectations of future income in Azerbaijan as CEOR2. Thus, CEOR1
and CEOR2 are the same until 1999. These are the two measures of expected future
8
The estimated oil reserves are taken from the energy information administration website
http://www.eia.doe.gov/emeu/cabs/azerproj.html.
9
For example, Karabagh (offshore) was signed in December 1995 with an estimated oil reserve of 1
billion barrels. It was closed in January 1999 as a result of finding no hydrocarbon reserves. In the
construction of CEOR, we use 1 billion of barrels oil from Karabagh in December 1995, but subtract the
same amount (1 billion) when it became known to have failed, i.e., in March (for quarterly data) 1999.
10
For example, the Araz, Alov, and Sharg project was signed in July 21, 1998, and ratified in December
1998, but exploration was later suspended due to a confrontation with an Iranian gunboat in July 2001,
pending the resolution of Caspian Sea borders between Azerbaijan and Iran.
13
income in Azerbaijan that we consider. However, we think that the second measure of the
expectations of future income in Azerbaijan, CEOR2, is more appropriate, as it excludes
the estimated oil reserves from all unsuccessful PSAs in the subsequent periods.
We then construct the measure of the realized (actual) future income of Azerbaijan
over the period 1992–2003. This measure is based on the cumulative estimated oil reserves
from the 14 successful PSAs only, and is referred to as CEOR-Adj (i.e., CEOR adjusted
for the unsuccessful PSAs).11 The CEOR1, CEOR2, and CEOR-Adj measures are shown
in Fig. 3 for the period from March 1996 to December 2003.
By the end of June 1999, the total oil reserves, from 25 PSAs, were estimated to be
around 27 billion barrels, which fed into expectations of an ever higher income in Azerbaijan. Some of the PSAs subsequently failed, which contributed to a loss of around 6 billion
(more than 12 billion) barrels of oil, according to the COER1 measure (CEOR2 measure),
which led to expectations of a lower future income over the period 1999–2003. The lost
amount of oil reserves constitutes around 29% or 47% (depending on which measure we
use) of the estimated oil reserves from the successful PSAs. We think that either of these
amounts would have been enough to cause a revision of the expectations of an ever higher
income in Azerbaijan in the period of 1999–2003, which seems to have caused the depreciation of its real exchange rate since 1999. This is the focus of the next section, where
we empirically test whether the COER explains the variation in the RER and could have
caused the depreciation observed since 1999.
The gap between CEOR1 (or CEOR2) and CEOR-Adj shows the effect of unsuccessful
PSAs on the expectations of future income. The CEOR (continuous black line in Fig. 3)
therefore reflects (optimistic) expectations of the future income, based on current period
information. In other words, expectations based on the continuous line are unrealistic,
as it considers the future income derived partly from the unsuccessful PSAs. CEOR-Adj,
however, measures realistic expectations (or actual/realized future income), as it includes
only those PSAs that were actually successful over the period 1996–2003.
11
For example, we did not consider the oil reserve from the Karabagh (offshore) project when we constructed the realized future income, as this was later found to be a failed PSA.
14
3.0
2.5
2.0
CEOR
CEOR1
1996
CEOR2
1998
2000
CEOR−Adj
2002
2004
Notes: CEOR1 and CEOR2 stand for the cumulative (over time) estimated oil reserves (in billions of
barrels) from two measures, based on production sharing agreements (PSAs) between the Azerbaijan
government and oil extracting companies. This is a proxy for the evolving expectations of future income
in Azerbaijan over the period 1996–2003. Of the 25 PSAs signed, 14 were successful, seven failed, and
another four were not ratified. CEOR1 is based on the estimated oil reserve from 18 PSAs (14 successful
and four not ratified) and seven failed PSAs, while CEOR2 is based on the estimated oil reserves from
the 14 successful PSAs and 11 unsuccessful PSAs. CEOR-Adj stands for the CEOR adjusted for the 11
unsuccessful PSAs, and is meant to capture the actual future income in Azerbaijan, as it is based on the
successful PSAs only.
Fig. 3: Expected future incomes (proxied by CEOR1 and CEOR2) and actual future
incomes (proxied by CEOR-Adj) in Azerbaijan from March 1996 to December
2003.
15
Table 1: Dickey-Fuller GLS unit root test.
Main variable
log(RER1)
log(RER2)
log(PSA)
Level series
Constant and Trend
−2.46 (1)
−2.14 (1)
−0.57 (1)
Control variable
log(TsOT)
log(Productivity)
Risk
Openness
FDI
Govt. expenditure
−3.51∗ (3)
−4.40∗∗ (0)
−1.44 (0)
−4.77∗∗ (1)
−2.15 (0)
−2.10 (0)
Constant
−1.49 (0)
−1.57 (1)
−0.39 (1)
−1.02 (0)
−1.38 (4)
−0.93 (0)
−2.78∗ (0)
−1.16 (0)
−1.37 (0)
Difference series
Constant and Trend
Constant
−5.68∗∗ (0)
−4.64∗∗ (0)
−5.25∗∗ (0)
−3.94∗∗ (0)
∗∗
−4.05 (1)
0.01 (4)
−4.21∗∗ (0)
−1.09 (3)
−5.13∗∗ (0)
−6.62∗∗ (1)
−7.07∗∗ (0)
−6.19∗∗ (0)
−4.18∗∗ (0)
−1.20 (3)
−4.85∗∗ (0)
−6.43∗∗ (1)
−7.06∗∗ (0)
−6.00∗∗ (0)
Notes: * and ** denote significance at the 5% and 1% levels, respectively, and ‘log’ stands
for the natural logarithm. The numbers in parentheses are optimal lag lengths, as chosen
by the Schwarz information criterion. The maximal lag length was set to seven quarters.
6. Empirical evidence
In this section, we describe our methodology and findings, and the implications of the
study.
6.1. Unit root testing
As we discuss in Section 3, the cointegration relationship in equation (6) requires the
linear combination of the variables to either be a stationary process or satisfy equation
(7). This can be the case if each of the variables in the cointegrating vector is integrated
of order 1 (I(1)). We test for the order of integration in the variables and present our
findings in Table 1. We conduct a unit root test from Elliott et al. (1996) because we are
uncertain about the mean of the data, i.e., whether the upward slope in the data is due
to a deterministic trend or the stochastic trend. The test statistic is a modified version of
the Dickey-Fuller t-test, but has a substantially improved power when an unknown mean
and trend are present. The test is therefore highly suitable, especially when it is hard to
assess the mean process in the data, which can be due to limited data observations. Like
the Dickey-Fuller (DF) test, it tests the null hypothesis of the series being a non-stationary
process, and is known as the DF Generalized Least Squares (GLS) test.
16
As per the DF GLS test, we find all of the variables to be I(1) processes in levels,
except for the Openness variable, which we find to be I(0), while the order of integration
for the log(Productivity) variable lies in the indecisive zone. This variable requires further
investigation because it is found to be I(0) in levels (with the constant and trend in an
auxiliary equation for the unit root test), but its further differences (up to of order 2,
not shown here) could not make it an I(0) process. There are two issues that we think
are important for the difference series of the log(Productivity) variable. One is serial
correlation, as it can be seen that the optimal lag length selected in the auxiliary equation
for the difference of log(Productivity) in the DF GLS unit root test is of order 3, and they
are individually significant (not shown here). The plot of the difference of log(Productivity)
in Fig. 4 highlights the point that its variance does not seem constant either, and if we take
care of any autocorrelation and heteroscedasticity, the series is likely be an I(0) process. An
ideal unit root test under a high degree of serial correlation and perhaps an unknown form of
heteroscedasticity is the test of Phillips and Perron (1988, PP). They propose an alternative
(non-parametric) method for controlling serial correlation when testing for a unit root, a
method which is robust to the general forms of heteroscedasticity. The PP method therefore
seems more appropriate for testing for a unit root in log(Productivity), which suffers from
serial correlation and heteroscedasticity. This is evident from the heteroscedasticity- and
autocorrelation-corrected (HAC) residual variance, which is under a quarter of the value
of the uncorrected residual variance of the difference series. As a result, the t-statistic is
inflated, and we therefore see the rejection of the null unit root hypothesis for the difference
series. According to the PP test, log(Productivity) may be either I(1) or I(0) (depending
on whether we believe it has a stochastic or deterministic trend), as its difference is a
stationary process.
6.2. Cointegration relationships
We test whether the CEOR variable explains the variation in RER (Eq. (6)) after
controlling for the effects of other variables such as TsOT, productivity, risk, openness,
FDI and government expenditures. While the relationship in equation (6) can explain the
17
Table 2: Phillips-Perron unit root test.
Control variable
log(Productivity)
Residual variance (no correction)
HAC corrected variance (Bartlett kernel)
Level series
Constant and Trend Constant
−6.03∗∗
−0.56
0.003296
0.006349
0.001226
0.002684
Difference series
Constant and Trend Constant
−10.95∗∗
−11.16∗∗
0.006095
0.006103
0.001343
0.001367
−10
0
%
10
20
Notes: see the comments for Table 1. HAC stands for the heteroscedasticity- and autocorrelation-corrected
residual variance, using non-parametric Bartlett kernel estimation.
1996
1998
2000
2002
Fig. 4: First difference of log(Productivity).
18
2004
real exchange rate in the long-run, the complete (short-run, long-run and lags)12 dynamics
of the system can be captured by the error correction model of Engle and Granger (1987),
or, more appropriately, by the Johansen (1995) vector error correction model (VECM).13
These models provide the long-run coefficients, along with the coefficients for the speed of
adjustment and the short-run deviations that together drive the real exchange rate. Such
a VECM can be written as
∆yt = µ + πyt−1 + γ1 ∆yt−1 + γ2 ∆yt−2 + · · · + γp ∆yt−p + ηt ,
(14)
where yt is a vector containing RER and the fundamentals (more precisely, yt = (RERt ,
′
COERt , F DIt , P roductivityt, T oTt , Riskt , Govtexpt ) ), p determines the number of lags,14
and ηt is the vector which is assumed to follow a normally distributed white noise process.
Since we find the Opennesst variable to be an I(0) process, we use it as an exogenous
′
variable to the system in equation (14). π = αβ contains the long-run (equilibrium)
relationships, which can be written as a product of two matrices (one for the speed of adjustment coefficients (α) and the other for the normalized cointegrating vector coefficients
(β)), and its rank determines the number of such long-run relationships. In addition, all
roots of the system are equal to 7 − r; i.e., we restrict the system to be at most integrated of order 1, as per the unit root test results. The rank of π can be tested using
the maximum eigenvalue statistic, −2ln(Q; r|r + 1) = −T ln(1 − λ̂r+1 ), and/or the trace
P
statistic, LRtr = −T ki=r+1 ln(1 − λi ), from Johansen and Juselius (1990), for testing
the existence of r vs. r + 1 cointegration relationships in π, where λ are the eigenvalues
(λ1 ≥ λ2 ≥ · · · ≥ λk ), and T is the number of observations. Critical values of the test
statistic are provided in the appendix of Johansen and Juselius (1990).
As we find all of the variables in the vector yt to be I(1), we now test for the number
of cointegrating vectors in the system. As we discuss earlier in the paper, we set the lag
order of the VECM in differences to be one, due to the small sample size. The trend
12
In the cointegration literature, the long-run relationship stands for the equilibrium (stationary) relationship and the short-run relationship defines deviations of the real exchange rate from the equilibrium
(stationary) level.
13
This is appropriate in cases where there is more than one cointegrated vector and/or the very likely
endogeniety issue among variables, which promotes the use of the VECM for RER.
14
For estimation, we only use p = 1, due to data size limitations.
19
variable is restricted in the cointegration space. Finally, the appropriateness of the model
is assessed based on measures of serial correlation, normality, heteroscedasticity, and the
model’s stability.15 We use the maximum eigenvalue and trace test statistics to determine
the number of cointegrating vectors in π, and the results of these tests are given in Table
3. When the trace test results are compared with the relevant limiting distribution, we
find that the data suggest using three cointegration relationships at the 5% level, and
two at the 1% level.16 However, several studies conclude that the trace test tends to be
over-sized in small samples; see for example Jacobson et al. (2001) and Toda (1995). The
maximum eigenvalue test suggests two and one cointegration relationships at the 5% and
1% levels, respectively. We use the Bartlett trace statistic for small sample correction with
the modified yt , and find evidence of one cointegration vector at the 5% level. Moreover, if
we allow two cointegration vectors, the signs in the two cointegration relationships appear
to be contrary to economic logic, while the VECM estimation with one cointegrating vector
resolves the wrong signs issue. Given all of this evidence and the economic rationales for
the cointegration of the variables in the vector yt , we believe that it is best to have one
cointegration relationship, namely that of RER with its fundamentals. Following this, we
estimate the VECM in equation (14) with one cointegration relationship. The results of
the VECM estimation for the variable RER are given in Table 4.
6.3. Findings
We select two models for each of RER1 and RER2, and call them RER1.1, RER1.2,
RER2.1, and RER2.2. The two models for each RER differ in that the FDI variable is
included in the vector yt along with CEOR1 in one model, while it is excluded from the
other, as it proves insignificant in both specifications of RER, with contradictory signs (see
column LR for RER1.1 and RER2.1 in Table 4). RER1.2 and RER2.2 are therefore the
focus of the discussions which follow. The findings are virtually the same qualitatively when
we replace CEOR1 with CEOR2 in the vector yt , and therefore we consider it redundant
15
We can check the stability of the estimated VECM based on the inverse kp roots of the characteristic
AR polynomial, where k is the number of endogenous variables and p is the largest lag. The stability
condition is verified if each of the roots has a modulus of less than one. In a VECM estimation with r
cointegration relationships, this implies that k − r roots should be equal to unity.
16
The p-values are computed using the simulated distributions of MacKinnon et al. (1999).
20
Table 3: Johansen unrestricted cointegration rank test.
No. of CE(s)
r
r
r
r
≤0
≤1
≤2
≤3
Eigenvalue
Trace statistic
p-value
Eigen-statistic
p-value
0.990
0.740
0.628
0.388
246.96
108.80
68.44
38.75
0.000
0.001
0.020
0.123
44.50
38.33
32.12
25.82
0.000
0.029
0.096
0.657
Notes: The trace test indicates two cointegrating equations and the eigenvalue
test one cointegration equation at the 1% level of significance. The critical values
assume no exogenous variable. We assume a linear deterministic trend (restricted).
The variables considered with each RER are CEOR1, FDI, Productivity, TsOT,
Govt. exp, and Openness (exogenous).
to report them here. However, we present the equilibrium relationships obtained from
CEOR1 and CEOR2 in Figs. 5 and 6 respectively.
For both specifications of RER (i.e., RER1.2 and RER2.2), the long-run coefficients
are all significant and have the right signs, including the exogenous variable Openness (see
Table 4). The error correction term α is negative, highly significant and large (greater
than 0.5), indicating a quick correction of RER to its equilibrium.
As expected, the CEOR establishes the equilibrium relationship with the real exchange
rate (as the CEOR1 coefficient is positive and significant), along with a significant negative
impact in the short-run,17 as do the control variables, which are found to be significant and
to have the right signs. The residual analysis at the bottom of Table 4 shows that the RER
models satisfy the basic assumptions of no residual autocorrelation, no heteroscedasticity,
and multivariate normal residuals. Moreover, the VECMs for RERs are found to be stable,
as all of the (non-unit) roots of its polynomial are well below one.18 The models thus appear
to be reasonable and appropriate, with good explanatory powers (see below).
In Fig. 5, we plot ERER1.2 and ERER2.2 (dashed lines) using only LR coefficients,19
17
Apart from its insignificance in the short-run for RER2.2.
If a VECM has k endogenous variables and r cointegrating vectors, there will be k − r unit moduli
in the companion matrix. If any of the remaining moduli are too close to one, either the cointegrating
equations are not stationary or there is another common trend and the rank of the system specified in the
VECM is too high.
19
For example, ERER1.2 = 2.6 + 0.31log(CEOR) + 0.37log(Productivity) – 0.074log(TsOT) –
0.161log(Risk) + 0.110log(Govt. exp) – 0.011Trend.
18
21
Table 4: Estimates of the error correction model in equation (14).
Main variable
α
CEOR1
RER1.1
LR
SR
−0.795∗∗∗
0.323∗∗∗
−0.161∗∗∗
RER2.1
LR
SR
−0.634∗∗∗
0.444∗∗∗
0.024
RER1.2
LR
SR
−0.818∗∗∗
0.308∗∗∗
−0.146∗∗∗∗
RER2.2
LR
SR
−0.624∗∗∗
0.353∗∗∗
0.039
Control variable
RER1
RER2
FDI
Productivity
TsOT
Risk
Govt. exp
Openness
Trend
Constant
−0.026
0.449∗∗∗
−0.101∗∗∗
−0.105∗
0.136∗∗∗
−0.011∗∗∗
2.055
0.036
0.639∗∗∗
−0.302∗∗∗
−0.065
0.032
−0.022∗∗∗
0.748
0.374∗∗∗
−0.074∗∗∗
−0.161∗∗∗
0.110∗∗∗
−0.011∗∗∗
2.593
0.626∗∗∗
−0.349∗∗∗
−0.276∗∗∗
0.064
−0.019∗∗∗
2.004
−0.065
0.019
−0.142∗∗
0.088∗∗∗
−0.163∗
0.029
−0.064∗∗
0.295∗∗∗
0.030
−0.019
−0.174
0.017
−0.146
0.019
−0.031
0.150
Residual analyses for models RER1.2 and RER2.2
LM test for residual autocorrelation for lags 2–6
Jarque-Bera test for the residual’s multivariate normality
χ2 test for residual heteroscedasticity
VECM stability condition (Roots of characteristic polynomial)
−0.055
−0.117∗∗
0.091∗∗∗
−0.096
0.032
−0.054∗∗
0.254∗∗∗
p-value
**
0.883
0.524
1,1,1,1,1,0.41,
0.40,0.31,0.31,
0.16,0.13,0.13.
0.007
−0.128
0.066
−0.107
−0.011
−0.018
0.094
p-value
**
1.000
0.659
1,1,1,1,1,0.54,
0.50,0.36,0.36,
0.23,0.23,0.12.
Notes: *, **, and *** indicate significance of the test statistic at the 10%, 5% and 1% levels of significance,
respectively. Since α is only significant for RER1 and the Productivity equations, we restrict α to be zero
for the other six variables before undertaking the residual analyses. However, the other coefficients (two
each of αs, βs, and γs) are not affected much by these restrictions, apart from their significance levels, which
increase. SR and LR stand for short-run and long-run coefficients, respectively. For SR, the variables are
differenced and have a one period lag.
22
along with RER1.2 and RER2.2 (solid lines). ERER1.2-Adj and ERER2.2-Adj use CEORAdj in the computations. For instance, ERER1.2-Adj = 2.6 + 0.31log(CEOR-Adj) +
3.8
RER1.2
ERER1.2
ERER1.2−Adj
1996
1998
2000
2002
2004
3.6 3.8 4.0 4.2 4.4 4.6 4.8
RER2
4.2
4.0
RER1
4.4
4.6
0.37log(Productivity) – 0.074log(TsOT) – 0.161log(Risk) + 0.110log(Govt. exp) – 0.011Trend.
RER2.2
ERER2.2
ERER2.2−Adj
1996
1998
2000
2002
2004
Notes: Over-shooting = RERi.2 – ERERi.2-Adj (with successful PSAs only), where i = 1, 2. The ERERs use the CEOR1
measure of expected future income in Azerbaijan. The overshooting seems to end as RERi.2 – ERERi.2-Adj converges to
zero.
Fig. 5: Oil euphoria and the overshooting of RER when CEOR1 measures expectations
of future income.
We can see that the ERER1.2 and ERER2.2 each explain more than 90% of the variation
in RER1 and RER2, respectively.20 Apart from the productivity variable, the CEOR explains most of the variation in RER. ERER1.2 and ERER2.2 explain why RER1 and RER2
appreciated until 1998 and depreciated thereafter: it is because the signing of numerous
PSAs fueled expectations of ever higher future incomes, which increased the consumption,
causing RER to appreciate over the period 1994–98, even though the actual oil income
growth remained small. The depreciation of RER over the period 1999–2003 was a result
of the revision of the expected future income estimates, driven by the realistic oil reserves
estimates, when some of the PSAs signed during the period 1994–99 failed or ran into dif20
The value of R2 using only the variables in levels can be obtained by regressing RER on its fundamentals.
23
RER2
1996
1998
2000
2002
2004
3.6 3.8 4.0 4.2 4.4 4.6 4.8
4.6
4.4
4.2
4.0
RER1
3.8
RER1.2
ERER1.2
ERER1.2−Adj
RER2.2
ERER2.2
ERER2.2−Adj
1996
1998
2000
2002
2004
Notes: Over-shooting = RERi.2 – ERERi.2-Adj (with successful PSAs only), where i = 1, 2. The ERERs use the CEOR2
measure of expected future income in Azerbaijan. The overshooting seems to end as RERi.2 – ERERi.2-Adj converges to
zero.
Fig. 6: Oil euphoria and the overshooting of RER when CEOR2 measures expectations
of future income.
24
ficulties (as virtually no oil was found). As a result, consumption growth decreased, which
removed the pressure from the prices of non-tradeable goods, meaning that RER depreciated to its long-term equilibrium level over the period 1999–2003, despite the actual oil
income growth having started to pick up. Thus, the expectations of future income, though
driven by oil euphoria, appear to have played an important role in the movements of RER,
as it initially appreciated when the actual oil income level remained low but expectations
of future income were high, and later depreciated when the actual income level started to
increase but expectations of future income started to drop.
We therefore do not see either the NBA’s loose monetary policy or its interventions in
the foreign exchange market as playing any significant role in driving the RER. Had the
NBA been successful, we would have seen the ERER in Figs. 5 and 6 responding to the
strengthening economic fundamentals and increasing from 1999 onward. The downward
direction of ERER since 1999 explains why the RER has likewise depreciated since then.
In addition, the 1994–99 oil euphoria caused an appreciation of RER in excess of what
we would have observed if the expectations of future incomes had been more realistic, i.e.,
if expectations had mimicked CEOR-Adj rather than CEOR (see for example Fig. 6).
ERER-Adj is therefore the realistic RER, as it is associated with CEOR-Adj (a proxy for
the actual future income in Azerbaijan, which only includes the estimated oil reserves from
the successful PSAs). The difference between RER and ERER-Adj can be termed an overshooting of RER, which is in excess of what should have been the case had the expectations
about future incomes been more realistic. We find that the ERER, ERER-Adj and RER
all converge towards zero (see Figs. 5 and 6).
Lastly, as per Fig. 6, the excessive appreciation over the period 1994–99 was reversed by
the depreciation over the period 1999–2003; we therefore see the RER starting to appreciate
again from 2004/05.
7. Discussion
We expect the RER in Azerbaijan to begin to appreciate again in 2005/06, and an
inspection of the actual RER data reveals that RER has indeed been appreciating since
2005. The importance of this point is that the depreciation of the RER in the period
1999–2003 helped Azerbaijan to achieve rapid growth in the non-oil tradable goods sector
25
in 2002–2003. The non-oil tradable sector grew at around 9.6% and 22.2% in 2002 and
2003, respectively. This contributed to the GDP growth rates of 10.6% and 11.2% in 2002
and 2003, respectively.
We admit that this paper’s findings for Azerbaijan should perhaps be treated with a
degree of caution, as we have only been able to collect a small data sample. However, we
have also redone many of the analyses, as well as performing robustness checks which are
not included here on other fundamentals, such as the interest rate differential vis-à-vis the
US, oil prices, NBA intervention in the foreign exchange market, the money supply, the
Russian crisis of 1999, etc. Of these additional results, none are significant and/or have the
expected signs to be able to explain the depreciation of RER. We therefore believe that our
model, with oil euphoria driven expectations of future income as the fundamental, is the
best model for explaining this phenomenon in RER, as suggested by the model in equation
(13). Similar phenomena appear to have occurred or to be occurring in other economies
that see boom and doom, as is discussed below.
Our finding of RER movements being caused by changing expectations of future income
is not limited to Azerbaijan. It applies to any country in which exogenous factors (a
discovery of oil reserves, an increased global demand for natural resources, a considerable
rise in world commodity prices, etc.) can cause substantial swings in the RER through
their effects on expectations of future income.
Table 5: RER and the GFCF growth.
Country
Time period
RER movements
GFCF growth
Australia
1980–1999
2000–2011
Depreciation
Appreciation
3.56%
6.39%
USA
1991–2000
2001–2010
Appreciation
Depreciation
8.71%
−0.68%
UK
1977–1980
1981–1984
Appreciation
Depreciation
16.54%
7.67%
Table 5 shows the average growth rates of gross fixed capital formation (GFCF) and
movements in the RER in Australia, the UK, and the USA in selected years. Since GFCF
26
depends on the expected future income, we use it as a proxy for expectations of future
income. As the table indicates, the expectations of future income were relatively low
and the RER depreciated in Australia in 1980–1999. An increased global demand for
Australia’s natural resources led to an upward adjustment in the expectations of future
income, and thus, to an appreciation of its RER in 2000–2010. Similarly, the discovery of
huge oil reserves in the North Seas just prior to the first oil shock of 1973/74, along with
higher oil prices, caused an increase in the expected future income and an appreciation
of the RER in the UK in the late 1970s. A substantial fall in oil prices and a slowdown
in the development of new oil fields led to a decline in the expected future income and
contributed to the depreciation of the RER in the early 1980s. The run of bad luck which
has plagued the US economy over the last decade or so relative to the decade of the 1990s
has contributed to a continuously depreciating RER. This has consistently built on the
low expected future income of the US, with a GFCF growth of only −0.68% over the last
decade. On the other hand, the good luck of the 1990s21 caused the RER to appreciate,
due to the high expected future income.22
As per the above analysis, high expected future income levels (as proxied by a high
growth in GFCF) are associated with an appreciation of the RER, and low expectations
of future income are associated with a depreciation of the RER. Alternatively, RER appreciates during boom times and depreciates during doom times. This is exactly what has
happened or is happening in Australia, the UK, and the USA.
8. Conclusion
We show that changes in exogenous factors (such as the discovery of substantial oil
reserves, an increased global demand for natural resources, or a substantial increase in
commodity prices) can cause—through their effects on expectations of future income—
21
Ahmed et al. (2004) report the stability in the US GDP growth over the period 1984Q1–2002Q1 as
being ‘good luck’ for the US economy.
22
Wei et al. (2010) study the effect of the IMF bailouts on international institutional investments in crisishit economies, and find that IMF bailouts encourage international institutional investments. Alternatively,
IMF bailouts help to raise investors’ confidence in countries hit by economic crises. We see a direct link
between raising business confidence and changing expectations of the future income of crisis-hit countries
due to bailouts, which can ultimately cause real exchange rate movements.
27
movements in the real exchange rate which are in excess of, or even in the direction opposite
to, what one would expect, given the actual changes in current income. We provide both a
theoretical model and empirical evidence to show that the real exchange rate appreciates
during boom times and depreciates during doom times.
In Azerbaijan, the oil euphoria and the herding behaviour demonstrated by oil companies (which signed numerous production sharing agreements with the government of
Azerbaijan) caused an overestimation of the oil reserves, unrealistically high expectations
of future income, and overshooting (that is, excessive appreciation) of the real exchange
rate (that is, the current real exchange rate appreciated more than the equilibrium real
exchange rate) in the first half of the sample period, even though the actual oil income
remained small. In the second half of the sample period, more accurate estimates of Azerbaijan’s oil reserves were made, which showed that the actual reserves were much lower
than the estimates in the first half of the sample period suggested. Accordingly, expectations of future income were revised downwards and the real exchange rate depreciated
(towards its long-term equilibrium level), even though the actual oil income started to
increase.
In Australia in 1980–1999, the expectations of future income were relatively low, and
the real exchange rate therefore depreciated. The subsequent increased global demand for
natural resources led to an upward adjustment in the expectation of future income, and
thus, appreciation of the RER in 2000–2010. Similarly, the discovery of huge oil reserves
in the North Seas just prior to the first oil shock of 1973/74, along with higher oil prices,
caused an increase in the expected future income and an appreciation of the RER in the
UK in the late 1970s. The substantial decline experienced in the oil prices at the time and
the discovery and development of new oil fields led to a decline in expectations of future
income, and contributed to the depreciation of the RER in the UK in the early 1980s. The
bad luck which has plagued the US economy over the last decade, relative to the decade
of the 1990s, has contributed to a continuously depreciated RER. This has consistently
built on the low expectations of US future income of only −0.68% in the last decade. On
the other hand, the good luck of the 1990s caused the RER to appreciate, due to the high
expectations of future income.
28
Acknowledgements
We would like to thank the anonymous referees, the editor, Professor Ike Mathur,
Professor Robert S. Chirinko (University of Illinois at Chicago, USA), Dr. Qingyuan Du
(Monash University, Australia), and Professor Shang-Jin Wei (Columbia University, USA)
for their constructive and useful comments and suggestions, which improved our revised
manuscript substantially.
29
References
Ahmed, S., Levin, A., Wilson, B. A., 2004. Recent U.S. macroeconomic stability: Good
policies, good practices, or good luck? The Review of Economics and Statistics 86,
824–832.
Bagella, M., Becchetti, L., Hasan, I., 2006. Real effective exchange rate volatility and
growth: A framework to measure advantages of flexibility vs. costs of volatility. Journal
of Banking and Finance 30, 1149–1169.
Beguna, A., 2002. Competitiveness and the equilibrium exchange rate in Latvia. EuroFaculty Working Paper in Economics No. 16.
Carroll, C. D., 2001. A theory of the consumption function, with and without liquidity
constraints. Journal of Economic Perspectives 15, 23–45.
Cashin, P., Céspedes, L. F., Sahay, R., 2004. Commodity currencies and the real exchange
rate. Journal of Development Economics 75, 239–268.
Chen, Y., Rogoff, K., 2003. Commodity currencies. Journal of International Economics 60,
133–160.
Chortareas, G., Kapetanios, G., 2009. Getting PPP right: Identifying mean-reverting real
exchange rates in panels. Journal of Banking and Finance 33, 390–404.
Clark, P., MacDonald, R., 1999. Exchange rates and economic fundamentals: A methodological comparison of BEERs and FEERs. In: Stein, J., MacDonald, R. (Eds.), Equilibrium Exchange Rates. Kluwer Academic Publishers, Boston, MA.
Coricelli, F., Jazbec, B., 2004. Real exchange rate dynamics in transition economies. Structural Change and Economic Dynamics 15, 83–100.
Dasgupta, P., Eastwood, R., Heal, G., 1978. Resource management in a trading economy.
The Quarterly Journal of Economics 92, 297–306.
De Broeck, M., Sløk, T., 2006. Interpreting real exchange rate movements in transition
countries. Journal of International Economics 68, 368–383.
30
Dornbusch, R., 1974. Tariffs and nontraded goods. Journal of International Economics 4,
177–185.
Edwards, S., 1989. Exchange rate misalignment in developing countries. The World Bank
Research Observer 4, 3–21.
Égert, B., 2002. Estimating the impact of the Balassa–Samuelson effect on inflation and
the real exchange rate during the transition. Economic Systems 26, 1–16.
Égert, B., 2005. Equilibrium exchange rates in South Eastern Europe, Russia, Ukraine and
Turkey: Healthy or (Dutch) diseased? Economic Systems 29, 205–241.
Égert, B., Lommatzsch, K., Lahreche-Revil, A., 2006. Real exchange rates in small open
OECD and transition economies: Comparing apples with oranges? Journal of Banking
and Finance 30, 3393–3406.
Elliott, G., Rothenberg, T. J., Stock, J. H., 1996. Efficient tests for an autoregressive unit
root. Econometrica 64, 813–836.
Engle, R., Granger, C., 1987. Co-integration and error correction: Representation, estimation, and testing. Econometrica 55, 251–276.
Frenkel, J. A., Mussa, M. L., 1985. Asset markets, exchange rates and the balance of
payments. In: Jones, R. W., Kenen, P. B. (Eds.), Handbook of International Economics.
Vol. 2. Elsevier, Amsterdam.
Halpern, L., Wyplosz, C., 1997. Equilibrium exchange rates in transition economies. International Monetary Fund Staff Papers 44, 430–461.
IMF, 2004. Azerbaijan country report. International Monetary Fund Staff Country Report
No. 04/9.
Jacobson, T., Jansson, P., Vredin, A., Warne, A., 2001. Monetary policy analysis and
inflation targeting in a small open economy: A VAR approach. Journal of Applied
Econometrics 16, 487–520.
31
Johansen, S., 1995. Likelihood-Based Inference in Cointegrated Vector Autoregressive Models. Oxford University Press, Oxford.
Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on
cointegration—with applications to the demand for money. Oxford Bulletin of Economics
and Statistics 52, 169–210.
Koranchelian, T., 2005. The equilibrium real exchange rate in a commodity exporting
country: Algeria’s experience. International Monetary Fund, Working Paper No. 05/135.
Korhonen, I., Juurikkala, T., 2009. Equilibrium exchange rates in oil-exporting countries.
Journal of Economics and Finance 33, 71–79.
Krueger, A. O., 1982. Analysing disequilibrium exchange-rate systems in developing countries. World Development 10, 1059–1068.
Kuralbayeva, K., Kutan, A. M., Wyzan, M. L., 2001. Is Kazakhstan vulnerable to the
Dutch disease? ZEI Working Papers, ZEI–Center for European Integration Studies,
University of Bonn.
Kutan, A., Wyzan, M., 2005. Explaining the real exchange rate in Kazakhstan, 1996–2003:
Is Kazakhstan vulnerable to the Dutch disease? Economic Systems 29, 242–255.
Lim, G., 1992. Testing for the fundamental determinants of the long run real exchange
rate. Journal of Banking and Finance 16, 625–642.
MacKinnon, J. G., Haug, A. A., Michelis, L., 1999. Numerical distribution functions of
likelihood ratio tests for cointegration. Journal of Applied Econometrics 14, 563–577.
Nikolaou, K., 2008. The behaviour of the real exchange rate: Evidence from regression
quantiles. Journal of Banking and Finance 32, 664–679.
Obstfeld, M., Rogoff, K., 1996. Foundations of International Macroeconomics. MIT Press,
Cambridge, MA.
Obstfeld, M., Rogoff, K., 2001. The six major puzzles in international macroeconomics: Is
there a common cause? In: Bernanke, B. S., Rogoff, K. (Eds.), NBER Macroeconomics
Annual. Vol. 15. MIT Press, Cambridge, MA.
32
Orlowski, L., 1998. Exchange-rate policies in central Europe and monetary union. Comparative Economic Studies 40, 58–78.
Paiva, C., 2001. Competitiveness and the equilibrium exchange rate in Costa Rica. International Monetary Fund, Working Paper No. 01/23.
Parsley, D. C., Wei, S.-J., 2007. A prism into the PPP puzzles: The micro-foundations of
Big Mac real exchange rates. Economic Journal 117, 1336–1356.
Phillips, P. C. B., Perron, P., 1988. Testing for a unit root in time series regression.
Biometrika 75, 335–346.
Pindyck, R. S., 1980. Uncertainty and exhaustible resource markets. Journal of Political
Economy 88, 1203–1225.
Sarno, L., Valente, G., 2006. Deviations from purchasing power parity under different
exchange rate regimes: Do they revert and, if so, how? Journal of Banking and Finance
30, 3147–3169.
Toda, H. Y., 1995. Finite sample performance of likelihood ratio tests for cointegrating
ranks in vector autoregressions. Econometric Theory 11, 1015–1032.
Van der Ploeg, F., 2011. Natural resources: Curse or blessing? Journal of Economic
Literature 49, 366–420.
Wei, S.-J., Zhang, Z., Du, Q., 2010. Does the global fireman inadvertently add fuel to the
fire? New evidence from institutional investors’ response to IMF program announcements. Journal of International Money and Finance 29, 728–741.
Zalduendo, J., 2006. Determinants of Venezuela’s equilibrium real exchange rate. International Monetary Fund, Working Paper No. 06/74.
Zhou, S., Kutan, A. M., 2011. Is the evidence for PPP reliable? A sustainability examination of the stationarity of real exchange rates. Journal of Banking and Finance 35,
2479–2490.
33