Sunil Kumar and S. M. Lokare*
The decade of 1990s saw far reaching economic reforms in India. Ever since the
switch over to market determined exchange rate that unfolded in 1993, the Indian economy
has witnessed massive transformation in terms of greater external sector openness and
higher degree of integration with the global economy. Against this backdrop, this paper
attempts to test the validity of purchasing power parity (PPP) condition in the case of India
during pre and post reform periods, using monthly data spanning over more than three
decades. The results from array of unit root tests unravel that real exchange rate (RER)
series remain non-stationary in the pre-reforms period and turn out to be stationary during
the post reforms period, indicating thereby that PPP condition holds during the post reforms
period, albeit, over the long run. These results are also corroborated by the cointegration test
and vector error correction model applied to the components of RER. This lends credence to
the fundamental principle that chances of holding PPP improve in an economy with greater
external sector openness and integration. Holding of PPP during post-reforms period also
provides support to the market determined exchange rate mechanism put in place in India
during the early 1990s.
JEL Classification : C23, E50, F31
Keywords : Real Exchange Rate; Bilateral Nominal Exchange Rate, Purchasing
Power Parity, Domestic Prices and US Prices
Prologue
The purchasing power parity theory is one of the oldest and most
luminous topics that continues to enfold the intellectual discourse in
the international economics. It constitutes a central building block in
the monetary model of exchange rate determination. Salamanca School
was the first to articulate the proposition of PPP in the 16th Century, while it was Gustav Cassel (1921), a Swedish economist who
championed the use of PPP as a model for setting relative gold parities.
However, the modern origins could be traced back to the restoration of
world financial system after its collapse during World War. I.
PPP is a condition of open economy general equilibrium model.
wherein, national price levels tend to be same under a common
currency. It states that arbitrage forces will lead to the equalisation of
goods prices internationally once the prices are measured in the same
currency. The underlying principle embedded in the model is that
goods market arbitrage enforces weeding out of differential in the
prices of traded commodities. Efficient arbitrage in goods market
across the countries emanates increasingly with external opening of
economies. PPP theory provided a point of reference for the long-run
exchange rate in many of the modern exchange rate theories.
The importance of PPP theorem could be gauged by its wide
applications in formulating the policy decisions. It is applied, interalia,
in choosing the right initial exchange rate for a newly independent
country; forecasting medium and long-term Real Exchange Rate; and
adjusting for price differentials in international comparisons of
income. The PPP condition not only helps in understanding the nature
of nominal and real shocks in the exchange rate models but also help
policy makers and researchers to compute RER misalignment. A
proper assessment of the deviation of the real exchange rate from its
equilibrium path can go a long way in enabling policy makers to
design an exchange rate policy which can achieve the long-term
sustainability of the balance of payments (Joshi, 2007). A common
method of determining the extent of misalignment of the exchange
rate is based on the principle of PPP theory for open economies,
which assumes that exchange rates adjust to offset the changes in
relative prices.
There is a growing body of empirical literature on PPP and a
consensus has emerged on a couple of facts. A number of studies
have weighed in favour of real exchange rate tending toward purchasing power parity in the very long run. However, the speed of
convergence to PPP is very slow. While few economists take PPP
seriously as a short-term proposition, most instinctively believe in
some variant of purchasing power parity as an anchor for long-term
real exchange rates (Rogoff K, 1996).
Few empirical studies have been undertaken to testify whether
PPP holds or not in Indian case also. In this paper, we have empirically
investigated PPP during pre-reforms and post reforms in the case of
India’s bilateral real exchange rate (RER) with USA. We have taken
monthly data from 1970 to 2008 and this was divided in two periods:
period I from 1970M01 to 1993M02 and period II from 1993M03 to
2008M08. The division of time period was drawn from the structural
break that took place with the introduction of liberalized exchange
rate mechanism (LERM) in India in March 1993.
The objective of this paper is to test the validity of PPP theorem
in an emerging Giant- India with reference to US. In this endeavor,
the paper attempts to test for the existence of any stable long-term
equilibrium relationship between the exchange rate and domestic/
US prices (WPI). This relationship is sought to be tested through the
procedure of Johansen Co-integration technique with Vector Error
Correction Model. This paper distinctly differs from the extant
literature by covering a long time horizon - monthly data spawning
for more than three decades (since 1970) and endeavors to test the
validity of PPP in the fixed and floating rate regimes.
The paper is structured into the following sections. Section II provides a review of the empirical literature on the PPP hypothesis.
Literature suggests a mixed result about holding of PPP; however,
recent studies have increasingly found results in favour of PPP in the
long run. Section III, delineates the process of opening up of external
sector and the evolution of exchange rate regime, while Section IV discusses model specifications, data sources, and empirical results.
Finally, the paper ends with the concluding remarks.
Section II
A Peep into the PPP Literature
Considering the central place that PPP occupies in the monetary
models of exchange rate determination, it is not surprising that
considerable amount of research has been devoted to its empirical
verification.
Law of One Price: Does it hold…?
It is the Law of One Price (LOP), which forms a central pillar of
PPP. LOP asserts that similar goods should be sold at similar prices
across countries. In the long-run, arbitrage ensures that the LOP
exists- that is identical goods denominated in a common currency
must sell for the same price in two separate markets, without
transportation costs and differential taxes thereby, causing intra
national price convergence.
LOP states that for any good i, if traded without frictions, the following should hold:
Pit = P*it Et (1)
In the equation (1), Pit and P*it are domestic price and foreign
currency price of i commodity at time t, respectively, while Et is the
nominal exchange rate at time t.
However, in practice tariffs, transportation costs and non-tariff
barriers, quality standards, taxes, profit margins, monopolistic
tendencies, etc., drive a wedge between prices in different countries.
Theoretical and empirical literature suggests that the PPP can be
violated in the short run. PPP may also not hold on account of factors,
viz., (i) if there are transaction costs and trade frictions, (ii) differential
baskets of goods are used to construct aggregate price indices, and
(iii) government intervenes in foreign exchange markets (Duncan,
Roberto, 2003).
Balassa- Samuelson (1964) provided an explanation as to why
the price levels differ across countries. They argued that when all countries price levels are translated to dollars at prevailing nominal
exchange rate (NER), rich countries tend to have higher price levels
than the poor ones. According to them, the presence of non-traded
goods, for which no international arbitrage exists, can lead to
systematic movements in real exchange rates inconsistent with PPP.
Productivity differentials in the traded and non-traded goods sectors
remain the underlying factor for the above hypothesis. Rich countries
have higher exchange rate adjusted price levels than poor countries
due to differing capital labour ratios (Bhagawati, 1984).
Empirical evidence showed that PPP performs better for those
countries that are geographically close to each other and where trade
linkages are high. Moreover, PPP holds better for traded goods
compared to non-traded goods (Officer, 1986). Reasons for failure of
PPP may be attributed to heterogeneity in the baskets of goods
considered for construction of price indices in various countries,
imperfect competition in goods market, and increase in the volume
of global capital flows during the last few decades, which led to sharp
deviation from PPP. Short-term exchange rate volatility emanates
from changes in portfolio preferences, short-term asset price bubbles
and monetary shocks. The failure of short-term PPP could be
attributed in part to stickiness in nominal prices (Rogoff, 1996).
Some believe PPP is a short-term proposition, while most believe
in some variant of PPP as an anchor for RER. However, the consensus
evidence indicates that- speed of convergence to PPP is extremely
slow (decay at 15 per cent per year); short-term deviations are large
and volatile; and real exchange rates tend toward PPP in the very
long run.
Many studies rejected PPP hypothesis on the ground of unit root
problem in RER. If the unit root model can characterise real exchange
rate behavior, then PPP does not hold because there is no propensity
to revert back to any equilibrium level (Cashin et, al, 2003).
Researchers found it difficult to reject the hypothesis that major
countries RER follow a random walk under floating rate regimes and hence difficult to prove convergence. Early tests include: Richard
Roll, 1979; Michael Darby, 1983; Michael Adlerand Bruce Lehmann,
1983 and Edison, 1985. Later papers incorporating standard unit root
tests include John Huizinga, 1987; Meese and Rogoff, 1988. The
tests using co-integration methods on modern floating rate have
failed too to reject random walk hypothesis (Bouncer, 1994).
The evidence is, however, inconclusive due to low power of unit
root tests in small samples. In the recent literature, some studies have
found evidence in favour of PPP- applying unit root tests (Frankel,
1986; Lothian and Taylor, 1996; Taylor, 2002; etc.).
Frankel was able to reject it using Dickey-Fuller tests. His
estimation showed a rate of decay for RER deviations of 14 per cent
per year (4.6 years) (Dickeyand Fuller, 1979). During 1990s, the long
horizon data studies tend to find evidence of mean reversion in real
exchange rates (Abuafand Jorion, 1990; Lothianand Taylor,
Cheungand Lai, 1994). Several studies found strong rejections of the
random walk model (Abuaf and Jorion, 1990; Glen, 1992; Diebold,
Husted, Rush, 1991).
A study by using the panel cointegration method for market
exchange rates of 17 African countries, supported the weak form of
the long-run PPP hypothesis (Nagayasu, 1999). Conversely another
study rejected the PPP hypothesis for Sri Lanka during the floating
exchange rate regime, thereby indicating the existence of market
frictions such as transaction costs prevalent in the international trade
(Wikremsinghe, 2001). There is also evidence supporting the
alternative hypothesis of acceptance of the PPP theorem for demeaned
data provided by Mohua Paul (2002), who tested the validity of PPP
for six South East Asian countries, including India, using panel unit
root test for multilateral RER based on dynamic export, import and
trade weights and concluded that PPP could be used to assess the
levels of exchange rate. Taking in to account the data from 1973 to
1997, for a sample of 30 developing countries, Holmes (2001)
confirmed the existence of PPP and proved that there was no evidence
to prove that PPP is confined merely to high inflation countries as established by some studies (McNown and Wallace, 1989; Liu, 1992;
Mahdavi and Zhou, 1994).
Similarly, Holmes (2002) tested non-linearity in US $/ Latin
American RER and found that non-linearity existed for 7 out of 13
countries and concluded that the identification of non-linearity should
offer some explanation as to why PPP does not exist in many cases.
While another study using non-linear tests of stationarity and
cointegration for a sample of 10 Africans countries for the Post-
Bretton Woods era found that long run PPP held in 8 out of 10
countries, if an explicit distinction were made between positive and
negative shocks (Holmes and Wang, 2004). A more recent study also
validated the applicability of PPP hypothesis for East Caribbean
Currency Union by finding that many real exchange rates are
cointegrated over the period of 1980s and 1990s. There is also a
study, which concludes that nominal exchange rate was consistent
with PPP hypothesis and underlying behaviour of exchange rate was
consistent with fundamentals (Schweigert, 2002).
While the PPP condition constitutes one of the fundamental but
testable theoretical benchmarks against a set of other financial
market conditions such as interest rate differential that become
important in a world of dynamic cross border capital flows in the
determination of real exchange rate adjustments, the issue of
evaluating the fundamental equilibrium exchange rate requires
probe in to the working of fundamental economic factors such as
supply, demand and nominal factors, which govern the eventual
outcomes of the external sector account of any country. Against this
backdrop, a more recent study focused on the task of evaluating the
applicability of PPP in the Indian context, while also positing a
broader framework, incorporating fundamental economic factors to
estimate the equilibrium real exchange rate and to identify factors
that could have determined its movements during the post-reform
years (Joshi, 2007). The study explored that real exchange rate in
India is predominantly determined by permanent real demand shocks
followed by nominal and supply shocks.
Another study in the Indian context, by Kohli (2002), using unit
root and cointegration tests found mean reverting tendencies in the
real exchange rate series constructed using the consumer price index
as the deflator as well as for series constructed using wholesale and
consumer price indices, suggesting thereby that monetary policy
impulses were the main cause of disturbance in real exchange rate.
Further, the evidence of non stationarity of the relative differential of
tradable and non-tradable goods suggested that real shocks such as
permanent changes in productivity or Government spending were
important for the determination of real exchange rate movements.
Aggarwal (2000) found that prices, interest rates, and money
supply in the home and foreign country are important variables in
explaining the behaviour of bilateral exchange rate between India
and USA. Other significant variables are found to be balance of
payments items like current and capital account balances and foreign
exchange reserves. The exchange rate is found to be determined by
the domestic price index with a positive coEfficient and wholesale
price index in the USA. with a negative coEfficient. Domestic price is
explained by the domestic and foreign rates of interest and the money
supply in the home country. Thus, the interest rates and money
supply1 explain the exchange rate indirectly through the domestic
price. Interest rate in the home market is explained by the interest
rate in the foreign country and the exchange rate between the two
countries. That is, the interest rate differential between the two
countries also affects the exchange rate between them.
It is also well documented in the literature that, compared to
fixed exchange rate, a flexible exchange rate arrangement, under
normal circumstances leads to quicker convergence towards the
equilibrium because of faster self-stabilising adjustments in the
nominal exchange rate in tandem with changes in fundamentals as
compared with slower convergence through changes in relative price
ratios, which remain sticky because of market rigidities (Joshi, 2007).
Thus, the review of literature shows that the evidence in favour
of PPP is inconclusive, nevertheless the comprehensive research on
the subject, especially, in the case of developing countries is rather scarce. The theory of exchange rate determination still lacks models
that are both theoretically interesting and empirically defensible
(Krugman, 1993). Against this backdrop, purpose of the present
study is to address the question whether or not relative prices
determine relative exchange rate positions in bi-lateral framework
(India and US), without getting into the fundamental factors
underlying the exchange rate behaviour.
Recent empirical tests of PPP have mainly focused on the long
run given that there are frequent large and persistent short run
deviations from PPP. Earlier empirical results on PPP have not been
very encouraging , one of the reasons being the test used in many of
the earlier studies is known to have low power in small samples
(Hakkio (1986), DeJong, Nankervis, Savin and Whiteman (1992)).
One way to increase the power of the empirical tests is to use longer
span of data. For instance, Diebold, Husted and Rush (1991) and
Lothial and Taylor (1996) found support to PPP. In view of the above,
this study attempts to cover long span data to test the PPP hypothesis
in the Indian context.
Section III
External Sector Openness in India
III. A. Opening up of India’s External Sector
The external payment crisis that India witnessed in 1991 called
for wide-ranging external sector reforms. These included a marketbased
exchange rate system, introduction of convertibility of the rupee
for external transactions on the current account, and a compositional
shift in cross-border capital inflows from debt-creating to non-debtcreating
flows. FDI is encouraged through a very liberal but dual
route: a progressively expanding automatic route and a case-by-case
route. Indian companies were also permitted to access international
markets through Global Depository Receipts/American Depository
Receipts (GDRs/ADRs) under an automatic route, subject to specified
guidelines. Foreign investment in the form of Indian joint ventures
abroad was also permitted.
Restrictions on outflows involving Indian corporates, banks and
those who earn foreign exchange (like exporters) have also been
liberalised over time, subject to certain prudential guidelines. As a
result of pursuing the above approach, India has attracted considerable
private flows, primarily in the form of FDI, portfolio investment,
ECB and NRI deposits. Indian companies have been permitted to
raise resources from abroad through the issue of ADRs, GDRs,
Foreign Currency Convertible Bonds (FCCBs) and ECBs. Foreign
companies are also allowed to tap the domestic stock markets. FIIs
have been permitted to invest in all types of securities including
government securities. The Indian stock exchanges have been
allowed to set up trading terminals abroad. The trading platforms of
Indian exchanges are now accessed through the internet from
anywhere in the world. The Reserve Bank of India (RBI) permitted
two-way fungibility for ADRs/GDRs, which meant that investors
(foreign institutional or domestic) who hold ADRs/GDRs can cancel
them with the depository and sell the underlying shares in the market.
With the introduction of LERM and other liberalisation measures
both in the case of current and capital account, Indian economy has
witnessed increasing openness during the post reforms period. This
fact is amply reflected by measures of external openness such as
exports plus imports and capital inflows plus capital outflows as a
percentage of GDP, respectively. The current account, as measured
by the sum of current receipts and current payments, as a percentage
of GDP improved significantly from 19 per cent in 1990-91 to about
53 per cent of GDP in 2007-08. Likewise, the sum of gross capital
inflows and outflows increased from 12 per cent of GDP in 1990-91
to around 64 per cent in 2007-08.
III. B. Evolution of Exchange Rate Regime
The period after Independence in 1947 was followed by a fixed
exchange rate regime where the Indian rupee was pegged to the
pound sterling on account of historic links with Britain and this was
in line with the Bretton Woods System prevailing at that time. With
the breakdown of Bretton Woods system in the early 1970s and the consequent switch towards a system of managed exchange rates, and
with the declining share of the UK in India’s trade, the Indian rupee,
effective September 1975, was delinked from the pound sterling in
order to overcome the weaknesses of pegging to a single currency.
During the period of 1975 to 1992, the exchange rate of rupee was
officially determined by the Reserve Bank within a nominal band of
+/- 5 per cent of the weighted basket of currencies of India’s major
trading partners. The exchange rate regime of this period can be best
characterised as an adjustable nominal peg with a band, with the
nominal exchange rate being the operating variable to achieve the
intermediate target of a medium-term equilibrium path of the real
effective exchange rate (REER).
At the beginning of the 1990s, the significant rise in oil prices and
suspension of remittances from the Gulf region in the wake of the Gulf
crisis led to severe problems in the balance of payments in India. A twostep
downward adjustment of 18-19 per cent in the exchange rate of the
Indian rupee was made on July 1 and 3, 1991 with a view to placing it
at an appropriate level in line with the inflation differential with major
trading partners so as to maintain the competitiveness of exports. This
provided the necessary impetus for a move towards greater exchange
rate flexibility. Consequently, following the recommendations of the
High Level Committee on Balance of payments (Chairman: C.
Rangarajan), the Liberalised Exchange Rate Management System
(LERMS) involving dual exchange rate system was instituted in March
1992 in conjunction with other measures of liberalisation in the areas
of trade, industry and foreign investment. Under the LERMS, 40 per
cent of exchange earnings had to be surrendered at an official rate
determined by the Reserve Bank, which in turn was obliged to sell
foreign exchange only for import of certain essential commodities
such as oil, fertiliser and life saving drugs besides the Government’s
debt servicing. The balance 60 per cent of exchange earnings was to be
converted at rates determined by the market. The LERMS was
essentially a transitional mechanism and a downward adjustment in
the official exchange rate took place in early December 1992 and
ultimate convergence of the dual rates was made effective from March 1, 1993. The unification of the exchange rate of the Indian rupee was
an important step towards current account convertibility, which was
finally achieved in August 1994. The experience with the market
determined exchange rate system in India has remained satisfactory.
India’s exchange rate policy of focusing on managing volatility
with no fixed rate target while allowing the underlying demand and
supply conditions to determine the exchange rate movements over a
period in an orderly way has stood the test of time. The foreign
exchange market has been characterised by orderly conditions for
most of the period, excepting a few episodes of volatility. The Reserve
Bank continues to follow the same approach of watchfulness, caution
and flexibility in regard to foreign exchange market. It co-ordinates
its market operations carefully, particularly in regard to the foreign
exchange market with appropriate monetary, regulatory and other
measures as considered necessary from time to time. The conduct of
exchange rate policy in India is guided by three major objectives.
First, to maintain orderly conditions in the foreign exchange market
by providing foreign exchange as considered necessary from time to
time, and to prevent the emergence of destabilising and self-fulfilling
speculative activities. Second, to help in maintaining an adequate
level of foreign exchange reserves. Third, to help eliminate market
constraints with a view to facilitating the development of a healthy
foreign exchange market. International research on viable exchange
rate strategies in emerging markets has also lent considerable support
to the exchange rate policy followed by India (RBI, 2002-03).
Section IV
Model Specification, Methodology and Empirical Observations
IV. A. Model Specification
PPP model articulated first by the scholars of the Salamanca school
in the sixteenth century provides that once converted in common
currency, national price levels should be equal. Nominal exchange rate
responds to relative changes in the price levels of different countries so
that the underlying principle of one price will hold. However, this
adjustment in three variables takes time and the principle of one price holds only in long run. Alternatively, real exchange rate attains
equilibrium in the long-run, suggesting thereby that PPP condition does
not hold in the short run. The fundamental PPP specification is delineated
below.
The above specification can be estimated using simple OLS
process if the variables are stationary, otherwise, regression would be
spurious. In case these variables contain non-stationary process, then
cointegration framework put forward by Engle and Granger (EG) in
1987 and the other by Johansen in 1988 could be used for estimation.
IV. B. Methodology
According to the co-integration theory, there may be co-integrating
relationship between the variables involved if they are of the order of
I(1) i.e., they are stationary at the Ist difference level and their lineal combination is I(0). A battery of unit root tests is available to find out
whether the series is stationary or not. In this paper, Augmented Dickey-
Fuller (ADF), Dickey-Fuller GLS, Phillips-Perron, and Kwiatkowski-
Phillips-Schmidt-Shin (KPSS) tests were applied to test the unit root in
time series.
The importance of cointegration tests in the modeling of
nonstationary economic series becomes clear in the so-called
Granger representation theorem, first formulated in Granger and
Weiss (1983). Nonetheless, the necessary techniques for testing for
cointegration were developed jointly by Engle and Granger (1987)1.
In this paper, we have used Johansen Maximum Likelihood Method
for estimation.
The components of the specification, i.e., nominal exchange
rate, Indian price level and USA price level were tested for stationarity
and these variables were found to be having unit root. However, the
linear combination of the components of real exchange rate was
found stationary, indicating thereby that these components are
cointegrated. Therefore, PPP model was specified in terms of Vector
Error Correction Model (VECM). Under VECM model, when there
is disequilibrium, some of the variables must respond to restore
equilibrium. The error correction representation of a cointegrated
system regarding PPP is set out below :
Note that the dependent variables are stationary. Hence, the
equations are meaningful only if right hand side variables are stationary. If there is cointegration, the term within the parenthesis
(et-1) is stationary. Nothing gets altered, if we specify ECM in more
general form :
IV. C. Data
In this paper, we have used data from 1970M01 to 2009M03 on
nominal exchange rate of Indian rupee vis-à-vis the US Dollar, wholesale
price index (WPI) of India and producers price index (PPI) of the USA.
PPI index mainly consists of trading commodities of USA. In order to
ensure consistency, data on all these variables was taken from
International Financial Statistics (IFS) of International Monetary Fund
(IMF). For applying econometric tests all these variables viz., RER,
exchange rate, WPI of India and PPI of USA have been used in log form.
In the remainder of this chapter lexr is used for log of nominal exchange
rate, lindwp for log of prices of India, and luswp for prices of USA.
IV. D. Empirical Observations
Firstly, a battery of unit root tests are applied to test for stationarity
in RER during both periods I and period II. Table 1 shows that
Augmented Dickey-Fuller, DF-GLS, and Phillips Perron tests
accepted the null hypothesis of unit root in RER and KPSS rejected the null hypothesis of RER being stationary during period I (prereform
period). While, during period II (post-reforms period),
Augmented Dickey-Fuller, and Phillips Perron rejected the null
hypothesis of unit root and KPSS accepted the null hypothesis of
RER being stationary (Table 1). The unit root tests applied on RER
showed that the PPP holds during period II (post-reforms period) and
not in period I (pre-reforms period).
Furthermore, the components of RER (nominal exchange rate,
prices in India and USA) were tested for unit root. These variables
were found to be non-stationary series. Hence, they were tested for
unit root in first difference. As shown in Annex 1-3, Augmented
Dickey-Fuller, DF-GLS, and Phillips Perron tests rejected the null
hypothesis of unit root at 1 per cent level, while KPSS accepted the
null of stationarity at 1 per cent level during both period I and period
II. All these variables are found to be integrated of order I(1) process.
In the case of combination of these variables (components of RER),
i.e., error term, Augmented Dickey-Fuller, DF-GLS, and Phillips
Perron tests rejected the null hypothesis of unit root at 1 per cent
level and KPSS accept the null of stationary at 1 per cent level only
during period II (Annex 4-5) indicating that the components of RER are cointegrated only during period II and not in period I. It may be
mentioned that 6 lags have been selected for VECM based on
sequential modified LR test statistics.
Table 1 : Unit Root Tests for Real Exchange Rate of India Rupee and US Dollar (LRER) |
Unit Root/ Stationary Test |
t-Statistics |
1% |
5% |
10% |
Period I (Pre-reforms) |
ADF |
-2.365 |
-3.992 |
-3.426 |
-3.136 |
DF-GLS |
-1.684 |
-3.468 |
-2.915 |
-2.613 |
PP |
-0.211 |
-3.454 |
-2.872 |
-2.572 |
KPSS |
0.241 |
0.216 |
0.146 |
0.119 |
Period II (Post-Reforms) |
ADF |
-3.206* |
-4.006 |
-3.433 |
-3.140 |
DF-GLS |
-2.076 |
-3.468 |
-2.937 |
-2.647 |
PP |
-3.257** |
-3.464 |
-2.876 |
-2.575 |
KPSS |
0.267^^^ |
0.739 |
0.463 |
-0.347 |
Period I contains observations from 1970M01 to 1993M02 and Period II from 1993M03 to
2009M03. LRER is the log of the bilateral (US$) RER.
*, **, *** rejects null hypothesis of unit root at 10%, 5%, and 1% level of significance,
respectively.
^, ^^, ^^^ accepts null of stationary at 10%, 5%, and 1% level of significance. |
Table 2 : Trace Statistics and Maximum Eigenvalue Statistics for Rank of Cointegration |
Null |
Alternate |
Trace Test |
Eigenvalue Test |
Trace Statistics |
0.05 Critical Value |
Prob.** |
Max-Eigen Statistics |
0.05 Critical Value |
Prob.** |
R=0 |
R ≥1 |
40.73280 |
29.79707 |
0.0019 |
26.67643 |
21.13162 |
0.0075 |
R≤1 |
R ≥2 |
14.05637 |
15.49471 |
0.0814 |
12.25915 |
14.26460 |
0.1013 |
R≤2 |
R ≥3 |
1.797218 |
3.841466 |
0.1800 |
3.841466 |
3.841466 |
0.1800 |
Both trace and max-eigenvalue tests indicate 1 cointegrating eqn(s) at the 0.05 level of significance. |
Going forward, we have examined the order of integration of the
variables using Johansen’s Trace statistics and Eigen value statistics.
Both these statistics furnished in Table 2 indicate the existence of 1
cointegrating equation at 0.05 level of significance. In the light of
rank of cointegration being one, we have normalised the cointegrating
vector with respect to lexr and the estimated cointegrating relation.
The results of the long-run cointegration model are given in
Table 3. It may be noted that normalised cointegrating parameters
bear theoretically predicted sign and are found to be significant. The
coEfficients of both prices in India and USA show that there is more
than proportionate relationship with nominal exchange rate.
PPP theory, however, states that the coEfficients of both domestic
and foreign prices should be proportionate, i.e., these coEfficients
take the value 1. The significance of the difference between estimated
coEfficients of LPIN and LPUS was statistically tested and t-statistics
accepted the null hypothesis (Table 4). It shows that PPP theory holds in this case as coEfficients of both domestic and foreign prices are not
statistically different from 1.
Table 3: Estimation of Long Term Cointegration Model |
|
LEXR(-1) |
LPIN(-1) |
LPUS(-1) |
C |
CointEq1 |
1.000000 |
1.116445 |
-1.298991 |
4.629428 |
|
|
(0.12885) |
(0.24583) |
|
|
|
(8.66469] |
[ -5.86929] |
|
Table 4 : Coefficient Restrictions Test of
Long Term Cointegration Model |
|
LEXR over LPIN)
Null hypothesis, β1≠1 |
LEXR over LPUS
Null hypothesis, β2≠1 |
t Statistics |
0.903725 |
1.350944 |
Further, these three variables were investigated for direction of
causality using pair-wise Granger Causality. The time series
pertaining to LEXR, LPIN, and LPUS are characterised with I(1)
process. Since Granger causality test requires the series to be
stationary, this test has been run with their first difference. Table 5 presents the results of Granger causality tests. It has been found that
unidirectional causality runs from PUS (the prices in PUS) to EXR at
10 per cent level of significance. The Granger Causality running from
PIN to EXR has been found to be weak and statistically insignificant.
At the same time, unidirectional causality has been found running
from PUS to PIN at significance level of 1 per cent.
Going forward, short-term dynamics, i.e., ECM has been found
to be working in the cointegration model. The speed of adjustment
parameters (coEfficients) of nominal exchange rate and US prices
have been found greater than zero with negative sign and significant
as manifested by t test2. This confirms the Granger representation
theorem that error correction model for I(1) variables necessarily
implies cointegration. The significant coEfficients in Table 6 further manifest that the disequilibrium among the variables from their
long-run trend is corrected through the dynamic adjustment of
exchange rate and US prices but not by Indian prices in the shortrun.
Table 5: Results of Granger Causality Test |
Null Hypothesis: |
Obs. |
F-Statistic |
Probability |
Result |
DPIN does not Granger Cause DEXR |
187 |
1.35135 |
0.23699 |
Accept H0 |
DEXR does not Granger Cause DPIN |
1.92893 |
0.07863 |
Reject H0 |
DPUS does not Granger Cause DEXR |
187 |
1.33304 |
0.24484 |
Accept H0 |
DEXR does not Granger Cause DPUS |
4.72014 |
0.00018 |
Reject H0 |
DPUS does not Granger Cause DPIN |
187 |
5.94883 |
0.000015 |
Reject H0 |
DPIN does not Granger Cause DPUS |
2.10506 |
0.05496 |
Reject H0 |
Table 6 : Error Correction Estimates of Components of Real Exchange Rate |
Error Correction : |
D(LEXR) |
D(LINDWP) |
D(LUSWP) |
CointEq1 |
-0.082889 |
-0.012722 |
-0.041028 |
(0.03299) |
(0.01123) |
(0.01712) |
[ -2.51248] |
[ -1.133061] |
[-2.39611] |
When there is positive deviation in the exchange rate from its
long run equilibrium level, both nominal exchange rate and US prices
respond negatively (represented by a negative adjustment
coEfficients). On the other hand, prices in India also respond
negatively to deviations in nominal exchange rate (represented by
negative adjustment coEfficient) but the coEfficient is not statistically
significant. This implies that disequilibrium in the real exchange rate
is being corrected by deviations in nominal exchange rate and US
prices in the short run.
Section V
Concluding Remarks
Exchange rate models in the literature assume that PPP holds in
the open market economies. The PPP condition not only helps in
understanding the nature of nominal and real shocks in the exchange
rate models but also help policy makers and researchers to compute
RER misalignment. Theoretical and empirical literature suggests that
the PPP can be violated in the short run. PPP may also not hold on
account of factors, viz., (i) if there are transaction costs and trade
frictions, (ii) differential baskets of goods are used to construct
aggregate price indices, and (iii) government intervenes in foreign
exchange rate markets.
The main objective of this paper is to examine the validity of
PPP condition in the case of India during pre and post reform periods
using monthly data from 1970 to 2009. The results of battery of tests
reveal that RER is non-stationary during the pre-reform period and
stationary in the post-reforms era, which underscores the fact that
RER condition holds during the post reforms period. This is validated
by the result of unit root tests wherein the components of RER
(exchange rate, prices of India and USA) have been found cointegrated
as the linear combination of RER components turned out to be
stationary during the post-reforms period and not during the prereforms
period.
Further, PPP condition has been estimated using VECM. The
results imply that long-run relationship between exchange rate and
prices of India and USA is more than proportionate but difference is
statistically insignificant. However, ECM is found to be working
with all the coEfficients having negative sign but statistically
significant only in the case of LEXR and LPUS. The ECM coEfficients
are very small, suggesting that these variables take quite a long time
in correcting the deviations in the long run equilibrium exchange
rate.
In sum, RER holds in the post-reforms period and not in the Prereforms
period, which is validated by the cointegration tests and
vector error correction model employed in this case. Thus, PPP
condition holds in the long run during the post-reforms period,
implying that any deviations in the equilibrium exchange rate remain
transitory and revert back to the underlying trend eventually. These
findings underscore and support the fundamental principle that
chances of PPP improve in an economy with its increased external
sector openness. Holding of PPP during post-reforms period is also
consistent with liberalised exchange rate mechanism (LERM)
introduced since March 1993, where exchange rate is determined by
demand and supply.
NOTES :
* The authors are working in the Department of Economic Analysis and Policy,
Reserve Bank of India. The views expressed in the paper are strictly personal. Errors
and omissions, if any, are the sole responsibility of the authors.
1 Engle and Granger framework is a two-step procedure wherein null
hypothesis of no cointegration is tested between a set of I(1) variables
applying unit root tests to the residuals. EG two-step procedure first applies
Ordinary Least Squares (OLS) estimation method to the variables, and then
tests the stationarity of the residual obtained from the regression equation. If
the residual is stationary, then there is a cointegrating relationship between
the variables. Another seminal work in the evolution of cointegration
techniques is by Johansen (1988, 1991). Johansen Maximum Likelihood
Method based on Vector Auto Regression (VAR) known as Vector Error
Correction Mechanism (VECM) deals with more than two variables.
2 In case of one cointegration relationship the significance of speed of
adjustment parameters can be determined by t statistics.
Select Reference :
Ahking. F. (2003), “Efficient Unit Root Tests of Real Exchange Rates in the
Post-Britton Woods Era”, University of Connecticut, Economics Bulletin.
Ahmad B., et al, (2004): “Re-examining Purchasing Power Parity for East-
Asian Currencies: 1976-2002” MPRA Paper No. 2025, (posted 07.
November 2007)
Abuaf, N and Jorion, P. (1990): “Purchasing Power Parity in the Long
Run,” Journal of Finance, 45(1 ), pp. 157-74, march.
Agarwal R.N. (2000) “Exchange Rate Determination in India Endogenising
Foreign Capital Flows and Some Entities of Monetary Sector, Discussion
Paper, Institute of Economic Growth, Delhi.
Bhagwati J. (1984): “Why Are Services Cheaper in Poor Countries?”
Economic. Journal, 94(374), pp. 279-86, June.
Barumshah et. al. (2004), ‘Re-examining Purchasing Power Parity for East-
Asian Currencies: 1976-2002”, University Putra Malaysia, MPRA.
Breuer, J (1994):”An Assessment of the Evidence on Purchasing Power
Parity,” in JOHN WILLIAMSON ed., pp. 245-77.
Balassa, B. (1964): “The Purchasing Power Parity Doctrine: A Reappraisal,”
J Pofit Econ., 72(6), pp. 584-96, December.
Dickey, D and Fuller W. (1979): “Distribution of the Estimators for
Autoregressive Time Series with a Unit Root,” Journal of American
Statistical Association, 74(366), pp. 427-31, June.
Diebold, Francis X.; Husted, Steve and Rush, Mark (1991): “Real Exchange
Rates under the Gold Standard,” Journal of Political Economy, 99(6),
1252-71, December.
DeJong, David N., Nankervis, John C., Savin, N. E., Whiteman, Charles
(1992) “Integration versus trend stationarity in time series” Econometrica
60, 423-433.
Frankel J. (1986): “International Capital Mobility and Crowding-out in the
U.S. Economy: Imperfect Integration of Financial Markets or Goods
Markets’ in How open is the U.S. economy?” Ed.: RIK W. HAFER,
Lexington Books, pp. 33-67.
Goenner C. F. and Sean. M. S. (2002), “Convergence to the Law of One
Price on the Internet”, University of North Dakota, University of the Pacific.
Glen, J. D. (1992):”Real Exchange Rates in the Short, Medium, and Long
Run,” Journal of International Economics, 33(1/2), pp. 147-66, August.
Hakkio, Craig S. (1986): “Does the exchange rate follow a random walk? A
Monte Carlo study of four tests for a random walk” Journal of International
Money and Finance 5, 221-229.
Joshi. H. (2007), “The Fundamental Equilibrium Real Exchange Rate in
India : An Approach to Estimation and Measurement of Misalignment”,
RBI, Occasional Papers, Vol. 27, No. 3.
Kohli.R. (2002), “Real Exchange Rate Stationarity in Managed Floats:
Evidence from India”, Indian Council for Research on International
Economic Relations, Working Paper No. 93
Lothian, J.R. and Taylor M.P (1996): “Real Exchange Rate Behaviour: The
Recent Float from the Perspective of the Past Two Centuries”, Journal of
Political Economy, 104, 488- 509.
McNown, R. and Wallace M. (1989): “National Price Level, Purchasing
Power Parity, and Cointegration: A Test of Four High Inflation Economies”,
Journal of International Money and Finance, 8, 533-45.
Nejib. H. (2008), “The Purchasing Power Parity and the Symmetry,
Proportionality Conditions: Panel Cointegration Evidence from Some
African Countries”, International Research Journal of Finance and
economics, ISSN, Issue 16.
Officer, L (1976): “The Purchasing-power-Parity Theory of Exchange
Rates: A Review Article,” IMF Staff Papers, 1, 23(l) ,pp. 1-60, March
Rogoff K. (1996): “The Purchasing Power Parity Puzzle”, Journal of
Economic Literature, Vol. 34, No. 2, pp. 647-668, June.
Annex 1 : Unit-Root Test for Bilateral Real Exchange Rate of Indian Rupee with US$ (LRER) |
Unit Root Test |
Period II(0) |
Period II I(0) |
Augmented Dickey-Fuller |
t-Statistics |
-2.364692 |
-3.206252* |
Critical Values % level |
-3.991656 |
-4.006311 |
5% level |
-3.426191 |
-3.433278 |
10% level |
-3.136301 |
-3.140478 |
DF-GLS |
t-Statistics |
-1.684368 |
-2.076163 |
Critical Values 1% level |
-3.467600 |
-3.468400 |
5% level |
-2.914800 |
-2.937000 |
10% level |
-2.613400 |
-2.647000 |
Phillips-Perron (Newey-West Using Bartlett Kernel) |
Adj. t-Statistics |
-0.210197 |
-3.257362*** |
Critical Values 1% level |
-3.453823 |
-3.464280 |
5% level |
-2.871768 |
-2.876356 |
10% level |
-2.572293 |
-2.574746 |
KPSS |
LM -Statistics |
0.241196 |
0.266712^^^ |
Critical Values 1% level |
0.216000 |
0.739000 |
5% level |
0.146000 |
0.463000 |
10% level |
0.119000 |
0.347000 |
Period I contains observations from 1970M01 to 1993M02 and Period II from 1993M03 to 2009M03. LRER is the log of the bilateral (US$) RER.
*, **, *** rejects null hypothesis of unit root at 10%, 5%, and 1% level of significance, respectively.
^, ^^, ^^^ accepts null of stationary at 10%, 5%, and 1% level of significance. |
Annex 2 : Unit-Root Test for Nominal Bilateral Nominal Exchange Rate of Indian Rupee with US$ (LEXR) |
Unit Root Test |
Period I |
Period II |
I(0) |
I(1) |
I(0) |
I(1) |
Augmented Dickey-Fuller |
t-Statistics |
1.857250 |
-12.31178*** |
-2.514754 |
-11.42035 *** |
Critical Values 1% level |
-3.453910 |
-3.453910 |
-3.464280 |
-3.464280 |
5% level |
-2.871806 |
-2.871806 |
-2.876356 |
-2.876356 |
10% level |
-2.572313 |
-2.572313 |
-2.574746 |
-2.574746 |
DF-GLS |
t-Statistics |
1.114642 |
-11.59819*** |
0.860337 |
-11.37020*** |
Critical Values 1% level |
-2.573367 |
-2.573367 |
-2.576999 |
-2.576999 |
5% level |
-1.941978 |
-1.941978 |
-1.942482 |
-1.942482 |
10% level |
-1.615931 |
-1.615931 |
-1.615606 |
-1.615606 |
Phillips-Perron (Newey-West Using Bartlett Kernel) |
Adj. t-Statistics |
1.997638 |
-12.54128*** |
-2.511417 |
-11.42212*** |
Critical Values 1% level |
-3.453910 |
-3.453910 |
-3.464280 |
-3.464280 |
5% level |
-2.871806 |
-2.871806 |
-2.876356 |
-2.876356 |
10% level |
-2.572313 |
-2.572313 |
-2.574746 |
-2.574746 |
KPSS |
LM -Statistics |
1.665699 |
0.605880^^^ |
1.184326 |
0.276564^^^ |
Critical Values 1% level |
0.739000 |
0.739000 |
0.739000 |
0.739000 |
5% level |
0.463000 |
0.463000 |
0.463000 |
0.463000 |
10% level |
0.347000 |
0.347000 |
0.347000 |
0.347000 |
Period I contains observations from 1970M01 to 1993M02 and Period II from 1993M03 to 2009M03. LEXR is the log of the bilateral (US$) nominal exchange rate.
*, **, *** rejects null hypothesis of unit root at 10%, 5%, and 1% level of significance, respectively.
^, ^^, ^^^ accepts null of stationary at 10%, 5%, and 1% level of significance. |
Annex 3 : Unit-Root Test for Indian Price Level (LPIN) |
Unit Root Test |
Period I |
Period II |
I(0) |
I(1) |
I(0) |
I(1) |
Augmented Dickey-Fuller |
t-Statistics |
-0.617532 |
-10.15714*** |
-1.776588 |
-9.781614*** |
Critical Values 1% level |
-3.453910 |
-3.453910 |
-3.464280 |
-3.464280 |
5% level |
-2.871806 |
-2.871806 |
-2.876356 |
-2.876356 |
10% level |
-2.572313 |
-2.572313 |
-2.574746 |
-2.574746 |
DF-GLS |
t-Statistics |
1.114642 |
-6.535593*** |
0.873862 |
-7.652702*** |
Critical Values 1% level |
-2.573398 |
-2.573398 |
-2.576999 |
-2.576999 |
5% level |
-1.941982 |
-1.941982 |
-1.942482 |
-1.942482 |
10% level |
-1.615929 |
-1.615929 |
-1.615929 |
-1.615929 |
Phillips-Perron (Newey-West Using Bartlett Kernel) |
Adj. t-Statistics |
-0.548653 |
-10.25036*** |
-1.898202 |
-9.781614*** |
Critical Values 1% level |
-3.453910 |
-3.453910 |
-3.464280 |
-3.464280 |
5% level |
-2.871806 |
-2.871806 |
-2.876356 |
-2.876356 |
10% level |
-2.572313 |
-2.572313 |
-2.574746 |
-2.574746 |
KPSS |
LM -Statistics |
1.924996 |
0.061479^^^ |
1.699344 |
0.225834^^^ |
Critical Values 1% level |
0.739000 |
0.739000 |
0.739000 |
0.739000 |
5% level |
0.463000 |
0.463000 |
0.463000 |
0.463000 |
10% level |
0.347000 |
0.347000 |
0.347000 |
0.347000 |
Period I contains observations from 1970M01 to 1993M02 and Period II from 1993M03 to 2009M03. LPIN is the log of the Indian WPI.
*, **, *** rejects null hypothesis of unit root at 10%,5%, and 1% level of significance, respectively.
^, ^^, ^^^ accepts null of stationary at 10%,5%, and 1% level of significance. |
Annex 4 : Unit-Root Test for US Price Level (LPUS) |
Unit Root Test |
Period I |
Period II |
I(0) |
I(1) |
I(0) |
I(1) |
Augmented Dickey-Fuller |
t-Statistics |
-2.702267 |
-6.038196*** |
-0.716393 |
-9.282352*** |
Critical Values 1% level |
-3.454085 |
-3.454085 |
-3.464280 |
-3.464280 |
5% level |
-2.871883 |
-2.871883 |
-2.876356 |
-2.876356 |
10% level |
-2.572354 |
-2.572354 |
-2.574746 |
-2.574746 |
DF-GLS |
t-Statistics |
0.982306 |
-6.018803*** |
0.392034 |
-9.257905*** |
Critical Values 1% level |
-2.573429 |
-2.573429 |
-2.576999 |
-2.576999 |
5% level |
-1.941986 |
-1.941986 |
-1.942482 |
-1.942482 |
10% level |
-1.615926 |
-1.615926 |
-1.615929 |
-1.615929 |
Phillips-Perron (Newey-West Using Bartlett Kernel) |
Adj. t-Statistics |
-2.490799 |
-14.00265*** |
-0.652394 |
-9.383623*** |
Critical Values 1% level |
-3.453910 |
-3.453910 |
-3.464280 |
-3.464280 |
5% level |
-2.871806 |
-2.871806 |
-2.876356 |
-2.876356 |
10% level |
-2.572313 |
-2.572313 |
-2.574746 |
-2.574746 |
KPSS |
LM -Statistics |
1.782120 |
0.783827 ^^^ |
1.462365 |
0.078552 ^^^ |
Critical Values 1% level |
0.739000 |
0.739000 |
0.739000 |
0.739000 |
5% level |
0.463000 |
0.463000 |
0.463000 |
0.463000 |
10% level |
0.347000 |
0.347000 |
0.347000 |
0.347000 |
Period I contains observations from 1970M01 to 1993M02 and Period II from 1993M03 to 2009M03. LPUS is the log of the US prices.
*, **, *** rejects null hypothesis of unit root at 10%,5%, and 1% level of significance, respectively.
^, ^^, ^^^ accepts null of stationary at 10%,5%, and 1% level of significance. |
Annex 5 : Unit-Root Test for Linear Combination (Error Term) of Nominal Bilateral Exchange Rate (LEXR), Price Level in India (LPIN) and Price Level in USA (LPUS) |
Unit Root Test |
Period I |
Period II |
I(0) |
I(0) |
Augmented Dickey-Fuller |
t-Statistics |
-1.419091 |
-2.933059** |
Critical Values 1% level |
-3.453910 |
-3.464280 |
5% level |
-2.871806 |
-2.876356 |
10% level |
-2.572313 |
-2.574746 |
DF-GLS |
t-Statistics |
-0.621099 |
-1.035102 |
Critical Values 1% level |
-2.573367 |
-2.576999 |
5% level |
-1.941978 |
-1.942482 |
10% level |
-1.615931 |
-1.615929 |
Phillips-Perron (Newey-West Using Bartlett Kernel) |
Adj. t-Statistics |
-1.475948 |
-2.576981** |
Critical Values 1% level |
-3.453823 |
-3.464280 |
5% level |
-2.871768 |
-2.876356 |
10% level |
-2.572293 |
-2.574746 |
KPSS |
LM -Statistics |
0.534756^^^ |
0.562622^^^ |
Critical Values 1% level |
0.739000 |
0.739000 |
5% level |
0.463000 |
0.463000 |
10% level |
0.347000 |
0.347000 |
Period I contains observations from 1970M01 to 1993M02 and Period II from 1993M03 to 2009M03.
*, **, *** rejects null hypothesis of unit root at 10%, 5%, and 1% level of significance, respectively.
^, ^^, ^^^ accepts null of stationary at 10%, 5%, and 1% level of significance. |
|