IN transition economies, as elsewhere, the choice of an exchange rate regime is controversial.1 Table 1 shows that all possible arrangements have been tried in the various countries in the few years before transition began. If the most popular arrangement is a managed float, the tendency has been toward increased exchange rate stability, especially as inflation declined.


IN transition economies, as elsewhere, the choice of an exchange rate regime is controversial.1 Table 1 shows that all possible arrangements have been tried in the various countries in the few years before transition began. If the most popular arrangement is a managed float, the tendency has been toward increased exchange rate stability, especially as inflation declined.

IN transition economies, as elsewhere, the choice of an exchange rate regime is controversial.1 Table 1 shows that all possible arrangements have been tried in the various countries in the few years before transition began. If the most popular arrangement is a managed float, the tendency has been toward increased exchange rate stability, especially as inflation declined.

This paper presents estimates of the equilibrium paths of the real exchange rate for countries for which sufficient data have been available (Croatia, the Czech Republic, Hungary, Poland, Russia, the Slovak Republic, and Slovenia, with partial results available for other countries). It represents an early attempt at drawing systematic lessons from the short experience accumulated so far. The usual signals for determining a benchmark exchange rate are missing: no track record of competition within an open trading system; a brief history of current accounts responding to exceptionally sharp short-term movements in production, consumption, and investment; quickly changing labor and other costs; and the redeployment of the entire structure of the economy.

Table 1.

Exchange Rate Regimes Since the Beginning of Transition

article image

The Czech Republic is considered as a continuation of Czechoslovakia.

Depreciations occurred in March 1989 (5 percent), May 1989 (6 percent), December 1989 (10 percent), February 1990 (5 percent), January 1991 (15 percent), November 1991 (5.8 percent) and then in more frequent smaller installments (three times and a total of 5.5 percent in 1992; 15 percent in five times over 1993; and 16.8 percent in seven times in 1994).

One devaluation (16.8 percent) in May 1992.

One devaluation (10 percent) in July 1993.

In challenging times, simple rules become naturally appealing and it is perhaps unavoidable that purchasing power parity (PPP) be invoked at this stage. Its logic is easily understood by policymakers and by the population at large given the widespread use of foreign currencies.2 Unfortunately, in this case more than ever, PPP is the wrong rule to follow. It is well known that PPP only holds under very restrictive conditions. These conditions may be approximately verified among countries that undergo broadly similar shocks3 but they are unlikely to be even remotely met in transforming countries. Yet, rapid real exchange rate changes—real appreciation of 10 to 20 percent a year—are quickly branded as a threat to competitiveness. We argue to the contrary that these exchange rate appreciations usually correspond to a return to a continuously appreciating equilibrium level.

So far there has been very limited effort to deal with questions such as, what is the “proper” level of the exchange rate, and what is its “normal” evolution over time? The answers have been mostly implicit, not based on the kind of formal and data-based analysis that is common practice for more economically stable countries. There are good reasons for this state of affairs: reliable quantitative information is not available or of poor quality, and covers a very short time period. It is also important that no theory of the exchange rate in transition has yet been worked out. These difficulties are obvious in the present paper and imply that the results presented here are not definitive. Our goal, rather, is to contribute to a debate that has been lively but, so far, mostly based on “soft” treatment of the data.

I. A Stylized Fact

Real Exchange Rate Behavior

Empirical studies of the exchange rate face two familiar hurdles. First, nominal shocks temporarily affect the real exchange rate and move it away from equilibrium. Second, the real equilibrium exchange rate itself may change in response to real shocks. Most transition economies have undergone massive nominal shocks; transformation itself proceeds through a succession of large real shocks. Our approach is first to identify the path of the real equilibrium exchange rate (REER) and then to explain the evolution of the actual real exchange rate about its equilibrium path.

Figure 1 presents the evolution of three measures of the real exchange rates for a number of transition economies. The first measure is the CPI-deflated real exchange rate (Λ = P/EP*, where E is the exchange rate defined as the domestic price of foreign currency, and P and P* are the domestic and foreign consumer price indices (CPIs), respectively). The nominal exchange rate is the geometric average of the U.S. and German exchange rates. Similarly, the foreign CPI is the geometric average of the U.S. and German CPIs, in both cases using equal weights. The second measure is the ratio of nontraded to traded good prices (M = PN/PT). It is approximated as the ratio of the CPI to the industrial price index, with the same weighting scheme. Our third measure is the dollar wage (Ω = W/E). The dollar wage is a frequently used measure of the real exchange rate4 because it avoids the “index problem,” the fact that price indices are not comparable across countries. In this figure and in the rest of the paper, given our definition of the exchange rate, an increase represents an appreciation.

The CPI-based real exchange rate (thereafter referred to as the real exchange rate) exhibits strikingly similar features across all countries, irrespective of the exchange regime. It appears to undergo an initial massive depreciation (where available data exist). Eventually, once the macroeconomic situation is brought under control, it follows a trend of appreciation. The total rate of real appreciation from the trough ranges from about 30 percent in Hungary to 1,000 percent or more in Russia. Albania, Romania, and Ukraine illustrate cases of setbacks along the way, with relapses into inflation and in the setting up of market mechanism. Slovenia emerges as an outlier with a stable real exchange rate.

The relative price of nontraded and traded goods generally follows the same pattern although the trend rate of appreciation is lower than for the real exchange rate. The exceptions are Russia, where the ratio actually depreciates, and Hungary and Slovenia, where the trend rate of appreciation is larger than for the real exchange rate. The dollar wage generally exhibits the same behavior as the real exchange rate, including in the special case of Slovenia once the dollar wage rebounded from its January 1992 collapse.5

Figure 1.
Figure 1.
Figure 1.

Real Exchange Rates

(average vis-à-vis the U.S. dollar and the deutsche mark)

Citation: IMF Staff Papers 1997, 004; 10.5089/9781451930962.024.A002

Sources: IMF; and Russian-European Center for Economic Policy.

The observed similarity across real exchange rate measures and across countries is striking and cannot be just a fluke. It suggests the stylized fact proposed in Figure 2. The figure presents our hypothesis. The actual real exchange rate initially depreciates and overshoots its equilibrium path so that there is at first a sizable undervaluation. Over time the real exchange rate appreciates for two reasons. First, the initial undervaluation is gradually corrected. Second, the REER itself appreciates as a result of the transformation process. The rate of equilibrium appreciation is higher the more complete the market system is and the faster capital is accumulated.

Initial Undervaluation

Undervaluation occurs when markets are liberalized. It frequently results from the abrupt depreciation that accompanies the end of a command economy and the dismantling of previously prevailing multiple exchange rates. Three explanations are possible.

First, a pent-up demand for foreign assets (previously reflected in the black market premium) faces a negligible supply. Since the available net stock of foreign assets cannot increase instantaneously in volume, market equilibrium requires a price increase. The undervalued exchange rate, in turn, allows for the net acquisition of foreign assets through current account surpluses. Over time, correction from undervaluation leads to the closing of the current surplus as the domestic stock of foreign assets rises to meet demand at the equilibrium exchange rate.

Second, the freeing of prices in the presence of a monetary overhang is met by a sudden burst of inflation. The associated flight from domestic currency further exacerbates the demand for foreign assets, mostly foreign cash. This disequilibrium situation also results in undervaluation.

Third, the return to some degree of convertibility raises a difficult policy issue for untested authorities lacking credibility. This is further aggravated by the ignorance of what is the appropriate—equilibrium—level of the exchange rate (Berg and Sachs, 1992). In such a situation, the authorities must decide whether to err on the side of over- or undervaluation. Prudence suggests taking the risk of an undervaluation rather than taking the risk of being unable to sustain convertibility.

Figure 2.

Process of Real Appreciation

Following the initial drop, the real exchange rate subsequently appreciates for two reasons. First is the correction of the initial undervaluation. Second, the REER itself embarks on a path of trend appreciation likely to last for several years. This prediction is based on the following six reasons.

First, rapid productivity gains are expected when formerly inefficient economies respond to market forces. Firms that used to maximize output and/or employment now shift toward profit maximization. The resulting deep overhaul of the economy includes an end to overmanning and the closure of activities that are not profitable at world prices. The visible outcome is a dramatic reduction in the size of industry and agriculture and the development of the service industry. The mere emergence of services (banking and finance, management consulting, marketing, etc.) is likely to raise aggregate effectiveness considerably. Once income is rising again, demand for nontradables increases and results in real appreciation.6

Second, if productivity gains are higher in the traded than in the nontraded good sector, the REER appreciates as predicted by Balassa (1964) and Samuelson (1964). This pattern of “normal” development may be seen as contradictory to the rapid emergence of a service sector, as just noted above. They are not mutually exclusive and may occur together or sequentially.

Third, as argued in Coorey, Mecagni, and Offerdal (1996), transition economies inherit a set of natural resource prices considerably below world prices. Similarly, public utility prices also used to be set low and governments are worried about upsetting unstable public support by raising these prices. This leads to low nontraded good prices. The situation is not sustainable, though. When these prices are raised to match production costs, the real exchange rate appreciates.

Fourth, public spending changes in its structure but does not necessarily decline in the aggregate, and an appropriate welfare system has to be built. As the private sector becomes more productive, social returns from public investment (infrastructure and environment, for example) rise and warrant higher public spending. Tax income must, at the very least, be maintained. Yet firms can no longer provide most of government revenue: the very high rates of corporate taxation inherited from central planning become highly inefficient and receipts quickly shrink.7 Where tax reform is implemented, personal income taxation and VAT become the dominant source of fiscal income. Such a deep overhaul is bound to affect a wide array of relative prices. The exact effect on the real exchange rate is ambiguous, though. Still, nonmonetary financing of public deficits is likely to lead to real appreciation via high real interest rates.

Fifth, very high potential returns on capital justify accumulation at a rate that exceeds domestic savings. Foreign investment tends to produce a real exchange rate appreciation best understood as a permanent or equilibrium change. Given the timing of effects, capital inflows and the attendant real appreciation occur before capital is put in place and productivity rises.

Finally, locally produced traded goods were initially of poor quality and poorly marketed. As firms learn to operate on world markets, the terms of trade are likely to improve.

II. Real Equilibrium Exchange Rate: A Framework

At a very general level, the exchange rate is at its equilibrium level when the economy is simultaneously in internal (output, employment, inflation) and external (current and capital account) equilibrium. Internal equilibrium can be said to occur when the nontradable good market clears in the current period and is expected to do so in the future (see, e.g., Edwards, 1989), while external equilibrium occurs when current account balances are compatible with sustainable capital flows (see, e.g., Williamson, 1985).

For the practical purpose of determining the REER, three main difficulties arise. First, there is no unique definition of the real exchange rate (Figure 1 proposes three different relative prices). Second, the REER matters so much because it interacts with a wide number of important variables: competitiveness, the allocation of resources across industries, the pattern of spending, and intertemporal transfers of incomes through current account imbalances. Data for many of these variables are not available for the transition economies, or exist only for very short samples. Third, and to make matters worse, these variables are better seen as jointly endogenous with the real exchange rate itself. As a consequence, the list of variables of potential interest, and the precise effect of each of them on the RBER, depends on a particular model. For example, a real appreciation may be explained by a loss of competitiveness if domestic costs and prices exogenously rise faster than the exchange rate. But it might just as well be the endogenous response to an improvement in competitiveness, such as an exogenous increase in world demand for domestic output.8

In the absence of an adequate model of the exchange rate during transition, we proceed by clarifying the links between our three measures of the real exchange rate and by providing a framework that accounts for the sources of changes in the REER. Using lowercase letters to represent logs (e.g., λ = lnΛ), we have


where p and p* are, respectively, the domestic and foreign consumer price indices; w and w*, the nominal wages at home and abroad: pN and pT, the price indices for nontraded and traded goods; and e, the nominal exchange rate defined as the foreign price of domestic currency. In the presence of price and wage stickiness, all these measures are likely to be affected by both nominal and real disturbances. Overlooking for now nominal disturbances, we concentrate on the sources of fluctuations of the REER.

Let the CPI be p = γpN + (1 − γ) pT. Initially at least, domestically produced traded goods are of poor quality and poorly marketed. As they sell at a discount on world markets, the law of one price does not apply. We assume that


where κ can be thought of as a measure of “quality,” presumably rising over time.9 “Quality” must be understood in a wide sense: it includes market power and the ability to differentiate domestic products on both domestic and foreign markets. Initially, we assume κ < 0. Then we obtain


Optimizing firms set the real wage equal to the marginal productivity of labor. As this may not be a realistic description of the early phase of transition, we allow for a more general formulation:


where ρi is a measure of excess wages and ai is (the log of) marginal productivity of labor in sector i (I = T, N).

Finally, we allow for wages to differ, possibly temporarily, across sectors:


With these notations, we get


The four terms in equation (8b) illustrate four reasons why we might observe a real exchange rate appreciation:

  • 1. Domestic producers of traded goods improve the quality κ of their products. This effect results in an improvement in the terms of trade.

  • 2. Wages in the nontraded good sectors rise relative to wages in the traded good sector. Initially, wages in the fast growing and often informal nontraded good sector (new services, imports of foreign goods) are lower than those in the traditional (industry-based) traded good sector. Over time, the gap should be closed under the combined pressure of trade union activity and competition for expert manpower in the previously neglected service industry.

  • 3. Initially, wages far exceed productivity in the traded good sector (ρT > 0), while the margin is nil, or even negative, in the nontraded good sector (ρN ≤ 0). The subsequent correction of these imbalances leads to a real appreciation as prices in nontraded goods absorb the rise in costs.

  • 4. The Balassa-Samuelson effect predicts a real exchange appreciation when productivity in the traded good sector increases faster than in the nontraded good sector.



where a and a* are, respectively, the aggregate marginal productivity of labor at home and abroad, and ρ is the aggregate excess of wages over labor productivity at home (it is assumed that p* = 0).10 The dollar wage is equal to 3 if we assume that the foreign wage is constant and normalize it to unity (so that w* = 0). An important implication of equation (9) is that ω and the dollar wage are functions of aggregate productivity, in contrast with price-based real exchange rates, which relate to sectoral productivity differentials.

III. Equilibrium Dollar Wages

Methodology and Data

This section provides estimates of the REER. Usually, the real exchange rate is found by determining a benchmark period during which the actual rate is believed to be in equilibrium. One may then track down the evolution of the equilibrium rate over time. The main advantage of this method is that it does not require international comparisons of price levels: easily available national price or wage indices can be used. In the case of transforming countries, this method cannot be used because of the absence of relevant history and because exchange rates and prices were not governed by market forces before 1989.

Our approach is based on international wage comparisons. We do not attempt price comparisons, even though some are available from the Heston-Summers (1991) “International Comparison Program” (ICP), first, because the dollar wage is closely monitored in all transition countries and is familiar to analysts and policymakers, and second, because it is available without delay at a monthly frequency. The ICP measures are unsuitable for policy purposes since they are only available at a five-year frequency and with a considerable lag. Yet, dollar wages suffer from a number of defects, in particular the fact that the definition of labor costs varies widely from one country to another. This is particularly so in transforming countries, where direct labor costs used to be only a portion of total costs given the “social function” of firms.11 These caveats must be kept in mind when assessing our results.

In order to search for estimates of the equilibrium dollar wage, we formalize the popular rule of thumb that consists of comparing a transforming country’s dollar wage to the dollar wage in relevant countries, where relevance is defined as broad similarity in terms of the stage of development. We make this process systematic by bringing to bear all the countries of the world for which we can assemble adequate data.

The choice of data is guided by equations (8) and (9). Attention is directed to (1) a number of indicators of economic effectiveness (“quality,” κ the gap between sectoral wages, θ; and aggregate excess wage, ρ); (2) aggregate productivity, a; and (3) differences in productivity and effectiveness across sectors (aT − aN and ρT − ρN). Sectoral productivity and effectiveness data are not available for most countries. Aggregate (average) productivity is trivially measured as GDP per worker. The choice of proxies for the other terms follows from the large recent literature on growth catch-up (see, e.g., Barro and Sala-i-Martin, 1995). Human capital (usually proxied by education, i.e., investment in human capital), the size of government, and the size of the agriculture sector have been found to be convincing explanatory variables of economic effectiveness.

The dollar wage is estimated pooling time series concerning 80 countries from all five continents listed in the Appendix. For each country we include, whenever available, five observations taken five years apart (1970, 1975, 1980, 1985, and 1990). This allows us to exploit the long-run growth effects on the dollar wage while keeping data collection manageable. The measure of aggregate (average) productivity, GDP per worker, is provided by the ICP database updated on the Internet:12 this source offers comparable PPP-adjusted measures of GDP.13 For the dollar wage we use wages and average hours of work from the ILO Statistical Yearbook (various issues) and convert into average monthly wages.14 The conversion to U.S. dollars is done using average exchange rates from the IMF’s International Financial Statistics. Education is measured as the proportion of the population of school age in secondary schools from the World Bank’s Social indicators of Development (1995). The same source is used for the ratio of agriculture to industrial output and the share of government spending in GDP. The number of observations (country and date coverage) is determined by data availability and is presented with the results.

The wide fluctuations of the dollar during the sample period have the effect of introducing spurious fluctuations in dollar wages. For example, at the height of the dollar’s value in 1985, dollar wages were much lower than in 1980 or 1990. For this reason we use year-specific dummy variables. In addition, to allow for a secular trend in dollar wages worldwide, we introduce a linear trend (all variables are estimated in log form).

Pooling data over a large number of countries assumes that the same process drives the dollar wage worldwide. To test for this assumption we have extensively searched for regional effects. We look for both fixed and variable effects using dummies for the following country groupings: OECD, Africa, Southeast Asia, Latin America, and transition economies. We have also explored individual country effects, with particular attention to formerly planned economies.


In interpreting the results, we consider that departures of actual dollar wages from equilibrium are temporary and captured by the error term. Equivalently, we assume that the fitted values of the regression provide an estimate of equilibrium dollar wages. This requires care because of the possibility of persistent or systematic deviations from equilibrium. In particular, many countries in our sample have experienced very high, and sometimes sustained, inflation during the period under review. To account for possible short-term nonneutralities, we have added the inflation rate as a regressor. However, we do not consider inflation as a determinant of the equilibrium dollar wage.

Table 2 presents the regression results. We only report the significant coefficients, once the nonsignificant ones have been purged. All variables are in logs except the school enrollment ratio and the inflation rate, which are in percent. The first column presents the raw results, which assume that all coefficients’ elasticities are the same in all countries. The second column allows for fixed effects. With (at most) five observations per country, it is impossible to allow for single country dummies. Instead, we looked for significant “regional dummies,” lumping together the OECD countries, the centrally planned economies, Southeast Asia, Africa, and Latin America, living out the Asian countries as the residual. Given the paper’s focus, we have scrutinized more carefully the role of country dummies for formerly planned economies. Other country dummies have been tested one at a time; several of them turned out to be significant but are not shown. We report the result of the search that exhibited the best overall performance in terms of the standard tests reported at the bottom of the table (normality, heteroscedasticity, stability). Column (3) reports tests of the assumption that elasticities are identical. For that purpose, we have interacted the regional and country dummies with each coefficient. Again, we only report the significant terms. Finally, column (4) tests for both fixed and slope effects.

Table 2.

Dollar Wage Equation

article image
article image
Notes: All variables are in logs except school enrollment and inflation, which are in percent. t- statistics are in parentheses. Heteroscedasticity-consistent standard errors and covariance. One asterisk and two asterisks indicate significance at the 5 percent and 1 percent confidence level, respectively. Not reported: constant, nonplanned-economy country dummies. Year dummies: (1) 1975: 0.21 (1.88), 1980: 0.41 (3.92); (2) 1975: 0.15 (2.04), 1980: 0.36 (6.29); (3) 1975: 0.19 (2.11), 1980: 0.40 (4.93); (4) 1975: 0.14 (2.07), 1980: 0.37 (7.09). In regression (3), regional dummies have been interacted with right-hand-side variables. The significant terms concerning GDP are reported in the table in the dummy lines. Other significant terms are school enrollment interacted with planned economy: 1.35 (2.11); with Latin America: −1.73 (−2.96); Agriculture/industry interacted with OECD: −0.17 (−2.27). In regression (4), regional dummies have been added as constant and interacted with right-hand-side variables. The significant interacted terms concerning GDP are reported in the table. Other significant terms are school enrollment interacted with planned economy: 1.42 (2.17); with Latin America: −1.42 (−3.89); Agriculture/industry interacted with OECD: −0.15 (−2.22). Fixed country effects are not reported but few were actually significant; the only planned economy dummy with a significant coefficient is Czechoslovakia, as reported in the table. Countries used are as follows: Algeria, Argentina, Australia, Austria, The Bahamas, Bangladesh, Barbados, Belgium, Bolivia, Botswana, Brazil, Bulgaria, Burundi, Canada, Cameroon, Chile, China, Colombia, Costa Rica, Côte d’Ivoire, Cyprus, Czechoslovakia, Denmark, Dominican Republic. Ecuador, Egypt, El Salvador, Fiji, Finland, France, The Gambia, West Germany, Ghana, Greece, Guatemala, Guyana, Honduras, Hungary, India, Ireland, Israel, Italy, Jamaica, Japan, Jordan, Kenya, Republic of Korea, Luxembourg, Malawi, Mexico, Morocco, Myanmar, Netherlands, New Zealand, Nicaragua, Nigeria, Norway, Pakistan, Panama, Papua New Guinea, Paraguay, Peru, The Philippines, Poland, Portugal, Romania, Sierra Leone, Singapore, South Africa, Spain, Sri Lanka, Sweden, Switzerland, Syrian Arab Republic, Tanzania, Thailand, Tonga, Turkey, United Kingdom, United States, USSR, Venezuela, Western Samoa, and Yugoslavia.

The results accord with common sense. The coefficients are generally estimated quite precisely and are robust from one specification to another. The main limitation comes from specification tests: for simulation purposes we will use regression (3), which passes at the 1 percent confidence level all tests except for normality.15

The size of agriculture relative to industry is a standard measure of development: a 10 percent decrease in this ratio raises the dollar wage by 1 to 2 percent. The growth-enhancing role of government spending has been recently studied in the endogenous growth literature (Barro and Sala-i-Martin, 1995). Our results are in line with previous estimates: a 10 percent increase in the size of government raises wages by 3 to 6 percent. This effect is interpreted as measuring the effect of public services and infrastructure on aggregate productivity. Yet, public spending could also be a source of inefficiency, especially in planned economies. We have tested and rejected the hypothesis that the coefficient of the government variable is different (and negative) for the planned economies.

In column (1) of Table 2, the estimated effect of GDP per worker on the dollar wage is large, but it declines when we use regional and country dummies, suggesting that this variable partly works as a proxy for country-specific effects. Conversely, the effect of investment in human capital—measured here as secondary school enrollment—rises as we introduce the regional and national dummies. Both effects are clearly established. Finally, it may be noted that 10 additional percentage points of inflation reduce the dollar wage by 1.5 to 2.5 percent, a powerful effect to which we return below.

The role of the regional dummies is interesting. 16 Unsurprisingly, we find that being in the OECD group brings in a dollar wage premium of about 90 percent (regression (2)) or raises by 6 percent the effect of productivity (regression (3)). Similarly, being a planned economy is found to bring a discount of about 70 percent, or to reduce the effect of productivity by 16 percent. Presumably, this captures unmeasured effects like infrastructure, the stock of human capital, possible network externalities, and inefficient use of resources, and intangible assets such as the legal and political systems. Regression (3) offers an interpretation of the channel through which these regional specificities operate. The positive effect of secondary education is stronger in the planned economies than in the rest of the world, which is compatible with the view that planned economies combined an inefficient capital stock (hence a comparatively low productivity of labor) with an efficient school system.17

Estimates of Equilibrium Dollar Wages

We use regression (3) of Table 2 to generate estimates of the equilibrium dollar wage. Under the assumption that temporary deviations of actual from equilibrium wages are captured by the error terms, it is appropriate to use the regression to predict a country’s equilibrium dollar wage using its indicators of economic and social development. Table 2 shows that inflation exerts a significant negative influence on the dollar wage. Under the assumption that inflation is neutral in the long run,18 it should not affect the equilibrium dollar wage. But inflation may exert permanent effects, for example through market inefficiencies and dollarization. The literature generally supports the view that inflation is nonneutral.19 Our procedure is to keep the inflation term in our simulations, but to set it at a constant benchmark level, a presumed permanent rate of inflation of 10 percent.20

The simulation results are shown in Table 3 for 1990, or the latest available data used for the estimation. The regional and country dummy variables capture special conditions that make a particular country different from the international norm. To predict the dollar wage at the outset of the transition period, we include in column (4) the planned economy dummy. To predict the dollar wage that would emerge under “normal” conditions, we set the dummy equal to zero in column (3). Finally, if we believe that the OECD countries represent the correct reference, we include in column (3) the OECD dummy variable.21 Taking as reference the estimates obtained using the planned economy dummy, in 1990 the actual wage (column (2)) was below equilibrium for most countries (Bulgaria, China, Croatia, Czechoslovakia, Poland, Romania, and Russia). This may reflect undervaluation, mismeasurement of labor costs, overestimation of the performance indicators (in particular the PPP-adjusted GDP), or a distribution of income skewed toward the shareholder, that is, the state. Hungary appears to be at equilibrium if we use the planned economy estimate; it is below equilibrium if we consider that it had already covered some of the distance between the command system and a market economy. Only Slovenia appears to be clearly overvalued, but it must be remembered that the estimate is based on data for the whole of the former Yugoslavia.

Table 3.

Actual and Estimated Dollar Wages

article image
Sources: Actual wage (average): IMF; Equilibrium: authors’ calculations from regression (3) in Table 2.Note: For Croatia and Slovenia, we use the equation estimated for Yugoslavia.

Figure 2 shows both in-sample (1970–90) and out-of-sample (1991–96) annual forecasts estimates for the period 1991–96. For the latest period, we use different data, some of which are forecasts from various sources.22 Along with the actual dollar wage (W) displayed in column (2) of Table 3, the figure presents two estimates of the equilibrium level: W1, which corresponds to the normal situation (no dummy variable as in column (3) of Table 3) and W2, which uses the planned economy dummy (as in column (4)). We regard Wl as an upper bound for the equilibrium wage, to be reached once most of the market economy institutions are in place and functioning smoothly. We regard W2 as a lower bound, which corresponds to the starting situation. The actual equilibrium dollar wage is presumed to lie in between, gradually rising from the lower to the upper bound as transformation takes place. Eventually, the equilibrium dollar wage will correspond to the OECD case, but we do not show the corresponding estimate.

According to these estimates, with the exception of Slovenia and Hungary, all countries started the transition process with a significantly undervalued exchange rate. Along with the ensuing gradual narrowing of the gap, this confirms our stylized fact presented in Figure 3. Those countries that have been more determined in adopting market-enhancing measures and stabilizing the macro economy (the Czech Republic and Poland) by now have dollar wages close to the lower bound, that is, not very far from equilibrium. The same applies to Croatia. Hungary has let the actual dollar wage trail behind its equilibrium level. In Russia, the massive fall of the actual dollar wage is entirely explained by inflation. More generally, if we reestimate the equilibrium wage using historical inflation rates instead of the “steady-state” rate of 10 percent (see the Appendix), we can explain the initial undervaluation with the burst of price increases in Bulgaria, Croatia, Poland, and Russia.

We conclude that by 1996 the real exchange rate is in the neighborhood of equilibrium in Croatia, the Czech Republic, Poland, Slovenia, and possibly Hungary. Undervaluation most likely still characterizes the other countries in our sample. Needless to say, the results presented in this section need to be considered with great prudence. This applies particularly in the successor countries of the former Yugoslavia and Czechoslovakia.

IV. Dynamics of the Equilibrium Exchange Rate


The previous section offers estimates of the path of the REER. It does not provide an explanation for the path of the actual real exchange rate during the first years of transition. This section uses higher frequency (monthly) data to test whether the real exchange rate has been moving toward equilibrium as postulated by the stylized fact. The strategy for studying the evolution of a variable about its equilibrium level is the error-correction-model approach, whereby a (cointegrating) equilibrium relationship is embedded into a dynamic specification. We adopt this modeling strategy but we emphatically do not pretend to capture cointegration over the brief sample period available (less than live years). In fact, we do not even test for stationarity of the variables.23 Accordingly, we assume that the long-run relationship characterizing the real exchange rate λ (or μ) is

Figure 3.
Figure 3.

Dollar Wages

Citation: IMF Staff Papers 1997, 004; 10.5089/9781451930962.024.A002

Sources: IMF and PlanEcon. W is actual dollar wage.Note: W1 is estimated equilibrium dollar wage, assuming no regional specificity (col. (1) of Table 3). W2 assumes planned economy specificity (col. (2) of Table 3).

where x vector of real variables explaining the REER as in equation (8). If in addition, in the short run, off-equilibrium, the real exchange rate responds to the vector z of nominal and real temporary disturbances, including tagged changes in λ and x, the error-correction-model (ECM) equation to be estimated is


The coefficient α, which comes in front of the once-lagged “cointegrating factor,” describes the speed of adjustment toward equilibrium and is the one of greatest interest.

We present results for all three measures of the real exchange rate shown in Figure 1 (the doubly deflated nominal exchange rate λ, the ratio of nontraded to traded good prices μ and the dollar wage). The choice of variables in x and z is dictated partly by data availability, and partly by the accounting framework presented in Section II.

The variables to be collected in x are those that affect the REER. Most of them are not directly observable or available. This is particularly the case of sectoral level data such as the Balassa-Samuelson term (aT - aN). Using equation (6) and the definition of aggregate productivity, equation (8b) can be rewritten (in logs) as


Two of the right-hand-side variables can be matched reasonably well by existing data: the real aggregate producer wage (w - p) is measured as (the log of) the ratio of the average wage to the wholesale or industrial price index (depending upon data availability), although the nontraded good sector is likely to be underrepresented. Marginal labor productivity in the traded good sector, aT, is approximated by average labor productivity in industry. The other terms capture distortions of the transition economy. In the present formulation we expect κ (“quality”) and 9 (the wage gap wN -wT, which we believe is negative) to increase as transition progresses. We have less intuition for the other terms, for which data do not exist, and resort to available data that may measure economic distortions, noting the risk of collinearity with productivity in industry.

Given the extreme brevity of the period under review, we use monthly data and increase the sample size by pooling data across all countries. Our choice of indicators of distortions is a compromise between theory and data availability. To capture distortions in the financial and banking system, we use the spread between the lending and deposit interest rates (iL - iD). In principle, a large spread corresponds to a poorly functioning financial system, although it may also signal a liberalization of the market. In fact, in many countries the spread has declined with liberalization and the elimination of inflation. Another potentially useful indicator is the velocity of money. This variable has been included but turned out never to be significant, maybe for the reasons discussed below. Distortions in the labor market are measured by the level of employment and by the rate of unemployment. Given the initially high levels of employment in centrally planned economies, labor force participation was expected to decline as personal preferences should lead some to choose to stay at home. Declines in employment, L, and the rise of unemployment, U, may also reflect the end of the practice of overmanning, which resulted from the employment guarantee offered by the state through firms that did not operate for profit.24 These indicators are bound to be noisy because they are also affected by traditional macroeconomic factors.

The variables collected in the vector z correspond to temporary influences on the real exchange rate. This suggests looking mainly at nominal variables, representing monetary and fiscal policy. The log of money was found not to enter significantly. There is no monthly indicator of fiscal policy so we use instead the lending interest rate, implicitly assuming a reduced form in the IS-LM tradition. The nominal exchange corresponds to the transitory effect of nominal exchange rate changes on the real exchange rate as in Edwards (1989). Finally, inflation was found in Section III to affect the real exchange rate. Yet we cannot use it as a regressor here. Looking at equation (11), we note that the regressand dλ includes both inflation and the rate of exchange rate depreciation. When the rate of exchange rate depreciation—de—enters as a regressor, inflation cannot be used simultaneously on the right-hand side.25


The data used in this section are unpublished, provided by the IMF’s European I Department, which forewarns us that reliability is not guaranteed. Cross-section time-series estimation requires that data be comparable across all countries in the sample. This limits our investigation to six countries: Croatia, the Czech Republic, Hungary, Poland, the Slovak Republic, and Slovenia. For each country, the starting date has been chosen to eliminate the initial high inflation episode. The end date is dictated by data availability.

A number of dummy variables are designed to account for sharp and clearly identified policy changes likely to affect the real exchange rate independently of the other variables: this concerns primarily major changes in either taxation or tariffs. The dates chosen are: October 1992 for Croatia (introduction of sales tax); January 1993 for the Czech Republic and Slovakia (introduction of VAT); July 1993 for Poland (introduction of VAT) and Slovakia (increase in indirect taxation); February 1992 for Slovenia (tax reform). We also use monthly dummies to remove seasonal factors.


Given our choice of variables, the equation to be estimated is found upon substitution of equation (12) into (11):26


where w is the log of the average wage; pY the log of the industrial price index; y, the log index of industrial production; l, the log of employment; u, the log of unemployment; de, the rate of appreciation, and iL and iD, respectively, the credit and deposit interest rates. Vector dzt-1 includes all lagged changes of the dependent and of all variables that appear inside the squared brackets.

The speed of adjustment term, α, and the other coefficients, ci, are expected to be positive. According to equation (12), increases in the real producer wage lead to a real depreciation, while productivity has the opposite elicci. Although we have found some theoretical ambiguity, we expect that the distortion lerms are such that when the corresponding distortion is reduced (l down, u up, iL-iD down), the REER appreciates. The terms de and diL capture monetary and fiscal policy disturbances. The estimated equation follows from equation (13):


where ci = α ci.

Table 4 presents the results. The coefficients for the constant and the various dummy variables are not reported. For each of the three real exchange rate measures we present two regressions: in columns (1), (4), and (6), we estimate equation (14) directly as shown and add country-specific dummy variables to test for fixed effects. In columns (2), (5), and (7), we further test for country-specific estimates by interacting the country dummies with the right-hand-side variables. In each case, as the result of an extensive search, we have eliminated the nonsignificant entries. At the bottom of the tables we report those country-specific slope coefficients that are significant. Because most of them concern Croatia, we present in column (3) a regression based on data excluding this country.

In a rigorous interpretation of the tests, only the regressions reported in columns (1) and (7) can be accepted as free of misspecification. Yet most parameter estimates are stable across specifications. The goodness of fit is highly satisfactory given that data are both cross sectional and time series. The residuals often exhibit heteroscedasticity and nonnormality.27 That a simple formulation, imposing identical coefficients on all countries in the sample, explains a surprisingly large proportion of total variance supports the working assumption that the same forces systematically affect the real exchange rate during the early transition phase under review.

The pattern of coefficient signs across the various regressions does not fully conform with our priors in equation (14). Productivity in the traded good sectors appears with the correct sign for λ and μ, but not for the dollar wage. Conversely, the coefficient of the real producer wage has the correct sign for μ and the dollar wage, but not for λ. Mismeasurement is a serious issue. Industry probably weighs excessively in the real wage so that the fast-growing nontraded sector may be increasingly underrepresented. Our measures of distortions are problematic and not directly related to the theoretical concepts. In any case, the theoretical effects of the distortions are open to some ambiguity.

Overall, the results from Tables 2 and 4 indicate that increases in labor productivity, and more generally gains in economic efficiency, result in a continuing appreciation of the REER. The exact channel of this effect is not elucidated, but its existence and strength warrant it being considered a stylized fact. The short-run effect of a change in the nominal exchange rate on the real rate λ is about 85 percent within a month (column (1)) and the speed of adjustment α indicates a half-life of 19 months.28 Exchange rate policy exerts a powerful and lasting influence on the real exchange rate.

Taken together, these results confirm the hypothesis that the CPI-adjusted real exchange rate appreciation observed in most transition economies corresponds to a correction from its initial undervaluation. Deviations from equilibrium are due mainly to this initial condition and to nominal exchange rate policies that may have mistakenly but successfully delayed the unavoidable return to equilibrium. On the other hand, the nominal exchange rate has a smaller effect on the dollar wage and no significant effect on the ratio of nontraded to traded goods prices. Thus, the high degree of absolute price stickiness implicit in the impact of the nominal on the real exchange rate translates into a high degree of relative price rigidity. In contrast, nominal wages are less sticky than prices: not only is the pass-through of the nominal exchange rate lower but the speed of adjustment is much higher, with a half-life slightly above two months (column (7)).

Using the information collected in Table 1, we have attempted to test formally the role of the exchange rate regime. The hypothesis that the results change when the floating countries are excluded from the sample is rejected at any conventional level for λ. but not for the ratio of traded to nontraded good prices nor for the dollar wage. However, these tests should be considered as provisional because the estimates over the subsamples are quite unstable and generally not very satisfactory.

Table 4.

Real Exchange Rate Dynamics

article image
article image
Notes: Heteroscedastic-consistent standard errors and covariances. t-statistics in parentheses. Seasonal, country, tax dummies not reponed. One, two, and three asterisks represent significance at the 10,5, and 1 percent level, respectively. Estimation periods: Croatia: February 1992–July 1994, Hungary: January 1990–April 1995, Slovak Republic: January 1992–September 1994, Czech Republic: January 1992–January 1995, Poland: January 1990–April 1995, Slovenia: February 1992–July 1994.

Column (2): country dummies interacted with coefficients. Significant coefficients: λi-l: Croatia -0.16 (-4.6); Yt - 1t-l Croatia -0.11 (-2.13); de: Croatia 0.013 (2.2); it1Lit1D, Slovak Republic 0.0035 (1.9).

Column (5): country dummies interacted with coefficients. Significant coefficients: Croatia: λt-l -0.47 (-5.8); wt-1;0.084 (2.2); diL. 0.000069 (7.3); lt-l. 1.25 (5.0); and Czech Republic: Ut-l,-0.022 (-1.7).

Column (7): country dummies interacted with coefficients. Significant coefficients: wts1 Poland, 0.44 (5.7); wt-l: Hungary, -0.15, (-2.2), Slovak Republic, -0.33 (-3.0); yt-l. - lt-l: Croatia. -0.66 (-3.3), Czech Republic, 0.18 (5.2), Slovak Republic, 0.16 (3.1):Lt-1 Poland, 0.55 (3.4); Ut-l: Croatia, -0.75 (-4.6), Poland. -0.05 (-2.5), Slovenia, -0.17 (-4.6).

V. Conclusion

This paper proposes an interpretation of the evolution of the real exchange rate over the first five years of transition. The striking feature has been the process of continuing appreciation that has followed the—often deep—initial jump in real depreciation. We hypothesize that, initially, the real exchange rate overshot its equilibrium level and that the subsequent depreciation corresponds to the combination of a return to equilibrium and of a continuing real equilibrium appreciation. The equilibrium appreciation is explained by rapid gains in efficiency once markets drive prices and the allocation of resources.

This hypothesis is tested in a number of ways. First, we use international data to estimate the equilibrium dollar wage, using data that precede the transition period.29 The equilibrium dollar wages estimated over the period 1991–96 support our hypothesis. Actual dollar wages have been converging to their equilibrium level, but this appears to be a slow process. Most important, the often very large increases in dollar wages observed over the first few years of transformation have not (yet) resulted in overvaluation.

We also explore in detail the dynamics of the actual exchange rate during the transition. In the absence of a theory of the exchange rate in transition economies, the results must be considered with more than usual caution. They confirm the longer-run role of gains in effectiveness. They also show that the speed at which the real exchange rate converges to its equilibrium level is quite slow, even though wages appear to be much less sticky than consumer prices.

A number of policy implications emerge from this analysis. First, PPP is not an appropriate benchmark in transforming economies. Real appreciation is the equilibrium outcome of a successful transformation. Indeed, transition will be complete when the real appreciation stops. By then, which may be decades away, price and wage levels in Central and Eastern Europe will have converged to levels not too different from those in Western Europe.

Second, the need for a continuous real appreciation has important implications for the choice of exchange rate policies and regime. If the exchange rate is pegged to a Western currency (dollar, deutsche mark, ECU, or a basket), real appreciation requires a higher inflation rate than abroad. Inflation can be brought down to low levels only if the nominal exchange rate is allowed to float and appreciate. Resisting a real appreciation is not only hopeless, it also leads to potentially speculative capital inflows and interventions that, if not sterilized, lead to faster money growth and eventually inflation. If sterilized, there can be a buildup of reserves fueling further inflows in an unending spiral. Even more destabilizing would be a policy of nominal depreciation, for example, based on a PPP rule, which leads to a dangerous cycle of inflation and depreciations.

Third, the choice of an exchange rate policy is lightly linked to the desired level of inflation. There are good reasons for not aiming at very low inflation rates in the early years of transition: public finance arguments in favor of a (moderate) inflation tax until the implementation of a tax reform that establishes a broad and fair base; macroeconomic arguments against a policy of nominal exchange rate appreciation given the uncertainty about the desired rate and the volatility of flexible rates; efficiency arguments in favor of sufficient inflation to allow for relative price changes without actually forcing some wages and prices to decline; and political economy arguments based on the likely difficulty of imposing continuous appreciation vis-à-vis such strong currencies as the deutsche mark.

Finally, the estimates presented here should not convey the feeling that the equilibrium exchange rate can be known with any degree of precision. The purpose of this paper is to provide ballpark estimates of the initial degree of undervaluation and a feel for what is a normal rate of real appreciation. The estimates rely on dramatically short time series and highly imperfect data.


Effect of Inflation on Estimated Dollar Wages

Figure A1.
Figure A1.

Dollar Wages Estimated Using Historical Inflation

Citation: IMF Staff Papers 1997, 004; 10.5089/9781451930962.024.A002

Sources: IMF and PlanEcon. W is actual dollar wage.Note: W1 is estimated equilibrium dollar wage, assuming no regional specialty (col. (1) of table 3). W2 assumes planned economy specificity (col. (2) of Table 3).


  • Asea, Patrick K., and Enrique G. Mendoza, 1994, “Do Long-Run Productivity Differentials Explain Long-Run Real Exchange Rates?” IMF Working Paper 94/60 (Washington: International Monetary Fund, May).

    • Search Google Scholar
    • Export Citation
  • Balassa, Béla, 1964, “The Purchasing Power Parity Doctrine: A Reappraisal,” Journal of Political Economy, Vol. 72 (December), pp. 58496.

    • Search Google Scholar
    • Export Citation
  • Barro, Robert, and Xavier Sala-i-Martin, 1995, Economic Growth (New York: McGraw-Hill).

  • Bélanger, Gérard, and others, 1994, “Eastern Europe: Factors Underlying the Weakening Performance of Tax Revenues,” IMF Working Paper 94/104 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Berg, Andrew, and Jeffrey Sachs, 1992, “Structural Adjustment and International Trade in Eastern Europe: The Case of Poland,” Economic Policy, Vol. 14 (April), pp. 11774.

    • Search Google Scholar
    • Export Citation
  • Bhagwati, Jagdish N., 1984, “Why Are Services Cheaper in the Poor Countries?” Economic Journal, Vol. 94 (June), pp. 27986.

  • Borensztein, Eduardo, and Paul R. Masson, 1993, “Exchange Arrangements of Previously Centrally Planned Economies,” in Financial Sector Reforms and Exchange Rate Arrangements in Eastern Europe, IMF Occasional Paper No. 102 (Washington: International Monetary Fund, February).

    • Search Google Scholar
    • Export Citation
  • Coorey, Sharmini, Mauro Mecagni, and Erik Offerdal, 1996, “Disinflation in Transition Economies: The Role of Relative Price Adjustment,” IMF Working Paper 96/138 (Washington: International Monetary Fund, December).

    • Search Google Scholar
    • Export Citation
  • De Gregorio, José, Alberto Giovannini, and Holger C. Wolf. 1994, “International Evidence on Tradables and Nontradabies Inflation,” European Economic Review, Vol. 38 (June), pp. 122544.

    • Search Google Scholar
    • Export Citation
  • Dobrinski, Rumen, 1994, “Exchange Rate Policy and Macroeconomic Stabilization in a Transition Economy: Lessons from the Bulgarian Exchange Rate Crisis of 1993,” paper presented at the European Economics Association Meeting, Maastricht, September.

    • Search Google Scholar
    • Export Citation
  • Dotsey, Michael, and Peter Ireland, 1996, “The Welfare Cost of Inflation in General Equilibrium,” Journal of Monetary Economics, Vol. 37 (February), pp. 2947.

    • Search Google Scholar
    • Export Citation
  • Edwards, Sebastian, 1989, Real Exchange Rates, Devaluation and Adjustment: Exchange Rale Policy in Developing Countries (Cambridge, Massachusetts: MIT Press).

    • Search Google Scholar
    • Export Citation
  • Froot, Kenneth A., and Kenneth Rogoff, 1994, “Perspectives on PPP and Long-Run Real Exchange Rates,” NBER Working Paper No. 4952 (Cambridge, Massachussetts: National Bureau of Economic Research, December).

    • Search Google Scholar
    • Export Citation
  • International Labour Organisation, Statistical Yearbook (Geneva: ILO, various issues).

  • International Monetary Fund, International Financial Statistics (Washington: IMF, various issues).

  • Lipschitz, Leslie, and Donogh McDonald, 1992, “Real Exchange Rates and Competitiveness: A Clarification of Concepts, and Some Measurements for Europe,” Empirica, Vol. 19, No. 1, pp. 3769.

    • Search Google Scholar
    • Export Citation
  • Portes, Richard, 1994, “Integrating the Central and East European Countries in the International Monetary System” (unpublished; London; Centre for Economic Policy Research).

    • Search Google Scholar
    • Export Citation
  • Richards, Anthony, and Gunnar Tersman, 1995, “Growth, Nontradables, and Price Convergence in the Baltics,” IMF Working Paper 95/45 (Washington: International Monetary Fund, April).

    • Search Google Scholar
    • Export Citation
  • Rogoff, Kenneth, 1996, “The Purchasing Power Parity Puzzle,” Journal of Economic Literature, Vol. 36 (June), pp. 64768.

  • Sahay, Ratna, and Carlos A. Végh, 1996a, “Inflation and Stabilization in Transition Economies: A Comparison with Market Economies,” Policy Reform, Vol. 1, pp. 75108.

    • Search Google Scholar
    • Export Citation
  • Sahay, Ratna, and Carlos A. Végh, 1996b, “Dollarization in Transition Economies: Evidence and Policy Implications,” in Currency Substitution and the International Use of Money, ed. by Paul Mizen and Eric Pentecost (London: Edward Elgar), pp. 193224.

    • Search Google Scholar
    • Export Citation
  • Samuelson, Paul A., 1964, “Theoretical Notes on Trade Problems,” Review of Economics and Statistics, Vol.46 (May), pp. 14554.

  • Summers, Robert, and Alan Heston, 1991, “The Penn World Table (Mark 5): An Expanded Set of International Comparisons, 1950–1988,” Quarterly Journal of Economics, Vol. 106 (May), pp. 32768.

    • Search Google Scholar
    • Export Citation
  • Williamson, John. 1985, The Exchange Rate System (Washington: Institute for International Economics).

  • World Bank, 1995, Social Indicators of Development (Washington: World Bank).

  • Young, Alwyn, 1994, “The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth Experience,” NBER Working Paper No. 4680 (Cambridge, Massachusetts: National Bureau of Economic Research, March).

    • Search Google Scholar
    • Export Citation

László Halpern is Senior Research Fellow and Deputy Director of the Institute of Economics of the Hungarian Academy of Sciences, and a Research Fellow of the Center for Economic Policy Research (CEPR) in London. He graduated from Karl Marx University of Economics (Budapest). Charles Wyplosz is Professor of Economics at the Graduate Institute of International Studies in Geneva, a Research Fellow of CEPR in London, and Managing Editor of Economic Policy. He received his Ph.D. from Harvard University. This paper was initially written while Charles Wyplosz was visiting the IMF’s European I Department, which provided invaluable support, advice, and help with data. László Halpern acknowledges support from the Hungarian Science Fund (OTKA 3231). Many individuals were involved in the data collection and advice regarding the data; the authors are deeply grateful to them all. They particularly thank Massimo Russo, Michael Deppler, and Gérard Bélanger for making this project at all possible and for useful advice, many IMF staff members for comments, Phi-Anh Plesch and Harmen Lehment for additional help with data, and David Maxwell and Nadine Orosa for excellent research assistance.


Sahay and Végh (1996b) discuss dollarization extensively.


For overviews of recent research, see Froot and Rogoff (1994) and Rogoff (1996).


Let N = W/EW* be the conventional wage-deflated real exchange rate, where W and W* are, respectively, the domestic and foreign nominal wages. W/E = N if we assume that W* is constant and normalized to unity. Note that we do not average dollar and deutsche mark wages, as we do with the other exchange rates. The reason is that the dollar wage has become a highly recognized measure in the transition countries.


Note that the figure exhibits the dollar wage itself, not an index. As a result, the scale (shown on the right vertical axis) is often different from that used for the two relative price indices (shown on the left axis).


For a formal analysis of demand effects on the real exchange rate, see De Gregorio, Giovannini, and Wolf (1994).


For an extensive analysis of the cyclical and structural reasons behind the rapid fall in tax income, see Bélanger and others (1994).


Lipschitz and McDonald (1992) correctly suggest that international competitiveness is best captured by measures of producer profitability (the ratio of relative producer prices or value-added deflators to relative labor costs). Unfortunately, such data are not available for the transition countries.


We assume away the distinction between traded and nontraded good prices abroad. A symmetric treatment would lead to the following version of equation (8b) below: λ = κ + γθ + θ(ρT − ρN) + γ(aTaN) − γ*(a*Ta*N) as in De Gregorio, Giovannini, and Wolf, 1994. We adopt the small country assumption because we do not have data on the “foreign country” to match the productivity terms a*T and a*N.


We define aggregate variables as x = γxN + (1 − γ) xT, for x = a, ρ. There is no compelling reason to use the same weight γ as for the CPI. The more general case is straightforward and qualitatively identical.


A “normalization” is under way but this contributes to pollution of the data. Furthermore, it proceeds at different speeds from one country to another.


See Summers and Heston (1991).

The use of PPP-adjusted GDP figures suggests another interpretation of our procedure. Our regression can be written as
where Y is nominal GDP, L is employment, η is the PPP exchange rate, and the dots represent the other variables suggested by equations (8) and (9). Let λ0 be the value of the real exchange rate in a henehmark period; then η = λ0(P/P*). If the wage share WL/Y is constant, the regression becomes
W/E= α ′(W/η)+,where α ′=α(WL/Y).
This looks like regressing 1/E on 1/η, that is, directly testing for PPP, which avoids introducing the often improperly measured wage variable. We did not adopt this approach because we are unwilling to assume a constant wage share in the early phase of the transition. We thank Timothy Lane for pointing out the similarity between the two procedures.

The ILO Statistical Yearbook does not always provide monthly wages: in some countries it publishes instead hourly, weekly, or annual wages. We therefore need information to convert them into monthly wages. In many cases, we have requested information from the IMF country desk economist.


Given the size of our sample, the rejection of normality is unlikely to indicate serious misspecification.


Once we allow for slope effects, there is no significant fixed effect.


The negative sign for Southeast Asia supports the result by Young (1994) that these countries’ output has grown fast not because of gains in total factor productivity but because of fast, and not particularly efficient, increases in the use of inputs. This interpretation applies even more strongly to the planned economies. The positive value for the African dummy is more difficult to interpret. It might correspond to the opposite effect: while inputs grow very slowly, scarcity leads to a more efficient use. Clearly, more research is called for to elucidate these results.


The measured short-run effect may correspond to such inflation effects as capital flight and exchange rate overshooting.


For a recent contribution and more references, see Dotsey and Ireland (1996).


Sensitivity analysis is straightforward: Table 2 shows that setting the benchmark inflation rate at 0 percent (20 percent) raises (lowers) the estimated equilibrium dollar wage by 1.4 percent, a trivial amount.


In regression 3, this dummy appears significantly when interacted with the GDP variable and with the relative shares of agriculture and industry in GDP (see the notes to Table 2 for details).


Most data have been provided by the IMF. For 1995 and 1996, as well as for PPP-adjusted GDP, we have used PlanEcon estimates.


As reported in Rogoff (1996), tests of PPP indicate that the convergence of the exchange rate toward its equilibrium (assumed to he PPP in most studies) is very slow, with a half-life of three to five years.


When the size of the labor force undergoes a rapid decline, employment and unemployment do not move systematically in opposite directions and can therefore be treated as two distinct regressors.


Of course, we could replace de with the rate of inflation, but this would be exactly the same regression. This observation may also explain why the velocity of money is not significant. Being strongly influenced by the rate of inflation, velocity may just be a proxy of inflation, which is highly collinear with the rate of depreciation.


Here we use λ, Looking at equations (8) and (9), it can he checked that the expressions for μ and ω are similar. Indeed, μ only differs from λ. by the “quality” parameter κ. while to includes aggregate productivity and the wage distortion factor ρ.


No reliable methodology is available to test stationarity in the panel estimation of an error correction model. For this reason it is not possible to properly test for cointegration. However, the real exchange rate, the real producer wage, labor productivity, employment, and unemployment are cointegrated, which gives some support for the estimation method used here. Columns (1) and (7) show that the intuitive panel estimation of the error correction mechanism should not be rejected. Misspecification systematically emerges when we interact the regressors with country-specific dummy variables, Heteroscedasticity must be assumed for our heterogeneous sample. The usual treatment—weighted ordinary least squares (OLS), using standard errors from single country estimations as weights—was tried but did not help.


The half-life is the lime it takes for a discrepancy between the actual and equilibrium exchange rate to be reduced by half, lt is computed as ln 2/α.


While we focus on the transforming economies, the results obtained here can be readily applied to any country.

IMF Staff papers: Volume 44 No. 4
Author: International Monetary Fund. Research Dept.