In light of the persistence of moderate inflation in many transition economies, this paper analyzes whether inflation resulted from insufficiently tight financial policies and wage pressures or from the protracted adjustment of relative prices. Using a new database for 21 countries, the effect of relative price variability on inflation is estimated within a framework controlling for nominal and real shocks. Money and wage growth were the most important determinants of inflation; relative price variability had a sizable effect at high inflation during initial liberalization and a small effect at moderate inflation. Cost recovery may contribute to variability, particularly in the advanced stages of the transition.

Abstract

In light of the persistence of moderate inflation in many transition economies, this paper analyzes whether inflation resulted from insufficiently tight financial policies and wage pressures or from the protracted adjustment of relative prices. Using a new database for 21 countries, the effect of relative price variability on inflation is estimated within a framework controlling for nominal and real shocks. Money and wage growth were the most important determinants of inflation; relative price variability had a sizable effect at high inflation during initial liberalization and a small effect at moderate inflation. Cost recovery may contribute to variability, particularly in the advanced stages of the transition.

I. Introduction

Although most formerly centrally planned economies experienced very high rates of inflation at the beginning of their transition, many have succeeded in lowering inflation to moderate levels of about 20-40 percent a year. However, there has been limited success in reducing inflation to relatively low levels of the order of about 10 percent a year. In some cases, inflation has declined to low monthly rates, but these reductions have, so far, rarely been sustained.

A key question for Fund-supported programs in these economies is whether the persistence of moderate inflation results from the traditional factors of insufficiently tight financial policies and wage pressures, or from factors peculiar to transition economies—specifically the sizeable adjustment of relative prices that is necessary for the transformation to a market economy. This paper attempts to shed light on this issue by analyzing the empirical evidence relating to the determinants of inflation in a group of 21 transition economies. It considers whether relative price adjustments take place over a prolonged period—even following rapid liberalization—and contribute to a period of moderate inflation, perhaps initially through an increase in velocity and, depending on exchange rate policy, over time through an increase in the money supply via the balance of payments. Some factors that could underlie the marked real appreciation experienced by many transition economies are also examined. The paper does not provide specific levels of feasible or appropriate inflation targets for Fund-supported programs. This would require an analysis of the output and other costs of disinflation as well as consideration of the relative priorities placed on disinflation versus other program objectives: subjects which are beyond the scope of this paper.

A review of inflation performance in transition economies under Fund-supported programs indicates that the overshooting of inflation targets has frequently been associated with money growth in excess of program projections (Table 1). However, targets have been overshot even when money growth projections were met (and vice versa), suggesting that the link between money growth and inflation may shift in ways that are difficult to anticipate and that money growth alone does not provide a complete explanation of inflation, particularly in the short run. These features point to a need to understand better some of the factors that could contribute to money growth and to unstable velocity in these countries, such as relative price adjustment.

Table 1.

Summary of Performance under Recent Fund-Supported Programs in Transition Economies 1/

(By number of program periods) 2/

article image
Source: Annex I.

For programs approved from June 1992 to June 1995 in the sample of 21 countries included in this study. Levels or targeted and actual inflation and broad money growth during each program period (defined as periods of at least half a calendar year, unless otherwise noted) are given in Annex I.

Inflation and money growth generally measured at end-of-period growth rates.

Includes two program periods of one calendar quarter.

Includes one program period of one calendar quarter.

Includes three program periods of one calendar quarter.

The paper is organized as follows. Section II briefly reviews the literature on the determinants of inflation, particularly that relating relative price variability to overall inflation, and provides a broad analytical framework for the subsequent empirical work. Section III analyzes the characteristics of the distribution of price changes in 21 transition economies using disaggregated CPI data.1 The results of econometric estimates of a semi-reduced form inflation model using panel data for the 21 countries are examined in Section IV. These results are complemented by Section V which takes a closer look at some issues related to relative price adjustment in five countries—Czech Republic, Poland, Estonia, Moldova, and Russia—selected to cover a range of progress with regard to stabilization and structural reform as well as different policy regimes. The concluding section summarizes the main findings of the paper.

II. Review of the Literature and Analytical Framework

The literature suggests four broad groups of factors that could explain the stickiness of inflation in transition economies after the initial monetary overhang has been dissipated: monetary growth fueled by fiscal obligations, often reflecting delayed structural reforms; wage increases out of line with productivity gains; underlying pressures for an appreciation of the real exchange rate coupled with a policy of stabilizing the nominal exchange rate; and relative price adjustment combined with downward price rigidities, or more generally, asymmetric price responses. The last three factors may be associated with unanticipated capital inflows and endogenous money growth through the balance of payments.

A. Fiscal Obligations and Money Growth

At the beginning of the transition, most centrally planned economies faced both a stock and flow macroeconomic disequilibrium. The former took the form of a monetary overhang which dissipated relatively quickly following an initial burst of inflation when prices were liberalized. The flow disequilibrium—mainly due to the monetization of explicit and implicit fiscal obligations—however, has been a driving force behind the persistence of inflation in many countries (Bruno (1993), Fischer, Sahay, and Vegh (1995), Scacciavillani (1994)). The weakening of the ability to expropriate enterprise surpluses and the collapse in output often contributed to a sharp decline in government revenue while subsidies and transfer payments increased. In many cases, fiscal deficits were financed by bank borrowing; however, even when these deficits were relatively small, banking system credit to public enterprises often increased sharply and fuelled rapid money growth. Indeed, in transition economies, the measured fiscal deficit vastly understates the true extent of fiscal obligations which mainly arise from unreformed institutional relationships, particularly with regard to public enterprises and the banking system (Dornbusch (1992), McKinnon (1992)).

An additional factor influencing the persistence of inflation even when some degree of monetary control has been established is the credibility of the fiscal adjustment which could contribute to shifts in velocity. Public perceptions of the sustainability of a fiscal and credit tightening are influenced by such factors as progress with regard to fiscal reform—particularly social safety net and pension reform and the adoption of an inflation-resilient tax system—public enterprise restructuring and privatization, and the health of the banking system (Bruno (1993)).

B. Wage Pressures

Wage policies can contribute to the stickiness of inflation in several ways. First, wage increases in excess of productivity gains can directly place upward pressure on prices. Second, higher wage bills are often responsible for the expansion of credit, and hence money, to state enterprises and to the government (Sahay and Vegh (1995a)). Third, explicit or implicit wage indexation can contribute to significant inflation inertia. Fourth, the lack of relative wage flexibility both at the sectoral and enterprise level can raise the output cost of reducing inflation and weaken the eventual economic recovery, thus making continued inflation a more politically acceptable alternative (Edwards (1992), SM/95/316). Although the empirical evidence is mixed on whether exogenous wage increases have driven inflation in transition economies, there are indications that large wage increases in reaction to an initial inflation shock have sustained and fueled inflationary pressure in a number of countries (Citrin and Lahiri (1995), Commander (1992), Dayal-Gulati (1996), SM/95/316).2

C. Real Exchange Rate Appreciation and Capital Inflows

Many transition economies have experienced strong upward pressure on their real exchange rates.3 In those countries that maintain a relatively stable nominal exchange rate, this real appreciation would be accompanied by capital inflows and monetary expansion and would be associated with a rate of inflation above that of trading partners. On the other hand, in countries with flexible exchange rates, real appreciation would be associated with nominal appreciation and downward pressure on the overall rate of inflation. The literature identifies at least three factors that could lead to real appreciation during transition.

  • Initial undervaluation of the real exchange rate relative to its equilibrium level. In transition economies, undervaluation usually arises when the nominal exchange rate of new currencies is initially set at excessively low levels (often to minimize risks to competitiveness or international reserves) or when the nominal exchange rate is influenced by temporary distortions in asset markets.4 Undervaluation may be manifested in different ways: first, the domestic prices of tradable goods may be lower than comparable goods in world markets because the arbitrage of traded goods prices takes place only gradually.5 Second, even after traded goods prices have equalized, real product wages may be lower than indicated by labor productivity in traded goods; these low wages would also be reflected in prices of services that are lower than in countries with comparable levels of per capita PPP-adjusted GDP.6 Evidence of very low prices of services have been found for Russia (De Masi and Koen (1995)) and the Baltics (Richards and Tersman (1995)). Although undervaluation is likely to have been significant in the early years of the transition (Saavalainen (1995)), there is little clear-cut evidence that it persists several years after liberalization, particularly when international differences in productivity in traded goods are taken into account.

  • Differential productivity growth between the tradable and nontradable sectors (Balassa (1964), Samuelson (1964)): higher productivity growth in tradables relative to nontradables may raise real wages and lead to an increase in the prices of nontradables relative to tradables, and hence, to a real appreciation. Although such productivity growth differentials have been observed in market economies,7 and suggested for the Baltics (Richards and Tersman (1995)), hard empirical evidence for transition economies is scant. Moreover, it is unclear whether productivity growth differentials in favor of tradables should be expected in transition economies since productivity gains in the relatively underdeveloped service sector (e.g., financial services, communications, retailing and wholesaling) could be substantial during transition (Halpern and Wyplosz (1996)).

  • Demand pressures associated with a perceived shift to a higher level of permanent income: the move to market-determined prices may increase permanent income in transition economies; while potential output takes time to expand, private consumption may adjust more rapidly through external borrowing. Similarly, an increase in investment financed by foreign savings may temporarily raise the demand for domestic resources.8 These increases in domestic absorption could give rise to a temporary appreciation of the equilibrium real exchange rate during transition. However, increases in private consumption or the fiscal deficit in excess of levels that would be supported by a higher permanent income could give rise to excessive real appreciation and capital inflows that would eventually be reversed.

D. Relative Price Adjustment

Evidence for transition economies

While in the long run inflation is determined by money supply growth, the adjustment of relative prices from a highly distorted to a market-determined structure could, under certain conditions, contribute to upward pressure on inflation in transition economies.9 Evidence for Russia and Kazakhstan indicates that price liberalization entails a measured increase in the variability of relative prices (De Broeck et al. (1995) and De Masi and Koen (1995)). Notably, in the case of Russia, although price variability subsided following a relatively rapid and comprehensive initial liberalization, it persisted at high levels in comparison with market economies. In addition, in both countries, variability is positively associated with inflation (and changes in inflation), although this finding is based on a simple correlation which does not control for common shocks to both variables.

The persistence of relative price variability in transition economies following even comprehensive initial price liberalization could be explained in terms of the unavoidably limited speed of structural reforms and the changing structure of output and demand associated with the gradual move to a market economy. Another explanation, which also has implications for the appreciation of the real exchange rate, is the cost-recovery hypothesis which argues that the adjustment of certain capital-intensive service prices (housing, utilities, transportation) must take place more gradually than the adjustment of other prices, particularly tradables which are subject to international competition (Zavoico (1995), Saavalainen (1995)). These services are distinguished by a capital stock that not only was inherited, with no associated debt, from the pre-transition era, but also is large relative to the PPP-adjusted per capita income of these countries. Initially, when consumer wage levels are low, such service prices would be set to cover only current costs, even in a fully liberalized environment, because there are no associated debt service costs. Maintenance costs may not be covered because it is optimal initially to consume the excessively large capital stock. As real incomes rise and the capital stock that can be supported by these incomes also rises,10 the prices of these services would be raised, at first to cover maintenance costs and then to cover (future) capital costs, until they reach a level at which new investment can take place.

The gradual nature of the capital stock adjustment and cost recovery process suggests that some service prices will continue to adjust over several years following even comprehensive price liberalization.11 It also implies that price levels in such economies would be lower than indicated by international comparisons of PPP-adjusted GDP, not due to undervaluation, but because of the pricing of capital.12 Thus the real exchange rate may also be expected to appreciate more steeply as real incomes rise (and the prices of these services are increased to permit new investment) than may normally be expected for market economies.13 Any attempt to bring about a more immediate adjustment in the relative prices of cost-recovery items—if successful—is likely to entail a substantial contraction in demand or—if unsuccessful—higher inflation.

Analytical approaches

What implications does the extensive theoretical literature have for an empirical investigation of the relationship between inflation and relative price variability? Classical theory suggests that relative price adjustment reflects real factors and would not affect the increase in overall prices, which depends on money growth. However, the empirical association in market economies between inflation and relative price variability—which can be measured in different ways (Box 1)—has long been noted (since the 1920s). The theoretical basis for this association is quite varied, particularly with regard to the implied direction of causality between inflation and variability—an important consideration for empirical work.

Since price liberalization and structural reform in transition economies continue to cause the realignment of relative prices, under what conditions could this realignment, as reflected by relative price variability, contribute systematically to inflation? According to one view (see Box 1), real sectoral shocks increase the dispersion of both desired and actual relative price changes and lead to an increase in inflation because firms (not affected by the shock) resist a decline in their own prices and, in general, respond asymmetrically to disturbances which induce a rise in desired prices than to those which induce a fall in such prices.14 This downward price inflexibility, or more generally asymmetric price response, implies that an increase in inflation would be associated not only with higher variance, but also a positively skewed distribution of individual price changes. In transition economies, the adjustment of relative prices to a new market-determined structure—including the adjustment of capital-intensive service prices according to the cost-recovery hypothesis—can be interpreted as a series of real sectoral shocks which, when combined with downward price rigidity and, over time, an accommodative monetary policy stance, can lead to higher inflation.

Relative Price Variability and Inflation

Changes in relative prices or relative price variability can be measured in terms of the characteristics—or shape—of the distribution of individual inflation rates of goods and services comprising a price index (such as the CPI), at a given point in time. The shape of such a distribution is commonly described by the measures of variance and skewness. An increase in relative price variability, for instance, would result in a greater dispersion of individual inflation rates and a broadening of the distribution—or an increase in variance (Diagram 1). The question whether relative price adjustment is associated with inflation can thus be seen in terms of whether changes in the shape of the distribution are systematically related to shifts in the mean of distribution—which correspond to changes in the average inflation rate (Diagram 2). Clearly, an increase in variance need not necessarily lead to an increase in inflation if price increases of some goods are matched by commensurate declines in others—as in a symmetric distribution. When a price distribution is positively skewed, however, large increases in the prices of a few goods are accompanied by small price increases (or even price declines) in other goods (Diagram 3). Since the latter are not commensurate with the large price increases, suggesting downward price inflexibility, an increase in positive skewness may be associated with an increase in average inflation.

An important consideration, which has implications for empirical work, is whether the direction of causality runs from variability to inflation or vice versa. Theoretical models in this area can be classified into three broad approaches according to the implied differences in these causal relations:

  • In the first approach, the dispersion of relative price changes and the unanticipated change in inflation (as well as the time variance of inflation) are determined simultaneously, and depend on the variance of unanticipated aggregate and relative demand shocks.1 Normality in the distribution of prices is important to the derivation of the results from these type of models. Since common exogenous shocks simultaneously increase relative price variance and inflation, which are both endogenous, the direction of causality is ambiguous (Cukierman (1979)). Increased relative price variability is associated symmetrically with unanticipated inflation or deflation.

  • In the second approach, the rate of inflation is treated as exogenous and causality runs from inflation or unexpected inflation to relative price variability.1 Because of differential costs of adjusting prices (“menu costs”), not all firms increase prices simultaneously when inflation (anticipated or not) rises, which leads to a greater dispersion of relative price changes. Since menu costs are symmetric, relative price variance rises with both increases or decreases in inflation. In transition economies, the administrative costs of adjusting prices of public services (rents, utilities, transport etc.,) have a similar effect, in terms of making prices sticky, as menu costs; thus when inflation rises, relative price variance could increase because administered prices are not adjusted in line with market conditions.

  • Under the third approach, real sectoral shocks increase the dispersion of prices and raise inflation because firms not affected by the shock resist a decline in their own prices. Causality thus runs from relative price variability to inflation; a positive relationship between the skewness of the distribution of price changes and the rate of inflation is also suggested. As indicated above (see Diagram 3), if individual prices are downwardly rigid, increases in a few prices in response to a real sectoral shock both increase the degree of skewness and shift the mean of the distribution to the right (and hence raise the inflation rate). In a more rigorous analysis, Ball and Mankiw (1995) also derive a positive relationship between skewness and inflation by introducing menu costs.

These analytical approaches to the relationship between inflation and relative price variability have two implications for empirical work: first, although much of the existing empirical evidence is based on simple bivariate correlations between inflation and relative price variability, the association between the two variables needs to be considered in a multivariate setting that controls for common shocks. Second, since causality can, in principle, run in either direction, it is important to examine the factors underlying increases in relative price variability.2

1 This group includes stochastic rational expectations models with imperfect information where agents cannot distinguish between general and relative price changes (Lucas (1973), Barro (1976)Hercowitz (1981)) and models with differential adjustment in individual markets due to staggered contracts (Blanchard (1983), Taylor (1981)). In a two-sector (controlled versus free goods) version of an imperfect information model, Leiderman (1987) shows that in addition to unanticipated aggregate and relative demand shocks, relative price variability within the free goods sector increases to the extent that price increases of controlled goods are not synchronized with (anticipated) nominal money supply growth.1 This group comprises models with inter-firm differences in costs of adjusting prices (“menu costs”) (Sheshinski and Weiss (1977)) and deterministic models of imperfect information (Parks (1978)). The latter differs from stochastic models (included in the first group) in that they do not explicitly model expectations formation, but rather take the expected overall inflation rate as a given.2 This is done, as an illustrative example, for five countries in Section V. Causality tests are not feasible in the current sample because of the short observation period for each country. In any case, such tests can capture only temporal precedence rather than logical causality. For market economies, causality tests find no clear patterns of leads and lags between inflation and relative price variability (see Appendix I).

III. Empirical Evidence on Relative Price Behavior

The preceding discussion suggests that the shape of the distribution of individual inflation rates of goods and services comprising a price index can provide important insights into factors underlying the relationship between inflation and relative price adjustment. This section, therefore, briefly reviews the main results of a detailed analysis of the characteristics of the inflation distributions (based on the CPIs) of the 21 transition economies included in the study.15 The sample of quarterly CPI data covers the period from 1991-92 to the third quarter of 1995 for most countries and typically has a level of disaggregation varying from 10-70 categories of goods and services (Appendix I, Table 13).

The analysis was based on four indicators characterizing the sample distributions of individual price changes, weighted by their contribution to the CPI or unweighted, for each time period and country. The results are broadly in line with findings for market economies (see Appendix I) and indicate that:

  • the sample distributions tend to shift substantially over time within countries indicating the presence of substantial relative price changes throughout the transition period—even well beyond comprehensive initial liberalization. This finding was confirmed by the lack of significant time persistence, in all but a few cases, in the different relative price variability measures.16 The evidence also suggests variations in the degree and nature of relative price adjustment across countries.17

  • The distributions also show frequent and significant departures from both normality and symmetry, consistent with the presence of downward inflexibility in individual prices (Appendix I, Table 14; see Box 1 for the implications of non-normality and asymmetry). The rate of rejection of normality and symmetry increases noticeably for weighted price changes, indicating that the weighting system exerts a significant influence on asymmetry.

  • When asymmetry prevails, the direction of skewness is frequently positive, suggesting that a small number of large relative price increases led the inflationary process and coexisted with a large number of small relative price reductions (Table 2).

  • A decomposition of variance indicates that most (some 65-75 percent on average) of the total variation in relative prices stems from within tradables, and to a lesser extent (about 15-25 percent), from within nontradables; the contribution from variance between tradable prices and nontradable prices is relatively minor.18

Table 2.

Share of Total Sample of Quarterly Observations with Positive Skewness

(In percentages)

article image

In addition, in keeping with the literature, the relationship between the average inflation rate and various indicators of relative price variability, were examined on a country by country basis (Appendix I, Tables 1518).19 The main result is that the empirical relationship between inflation and relative price adjustment could be sensitive to the choice of variability indicator, particularly between variance and skewness, suggesting that both measures of relative price variability should be included in the panel regressions. While variance appears to be somewhat positively correlated with inflation, skewness in the distribution of price changes appears to be only weakly associated with the level of inflation, although when significant, it too tends to be positively correlated.20

IV. Empirical Evidence on the Determinants of Inflation

In view of the discussion in the preceding sections, what light does an econometric analysis shed on the sources of inflationary pressure in transition economies? This section discusses the underlying methodology and results of such an analysis for a panel of 21 transition economies. These regressions are intended to capture the impact on inflation of relative price adjustment, including those giving rise to real appreciation, when nominal shocks such as money and nominal wage growth are controlled for.

A. The Basic Model for Inflation

The factors explaining inflation discussed in Section II were brought together in a simple static two-sector model of traded and nontraded goods and money market clearing to derive and interpret an estimated equation for inflation (Appendix II). Dynamics were not included because the limited time coverage of the data (on average about 12 quarters per country) permits the estimation of only short-run (mainly within-quarter) effects, with no distinction made between short- and long-run parameters. A semi-reduced form for inflation was selected for estimation rather than a reduced form version (Equations (11) or (11’) in Appendix II) because of the relevance of the regressors, particularly the real exchange rate, to the policy debate. The basic equation derived from the analytical model can be expressed as:

π=γ0+γ2m+γ3w+γ4(πNTπT)+γ5Vπγ2>0;γ3>0;γ4<0;γ5>0

where the γ's are functions of the structural parameters of the analytical model as shown in Equation (8) of Appendix II.21 Hence, the initial specification was of the form:22

π=γ0+γ1π(t1)+γ2m+γ3w+γ4(πNTπT)+γ5TVAR+γ6TSK+seasonaldummies

B. Estimation Methodology23

With the dependent variable defined as the quarterly end-of-period inflation rate, on the basis of the above equation, the specification initially included: 1) the lagged inflation rate, to account for inflation inertia; 2) the growth rate of broad money, both contemporaneous and up to a two-quarter lag; 3) an indicator of labor cost pressure (unit labor cost growth, when available, or nominal wage growth), both contemporaneous and with one-quarter lag;24 4) an indicator of real exchange rate behavior (the differential growth of nontraded and traded goods prices), both contemporaneous and with a one-quarter lag;25 5) and indicators of relative price adjustment (the Theil variance and skewness analyzed in Appendix I). Additive dummies were included to control seasonal effects. The differential impact of the exchange rate regime on inflation was tested through additive and multiplicative dummies (on both the inflation inertia and the real exchange rate terms) for countries and periods where the exchange rate was used as an explicit nominal anchor (see Appendix III, Table 19).26

The equations were estimated by ordinary least squares with corrections for the bias in standard errors (and hence t-statistics) due to heteroschedasticity, as indicated by diagnostic tests.27 Using a “general to specific” modelling strategy, a more parsimonious final equation was derived from a larger initial set of explanatory variables following a specification search which eliminated statistically insignificant regressors in order of least significance (Appendix III, Table 20). The final specification estimates were then tested for robustness to different definitions of the most important regressors: liquidity and relative price variance.

Based on evidence of parameter instability across regions (Table 3; Appendix III, Table 22),28 three regional specifications were estimated separately for Eastern Europe, the Baltics, and FSU, respectively.29 The robustness of the estimated coefficients was assessed by re-estimating the regional specifications by weighted least squares (Appendix III, Tables 2325). In addition, the sensitivity of the results to changes in the sample period was examined by re-estimating the regional specifications for Eastern Europe and the Baltics excluding the initial period of liberalization.30 Since consumer expenditure weights can give a distorted view of the relative importance of certain sectors (public services, for instance, tend to have a very small weight in the CPI of many transition economies), the sensitivity of the results to unweighted indicators of relative price variability was also examined.31

Table 3.

Sensitivity to Specification of Regional Subsamples 1/

(OLS-HCSE estimates; dependent variable: quarterly CPI inflation rate)

article image

For the t-statistics reported in parenthesis, three asterisk indicates statistical significance at 1 percent level; two asterisks at the 5 percent level; one asterisk at the 10 percent level; - indicates that the variable was not included because of lack of significance in the specification search. OLS-HCSE refers to correction of bias in least squares standard errors and t-statistics using a heteroschedasticity-consistent variance-covariance estimator.

Measured by the Theil variance.

To address the issue of simultaneity, a full instrumental variables procedure was attempted, but the estimates proved highly sensitive to the choice of instruments. Since the available instruments (which are limited by the availability of data) were only very poorly correlated with the two relative price variability terms, a more limited exercise was undertaken for each of the three regions: in order to account for one obvious source of simultaneity arising from the presence of price-controlled goods in the CPI,32 the variance and skewness terms were recalculated only on the basket of liberalized goods.33 These new terms were substituted for the two contemporaneous relative price variability indicators and the equations were re-estimated, instrumenting for all other contemporaneous terms—i.e., money and nominal wage growth.

C. Estimation Results

The specification search yielded equations with high explanatory power (an adjusted R-squared of over 0.9 for the pooled sample of 21 countries and the FSU and about 0.7 for the other two regions) and significant, and generally plausible, parameter estimates (Table 4).34 Variations in inflation in the post-liberalization period in Eastern Europe and the Baltics proved more difficult to capture and the corresponding equations have somewhat lower explanatory power (an adjusted R-squared of 0.5-0.6). The estimations for all three regions were robust to the correction of standard errors for bias due to heteroschedasticity and to re-estimation using weighted least squares (Appendix III, Tables 2325).35 The results for Eastern Europe and the Baltics were generally robust to estimation by instrumental variables, although the statistical significance (but not the size) of the coefficient on wage growth tended to diminish. For the FSU, the coefficient on wage growth increased and that on money growth became statistically insignificant. The latter appeared to be due to the lack of a good instrument for money growth in the available data set, perhaps reflecting the fact that money growth in the FSU has resulted mainly from discretionary credit expansion, and hence, would be difficult to instrument. In addition, as indicated above and is well-known in the literature, these estimates are sensitive to the choice of instruments.

Table 4.

Final Specification: Pooled and Regional Estimates 1/

(Dependent variable: quarterly CPI inflation rate)

article image

For the t-statistics reported in parenthesis, three asterisks indicate statistical significance at 1 percent level; two asterisks at the 5 percent level; one asterisk at the 10 percent level; - indicates that the variable was not included because of lack of significance in the specification search. OLS-HCSE refers to correction of bias in least squares standard errors and t-statistics using a heteroschedasticity-consistent variance-covariance estimator.

Measured by the Theil variance and skewness, respectively.

The specification search also revealed collinearity between nominal wage growth and relative price variance in the pool and the FSU block (the correlation coefficient in the two samples is almost 0.8), so that both terms could not be individually significant at the same time. Although the pooled sample results are driven mainly by the FSU, the procedure followed in the specification search resulted in the wage cost term being eliminated from the pooled sample (and the Theil variance retained) and vice versa for the FSU. Hence, alternative specifications were derived retaining the wage term in the pool but eliminating it in the FSU (Table 5).

Table 5.

Relative Price Indicators: Alternative Specifications in Pooled Sample and FSU 1/

(Dependent variable: quarterly CPI inflation rate)

article image

For the t-statistics reported in parenthesis, three asterisks indicate statistical significance at 1 percent level; two asterisks at the 5 percent level; one asterisk at the 10 percent level; - indicates that the variable was not included because of lack of significance in the specification search. OLS-HCSE refers to correction of bias in least squares standard errors and t-statistics using a heteroschedasticity-consistent variance-covariance estimator.

If nominal wage growth term is retained at an earlier stage of the specification search (despite being statistically insignificant, as indicated by the t statistic.)

If nominal wage growth term is eliminated (despite being statistically significant) and relative price variance is retained at an earlier stage of the specification search.

Measured by the Theil variance and skewness, respectively.

What light do the results shed on the sources of inflationary pressure in transition economies, particularly the role of relative price adjustments? Do the estimates suggest significant regional differences in these sources and whether their relative influence may shift over time? The main messages that emerge from the empirical analysis can be summarized as follows:

  • Inflation is strongly and positively correlated with broad money growth and displays a relatively rapid response to a monetary shock. Contemporaneous and lagged money growth has an elasticity of about 0.5-0.7 in the pooled sample and the FSU and about 0.3 in the other two regions (Table 4).36 Evaluated at the sample mean, it contributes on average about one half of inflation in the pool and over one third in each of the regions (Table 6). In Eastern Europe and the FSU, the relative contribution of money growth to inflation is higher in 1995 (the final year of the estimation) than in the sample mean. The money variable may reflect accommodation—either exogenously through credit creation or endogenously through capital inflows and reserve accumulation—of wage and relative price shocks, including real appreciation. This suggests that better monetary control, including by allowing nominal appreciation, could help bring about greater disinflation.

  • Nominal wage pressures appear to have a significant impact on inflation, with an elasticity of about 0.2-0.3, accounting on average for about a fifth to a fourth of quarterly inflation in Eastern Europe and the FSU; wage pressures do not appear to be a significant factor in the Baltics (Tables 4 and 6). However, since these labor cost indicators capture only pressures arising from monetary remuneration, the estimated coefficients may understate pressures arising from non-wage benefits (most likely monetized, and hence, reflected in the money growth variable) which were considerable, especially in the FSU. In Eastern Europe, the relative contribution of wage growth to inflation is higher in 1995 than in the sample mean.

  • The empirical significance of the impact of relative price adjustment is sensitive to the indicator used to measure variability—particularly between variance and skewness—and to the sample period. Except in the case of the FSU, the results are generally robust to the substitution of unweighted variance and skewness for the corresponding Theil measures (Tables 4 and 7). In particular, although both the Theil and unweighted variance are significant in the full sample estimation for Eastern Europe, they become statistically insignificant in the post-liberalization period; the unweighted skewness, on the other hand, captures some effect of relative price variability on inflation during this period.

  • Overall, the results suggest that relative price adjustment has a significant impact on inflation, although the size of this effect and the indicator capturing it varies by region and over the sample period. Since the estimations reflect only the partial, impact effects on inflation during a quarter (holding other factors constant), the size of the estimated effect may be understated to the extent that pressures on inflation stemming from relative price adjustments are accommodated by money growth—and thus captured by the money variable in the equation.

  • In the pooled sample and the FSU, variability is associated with nominal wage shocks. In the pool, relative price variance is estimated to contribute slightly less than a third of inflation (Tables 4 and 6). Due to collinearity in the data, the estimated contribution declines when nominal wage growth is included in the specification (Tables 5 and 6). Similarly, in the FSU, when wage growth is included, the estimated contribution from variability appears small, on average, although it picks up quite substantially in 1995; when the wage term is excluded, the estimated contribution—from both variance and skewness—rises to about a fifth of inflation.37

  • In Eastern Europe variance contributes, on average, about a sixth of inflation (Table 6). As indicated above, the impact of variance becomes insignificant in the post-liberalization period although the unweighted skewness suggests a small, but significant, relative price effect (Tables 4 and 7). Similarly, in the Baltics, variability is estimated to make only a small contribution to inflation. The significance of skewness rather than variance in the post-liberalization periods in Eastern Europe and the Baltics (which may be considered more advanced reformers with lower inflation rates), and the marked increase in the contribution of skewness when inflation declined sharply in the FSU in 1995, suggests that downward price rigidity may be a factor at moderate levels of inflation.38

  • The results do not show a significant impact of real appreciation on inflation, except in Eastern Europe where it has a negative elasticity of about 0.2 and, evaluated at the sample mean, shows a small dampening effect on inflation for a given money growth (Tables 4 and 6). The negative impact is consistent with nominal appreciation (see the analytical model derived in Appendix II), although the small estimated effect may reflect the tendency for countries in the sample to resist nominal appreciation through intervention and endogenous increases in money via the balance of payments. Hence, as in the case of relative price variability, some of the impact of real appreciation may be captured by the money growth term. Since strong real appreciation is a feature of many transition economies, a few related issues are examined in Section V.

  • Inflation inertia—reflecting backward-indexation or backward- looking expectations in wage and price formation—may become more important as moderate levels of inflation persist, suggesting that the output costs of reducing inflation tend to increase at these levels. Inertia (as measured by lagged inflation) appears to have contributed about one quarter to one third of the inflation in Eastern Europe and the Baltics, but is not significant in FSU countries where the level of inflation is on average much higher (Tables 4 and 6). In the post-liberalization period, however, the experiences of Eastern Europe and the Baltics diverge, with inertia effects becoming stronger in Eastern Europe, but insignificant in the Baltics.

  • Explicit exchange rate anchors appear to have only a marginal—and statistically weak—dampening effect on inflation. This does not, however, necessarily imply that exchange rate anchors are ineffective since they can contribute to lower money growth and nominal wage pressure by disciplining financial policies and dampening inflation expectations. While in the alternative specification for the pool, an exchange anchor enters as an additive dummy, in Eastern Europe it is manifested as a multiplicative dummy on the lagged inflation term, suggesting a reduction in inertia through a beneficial effect on expectations formation.39 An exchange regime effect could not be observed in the Baltics, most likely reflecting the somewhat similar experiences and policy actions of Estonia, which maintained a formal peg, and Latvia, which did not.40

Table 6.

Inflation Decomposition at Sample Mean and Final Year (1995) 1/

(In percentage points of inflation) 2/

article image

Final specification from Table 4, full sample period; alternative specification from Table 5.

“-” indicates that the variable was not included because of lack of significance in the specification search.

Includes regression residuals in 1995; residuals are zero at the sample mean.

Combined effect of contemporaneous and lagged variables (see Tables 4 and 5).

Table 7.

Sensitivity to Relative Price Indicators: Unweighted Indicators 1/

(Dependent variable: quarterly CPI inflation rate)

article image

For the t-statistics reported in parenthesis, three asterisks indicate statistical significance at I percent level; two asterisks at the 5 percent level; one asterisk at the 10 percent level; - indicates that the variable was not included because of lack of significance in the specification search. OLS-HCSE refers to correction of bias in least squares standard errors and t-statistics using a heteroschedasticity-consistent variance-covariance estimator.

Unweighted variance and skewness, respectively.

V. Relative Price Adjustment in Five Transition Economies

In light of the fact that the econometric results can provide only a broad indication of the influence of relative price adjustment on inflation, as an illustrative exercise, this section examines the evidence from five selected countries—the Czech Republic, Poland, Estonia, Moldova, and Russia—to obtain more insights into the role of relative price adjustment. The analysis is framed in terms of a few specific issues which are relevant to the policy debate:

  • What factors underlie variability in relative prices as measured by the variance and skewness of the price distributions?

  • Do the sources of money growth give some indication of how relative price and other shocks are accommodated under different policy regimes?

  • Does the analysis of relative price adjustment shed some light on the factors that could explain the marked real appreciation in transition economies?

The five selected countries have all achieved an initial macroeconomic stabilization and implemented some amount of liberalization and adjustment of relative prices. Nevertheless, they represent a range of progress in these areas, particularly with regard to inflation reduction, and also reflect some diversity in macroeconomic and structural policies (Table 8). The timing of their initial liberalization varies and they also adopted somewhat different strategies: the Czech Republic, Poland, and Estonia followed a “big-bang” approach, undertook extensive liberalization early, and used an explicit exchange rate anchor, while Moldova and Russia adopted a gradual approach to both stabilization and structural reform and relied on a more flexible exchange rate policy (see Annex II).

Table 8.

Country Performance in Key Areas

article image

Excludes proceeds from privatization as fiscal revenues.

Includes proceeds from privatization as fiscal revenues.

June-December 1992.

Liberalization index and reform grouping from De Melo, Denizer, and Gelb (1995) based on rankings in three areas: internal markets, external markets, and private sector entry; reflects the judgment of the authors, World Bank staff, and work by the EBRD.

A. Factors Underlying Relative Price Variability

A common feature of the present sample is the periodic sharp peaks in the measured variance of relative prices, concurrent with peaks in inflation, particularly in the early stages of transition (Chart 1). For the five countries, these peaks largely coincide with episodes of trade and price liberalization, wage and administered prices increases, tax reform, and terms of trade shocks, in some instances accompanied by monetary accommodation (particularly in Russia), and suggest that spikes in variance largely reflect exogenous relative price shocks stemming from structural change, at times accommodated by money growth. In terms of the models discussed in Section II (Box 1), this would suggest either a simultaneous effect or a direction of causality running from relative price variability to inflation.41

CHART 1
CHART 1
CHART 1
CHART 1

Inflation and Relative Price Variance

Citation: IMF Working Papers 1996, 138; 10.5089/9781451930061.001.A001

Although variance declines sharply following the initial bouts of price liberalization in all five countries, it appears to remain high relative to market economies. Comparisons of variance across countries need to be interpreted with caution because the measures also reflect the weights and level of disaggregation of the CPI data. Nevertheless, as a rough indication, a comparison of the Theil variance and unweighted variance for 1995 with corresponding indices for Argentina, Greece, Italy, and the United States suggests that relative price variance in these transition economies—with the exception of the Czech Republic—is substantially higher than in the market economies (Table 9). In addition, the contribution from variance within tradables tends to be somewhat greater in the transition economies, although this may reflect the relatively low weight given to nontradables in the CPI (particularly for Moldova and Russia).

Table 9.

Relative Price Variance in Some Transition and Market Economies

article image

All figures multiplied by 10,000.

Given previous indications that downward price rigidity may be an important factor in translating relative price shocks to an increase in overall inflation, can an analysis of skewness reveal any insights about factors—such as the cost-recovery hypothesis—which could explain a prolonged period of relative price adjustment? A comparison of inflation rates of cost-recovery items with overall inflation indicates that prices of these items tend to be adjusted discontinuously rather than smoothly (Chart 2). This may reflect the administrative and political costs of changing such prices which induce the authorities to recover the deterioration of these relative prices (during periods when prices are not changed) by periodically raising them well in excess of average inflation.42 Given this pattern of periodic price adjustments, goods for which cost-recovery is a consideration may be expected to dominate as outliers in periods when the distribution of prices is positively skewed. Moreover, if these prices increase in relative terms on a sustained basis as argued by the cost-recovery hypothesis, rather than simply being adjusted periodically to keep up with average inflation, they would also tend to dominate the distribution of prices when inflation rates are calculated on a cumulative basis.43

CHART 2
CHART 2

Consumer Price Inflation

(In percent)

Citation: IMF Working Papers 1996, 138; 10.5089/9781451930061.001.A001

An analysis of the distribution of individual inflation rates calculated on a cumulative and quarterly basis suggests that in three countries—the Czech Republic, Poland, and Estonia—cost-recovery items indeed dominate the distributions of cumulative inflation rates; in addition, on average they frequently rank among the ten most extreme outliers in periods when the distributions of quarterly inflation rates are positively skewed (Chart 3): a pattern which is evident in both the early and later years of the transition.44 In Russia, cost-recovery items are less dominant, although this relates to the much higher degree of disaggregation of the CPI data in which items with strong price seasonality (e.g., fruits and vegetables) appear individually.45 In Moldova, cost-recovery items dominate the distribution of cumulative inflation rates in the early years, but are strikingly absent from the later years—suggesting that these prices, which are largely administratively set, have not been adjusted during 1994-95 when overall inflation declined markedly.

CHART 3
CHART 3
CHART 3
CHART 3
CHART 3
CHART 3

Relative Price Changes in Cost-Recovery Items

Citation: IMF Working Papers 1996, 138; 10.5089/9781451930061.001.A001

1/ Bars indicate the ratio of individual inflation rates to the mean unweighted inflation rate, both calculated on a cumulative basis.2/ Bars indicate the ratio of individual inflation rates to the mean unweighted inflation rate, averaged over the quarters in which the distribution (of individual inflation rates) was positively skewed. Numbers in parentheses indicate percentage of quarters when individual inflation rates were at least 1.5 times the mean.3/ In the “cost-recovery” category.

B. Sources of Broad Money Growth

The impact of relative price adjustments, including those that contribute to real appreciation, on inflation clearly depends on the extent to which these adjustments are accommodated by money growth. A decomposition of broad money growth reveals some marked differences in the sources and extent of money growth (Table 10). In the Czech Republic and Estonia, base money growth almost entirely reflects increases in NIR under a fixed exchange rate regime, notwithstanding some sterilization in the former case. In Poland, both NIR and NDA play a role, consistent with a crawling peg accompanied by periodic devaluations. By contrast, in Moldova and Russia, not only does the expansion in NDA almost entirely dominate the growth in base money, but the growth in both variables is much larger than in the other cases. These patterns are suggestive of two aspects of the role of policies in transmitting relative price shocks to inflation. On the one hand, a fixed or crawling peg exchange regime can result in quite rapid and significant endogenous money growth—frequently at annual rates in excess of 20 percent in one quarter—which could accommodate inflationary pressures from relative price adjustments.46 Hence, other things equal, when relative price adjustments are significant, a money anchor could result in lower inflation because of stronger monetary control, including through nominal appreciation. On the other hand, in the absence of an exchange rate anchor, the accommodation of relative price shocks through discretionary credit creation can result in even greater money expansion because of the absence of a deterrence effect on lax financial policies and because of the difficulties of targeting money when money demand is unstable. Hence, countries with exchange rate pegs may end up with lower levels of inflation because monetary accommodation is more limited.47

Table 10.

Decomposition of Changes in Broad Money (1993-95)

(End-of-period rates of change, unless otherwise noted)

article image

Broad money growth equals ((1+percentage change in money multiplier)*(1 + percentage growth in base money))-1. For small changes, this is approximately equal to the sum of the rates of change in the money multiplier and base money.

Calculated residually as the difference between base money and NIR.

Includes only assets and liabilities in convertible currencies. Evaluated at exchange rates of end-December 1992.

C. Some Factors Underlying Real Appreciation

All five countries have experienced considerable real appreciation since the beginning of the transition, although the measured degree of appreciation is sensitive to the choice of index (Chart 4).48 Undervaluation is frequently cited as an explanation for this appreciation, even in the later stages of transition. Although international price comparisons are often used as evidence, direct comparisons can be misleading if differences in PPP-adjusted GDP are not taken into account (see Section II). A price comparison for 1993 suggests that price levels in all five countries, except Poland, were significantly lower, relative to a comparator Western European country (Austria) than would be indicated by differences in PPP-adjusted GDP (Table 11). The discrepancy is especially marked for Moldova and seems surprisingly large for the Czech Republic.

CHART 4