Czech Republic
Selected Issues and Statistical Appendix
Author: International Monetary Fund
Publication Date:
25 Jul 2001
eISBN:
9781451810141
Language:
English
Keywords:
ISCR; CR; price; CNB; exchange rate movement; inflation dynamics; price deregulation; CNB's inflation targeting framework; inflation in the Czech Republic; exchange rate change
Download PDF (2.8 MB)

This study addresses the following questions: (i) what is the process driving inflation in the Czech Republic? and (ii) in light of the ongoing structural transformation of the economy, can a reliable model of inflation be formulated for analysis and forecasting? The following statistical data are also included in detail: composition of GDP, agricultural production, average monthly earnings, electricity production and consumption, transfers to households, monetary survey, structure of domestic currency deposits, selected interest rates, lending and deposit rates of commercial banks, and so on.

Abstract

This study addresses the following questions: (i) what is the process driving inflation in the Czech Republic? and (ii) in light of the ongoing structural transformation of the economy, can a reliable model of inflation be formulated for analysis and forecasting? The following statistical data are also included in detail: composition of GDP, agricultural production, average monthly earnings, electricity production and consumption, transfers to households, monetary survey, structure of domestic currency deposits, selected interest rates, lending and deposit rates of commercial banks, and so on.

Modeling Inflation in the Czech Republic: Short-Run and Long-Run Dynamics1

A. Introduction

1. Over the past several years, consumer price inflation in the Czech Republic has come down from an average of over 20 percent in 1993 to around 4–5 percent at present (Figure 1). Despite the ongoing process of price deregulation, the decline in inflation reflected to a considerable extent the slack in the economy created since the recession of 1997–99, wage demands that were more in line with productivity developments, and the disinflation efforts of the CNB in the context of an inflation-targeting framework (see paragraphs 7 and 8). In fact, during the recession of 1997–99, inflation reached a low of 1.1 percent (year-on-year) in July 1999, but has since climbed higher, reflecting the economic recovery and the influence of special factors such as food and fuel price increases. Excluding food items and administered prices, year-on-year inflation remains at around 2¼ percent (see Figure 1). Against the background of a general absence of underlying inflationary pressures, the result has been a growing perception by market participants that the Czech Republic has entered an environment of lower and more stable inflation.

Figure 1.
Figure 1.

Czech Republic: Price, Wage, and Productivity Developments, 1993-2001

Citation: IMF Staff Country Reports 2001, 112; 10.5089/9781451810141.002.A001

Sources: Czech National Bank; Czech Statistical Office; and Fund staff calculations.1/ Net inflation is CPI inflation excluding administered prices and the effects of changes in indirect taxes.2/ Adjusted inflation is net inflation excluding the effects of changes in food prices.3/ In industrial enterprises with 20 or more employees.

2. This study addresses the following questions: (i) what is the process driving inflation in the Czech Republic; and (ii) in light of the ongoing structural transformation of the economy, can a reliable model of inflation be formulated for analysis and forecasting? In particular, the study estimates a model of consumer price inflation that encompasses both short- and long-run dynamics, thereby acknowledging that inflation is a dynamic process responsive to both temporary and permanent influences. Short-run forecasts are generated and the findings are used to highlight issues that the CNB could consider in operating under the new inflation-targeting framework announced in April 2001 (see paragraph 9).

3. There are several possible causes of inflation in an open economy, and no single cause is adequate to explain the data. Therefore, models of inflation that attempt to encompass several theories have a better chance of empirical success. Recognizing the trade-offs between model structure and simplicity, this study formulates a model for inflation in the Czech Republic that accounts for several influences, but relies on a parsimonious single-equation representation to describe the inflation process. To put recent inflation developments in perspective, Section B provides a historical overview and briefly discusses the Czech Republic’s monetary policy regime. Section C reviews the challenges posed by the available data. Section D derives equilibrium long-run relationships in markets that have a direct influence on inflation, and in the process evaluates the impact of individual factors on inflation. Section E describes an inflation equation that captures both short- and long-run dynamics, and assesses the stability of forecasts generated by the estimated equation. Finally, Section F contains concluding remarks and discusses policy issues.

4. The main results can be summarized as follows:

  • Inflation in the Czech Republic appears to be driven primarily by foreign exchange, money and labor market developments over the long run, and by the exchange rate, markup behavior in the economy, and changes in administered prices in the short run. Inflation trends in recent years reflected to a significant extent wage dynamics (some of them were passed through producer prices) and the transmission of external influences through the exchange rate channel.

  • The model suggests that the speed of adjustment to disequilibria in the long-run influences is very slow, and that most of the variation in inflation can be explained by short-run factors.

  • Despite the ongoing structural changes in the Czech economy, the estimation was able to derive stable structural relationships from the available data. However, forecasting has been challenging: while the model’s within-sample forecasts were very good, out-of-sample forecasts were very sensitive to the choice of path for exchange rate movements.

  • The results confirm the empirical experience that forecasts of Czech inflation are not very accurate. Thus, the CNB should rely on a range of methodologies to model the inflation process and produce forecasts.

B. Historical Perspective

Inflation developments

5. The high inflation rates in the Czech Republic during the first half of the 1990s reflected primarily the process of price deregulation and the adjustment of relative prices as the economy was opening up to the rest of the world.2 Inflation rates surged initially and remained at high levels as the authorities embarked on a program of price deregulation to bring administered prices closer to market levels. Furthermore, the increasing exposure of the Czech economy to the rest of the world facilitated the adjustment of relative prices, which, coupled with nominal price stickiness, contributed to the high inflation rates.

6. Labor market developments and external influences also played a role. Up until the onset of the recession in 1997, governance problems in the Czech corporate sector facilitated large nominal wage increases that were not supported by productivity trends. As a result, nominal wage growth was kept in the double digits, surpassing productivity (lower panel of Figure 1). Against the backdrop of an exchange rate peg and limited support from fiscal policy, the task of controlling inflation was complicated by large capital inflows in 1994–95, which rendered monetary management increasingly difficult.

7. The Czech Republic saw its inflation rate fall toward the end of the 1990s. Inflation began falling to lower levels as the 1997–99 recession eased demand pressures and exposed problems in the corporate sector, which could no longer accommodate high wage demands. The switch to inflation targeting beginning in early 1998 provided an anchor for inflation expectations after the abandonment of the peg in the wake of the May 1997 currency crisis. The monetary policy tightening that ensued controlled the crisis successfully, but at a cost to activity. Indeed, the disinflation process has been more rapid than the CNB had envisaged, reflecting in part the severity of the recession: net inflation, which the CNB has been targeting, fell short of the target three years in a row (Figure 2).3 Against the background of very weak domestic demand, the CNB began to ease monetary policy in July 1998. In addition, the appreciation of the koruna in recent years—mainly reflecting buoyant foreign direct investment (FDI) inflows—has helped ease inflation pressures further.

Figure 2.
Figure 2.

Czech Republic: The CNB’s Inflation Targeting Framework, 1997-2005

Citation: IMF Staff Country Reports 2001, 112; 10.5089/9781451810141.002.A001

Sources: Czech National Bank; and Fund staff estimates.

8. The adoption of inflation targeting provided the means for guiding macroeconomic policies and signaling the markets. The CNB announced a medium-term target—in terms of net inflation—of 1–3 percent by end-2005. Intermediate targets were defined based on the 12-month, end-December net inflation rate, consistent with a process of gradual disinflation. Quarterly inflation reports and regular communication with the public have helped convey the intentions of monetary policy.

9. Net inflation targeting was not without problems. Net inflation consisted of about 82 percent of the CPI basket and was a rather abstract concept for the general public to understand. In addition, movements in net inflation were dominated by the volatile food component, which accounted for over one-third of net inflation.4 Finally, the CNB found challenging the task of formulating reliable models for net inflation, owing mainly to the fact that the modeled relationships described the behavior of economic agents who optimize based on overall, not net, inflation. In view of these shortcomings, the CNB recently switched to targeting headline inflation (Box 1).

New Inflation Targeting Framework

In April 2001, the CNB announced changes to its inflation-targeting framework. Under the existing framework, the CNB has an end-year target range for net inflation (i.e., headline CPI excluding administered prices and the impact of indirect tax changes), which is announced about 20 months in advance. The new framework shifts to targeting headline inflation and specifies a continuous linear and declining target band from January 2002 through end-2005 (see lower panel of Figure 2). The changes were primarily motivated by the desire to increase the transparency of monetary policy and help better anchor inflation expectations. The target band for headline inflation was specified at 3–5 percent in January 2002, declining to 2–4 percent in December 2005. This range is consistent with the CNB’s medium-term target for net inflation of 1–3 percent announced in 1999, and is based on the assumption that the impact on headline inflation of changes in administered prices will not exceed 1.5 percentage points at any point in time. Should the effect of changes in administered prices exceed the amount assumed, the CNB intends to invoke escape clauses. The CNB has yet to announce details of the operational aspects of the new framework, including the design and use of escape clauses.

C. Data Issues

10. The quality of economic statistics in the Czech Republic is generally good and sufficient for the assessment of the macroeconomic situation, In general, series are available with reasonable lags for all principal economic indicators. However, the economic statistics of the Czech Republic are not without shortcomings, mainly reflecting problems encountered in transition economies, such as the limited information content of the data and the short statistical series. In particular, the ongoing structural transformation of the economy has made it difficult to capture stable structural relations in the data. Moreover, existing economic series are short (generally beginning with the formation of the Czech Republic in January 1993, but later in several instances) and close to half of the series length covers the recession of 1997–99, a rather atypical period. As a result, econometric estimation becomes particularly problematic. Given the very limited number of available annual observations, estimation in this study is carried out using monthly series.5

D. Influences on Inflation

11. This section examines the structure of various markets that influence inflation, and derives stable long-run relationships. The objective is to derive the determinants of inflation by examining each market individually, and to subsequently find how disequilibria in these markets (i.e., deviations from the estimated long-run relationships) affect Czech inflation.

Long-run influences on inflation

12. The analysis sought to establish long-run cointegrating relationships (interpreted as steady-state equilibria) in a range of markets. Regressions using monthly data for 1994:1–2001:3 found empirical long-run links between Czech inflation and the following fundamental determinants:

  • Wage inflation, capturing the impact on inflation of wage growth in excess of the long-run equilibrium (where wages, consumer prices, producer prices, productivity, and unemployment enter the equilibrium).

  • Imported inflation, capturing the influence of external factors—most notably, foreign prices—on inflation through a modified version of the Purchasing Power Parity (PPP) hypothesis to account for Balassa-Samuelson effects.

  • Other external influences, such as exchange rate movements, derived from the Uncovered Interest Parity (UIP) hypothesis. The hypothesis defines the interest differential between the Czech Republic and the rest of the world as being equal to the expected change in the exchange rate. This term represents the main transmission channel of monetary policy changes.

  • Monetary inflation, capturing the impact on inflation of excess demand for money.

13. The statistical evidence suggests that:

  • There is a very stable long-run relationship between the fundamental determinants of inflation in the labor market. The labor market relationship shows that real wages in the Czech Republic are: (i) positively related to producer prices (capturing the ability of producers to pass on wage increases by raising their prices) and labor productivity; and (ii) negatively related to the unemployment rate (capturing outside opportunities) The most important factor appears to be producer prices. Unemployment does not appear to have played an important role in the determination of Czech wages over the sample period.6

  • The PPP relationship appears to hold only when enhanced to capture Balassa-Samuelson effects. In particular, the PPP relationship—in other words, the real effective exchange rate (REER)—is a positive function of the productivity differential between the Czech Republic and its trading partners.

  • A pure UIP relationship could not be supported by the data. The estimation suggested that changes in the exchange rate of the koruna are not primarily related to interest-sensitive capital flows, but to other factors, including FDI and expectations about future inflows.

  • The money demand relationship was unstable in the sample. Nevertheless, it appears to have a significant effect on inflation over the long run. In particular, the relationship suggested a very inelastic demand for money with respect to real output and interest rates.7

14. All estimated coefficients in the long-run relationships have appealing economic interpretation and the expected signs.8 The next step is to combine these long-run relationships with short-run influences in a single equation for inflation.

E. Model of Inflation

15. Deviations of inflation from its long-run equilibrium levels do not always reflect disequilibria; they could also reflect cyclical or temporary factors. A model that explicitly separates short- from long-run dynamics has analytical appeal and allows the generation of short-run forecasts (see the Appendix). Inflation was modeled in an error-correction representation, and the estimated relationship was used to derive short-run forecasts.

16. The main results can be summarized as follows:

  • Short-run influences dominate inflation dynamics in the Czech Republic. By far, the most important short-run determinant was exchange rate movements, The other significant short-run determinants were producer and administered prices, the former possibly capturing a markup behavior at the retail level, with the latter capturing the direct impact of administrative measures on Czech inflation.

  • The response of inflation to disequilibria (i.e., deviations from the estimated long-run relationships) is small and prolonged. Among the long-run influences, the UIP relationship (namely, monetary policy and exchange rate movements), and monetary and wage inflation were the most important.

  • External effects dominate the determination of inflation in both the short and the long run. Exchange rate movements were also the most important influence on inflation over the long run. Labor market developments, especially the ability of producers to pass on wage increases, were another important factor. Imported inflation was not found to be a significant influence, suggesting that some price stickiness exists, and that the adjustment of the REER to disequilibria is achieved primarily through the exchange rate.9 The model showed a rather weak direct impact of monetary policy changes—captured through a policy-driven interest rate used in the estimation—-in the long run, but a quite substantial impact when accounting for the indirect effect through exchange rate movements. Changes in monetary policy also appeared as a significant determinant of short-run inflation dynamics, but with a sign opposite to that expected.10

  • It is a challenging task to generate accurate forecasts of inflation in the Czech Republic. Exchange rate movements dominate inflation dynamics in both the short and the long run, thus making conditional forecasts highly sensitive to different paths for the exchange rate. Small changes in assumptions about the exchange rate resulted in substantially different paths for inflation. In addition, when dynamic forecasting was performed (in which case a vector autoregression generated forecasts for the variables chosen to be determined endogenously), the resulting forecasts involved a trajectory that was below the latest forecast published by the CNB.

F. Concluding Remarks and Policy Implications

17. This study estimated a model of inflation for the Czech Republic and assessed its forecasting ability. A number of theories of inflation were used to gain a better understanding of developments and to increase confidence in the findings.

18. The results suggest that inflation in the Czech Republic is primarily driven by (i) developments in the foreign exchange, money, and labor markets over the long run; and (ii) external influences channeled through the exchange rate, domestic pricing factors (markup behavior at the retail level), and the impact of administered measures (i.e., changes in administered prices) in the short run.11 The impact of changes in monetary policy appears stronger through the exchange rate channel. Czech inflation is very slow to respond to disequilibria in its long-run determinants. Most of the immediate adjustment is captured by the exchange rate (which adjusts almost instantaneously to external disequilibria) and by producer prices (which presumably reflect wage developments at the firm level). However, the results, based on a backward-looking assessment, should be interpreted with caution, given the data limitations and the fact that more recent events—most prominently the rebound in activity and the gradual unblocking of the credit channel—are not yet fully captured in the data.

19. Short-run factors are by far the most important determinants of Czech inflation dynamics. In decreasing order of importance, exchange rate movements, producer prices, and administered prices account for most of the short-run variation in inflation. Long-run influences include monetary policy variables and the exchange rate, labor market developments, and monetary inflation. The data do not support the UIP theory of exchange rate changes: interest-sensitive capital flows do not appear to be the main driving forces of exchange rate changes.

20. The econometric results provide sufficient support to the hypothesis of Balassa-Samuelson effects over the sample period. The results suggest that productivity gains in the tradable goods sector stemming from the increasing openness of the Czech economy may have played a significant role in the appreciation of the REER in recent years.

21. Forecasting performance is tied to the ability to forecast exchange rate changes. The predominance of short-run factors (which appear with large and significant coefficients) in the error-correction representation makes forecasts very sensitive to the forecast path for each of the main determinants, especially the exchange rate.

22. The results have implications for monetary policy. The model sheds some, though not enough, light on the transmission lags of monetary policy. Most of the immediate impact of changes in policy instruments is felt through the exchange rate almost contemporaneously (with a lag of one month), The appearance of the (lagged) policy variable with a positive sign could be interpreted as a reaction function of the CNB, with the bulk of the effect of rising interest rates to be felt later. As regards the full impact (long-run effect) of changes in policy interest rates, the model shows a rather small and prolonged direct effect through the UIP relationship, but a more substantial effect through the indirect channel of the exchange rate.

23. The results also have implications for structural policies. The exposure of problems in the enterprise sector prompted the authorities to embark on a program of enterprise restructuring, Changes in management, operational and balance-sheet restructuring, or even outright bankruptcy have been accompanied by a general containment of wage increases in the traditional enterprise sector, a source of earlier inflation pressures. By persevering and completing enterprise reform, the Czech authorities will also ensure that wage developments remain in line with productivity.

24. Finally, the results have implications for the operational aspects of the new inflation targeting framework. As regards the ability to forecast inflation under the new inflation targeting regime, the methodology employed in this study suggests that forecasting inflation in the Czech Republic is a very challenging task. The CNB will therefore need to develop reliable modeling and forecasting techniques to assess inflation dynamics and to communicate forecasts to the public that will not risk damaging its credibility. No single methodology can capture all aspects of the inflation process. Given the caveats concerning this econometric estimation, forecasting, and the available data, it is appropriate to consider a range of methodologies (including structural VARs, and calibrated models) to model the inflation process in the Czech Republic and to generate forecasts. As the transformation process of the Czech economy advances and statistical series become longer, modeling and forecasting inflation in the Czech Republic should become an easier task. In the meantime, the CNB could explain more clearly to market participants the main assumptions behind its forecasts and the reasons for ex post deviations of inflation from its targeted path. After all, it is important not only to derive relatively good forecasts of inflation, but also to explain clearly to the markets the reasons for deviating from the targeted band.

References

  • Coorey, S., M. Mecagni, and E. Offerdal, 1996, “Disinflation in Transition Economies: The Role of Relative Price Adjustment,” IMF Working Paper 96/138, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • De Brouwer, G., and N. R. Ericsson, 1998, “Modeling Inflation in Australia,” Journal of Business and Economic Statistics, Vol. 16(4), pp. 433–449.

    • Search Google Scholar
    • Export Citation

  • Hendry, D. F., and J. A. Doornik, 1999, Empirical Econometric Modeling Using PcGive, (West Wickham: Timberland Consultants Ltd.).

  • Johansen, S., 1988, “Statistical Analysis of Cointegration Vectors,” Journal of Economic Dynamics and Control, Vol. 12, pp. 231–254.

    • Search Google Scholar
    • Export Citation

  • Johansen, S., 1995, Likelihood-based Inference in Cointegrated Vector Autoregressive Models, (Oxford: Oxford University Press).

  • Juselius, K., 1992, “Domestic and Foreign Effects on Prices in an Open Economy: The Case of Denmark,” Journal of Policy Modeling, Vol. 14(4), pp. 401–428.

    • Search Google Scholar
    • Export Citation

  • Laursen, T., 1998, “Inflation and its Determinants in the Czech Republic,” in Chapter II of Czech Republic—Selected Issues, SM/98/30, 1/30/98, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

APPENDIX: Determinants of Inflation in the Czech Republic

1. This appendix describes a model of inflation in the Czech Republic based on monthly data for 1994:1-2001:03, and derives monthly forecasts through end-2002.

A. Conceptual Issues

2. A number of factors are expected to affect inflation in the long run. Taking into account data availability, the treatment of these factors in this study is discussed below:1

  • Labor-market wage determination: On the worker side, wage demands are based on the cost of living, worker productivity, and the presence of outside opportunities (proxied by the unemployment rate). On the employer side, wage offers are based on the productivity of labor, and the ability of producers to raise their own prices (the easier it is for producers to raise their prices, the more they are inclined to accommodate higher wages). In equilibrium, real wages are expected to be a positive function of producer prices and productivity and a negative function of unemployment. Wages in excess of the long-run equilibrium are thus expected to exert a positive influence on inflation in the long run.

  • The external environment: This is modeled as two separate, but interacting, markets: (i) the goods market, described by a PPP relationship that is enhanced to take into account the secular rise in the Czech Republic’s REER; and (ii) the external capital market, described by a UIP relationship. In an equilibrium in the goods market, the REER is expected to be a function of the productivity differential between the Czech Republic and its trading partners. The higher the differential, the higher the equilibrium REER appreciation that should be expected. REER appreciation in excess of the long-run equilibrium is expected to exert a negative influence on inflation in the long run. In the external capital market, the differential in bond rates is a function of the expected rate of change of the exchange rate. With the expected change in the exchange rate equaling, on average, the actual change over the long run, the interest rate differential in the previous period is equal to the rate of change of the exchange rate at present. The PPP and UIP relationships allow for interactions between the two, where a deviation from PPP (as defined to account for the secular appreciation of the REER) can be corrected not only through adjustments in prices, but also through the exchange rate (in the case of sticky prices). Interest rate differentials in excess of the long-run equilibrium are expected to exert a negative influence on inflation in the long run.

  • Money market: Demand for real money balances is modeled as a positive function of real output and the deposit rate and a negative function of the bond rate and inflation. Excess money balances are expected to affect inflation positively in the long run.

B. Estimation Issues

3. Estimation of the above long-run cointegrating relationships is complicated by a number of factors: (i) not all the explanatory variables prescribed by theory are available in monthly series (namely real GDP), and consequently the series were interpolated. This reduces the precision of econometric estimates because they are based on inferred rather than actual data; (ii) the ongoing process of structural transformation of the economy over the sample period makes the estimation of structural relationships problematic; (iii) the clogging of the credit channel and the problems in the banking system have destabilized monetary aggregates, have affected the transmission mechanism of monetary policy, and have complicated the estimation of relationships that deal with money demand, and (iv) the changes in the exchange and monetary policy regimes over the sample period created structural breaks in the data and weakened the relation between policy variables, the exchange rate, and inflation.

C. Data

4. The following series were used in this study (notation in parentheses):

Average nominal wages in industry, expressed in koruni per hour (WAG); the consumer price index (CPI); the producer price index (PPI); productivity, measured by the value of output per unit of labor in the Czech industry (PROD); the unemployment rate (UNE); foreign prices, proxied by the German CPI (GCPI);2 the exchange rate of the koruna to the DM (EXC); the productivity differential between the Czech Republic and Germany, proxied by the ratio of nominal productivity in the Czech industry deflated by the Czech PPI to real productivity in German industry from the OECD data base (PROD DIFF); the Czech bond rate, proxied by the three-month PRIBOR rate (CINT);3 the German Treasury rate (GINT); real money, proxied by M2 deflated by CPI (MON); real output, proxied by real GDP (RGDP); the Czech deposit rate (CDEP); and the index of administered prices (ADM). All data were expressed in natural logarithms, except for interest rates. In contrast to the conventional notation, upper case letters denote the natural logarithm of the respective variable. Furthermore, variables preceded by D_denote first differences.

D. Integration

5. To determine the appropriate estimation procedure, tests for nonstationarity of the variables listed above were carried out using the Augmented Dickey-Fuller (ADF) method, which looks for the presence of unit roots in the series. Table 1 lists the up-to-twelfth-order ADF statistics for the variables above. According to the ADF tests, the null hypothesis of a unit root in the log-levels of the series—that is, each variable is integrated of order one: 1(1)—cannot be rejected for any of the variables. ADF tests (not shown) of the null hypotheses of integration of higher order rejected the null. All variables are thus stationary in first differences, and cointegration analysis among the level variables is required.

Table 1.

ADF Statistics Testing for a Unit Root

article image
Note: Critical values are -2.9 (5 percent level), and -3.5 (1 percent level). Lag length was determined by the choice of the lag with the highest ADF statistic in absolute value; see Hendry and Doornik (1996), page 42.

E. Cointegration

6. The Johansen procedure is used for the cointegration analysis.4 The Johansen procedure is a foil information maximum likelihood estimate for vector autoregressive systems, and, as such, it is not concerned about the endogeneity of the explanatory variables. Nonetheless, the procedure imposes a heavy toll on the degrees of freedom and on the precision of the econometric estimates in small samples because it uses a lag structure. The procedure searches for the existence of one or more long-run cointegrating relationships between the selected variables.

7. Table 2 reports the statistics from the Johansen procedure for the labor market.5 Only the final specification is reported.

Table 2.

Johansen Test of Existence of Long-run Relationships in the Labor Market

article image

Double asterisks denote significant test statistics at the 1 percent level. Critical values at the 5 percent level are 68.5 for the first hypothesis, and 47.2 for the second hypothesis.

Note: The lag length in the vector autoregression was set to two months on the basis of the Schwarz criterion.

8. The Johansen procedure found evidence of the following long-run relationship between the fundamental variables in the labor market:6

ECM(Wage)t: WAGt - CPIt = 0.7377*PPIt + 0.1168*PRODt - 0.0063*UNEt

9. The estimation found evidence of a long-run cointegrating relationship between the real wage in industry and its fundamental determinants. However, one of the criteria for the presence of cointegrating vectors suggests, albeit very weakly, a second cointegrating vector (see Table 2), The first cointegrating vector (shown above) describes the determination of equilibrium wages in the labor market. The second cointegrating vector (not shown) suggested a positive association between the unemployment rate and the wedge between consumer and producer prices, and a positive association between unemployment and productivity. Historically, such an association is found in the data (Appendix Figure 1). The second cointegrating vector can be interpreted as describing the processes of firing/hiring and price setting by producers, where by firing the least productive, producers can raise overall productivity and are therefore able to reduce (or contain the rise of) producer prices. Hence, this vector does not describe wage determination and we can ignore it in the remainder of this study.

Figure 1.
Figure 1.

Czech Republic: Inflation and Labor Market Conditions, 1994–2001

Citation: IMF Staff Country Reports 2001, 112; 10.5089/9781451810141.002.A001

Sources: Czech Statistical Office; Czech National Bank; and Fund staff estimates.1/ Defined as the difference between the CPI (in logarithms) and the PPI (in logarithms).

10. All estimated coefficients of the selected vector have the anticipated signs. The coefficient for PPI is relatively large, reflecting the importance of pricing power by producers in the wage determination process.7 The feedback coefficients suggest a relatively slow adjustment from wage disequilibrium to the price equation (see Table 2); indeed, the error-correction representation (see Section F) confirmed that the impact on inflation of a disequilibrium in the labor market is quite small and prolonged.

11. Table 3 reports the statistics from the Johansen procedure for the external goods market (PPP relationship). Only the final specification is reported.

Table 3.

Johansen Test of Existence of Long-run Relationships in the External Goods Market

article image

Double asterisks denote significant test statistics at the 1 percent level Critical values at the 5 percent level are 47.2 for the first hypothesis, and 29.7 for the second hypothesis.

Note: The lag length in the vector autoregression was set to one month on the basis of the Schwarz criterion.

12. The Johansen procedure found evidence of the following long-run relationship between the fundamental variables in the PPP relationship:

ECM(PPP)t: CPIt - EXCt - GCPIt = 4.9136*PROD_DIFFt

13. The estimation found evidence of a long-run cointegrating relationship between the REER of the Czech Republic (where Germany represents the rest of the world) and the productivity differential between the Czech Republic and Germany. All estimated coefficients have the anticipated signs.8 The coefficient for the productivity differential is relatively large, reflecting the presence of substantial Balassa-Samuel son effects in the sample period. The feedback coefficient for CPI suggests a relatively slow adjustment from an REER disequilibrium to the Czech price equation. The feedback coefficient for the exchange rate is much larger, suggesting price stickiness, where the required adjustment takes place through the exchange rate rather than prices. The error-correction representation confirmed this finding (see Section F).

14. Table 4 reports the statistics from the Johansen procedure for the external capital market (UIP relationship).

Table 4.

Johansen Test of Existence of Long-run Relationships in the External Capital Market

article image

Double asterisks denote significant test statistics at the 1 percent level. Critical values at the 5 percent level are 29.7 for the first hypothesis, and 15.4 for the second hypothesis.

Note: The lag length in the vector autoregression was set to three months on the basis of the Schwarz criterion.

15. The Johansen procedure found evidence of the following long-run relationship between the fundamental variables in the UIP relationship:

ECM(UIP)t: CINTt-1 - GINTt-1 = 26.711*D_EXCt

where CINT and GIN are as defined in Section C, and D_EXC is the change in the koruna/DM exchange rate at time t. A positive D_EXC indicates depreciation of the koruna

16. The results suggest that the exchange rate of the koruna reacts only marginally to deviations of the interest rate parity from long-run equilibrium (i.e., a 1 percentage point increase in the interest differential in the previous period would require a 0.04 percent—i.e., 1 percent times 1/26.711—depreciation of the exchange rate for equilibrium to be restored).9 This is strong evidence that the main contributing factor to movements in the exchange rate of the koruna is not interest-sensitive capital flows, but other determinants, including FDI. Irrespective of the appeal of the UIP hypothesis as an accurate description of exchange rate dynamics in the Czech Republic, the long-run relationship shown above is meaningful and sufficient to capture the interaction between the external capital market and exchange rate changes on the one hand, and the inflation rate on the other. The feedback coefficient for the exchange rate is the largest of all, suggesting that deviations from this long-run equilibrium are corrected mainly through adjustments—albeit slow—of the exchange rate (see Table 4).

17. Table 5 reports the statistics from the Johansen procedure for the money demand relationship.

Table 5.

Johansen Test of Existence of Long-run Relationships for Money Demand

article image

Double asterisks denote significant test statistics at the 1 percent level. Critical values at the 5 percent level are 68.5 for the first hypothesis, and 47.2 for the second hypothesis.

Note: The lag length in the vector autoregression was set to one month on the basis of the Schwarz criterion.

18. The Johansen procedure found evidence of the following long-run relationship between the fundamental variables in the money demand relationship:

ECM(Mon)t: MONt = + 0.0070*RGDPt - 0.2581*CINTt + 0.2901*CDEPt - 0.6469*D_CPIt

19. The estimated relationship is a meaningful money demand function: all estimated coefficients have the anticipated signs, although real money demand appears very inelastic with respect to real output and the interest rates.10 The feedback coefficients suggest a relatively slow adjustment from money demand disequilibrium to the inflation equation (see Table 5). The error-correction representation confirmed that the impact of a disequilibrium in money demand on inflation is considerably small and prolonged (see Section F).

F. Single-equation Modeling (Error Correction)

20. This section combines the relationships established above (long-run dynamics) to describe a single-equation, conditional, and parsimonious model for Czech inflation that encompasses both short- and long-run dynamics. The short-run dynamics are derived from the following vector of changes in the log-levels of the variables used so far in the analysis:

Is={Δvt-j}; j = 0, 1…

where vt-j is the vector of all variables used in the analysis so far, with the addition of one new variable, ADM, which denotes administered prices.11

21. An error correction model for Czech inflation was estimated using monthly data over 1994:06-2001.12 Estimation began from a general specification, which included the long-run relationships established above, rearranged so that all variables appear on the left-hand side, and the vector Is. Subsequently, parameter tests were performed (mostly zero restrictions) to reduce the model to a more manageable form. Below is the resulting specification for inflation and the relevant statistics (t-ratios are shown in parentheses below the estimated coefficients) and diagnostic tests (p-values are shown next to the estimated statistics):

D¯CPIt=0.0536(1.393)+0.2456(11.906)*D¯ADMt+0.4279(4.612)*D¯PPIt+1.4340(2.399)*D¯EXCt−1++1.4340(3.340)*D¯CINTt−1+0.0136(2.393)*ECM(Wage)t−1−0.0020(−1.322)*ECM(PPP)t−1−−0.0528(−2.376)*ECM(UIP)t−1+0.0022(2.313)*ECM(Mon)t−1

R-BAR2 = 0.821, DW = 1.52, σ^ = 0.00317

AR 1-5 F(5, 68) = 1.0287[0.4079]

Normality Chi^2(2) = 5.8719[0.0531]

Xi*Xj F(44, 28) = 0.78545[0.7680]

ARCH 5 F(5, 63) = 0.12103[0.9872]

Xi^2 F(16, 56) = 0.85905[0.6163]

RESET F(1, 72) = 2.4519[0.1218]

22. The model passes a battery of diagnostic tests. The hypotheses of no serial autocorrelation, homoskedasticity, and normality of the residuals cannot be rejected. The estimated standard error of the regression (0.00317) is very low. In addition, the tests for parameter constancy show that the estimated coefficients are very robust (Appendix Figure 2).

Figure 2.
Figure 2.
Figure 2.

Czech Republic: Parameter Stability of Estimated Coefficients, 1996–2001

Citation: IMF Staff Country Reports 2001, 112; 10.5089/9781451810141.002.A001

Source: Fund staff estimates.

23. All estimated coefficients are statistically significant (with the exception of the coefficients for ECM(PPP)t-1 and the constant) and have the anticipated signs.13 The insignificant coefficient of ECM(PPP)t-i suggests that deviations of the REER from its long-run equilibrium are not corrected through adjustment of relative prices but through the exchange rate. All ECM terms have small coefficients, suggesting a slow adjustment of inflation to deviations from the long-run equilibria in the respective markets. Most of the inflation dynamics are generated by the short-run influences.

24. The presence of the D_PPI term in the short-run influences provides evidence of a markup behavior at the retail level, about half of producer price inflation is transmitted to consumer price inflation within a month. The estimated coefficient of administered price changes is consistent with the share of the administered component in the Czech CPI, which is around 18 percent. Given that the estimated coefficient is the elasticity of consumer price inflation with respect to administered price changes, a 1 percent increase in administered prices should result in an immediate increase in the CPI of 0.25 percent. In fact, the restriction that the coefficient of D_ADM is equal to 0.18 (the share in the CPI) could not be rejected at the 5 percent level, and was barely rejected at the 1 percent level Finally, the coefficient of the lagged changes in the exchange rate suggests a very large elasticity over the sample period of the inflation rate with respect to exchange rate movements in the previous month: a 1 percent depreciation/appreciation of the exchange rate would result in a 1.4 percentage point rise/fall in inflation.14 It is unclear what factors may account for this large elasticity. In an effort to investigate whether the variable for exchange rate movements was also capturing other effects, omitted variable F tests were performed. However, the tests did not indicate a specification problem. Indeed, the parameter stability exhibited by the model confirms this result (see Appendix Figure 2).

G. Forecasting

25. This section describes the steps involved in deriving short-run monthly forecasts (through end-2002) for inflation. Two types of forecasts were employed: dynamic and conditional. Dynamic forecasts rely on a vector autoregression to generate forecasts for the variables chosen to be determined endogenously. The resulting inflation forecast (not shown here) involved a trajectory that was below the latest forecast published by the CNB. Dynamic forecasts are a good means of generating the future path of variables irrespective of preconceived notions about their likely evolution. However, they are not particularly relevant when the objective is the study of possible future developments in inflation based on a conditional distribution of the explanatory variables. Such a conditional distribution typically involves unchanged monetary policy instruments. In the case of this study, and in light of the heavy dependence of the estimated model on exchange rate dynamics, the conditional distribution also involved prescribed paths for the exchange rate.

26. The conditional inflation forecasts showed that the future path of inflation is very dependent on the chosen path for the exchange rate. Appendix Figure 3 shows three conditional forecasts to illustrate the point. The first forecast assumed an annual depreciation of the exchange rate of 1 percent; the rate of depreciation for 2001 was prorated based on the last available monthly observation. The second forecast assumed no change in the exchange rate from the last available monthly observation. Finally, the third forecast assumed an annual appreciation of the exchange rate of 1 percent; again the appreciation was prorated in 2001. All three forecasts maintained the assumption of unchanged monetary policy rates. The forecasts confirm the dependence of the forecasting performance of the model on the ability to accurately forecast exchange rate changes.

Figure 3.
Figure 3.

Czech Republic: Conditional Forecasts of Inflation, 2001–2002 1/

(In percent)

Citation: IMF Staff Country Reports 2001, 112; 10.5089/9781451810141.002.A001

Sources: Fund staff projections.1/ Forecast mean plus/minus 2 standard errors.2/ The conditional model assumes constant interest rates and a 1 percent annual rate of depreciation of the exchange rate.3/ The conditional model assumes constant interest rates and a constant exchange rate.4/ The conditional model assumes constant interest rates and a 1 percent annual rate of appreciation of the exchange rate.

STATISTICAL APPENDIX

Table A1.

Czech Republic: Gross Domestic Product, 1996–2000

article image
Source: Data provided by the Czech Statistical Office.
Table A2.

Czech Republic: Composition of Gross Domestic Product, 1996–2000

article image
Source: Data provided by the Czech Statistical Office.

Contribution to growth.

Table A3.

Czech Republic: Gross Domestic Product by Origin, 1996–2000

article image
Source: Data provided by the Czech Statistical Office.

Includes mining, manufacturing industry, electricity, gas and water.

Financial intermediation services indirectly measured, calculated as the difference between interest received and paid by banks.

Table A4.

Czech Republic: Industrial Production, 1996–2000 1/

(Annual percent change, at constant prices)

article image
Source: Czech Statistical Office.

Changes in index of physical production (IPP).

Table A5.

Czech Republic: Civil Employment by Sector, 1996–2000 1/

article image
Source: Czech Statistical Office.

Includes private entrepreneurs and workers in enterprises of all size groups.

Preliminary data.

Table A6.

Czech Republic: Civil Employment in Large Enterprises by Sector, 1996–2000 1/

(In thousands, annual average)

article image
Source: Czech Statistical Office.

In 1996, firms in industry, trade, and catering only with 100 employees or more; all others, with 25 or more employees. From 1997, firms in the business sphere with 20 employees or more (all financial intermediaries and all non-business sphere organizations).

Excluding foreign-owned enterprises.

Table A7.

Czech Republic: Average Monthly Earnings, 1996–2000 1/

(In koruny)

article image
Source: Czech Statistical Office.

In 1996, firms in industry, trade and catering only with 100 employees or more; all others, with 25 or more employees. From 1997, firms in the business sphere with 20 employees or more (all firms of financial intermediation and all non-business sphere organizations). Excluding armed forces.

Preliminary data

Excluding foreign-owned enterprises.

Table A8.

Czech Republic: Average Monthly Earnings per Quarter, 1998–2000 1/2/

(In koruny)

article image
Source: Czech Statistical Office.

Firms in the business sphere with 20 employees or more (all firms of financial intermediation and all non-business sphere organizations).

Excluding armed forces.

Preliminary data.

Excluding foreign-owned enterprises.

Table A9.

Czech Republic: Agricultural Production, 1996–2000

(Annual percentage change; at constant prices)

article image
Source: Czech Statistical Office.
Table A10.

Czech Republic: Electricity Production and Consumption, 1996–2000

(In millions of kilowatt hours)

article image
Source: Czech Statistical Office.
Table A11.

Czech Republic: Developments in Wholesale and Consumer Prices, 1997–2000 1/

(Average 1994 = 100)

article image
Source: Czech Statistical Office.

Period average.

1993 constant weights.

1999 contain weights.

Including restaurant meals, but excluding beverages and tobacco.

Net inflation is derived from CPI inflation by excluding goods subject to administered prices, and the effect of changes in fees and indirect taxes.

Table A12.

Czech Republic: Share of Non-State Sector in Output and Employment, 1996–99 1/

(In percent of total)

article image
Source: Czech Statistical Office.

Includes private cooperatives and nonprofit corporations serving households.

Employment in mixed-ownership is included. Preliminary data for 1998; estimates for 1999.

Table A13.

Czech Republic: Operations of the Consolidated General Government, 1997–2001

article image
Source: Czech Ministry of Finance

Privatization revenues; are recorded primarily as negative net lending and capital revenue (sales of voting rights from year 1999).

Used to cover costs associated with the rehabilitation of banks and state-owned enterprises.