Hungary—Selected Issues

International Monetary Fund

doi:10.5089/9781451817829.002.A001

Jump to Content Jump to Main Navigation

Hungary—Selected Issues

Author: International Monetary Fund

Publication Date:: 15 Oct 1996

eISBN:: 9781451817829

Language:: English

Keywords:: ISCR; CR; privatization; firm; Hungary; I. inflation inertia; inflation process; exchange rate policy; inflation performance; price liberalization; Inflation; Consumer prices; Price controls; Exchange rate policy; Nominal effective exchange rate

Download PDF (1.1 MB)

This Selected Issues paper analyzes empirically the main determinants of Hungary’s inflation rate during 1990–96. Although there exist a number of possible methodologies to analyze this issue, the one proposed in the paper takes explicit account of the time-series properties of the variables that are potential candidates for explaining Hungary’s inflation performance. This leads to the specification of a long-term equation, linking consumer prices to a number of macroeconomic variables as well as to proxies for relative price shocks. The paper also examines the external current account and net foreign assets in Hungary.

Abstract

I. Inflation Inertia in Hungary ^1/

1. Introduction

Unlike most other European transition economies, Hungary did not experience a surge in inflation—in some cases to several hundred percent on an annual basis—when, in the early 1990s, its transition to a market economy began to accelerate. Price liberalization, which had been implemented in an abrupt and comprehensive fashion in the context of the “big-bang” reforms elsewhere in the region, was instead instituted in a much more gradual way in Hungary’s case. As a result, rather than rising—then falling—sharply, Hungary’s inflation rate over the past decade has had a relatively inertial character, and its level rarely reached above 30 percent, or fell below 20 percent, on an annual basis (Chart 1).

What factors can explain Hungary’s inflation performance over the past five years? As in other countries, an explanation is likely to be found in the behavior of underlying monetary and exchange rate policies and, on the “supply” side, in the behavior of wages and profit margins. In addition, however, as a transition economy, Hungary faced a number of significant relative price shocks associated, inter alia, with reductions in price subsidies, changes in the structure of indirect consumption taxes (the VAT), and the gradual decontrol of energy prices. The prevalence of these relative price shocks reflected the existence of structural distortions similar to those that prevailed under central planning systems elsewhere in the region: energy products were heavily subsidized prior to liberalization and wages and prices in the service sector were held down in order to promote industrialization; as a result, the relative prices of energy and services rose sharply once prices were freed (Chart 2). In addition, another significant relative price shock related to the more rapid rise in consumer prices than producer prices, which reflected the larger weight of services in the CPI than in the PPI together with the sectoral price shocks described above; a further factor affecting relative CPI/PPI movements was the modification of the rate structure of the value added tax (VAT) on consumption (Chart 3).

This paper sets out to analyze empirically the main determinants of Hungary’s inflation rate over the last five years. While there exist a number of possible methodologies to analyze this issue, the one proposed here takes explicit account of the time-series properties of the variables that are potential candidates for explaining Hungary’s inflation performance. This leads to the specification of a long-run equation linking consumer prices to a number of macroeconomic variables as well as to proxies for relative price shocks. In a second step, and given the existence of a long-run or cointegrating relationship between prices and its determinants, the short-run dynamics link the inflation rate to the deviation between the current-period’s price level and its long-run equilibrium (specified in the first stage), as well as to other variables that theory or observation suggest are likely to exert a short-run impact on inflation.

The paper is organized as follows. In Section 2, the recent behavior of Hungary’s inflation performance and that of a number of related variables is described, in order to set the stage for the empirical analysis which follows in Section 3. The main conclusions are contained in Section 4.

2. Possible factors affecting inflation

Over the past five years, 12-month inflation in Hungary has averaged about 25 percent, with a standard deviation of 6 percentage points. Moreover, out of 60 monthly observations, nearly two thirds have fallen within the range of 20-30 percent. In other words, inflation in Hungary has had an inertial character. The key to explaining this behavior has traditionally been taken to reside in three broad areas: monetary and exchange rate policies, or “demand” factors; wage developments and changes in profitability or markups, or “cost” factors; and relative price shocks associated, inter alia, with price liberalization. The importance of these factors has undoubtedly varied over time. The rest of this section examines each of these factors in turn, together with their possible connection with inflation.

a. Monetary and exchange rate factors

Until the introduction of the March 1995 austerity package, exchange rate policy in Hungary was, in principle, devoted primarily to maintaining international competitiveness. In practice, however, this policy failed to prevent large swings in real exchange rates (Chart 4). ^1/ Prior to 1992, exchange rate changes took place in a predictable fashion, usually once a year. This predictability, however, led to substantial speculative pressures. In response, the authorities decided to increase the frequency of exchange rate realignments in order to generate greater uncertainty for speculators.

The practice of frequent ad hoc adjustments of the exchange rate failed to provide a nominal anchor for inflationary expectations and also was unsuccessful in preventing large short-term speculative flows at end-1994 and early 1995. In response, the authorities introduced a preannounced crawling peg exchange policy in March 1995, following an 8.3 percent step devaluation. In order to dampen external sources of inflation, the monthly rate of crawl was steadily reduced from 1.9 percent in the second quarter of 1995 to 1.2 percent in 1996. The credibility of this exchange rate policy is evidenced by the fact that the market exchange rate has remained close to the more appreciated end of the exchange rate band since mid-1995.

Since exchange rate policy was only loosely geared to maintaining international competitiveness and capital mobility was limited, monetary policy enjoyed a relative degree of independence. However, this independence was not used to target a sizeable or rapid reduction in inflation, particularly in the early years of the transition, when money growth outstripped growth in nominal activity (Chart 5). Annual inflation targets were not very ambitious, never falling much below 20 percent on a year-average basis (Table 1). This being said, the authorities were nevertheless successful in achieving these modest targets. ^1/ The clear exception, however, was in 1993 when actual inflation surpassed the midpoint of the target range by 6½ percentage points due, primarily, to the decision to raise VAT rates in order to shore up falling fiscal revenues.

Table 1.

CPI Inflation: Targets and Outcomes

^{^1/}Revised target, consistent with the March austerity package.

Table 1.

CPI Inflation: Targets and Outcomes

Year	CPI Inflation Target	CPI Inflation Outcome
1990	19-20	28.9
1991	34-36	35.0
1992	20-25	23.0
1993	15-17	22.5
1994	16-22	18.8
1995	26-27 ^1/	28.2

^{^1/}Revised target, consistent with the March austerity package.

Table 1.

CPI Inflation: Targets and Outcomes

Year	CPI Inflation Target	CPI Inflation Outcome
1990	19-20	28.9
1991	34-36	35.0
1992	20-25	23.0
1993	15-17	22.5
1994	16-22	18.8
1995	26-27 ^1/	28.2

^{^1/}Revised target, consistent with the March austerity package.

b. Cost factors; wages and profits (markups)

Between 1991 and 1994, gross wages increased by 140 percent, about the same rate as the CPI. While real consumption wages were therefore relatively stable, real product wages rose sharply, reflecting the slower increase in producer prices than consumer prices (Chart 6). ^1/

What accounts for the relative constancy of real consumption wages in Hungary where no formal system of wage indexation exists? A contributing factor is the fact that the minimum wage, which is centrally determined, has been adjusted over time in order to preserve its purchasing power in terms of (projected) consumer prices. ^2/ While the minimum wage applies directly to only a small fraction of the workforce, in practice its influence is much broader since it acts as a floor for wage agreements at the branch and enterprise levels. A recent World Bank study (Commander et. al. (1994)) revealed that wage increases granted at the branch and firm levels tended to be uncorrelated with sales and ability to pay, especially among large state-owned firms (with more than 2,000 employees). ^3/

The sharp increase in real product wages, together with labor hoarding facilitated by soft budget constraints, led to an erosion of enterprise profitability/markups in the early years of the transition (Chart 7). With the sustained decline in economic activity and increased enterprise financial discipline, the pace of labor shedding gradually quickened however, and profit margins began to recover in 1994. The evolution of markups over labor costs is an additional potential determinant of consumer price inflation. While the erosion of enterprise profitability/markups early in the transition dampened inflationary pressures arising from higher labor costs, the subsequent reversal of this tendency had the opposite effect.

c. Relative prices

While most theories of inflation assign an important role to macroeconomic factors such as those discussed in the previous two subsections, in transition economies the process of price liberalization is also likely to affect inflation. In Hungary over the past five years, price decontrol, subsidy reductions, changes in the rates and structure of indirect taxes and trade taxes have all contributed to substantial realignments of relative prices. Reflecting these changes, household energy and services prices rose by 350 percent and 240 percent, respectively, between 1990 and 1995, compared to an increase in overall consumer prices of 200 percent over the same period. The gradual increase in selected controlled prices is likely to have exerted an independent influence on inflation, particularly in an environment in which monetary policy was accommodating.

Following two decades of gradual price reform, ^1/ price liberalization in Hungary was stepped up in 1991 with the freeing of retail prices of petrol and diesel fuels, and increases (80 percent) in the prices of household electricity and gas. In early 1992, the prices of all food products, including milk and bread, were liberalized, and budgetary subsidies on dairy products and household energy were eliminated. As a result, the proportion of centrally-controlled prices at the consumer level fell to 7 percent by 1992, and to 6 percent by 1993. Central administration of prices remains only for household electricity and gas, and some services (notably postal rates, telephone tariffs, television licenses and long-distance train transportation). However, local governments set tariffs for local public transportation and publicly-owned apartments; these goods and services comprise an additional 5 percent of the consumption basket.

As regards energy products, following a substantial increase in 1991, energy prices underwent a series of step adjustments beginning in January 1995 in order to guarantee sufficient profit margins in the energy sector to facilitate its privatization (see Chapter V). The household prices of electricity and gas were raised by 65 percent and 53 percent, respectively, in January 1995; by a further 8 percent each in September 1995; and by 25 percent and 18 percent, respectively, in March 1996.

Substantial changes in indirect and trade taxes have also affected relative prices during the transition period. Under the initial structure of the VAT (1988), three tax rates were established: a zero (preferential) rate; a 15 percent rate for services; and a basic rate of 25 percent. About 40 percent of household consumption expenditure was zero-rated at the time. In January 1993, the preferred rate was raised to 6 percent, while services (previously in the 15 percent bracket) were shifted to the basic-rate category. The preferential rate was raised again in August 1993 to 10 percent, with a further upward revision to 12 percent in January 1995. At that time, a number of goods and services, including telecommunications and housing construction and refurbishment were reclassified to the basic-rate category. Following these modifications, about 60 percent and 35 percent of private consumption was taxed at the basic and preferential rates, respectively, compared with 45 percent and 46 percent, respectively, in 1992. As regards trade taxes, the gradual liberalization of external trade during the transition period is likely to have exerted a restraining effect on inflation. However, as part of the March 1995 austerity package, the authorities introduced an 8 percent surcharge on all imports excluding capital equipment and energy products. This move coincided with a large jump in the CPI.

The medium-term impact of relative price shocks—such as the ones described above—is likely to depend on a number of factors, including the stance of monetary policy, and the degree to which other market-determined nominal prices are rigid downwards. More generally, however, a substantial body of literature argues that greater variability of relative price disturbances is in fact positively correlated with inflation. ^1/ More recently, Ball and Mankiw (1994, 1995) have focussed on the role of the higher moments of the distribution of relative prices in the inflation process. In their framework, which is based on a menu cost model of nominal price adjustment, a firm will change its price only if the desired (absolute) price adjustment exceeds the menu cost, which is assumed to be invariant to the size of the desired price change. Therefore, firms will respond to large relative price shocks, but not to small ones. If the distribution of relative price shocks is asymmetric, therefore, the price level will be affected by relative price disturbances, at least in the short run. Thus, a positively skewed distribution of relative price shocks will be inflationary.

Of course, while changes in relative prices—including of the type described above—may contribute to short-run changes in the price level, a sustained impact requires an accommodating monetary policy. The usual story is that, in the absence of an increase in the quantity of money, as some nominal prices rise, the real value of money balances declines, thereby contributing to a reduction in aggregate demand. As a result, prices of commodities not subject to positive relative price shocks decline, and the aggregate price level returns to its initial equilibrium. However, in Hungary (as in other transition economies), policy-induced changes in relative prices have in fact coincided with sharp increases in the general level of prices, which have apparently been sustained. ^2/ This indicates the presence of monetary accommodation (either full or partial) of policy-induced price changes. The relevant point for our purposes, though, is that the observation of an empirical relationship between relative price shocks and inflation suggests that one cannot ignore such disturbances in analyzing the factors underlying Hungary’s recent inflation performance, particularly—as is the case—when relative price changes have been so large.

3. Empirical Analysis

In this section, an econometric model of inflation in Hungary is developed and estimated in order to gauge the relative importance of the various determinants of inflation. Based on the empirical results, consumer price inflation for 1996 is projected under the assumption that the present policy stance is maintained. This projection is then compared with the official forecast.

a. Time series properties of the variables

The first step in the empirical analysis is to establish the time-series properties of the relevant variables. This is done by employing Augmented Dickey-Fuller (ADF) tests for non-stationarity of each of the variables. Based on the discussion in the previous section, the variables included in the study are the log of consumer prices (1p); the log of broad money (1m); the log of gross wages in industry (1v); ^1/ the log of the nominal effective exchange rate (1e), defined as the foreign exchange price of domestic currency; the log of administered consumer prices (1pa); and the asymmetry of relative price changes (varsk), defined as the product of the variance and skewness of the distribution of monthly inflation rates for seven main categories of consumer products. In addition, the total trade balance and the convertible currency trade balance (measured in U.S. dollars and denoted by ttb and ctb, respectively), are included as proxies for the level of demand pressures in the economy. The paucity of alternative monthly indicators of excess demand pressures—e.g., a measure of the output gap—suggested the use of these proxies.

Table 2 presents the ADF statistics for each of the macroeconomic variables that is a candidate for inclusion in the model. ^2/ The test statistics reported in the table refer to the preferred specification of the model, which may or may not contain a constant or trend. The null hypothesis of nonstationarity is not rejected at the 5 percent level for any variable except ttb, varsk and lpa. The second column confirms that each series is stationary in first differences. Thus we conclude that the series lp, lm, lw, le, and ctb are integrated of the first order (I(1) variables).

Table 2.

HUNGARY: Time-Series Properties of the Macroeconomic Variables

^{^1/}The null hypothesis being tested is that the variable is nonstationary. c (nc) denotes that a constant was (not) included in the preferred specification; while t (nt) denotes the inclusion or not of a trend.

Notes: Sample period is 1990: 12 to 1995: 12 (monthly data). The augmented Dickey-Fuller (ADF) statistic is the t-statistic on the lagged level of the variable in a regression of the first difference of the variable on lags of the first difference of the variable and the lagged level of the variable.

^*Indicates significance at the 5 percent level.

^{^**}Indicates significance at the 1 percent level.

Table 2.

HUNGARY: Time-Series Properties of the Macroeconomic Variables

	ADF-Statistic ^1/
Variable	Levels	First Difference
Log consumer prices (1p)	-2.12 c,t	-6.00 c,t
Log broad money (1m)	-2.91 c,nt	-8.35 c,nt ^**
Log wages (1w)	-2.19 c,t	-8.01 c,nt ^**
Log nominal effective exchange rate (1e)	-0.74 c,t	-8.04 c,t ^**
Trade balance (ttb)	-4.47 c,t ^**	-7.14 nc,nt ^**
Convertible-currency trade balance (ctb)	-1.08 nc,nt	-14.24 nc,nt ^**
Asymmetry of relative price changes (varsk)	-6.48 nc,nt ^**	-7.08 nc,nt ^**
Log administered prices (1pa)	-4.31 c,t ^**	-6.55 c,nt ^**

^*Indicates significance at the 5 percent level.

^{^**}Indicates significance at the 1 percent level.

Table 2.

HUNGARY: Time-Series Properties of the Macroeconomic Variables

	ADF-Statistic ^1/
Variable	Levels	First Difference
Log consumer prices (1p)	-2.12 c,t	-6.00 c,t
Log broad money (1m)	-2.91 c,nt	-8.35 c,nt ^**
Log wages (1w)	-2.19 c,t	-8.01 c,nt ^**
Log nominal effective exchange rate (1e)	-0.74 c,t	-8.04 c,t ^**
Trade balance (ttb)	-4.47 c,t ^**	-7.14 nc,nt ^**
Convertible-currency trade balance (ctb)	-1.08 nc,nt	-14.24 nc,nt ^**
Asymmetry of relative price changes (varsk)	-6.48 nc,nt ^**	-7.08 nc,nt ^**
Log administered prices (1pa)	-4.31 c,t ^**	-6.55 c,nt ^**

^*Indicates significance at the 5 percent level.

^{^**}Indicates significance at the 1 percent level.

Given the time series properties of prices and its likely determinants, the strategy will be to search for a possible long-run relationship between the price level and other I(1) variables. In addition, if such a long-run relationship is uncovered, the next step will be to consider the short-run adjustment of inflation (i.e., the change in the price level) to shocks that cause prices to deviate from their long-run equilibrium.

b. Long-run relationships: cointegration

Having established the univariate properties of the variables, the next question is to determine how the variables are related. Our strategy in this subsection is first to test for the existence of a long-run relationship between consumer prices and the other macroeconomic variables discussed in Section 2, namely money, wages, the nominal effective exchange rate, and the convertible currency trade balance. Since the existence of multiple relationships or cointegrating vectors between different (and even the same) combinations of variables cannot be refuted ex ante, the approach followed here first involves testing for cointegration among all possible combinations of I(1) variables in the data set before settling on a preferred specification based on the results obtained.

To test for cointegration, we use the Engle-Granger procedure, which tests for the presence of a unit root in the residual term of a static regression linking the I(1) variables that are thought to be cointegrated. The methodology consists of two steps. First, a static equation between prices and its determinants is estimated using ordinary least squares (OLS). ^1/ In the second stage, the first difference of the residual from the static equation is regressed on the lagged residual and lags of the first difference of the residuals. The cointegration test involves comparing the t-statistic on the lagged residual with critical values which depend on the number of variables in the static model and the sample size.

The results of the Engle-Granger cointegration tests for alternative specifications of the model are shown in Table 3. As can be seen, prices are not cointegrated either with wages, the exchange rate or money individually; nor are there any combinations of three I(1) variables that form a cointegrating vector. However, a long-run relationship does exist among the variables prices, money, wages and the exchange rate, with the null hypothesis of no cointegration being comfortably rejected at the 5 percent level, and nearly at the 1 percent level. The only other combination of variables that was close to rejecting the null hypothesis involved the inclusion of the convertible currency trade balance (ctb), in addition to prices, money, wages and the exchange rate. However, the inclusion of this additional variable had almost no effect on the values of the remaining coefficients, and its own coefficient was very small. We conclude that the long-run equilibrium of the system may be adequately described by the equation: ^1/

Table 3.

HUNGARY: Engle-Granger Cointegration Test

Notes: The null hypothesis being tested is no cointegration. An asterisk denotes rejection of the null hypothesis at the 5 percent level.

Table 3.

HUNGARY: Engle-Granger Cointegration Test

Variables	t-statistic
Lp, lm, lw, le, ctb	-4.80 *
Lp, lm, lw, le	-4.80 *
Lp, lw, le	-3.01
Lp, lm, le	-2.72
Lp, lm, lw	-1.03
Lp, lm	-0.25
Lp, le	-1.50
Lp, lw	-1.08

Notes: The null hypothesis being tested is no cointegration. An asterisk denotes rejection of the null hypothesis at the 5 percent level.

Table 3.

HUNGARY: Engle-Granger Cointegration Test

Variables	t-statistic
Lp, lm, lw, le, ctb	-4.80 *
Lp, lm, lw, le	-4.80 *
Lp, lw, le	-3.01
Lp, lm, le	-2.72
Lp, lm, lw	-1.03
Lp, lm	-0.25
Lp, le	-1.50
Lp, lw	-1.08

Notes: The null hypothesis being tested is no cointegration. An asterisk denotes rejection of the null hypothesis at the 5 percent level.

\begin{matrix} lp = - 1.19 + 0.441 m + 0.381 w - 0.43 le & ADF = - {4.80}^{*} \end{matrix} (1)

Thus, consumer prices are positively related to broad money and wages, and negatively related to the nominal effective exchange rate (recall that an increase in the exchange rate indicates an appreciation), with the magnitude of the effect of each of these macroeconomic variables on prices being roughly equal. A 1 percent increase in money and wages, together with a 1 percent decrease in the exchange rate, produce an increase in the price level of slightly more than 1 percent in the long run. Finally, since in theory there could be more than one cointegrating vector between lp, lm, lw, and le, we used Johansen’s (1992) procedure to determine the number of cointegrating vectors among this set of variables. ^2/ Based on Johansen’s trace test, exactly one cointegrating vector among these variables is found, confirming the results obtained using the Engle-Granger procedure. ^3/

c. Short-run dynamics: error correction model

The dynamics of the model involving prices, money, wages and the exchange rate away from its long-run equilibrium may be characterized by an error correction model (ECM). Such a model relates inflation to the lagged first difference of prices, money, wages and the exchange rate, as well as to the lagged error correction (EC) term, representing the deviation of prices from their steady-state level. The general error correction model is therefore of the form:

\begin{array}{l} A (L) Δlp (t) = B (L) Δlm (t) + C (L) Δlw (t) + D (L) Δle (t) \\ + α [lp (t - 1) - {\hat{β}}_{0} - {\hat{β}}_{1} lm (t - 1) - {\hat{β}}_{2} lw (t - 1) - {\hat{β}}_{3} le (t - 1)] + μ (t), (2) \end{array}

where A(L) is the polynomial lag operator 1-a₁L-a₂L²_…-a_pLP; B(L), c(L), and D(L) are polynomial lag operators of the form K₁L+K₂L²+…+K_qL^q for k = b, c, d; and the $\hat{β}$ represent least square estimates of the parameters of the static model, so that the term in square brackets is the error correction term. Since (2) contains only stationary variables, it can be estimated by OLS. The coefficient on the error correction term, α, is inversely related to the speed of adjustment to the long-run equilibrium.

Model (2) describing the short-run dynamics may be augmented by the addition of stationary variables which are thought to affect inflation but not the long-run level of prices. Possible candidates might be any variable in Table 2 that was found to be stationary and so would be inappropriate for inclusion in the long-run model, but might exert some temporary influence on price behavior during the sample. Here we consider two indicators of relative price adjustment: the rate of change of administered prices (Δlpa); and the asymmetry of relative price changes (varsk). ^1/

We begin first by considering an ECM in which only the variables that entered the long-run model are included as explanatory variables for inflation. The methodology followed is the “general to specific” procedure suggested by Hendry (1986). Initially, inflation is regressed against the EC term and several lagged first differences of each of the variables that enter the long-run model, as in (2). A parsimonious representation of the inflation process is chosen by eliminating statistically insignificant variables in a step-by-step fashion until all the remaining variables are statistically significant. The resulting error correction model is:

\begin{matrix} Δ 1 p (t) = - \underset{(- 2.91)}{0.17} E C (t - 1) + \underset{(4.71)}{0.54} Δ 1 p (t - 1) + \underset{(3.53)}{0.39 Δ 1} p (t - 2) & (3) \end{matrix}

R² = 0.18,

where the terms in parentheses below equation (3) are the t-statistics. As can be seen, only the error correction term and the first two lags of inflation enter the parsimonious representation. As required for convergence to equilibrium, the coefficient on the EC term is negative and statistically significant. The magnitude of the coefficient implies that 17 percent of the deviation from long-run equilibrium is eliminated in the first month.

As can be seen from the R² statistic, however, the model in equation (3) explains only about 20 percent of the variation in inflation during the sample period. Moreover, as can be seen in Chart 8, large positive deviations between actual and predicted inflation are present in January 1993 and March 1995. The timing of these outliers coincides with major policy-induced changes in administered prices, including modification of VAT rates and the introduction of the import surcharge. These results suggest the need to consider alternative sources of inflation in the short-run model, including for example Alpa and varsk, as discussed earlier.

Whereas the skewness proxy, varsk, lacked any significant explanatory power for the short-run dynamics of inflation, the rate of change of administered prices, Alpa, was highly significant. The final parsimonious specification of the ECM was given by:

\begin{matrix} Δ 1 p (t) = - \underset{(- 2.45)}{0.09} E C (t - 1) + \underset{(5.31)}{0.38 Δ 1} p (t - 1) + \underset{(3.20)}{0.22 Δ 1} p (t - 2) + \underset{(9.64)}{1.00 Δ 1} p a & (4) \end{matrix}

R² = 0.73.

As can be seen, the coefficients on the EC term and lagged inflation are broadly similar to those obtained in the restricted version of the ECM given previously in (3). Administered price inflation enters (4) with a positive coefficient which is highly significant. ^1/

The goodness of fit of the ECM is substantially improved by the inclusion of administered price inflation in the model, which raises the R² from 18 percent to 73 percent. Moreover, as can be seen in Chart 8, including administered price changes in the short-term model greatly reduces the unexplained residual in January 1993 and March 1995, as well as over the entire sample more generally.

The stability over time of the parameters in equation (4) may be assessed by using a CUSUM test. This test involves sequential estimation of ex post prediction errors, where the prediction error for the n-th observation (w_n) is based on coefficients estimated with the first n-1 observations, adding a new observation for each estimation. The CUSUM test statistic is:

\begin{matrix} W_{t} = \underset{n = k + 1}{\overset{t}{Σ}} w_{n} / s, & where & s^{2} = 1 / (T - k) \end{matrix} \underset{t = 1}{\overset{T}{Σ}} {\hat{μ}}_{t^{2}}, (5)

where k is the number of parameters in the model, and T is the number of observations. The intuition of the test is that systematic changes in the estimated parameters over time would give rise to a disproportionate number of the prediction errors having the same sign. The CUSUM statistic for ECM and the corresponding 5 percent confidence bands are shown in Chart 9. As can be seen, the test statistic remains within the confidence band throughout the sample period and, therefore, the stability of the parameters over time cannot be rejected.

d. Forecasting inflation

Using the estimated parameters of the cointegrating vector (1) and the parsimonious ECM (4), a projection of consumer price inflation for 1996 is undertaken on the basis of announced monetary, wage, and exchange rate policies. The projection is based on actual data through March 1996. Thereafter, broad money is expected to increase by 1.1 percent per month, consistent with a 12-month growth rate of 14 percent. Wages in the last three quarters are forecast to grow by 20 percent relative to the corresponding month in the previous year. The nominal effective exchange rate is assumed to move in line with the 1.2 percent monthly rate of crawl. As to administered price changes, the projection reflects the 20 percent increase in energy prices in March and the further increase expected in September, together with the two-step reduction in the import surcharge by 1 percentage point in July and October.

The inflation forecast for the last nine months of 1996 implies a year average inflation rate of about 23 percent. This projection is at the lower bound of the range currently projected by the authorities (23-24 percent). However, since wage increases (adjusted for the number of working days) are currently running ahead of the 20 percent assumed in the projections, actual inflation in 1996 may be closer to the upper bound of the authorities’ range in the absence of wage moderation.

4. Conclusion

This paper has examined empirically the main determinants of Hungary’s inflation performance over the past five years. A cointegrating relationship was found to exist among the price level, the exchange rate, the quantity of money, and wages. This relationship defined the long-run equilibrium of the system, while the corresponding error correction model (ECM) provided the dynamics of inflation when the price level differed from its long-run equilibrium. The preferred ECM contained, in addition to the lagged value of the variables in the long-run relationship, the rate of increase of administered prices.

The relatively inertial character of inflation during Hungary’s transition process appears to be related, inter alia, to the failure to use monetary and exchange policies to target a sizable or rapid reduction in inflation, to rigidities in the labor market combined with soft budget constraints which served to thwart real wage adjustment, and to ongoing increases in key relative prices. The evidence would suggest that supply-side factors—such as the behavior of wages and administered prices—and demand side factors—including monetary and exchange rate policies—are likely to continue to exert an important influence on the future course of Hungary’s inflation rate.

References

Ball, Lawrence, and Mankiw, N. Gregory, “Asymmetric Price Adjustment and Economic Fluctuations,” The Economic Journal, Volume 104, pp. 247-261, March 1994.
- Search Google Scholar
- Export Citation
Ball, Lawrence, and Mankiw, N. Gregory, “Relative-Price Changes as Aggregate Supply Shocks,” The Quarterly Journal of Economics, Volume 110, pp. 161-193, February 1995.
- Search Google Scholar
- Export Citation
Commander, S., Kollo, J., and Ugaz, C. “Firm Behavior and the Labor Market in the Hungarian Transition,” World Bank Policy Research Working Paper Series, No. 1373, 1994.
- Search Google Scholar
- Export Citation
Fischer, Stanley, “Relative Shocks, Relative Price Variability, and Inflation,” Brookings Papers on Economic Activity, No. 2, pp. 381-441, January 1981.
- Search Google Scholar
- Export Citation
Hendry, David, “Econometric Modelling with Cointegrated Variables: An Overview,” Oxford Bulletin of Economics and Statistics, No. 48, pp. 201-12, August 1986.
- Search Google Scholar
- Export Citation
Johansen, Soren, “Statistical Analysis of Cointegration Vectors,” Journal of Economic Dynamics and Control, Volume 12, pp. 231-54, September 1988.
- Search Google Scholar
- Export Citation
Johansen, Soren, and Juselius, Katarina, “Testing Structural Hypotheses in a Multivariate Cointegration Analysis of the PPP and the UIP for the U.K.,” Journal of Econometrics, Volume 53, pp. 211-44, July-September 1992.
- Search Google Scholar
- Export Citation
Pujol, Thierry and Mark Griffiths, “Moderate Inflation in Poland: A Real Story” IMF Working Paper Series, No. WP/96/57, June 1996.
- Search Google Scholar
- Export Citation

Prepared by Rachel van Elkan.

This reflected various factors, including attempts to cool price pressures through the exchange rate. For example, in 1992, when Hungary faced a better-than-expected current account performance and large inflows of foreign direct investment, the authorities permitted a substantial erosion of external competitiveness in order to dampen inflationary pressures.

This was achieved, moreover, despite the fact that money demand was unstable—reflecting the impact of financial liberalization, heightened macroeconomic uncertainty, and financial disintermediation associated with high interest rate spreads—and in the absence of timely indicators of real activity.

However, in contrast to the experience of the preceding four years, real consumption wages fell sharply in 1995, reflecting the imposition of strict limits on pay increases in the state and public enterprise sectors as part of the March 1995 austerity package. This compression of real wages was subsequently partially reversed, however, as sizable wage bonuses—postponed from the previous year—were granted in the first months of 1996.

Another contributing factor was the still relatively soft budget constraints faced by public sector enterprises.

In the past few years, most decentralized wage negotiations have taken place at the enterprise level.

Beginning in 1968, the Government introduced a series of price reforms that gave enterprises some flexibility in price determination. By 1989, these reforms had already freed the prices of some 77 percent of consumer products.

See Fischer (1981) for a survey of the literature.

See Pujol and Griffiths (1996) for an analysis of the correlation between relative price variability and inflation in Poland.

Owing to the shorter time series available for economy-wide wages, data on industrial sector wages were used.

The sample was 1990:12-1995:12. Major breaks in the series precluded using a longer sample. All data are seasonally adjusted.

The use of OLS is justified by the “superconsistency” of the parameter estimates in the presence of cointegration.

An asterisk denotes stationarity of the residual from the cointegrating equation at the 5 percent level.

The Johansen procedure (Johansen and Juselius (1992)) is based on a vector error correction model which incorporates both long-run and short-run behavior in a single equation. This procedure has greater power than the Engle-Granger methodology to reject the null hypothesis of no cointegration. Johansen’s approach also provides a test for the number of cointegrating vectors among a set of variables.

The probability value of the trace test statistic under the null hypothesis of no cointegrating vector was 0.5 percent, while the probability value under the null hypothesis of at most one cointegrating vector was 36 percent.

Pujol and Griffiths (1996) investigate the relationship between inflation and measures of variance and skewness of the distribution of relative prices in Poland. While their results are suggestive, the estimations ignore any possible long-run relationship between the variables. If such a long-run relationship does exist, its inclusion is necessary for obtaining consistent parameter estimates.

The unit coefficient on Alpa reflects the fact that the series provided by the authorities has been multiplied by the weight of administered prices in the CPI.

Other Publishers

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/002/1996/109/article-A001-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully