A CAUSAL RELATIONSHIP BETWEEN INFLATION AND PRODUCTIVITY: AN EMPIRICAL APPROACH FOR ROMANIA

This paper attempts to analyze the relationship between the productivity and the inflation for a transition country of European Union as Romania. For this purpose we use quarterly data from 1990:IV to 2003:I and the causality analysis, which is based on an error correction model. The results of the empirical analysis showed that there is a causal relationship between inflation and productivity in Romanian economy. key words: productivity, inflation, Granger causality, cointegration, error correction models, Romanian economy. JEL Classification A10, C22 ∗ Dr. Nikolaos Dritsakis Associate Professor University of Macedonia Economics and Social Sciences Department of Applied Informatics 156 Egnatia Street P.O box 1591 540 06 Thessaloniki, Greece FAX: (2310) 891290 e-mail: drits@uom.gr


The relationship between Romania and European Union and its economy route during the period from 1990 to 2003.
The starting point for the transition process in Romania was more difficult than in other countries in Central and Eastern Europe. Pre-transition policies emphasized self-dependence, putting excessive focus on heavy industry and large infrastructure projects. This strategy led to the depletion of domestic energy sources and induced costly dependence on imports of energy and raw materials. During 1980s there was no growth in exports in order to repay the debt imports from the West. The technological lag increased significantly as a result. Towards the end of the 1980s the Romanian economy was on the verge of collapse and, unlike other transition economies, no attempts to reform had yet been tried.
Given this difficult legacy, the dominant political forces in place since the early 1990s advocated a gradualist approach, seeking to minimize the social costs associated with the transformation to market. The 1993 OECD Assessment of the Romanian economy pointed out clearly the risks associated with the delaying structural reforms. A key point in the Assessment was that, without deep restructuring of the economy, macroeconomic stabilization could not be sustained.
Therefore, since 1993 the boost in exports and the apparent success in reducing inflation under the stabilisation policy of Romanian economy were noted.
The Seville European Council (1996) encouraged Romania to pursue its efforts for accession in European Union and also reiterated its commitment to provide full support to this candidate country.
Romania is expected to join the European Union on the basis of the same economic and political criteria that had been set by the Copenhagen andMadrid European Councils (1993, 1995)  While output continued to grow in 1996 fiscal policy was derailed under the impact of a largely unrestructured economy. The official budget deficit was increased by quasi-fiscal items, such as the National Bank of Romania (NBR) refinancing of credits to the agricultural sector. This slippage resulted from preelection policies in support of output and demand and a pervasive lack of financial discipline in large state-owned companies. With rapid growth of the money base, inflation accelerated readily in such a point that World Bank halted their financial support.

Theoretical and empirical approaches
Inflation is always a monetary plenomenon, and productivity is a purely real occurrence. But upon reflection, we may reasonably think that inflation or at the least things associated with it must matter for firms ability to improve their first group of candidate countries, the so called ''front runners'' in 1998. Latvia, Lithuania, Slovakia, Bulgaria and Romania were invited to start EU accession negotiations in Helsinki. productivity. In considering a link between inflation and productivity there are two possible causal directions: productivity affects inflation or inflation affects productivity. The first generally has higher productivity allowing cost reductions that flow through to product prices and thereby reduce inflation. Higher productivity growth thus represents a positive supply shock that lowers inflationary pressures. The second effect posits that inflation affects productivity growth. From first principles, prices matter because they are a highly efficient means of transmitting the myriad of individual demand and supply decisions that occur throughout the economy (Bulman and Simon 2003). In an inflationary environment, the price mechanism loses its efficiency. It seems plausible then, that when prices are changing frequently, firms may find it more difficult to distinguish an increase in the relative scarcity of their inputs from an across the board increase in prices.
Similarly, the reduced certainty brought about by inflation increases the risk of entrepreneurial errors and would potentially induce lower levels of investment. This would all lower the overall productivity of the firm.
Early research into the inflation-productivity nexus was stimulated by the experience of high inflation of the 1970s and the subsequent fall in productivity growth. Most of the literature has debated the statistical question of whether the data support any relationship, and if so, the causal direction. Minimal work explores the theoretical side, or how inflation may be transmitted into slower productivity growth and vice versa. The view was a little circumspect about the nature of any relationship between productivity growth and inflation. Nonetheless, both Keynesian and neoclassical theory suggest a negative relationship (Lucas 1973). It is recognized that inflation has adverse effects on macroeconomic variables such as output and productivity growth (Bitros andPanas 2001, Dritsakis 2003).

Gillman, Harris and Matyas (2004) use a monetary model of endogenous
growth and based on a panel of OECD and APAC countries using annual data for the period 1961-1997, they found that there is a negative effect of inflation and productivity for these countries.
US data over the period 1948 -1981 demonstrate a similar correlation, with causation running one-way from higher inflation to slower productivity growth (Clark 1982, Buck and Fitzroy 1988, Saunders and Biswas 1990. Methodologically, these studies apply Granger-type causality test to OLS (Clark 1982, Ram 1984 or Full Information Maximum Likelihood (Jarret and Selody 1982) estimations. Gillman, and Nakov (2003) investigate the causal relationship between inflation and growth in two accession countries of EU, Hungary and Poland. Using exogenous variables such as money supply they concluded that there is a causal relationship with direction from money to inflation and from inflation to growth for both accession countries. Jarrett and Selody (1982) proposed two rationales for this occurrence: that the tax system's lack of neutrality during periods of inflation increases the private sector's tax burden, and that inflation's increasing variance with higher levels of inflation would cause sub-optimal resource allocations and increase the probability of entrepreneurial error, hence reducing investment. Using 1963-1979 Canadian data, Jarrett and Selody found a bi-directional relationship, with the rise in inflation explaining nearly the entire slowdown in productivity growth.
Another group of papers took up the debate in the mid 1990s. These had the advantage of being able to observe the productivity growth inflation relationship after the 1980s' disinflation, and also draw on the experience of a wider range of G7 economies. Smyth (1995aSmyth ( , 1995b using annual data for the period 1951-1991 for Germany and1955-1990 for USA respectively, suggested that there is not any causal relationship between productivity and inflation for both examined countries. Chowdhury and Mallik (1998) resulted in the same conclusion in their research for Australia and New Zealand, and also Cameron, Hum and Simpson (1996) in their research for USA, United Kingdom, Kanada, West Germany.
A further group of papers is skeptical of any inflation productivity growth relationship. These papers take two tacks. One approach is to argue that the results show that the business cycle drives simultaneous variations in both productivity growth and inflation, not a long run relationship (Sbordone and Kuttner 1994, Freeman and Yerger 1997, 2000. The stylized facts have productivity growth peaking ahead of the business cycle, with inflation then accelerating. In response, the monetary authorities increase interest rates, thus slowing output growth hence productivity growth through the effects of labour hoarding. Inflation's slow-down lags that of the real economy. Thus, an appropriate model of the productivity growth inflation relationship must absorb the business cycle through variables such as real interest rates, the output gap, or variations in GDP growth.
The other critique argues the statistical point that productivity growth and inflation have different orders of integration (Sbordone and Kuttner 1994, Cameron, Hum and Simpson 1996, Tsionas 2001. These studies claim inflation is nonstationary while productivity growth is stationary, and therefore there cannot be long run relationship. Almost all the paper run Granger causality tests, or a close relative, VAR models. There does appear to be a relationship between productivity growth and inflation, and where it is determinable, the causality appears to flow from inflation to productivity growth. This paper tries to investigate the direction of causality between inflation and productivity in Romania. It seems that this country has the most problems from all the other countries in transition, and that's the reason why we chose it. In the empirical analysis we used quarterly data for the period 1990:IV to 2003:I for the variables used. The remainder of the paper proceeds as follows: Section 1 is referred to the economic evolution of Romania during its transition period until to the decision of European Union in order to be a candidate accession country.
Section 2 employs with the theoretical and empirical approaches. The variables data with the methodology of VAR model are presented in section 3 of this paper.
Section 4 applies the Dickey-Fuller and Phillips Perron tests and investigates the stationarity of the used data. The analysis of cointegration between the used variables is implied in section 5. Section 6 reports the estimations of error correction models, and deploys the Granger causality tests. Final, section 7 presents the conclusions of this study.

Data and Methodological Issues
In order to test the causal relationship between the price level and the productivity of Romania, we use the following VAR model: where: All data are expressed by logarithms in order to include the proliferative effect of time series and are symbolized with the letter L preceding each variable name while ∆ denotes the first differences of these variables.
If these variables share a common stochastic trend and their first or second differences are stationary, then they can be cointegrated. Economic theory scarcely provides some guidance for which variables appear to have a stochastic trend and when these trends are common among the examined variables as well.
Initially, a bivariate VAR model of prices and productivity is estimated. Then, a four variable VAR model is introduced in order to account for potential influences of cyclical factors and changes in monetary policy on the price level-productivity relationship, two variables the real gross domestic product and the interest rate were added.
In order to test the existence of the statistical relationship among the examined variables, we pursue the following steps: The first step is to verify the order of integration of the variables, since the causality tests are valid if the variables have the same order of integration. For the integration of these variables we used ADF test (Dickey-Fuller 1979, 1981 and PP test (Phillips -Perron 1988).
The second step involves testing for the existence of cointegration between the price level and the productivity level by using the Engle -Granger (1987) method, the error correction model and the Johansen maximum likelihood approach (Johansen 1988, Johansen and Juselius 1990, 1992.  (Granger 1986(Granger , 1988. Thus, the third step involves utilization of the vector error correction model for testing the causality among the model variables. Engle-Granger (1987) claim that in the presence of cointegration, there always exists a corresponding error correction representation, which implies that changes in the dependent variable are a function of the level of imbalance in the cointegrated relationship captured by the error correction term (ECT). Thus, through the error correction term (ECM), model VECM establishes an additional way to examine the Granger causality. The non-significance of ECT is referred to as a long-run non-causality. The absence of short-run causality is established by the non-significance of the sum of the lags of each explanatory variable.
Finally, the non-significance of all the explanatory variables including the ECT term in the VECM indicates the absence of Granger causality.

Data stationary tests
To examine the stationarity of the mentioned variables of the above model (1), we have used the Dickey-Fuller (DF) and Augmented Dickey-Fuller (ADF) tests (1979,1981), but also Phillips-Perron (1988 including an intercept and trend and the corresponding numbers of lagged terms. As far as the autocorrelation disturbance term test is concerned, the Lagrange Multiplier LM(4) test has been used.

INSERT TABLE 1
The results of Table 1 suggest that the null hypothesis of a unit root in the time series cannot be rejected in variable levels at a 1%, 5% and 10% levels of significance. Therefore, no time series appear to be stationary in variable levels.
When the time series are transformed into first differences they become stationary and consequently the related variables can be characterized integrated of order one, i.e they are I (1). Moreover, for all variables the LM(4) test first differences shows that there is no serial correlation in the disturbance terms.

Cointegration test
In this section, by applying the Engle-Granger (1987) method and estimating the error correction model, it has been examined if there is any cointegration relationship between the productivity and the price level in the examined country since these two variables are integrated of order one.
The results of cointegration analysis using the Engle-Granger method and an error-correction model are presented in Table 2 testing for the significance of the coefficient of lagged level of the dependent variable. The results suggest that the hypothesis of no cointegration for the two variables, namely the price level and the productivity, is rejected.
INSERT Given the fact that in order to apply the Johansen approach a sufficient number of time lags is required, we have followed the relative procedure, which is based on the calculation LR (Likelihood Ratio) test statistic (Sims 1980). The results showed that the value ρ=2 is the appropriate specification for the above relationship. In addition, each equation of the VAR system passes a series of diagnostic tests including serial correlation, ARCH(4), normality and heteroskedasticity tests. Table 4 reports the specification tests for the VAR (4) system. The tests do not reveal any misspecification except the rejection of normality for price level and interest rate. From the above results we can infer that there is a long run relationship among the price level, the productivity, the real production, the interest rate for Romania for the examined period. Therefore, the above relationships can be used as an error correction mechanism in the VAR model.

VAR model with an error correction mechanism
After determining that the logarithms of the model variables are cointegrated, we must then estimate a VAR model in which we shall include a mechanism of error correction model (MEC). The error-correction model derived from the long run cointegration relationship, has the following form: where ∆ is reported to all variables' first differences u t-1 are the estimated residuals from the cointegrated regression (long-run relationship) and represent the deviation from the equilibrium in the time period t.
-1<λ<0 is the short-run parameter which expresses the response of the dependent variable in every period which starts from the equilibrium state.
V t is a 4X1 vector of white noise errors. Granger (1988) suggested that there are two channels of causality, the first one is obtained through the lagged variables (∆LPROD, ∆LGDP, ∆LINTER t ), when the coefficients of all these variables are statistically significant (F-distribution) and the second channel is raised in case the (λ) coefficient of the variable u t-1 is statistically significant (t-distribution). If λ is statistically significant in equation (5)  INSERT TABLE 6 From the results of table 6 we can infer that there is a unidirectional Granger causality between the price level and the productivity with direction from the price level to the productivity. This result is in accordance with the papers of Clark (1982), Ram (1984), and Buck and Fitzroy (1988) and Saunders and Biswas (1990) as well.
The price level and the productivity cause the gross domestic product for the examined period, while there is a bi-directional causal relationship between the gross domestic product and the interest rate. Finally, we can see that there is a dynamic causal relationship between the real gross domestic product and the productivity, and also between the interest rate and the productivity.

Conclusions
The purpose of this paper was to examine the subject of Granger causality between the price level and the productivity in a transition country to European Union such as Romania using quarterly data for the period 1990: Although Romania has high relatively inflation rates for the studied period, the results of empirical analysis suggested that there is a long run relationship between the productivity and the price level in both techniques of cointegration analysis which have been used, as well as in the multivariate cointegration analysis adding two more variables, which consist changes in real production such as gross domestic product and also in monetary policy such as the interest rate.
Then an error correction model's methodology has been used to estimate the short run and long run relationships. The selected vectors gave us the error correction terms, which proved to be statistically significant in 5% and 10% levels of significance respectively for the variables of the productivity and the real gross domestic product.
The results of causality analyses suggest that the Granger price level causes productivity. This result is consistent with the studies of Clark (1982), Ram (1984), Buck and Fitzroy (1988) and Saunders and Biswas (1990). Also, the price level and productivity cause the gross domestic product, while there is a bilateral causal relationship between gross domestic product and interest rate. Finally, there is a dynamic causal relationship between the gross domestic product and the productivity, but also between the interest rate and the productivity for the examined period.
τ µ is the t-statistic for testing the significance of δ 1 when a time trend is not included in equation 2 and τ τ is the t-statistic for testing the significance of δ 1 when a time trend is included in equation 2. The calculated statistics are those reported in D-F (1981). The critical values at 1%, 5% and 10% for N=50 are -3.58, -2.93 and -2.60 for τ µ and -4.15, -3.50 and -3.18 for τ τ respectively. The critical values for the P-P (1988) unit root tests are obtained from D-F(1981). The lag length structure of Φi of the dependent variable Xi is determined using a recursive procedure in the light of a Lagrange multiplier (LM) autocorrelation test (for orders up to four) which is asymptotically distributed as chi-squared distribution and the value of t-statistic of the coefficient associated with the last lag in the estimated autoregression. Numbers inside the brackets indicate significant levels. ***,**,* indicate significance at the 1, 5 and 10 percentage levels. Notes: The augmented D-F test is based on the equation (2) with constant and without trend, where u t is the estimated residual from the long-run model LCPI t = α 0 + α 1 LPROD t + u t (3) The lag length k is chosen so the estimated residuals of equation (2) will be without autocorrelation. The critical values for the rejection of null hypothesis of no coitegration between the two variables at 1%, 5% and 10% are -3.90, -3.33 and -3.04, respectively. The single equation error correction model is estimated for LCPI and LPROD The reported values are t tests for the estimated coefficient α 1 . The critical values for α 1 at 1%, 5% and 10% for N=50 are -4.32, -3.67 and -3.28, respectively.