Romania: Selected Issues and Statistical Appendix
Author: International Monetary Fund
Publication Date:
26 Jul 2004
eISBN:
9781451832792
Language:
English
Keywords:
ISCR; CR; interest rate; rate; market loan rate; rate equation; cause market rate; pass-through in Romania; pass-through from policy; Market interest rates; Deposit rates; Loans
Download PDF (717.8 KB)

This Selected Issues paper for Romania reports that the practice of nonpayment and arrears accumulation has been widespread in Romania. Managers of enterprises that remain in the pipeline for privatization for long periods of time have little incentive to reduce arrears. The state contributed to growth of arrears by accepting nonmonetary tax and utility payments, using tax offsets in procurement, and tolerating payment arrears. These practices have been prevalent at all levels of state and local government, as well as state utility companies.

Abstract

This Selected Issues paper for Romania reports that the practice of nonpayment and arrears accumulation has been widespread in Romania. Managers of enterprises that remain in the pipeline for privatization for long periods of time have little incentive to reduce arrears. The state contributed to growth of arrears by accepting nonmonetary tax and utility payments, using tax offsets in procurement, and tolerating payment arrears. These practices have been prevalent at all levels of state and local government, as well as state utility companies.

III. Romania: Transmission of Policy Interest Rate to Market Rates31

A. Introduction

1. This paper aims to test the hypothesis that the interest rate pass-through from policy to market rates plays a lesser role in Romania than in other transition economies in the region. The policy interest rate pass-through is claimed to be more slow and limited as a consequence of specific features of the Romanian monetary policy framework. The transmission from the policy interest rate to the lending and deposit rates studied here is part of the broader issue of the effectiveness of interest rate policy in controlling inflation and affecting aggregate demand, which, however, goes beyond the scope of this paper.

2. Several factors are usually referred to as explaining the ineffectiveness of interest rate policies. Those that Romania shares, to a larger or smaller extent, with other countries in the region are a low degree of monetization, underdeveloped financial markets, and capital controls. In addition, the lending policies of banks are often found to be price inelastic with respect to interest rates in the short run, because other, non-interest rate factors, like adjustment costs and, sometimes, directed lending, play a substantial role (see e.g. Cottarelli and Kourelis (1994), Schaechter, Stone, and Zelmer, 2000 or Carare et. al. (2002)). The balance sheet problems in the banking and corporate sectors are also frequently mentioned, but in the case of Romania they do not seem to be of critical importance.

3. The Romanian monetary system has, however, some specific characteristics that could potentially further weaken the interest rate instrument. Starting with 1997, the Romanian economy exhibited a strong and consistent increase in structural excess liquidity (Anthoni, Udea, and Braun, 2003). As the NBR has been increasing its reserves sharply, it had to control high-powered money by accepting deposits from the commercial banks. Hence instead of borrowing from the central bank, commercial banks typically have substantial deposits over and above their reserve requirements at the NBR. Therefore, instead of reflecting the marginal costs of funding for the commercial banks, the policy interest rate it merely reflects an opportunity cost. Since empirical evidence suggests that commercial banks react differently to cost increases than to revenue decreases, the question arises whether the Romanian situation of excess liquidity could cause such asymmetric behavior of banks, rendering policy interest rate less effective.

4. After estimating interest rate pass-through coefficients for several CEE economies, the paper concludes that the pass-through in Romania is in line with that in other countries in the region. Further research would be needed to estimate the contribution of various factors to the effectiveness of the policy interest rate.

B. The Model

5. The paper measures the interest rate pass-through from the policy rate to market rates by employing an error-correction framework. Assuming perfect competition in the loan market, the relation between market and policy rates can be described by

im = α+β⋅ip,(1)

where im is the market loan rate, ip is the policy rate, α is a mark up, and β reflects the demand elasticity of market rates with respect to policy rates. Relatively inelastic demand (an elasticity β lower than 1) is likely to be found when banks have substantial market power, either because no close substitutes for bank loans exists, i.e. when capital markets are underdeveloped, or because of the structure of the market for bank loans (De Bondt, 2002). A wide range of factors influence the structure of the market, such as the degree of stateownership of the banking sector, and the degree and form of regulation, including market entry restrictions and menu costs. Relatively elastic demand would signal that bank credit is not-rationed. In such a setting, banks would want to lend money to both low and high risk borrower, equalizing returns on both types of lending by charging risk-adjusted rates to the high-risk borrowers. Hence, the risk adjustment in the rate might on average cause market rates to react more than one-to-one to changes in the policy rate.

6. Relationship (1) does not touch upon the issue of timing. Market interest rates will not react instantly to changes in the policy rate. Even though bank will quickly adapt their short-term lending rates, medium and long-term rates will react more slowly, or not at all, as they are primarily guided by expectations of future short-term rates. Moreover, average lending rates will adapt only gradually, as new loans replace old ones. These considerations point to a gradual adjustment of market rates to the new policy rates. Therefore, equation (1) should be interpreted as valid only in the long run.

7. The long-run nature of equation (1) suggests a model in which equation (1) can be seen as a long-run equilibrium relationship, around which short-term dynamics abound. Such an approach is well-established in the literature. Engle and Granger (1987) suggest a two-step approach in which the long-run relationship is fitted in levels, while in the second step involves regressing the first differences of the dependent variables on their lagged values and lagged deviations from the long-run equilibrium relationship. This approach, labeled error-correction, is warranted as long as the dependent and explanatory variables are cointegrated, i.e. both are non-stationary, but there exists a linear combination of these series which is stationary. In general, interest rates series would not be expected to be non-stationary, as they normally do not exhibit a long-term trend. In transition economies, however, one might expect interest rate series to exhibit a declining trend as the transition takes hold and the problem of inflation is reigned in. This would imply these series to be integrated of order 1 (I(1)). To establish this hypothesis, the paper performs unit root tests on the series by applying the augmented Dickey-Fuller (1981) test on the individual series. In case both the policy rates and the market rates are I(1), the series might be cointegrated, which is subsequently tested using standard Johansen (1988, 1991) statistical tests. When a cointegrating relationship is found, the suggested interpretation of equation (1) as a long-run equilibrium relationship, around which short-term dynamics abound, is justified from a statistical point of view.

An error-correction model (ECM) of interest rate pass-through can be specified as

Δitm = γ1 + γ2Δit−1m + γ3(it−1m − β⋅it−1p − α) + ηt.(2)

Here, Δ is the difference operator, and the equation states that is the first difference of market interest rates, Δimt, depends on its own one-period lag, Δimt-1, the deviation from the long-run relationship in the last period, imt-1-β.ipt-1-α, and a constant, γ1. In such an ECM, the coefficient γ3 indicates the speed of adjustment of the short-run dynamics to the long-run equilibrium relationship. This coefficient hence can be interpreted to signal the effectiveness of the interest rate instrument of monetary policy; a higher value of the coefficient signals a faster market response and hence a more effective first step in the interest rate channel of monetary transmission.

8. This paper employs ECM (2) to test the whether the interest rate pass-through in Romania is low compared to other transition economies in the region, as claimed previously due to the nature of the monetary policy regime. This is done by a simple comparison and statistical testing of estimation results from different transition economies in the region.

C. The Data

9. For the purpose of estimation, data from a wide range of transition economies in Central and Eastern Europe is collected. The countries included are Romania, the Czech Republic, Hungary, Poland, the Slovak Republic, and Slovenia. The period under consideration is January 1995 - February 2004, and the frequency of data is monthly. Because of the transition most economies in the region have experienced, data problems abound: The Baltic states were not included owing to the lack of data, while Bulgaria was not included owing to its currency board arrangement. The sample for Slovak Republic has been limited to 2000–04 period, owing to the switch in the monetary regime from an exchange rate peg to a disinflationary regime with a floating exchange rate in 199832. The remaining countries each have broadly comparable monetary policy regimes, with inflation as the primary, or in some cases the sole target of monetary policy.

10. For these countries, the monthly data consists of average short and long-term loan rates, deposit rates, and the central bank policy rates. The period for which all data are available vary by country, but even the shortest series still has at least three years of monthly data available. In addition, short series of monthly interest rates on new loans (as opposed to all loans) are available for the Czech and Slovak Republics, and Romania.

D. Results on Outstanding Loan Rates: Equilibrium Equation and Basic ECM

11. Estimations results for the series on outstanding-loan rates are in Table 1. The table contains results on, first, equation (1), which is estimated for all short- and long-term lending rates on the outstanding stock of loans. Second, unit roots test are performed on all data series, using the standard augmented Dickey-Fuller test at the 5 percent uncertainty level. All policy rates and long- and short-term lending rates are found to be integrated of order 1, with the exception of the short term rate for Romania (which is found to be I(2)) and the short-term and policy rates for Slovenia (which are found to be I(0)). Third, to test for cointegration between the market and policy rates, standard Johansen cointegration tests are performed on the pairs of series.

12. In all countries in the sample, the policy rate is a highly significant explanatory variable for both the short- and the long-term market rates. Significance is lowest (but still high) in Hungary and Slovenia, presumably because of the small length of the time series in the case of Hungary (data from January 2000 onwards only), while the Slovenian policy rate is characterized by only a few movements since 1995. The magnitude of the estimated coefficients varies between 0.67 and 2.07, with most estimates being close to 1. Coefficient estimates below 0.8 are found for the Czech Republic (short- and long-term rates), Hungary (long-term rate), Romania (short- and long-term rates), and the Slovak Republic (long-term rate). This point to substantial market power of commercial banks, be it because no close substitutes for bank lending exists, or because of the limited competition in the banking market. In contrast, the banking markets in Poland and especially Slovenia exhibit relatively elastic demand, which hints at a market where credit is not rationed and banking competition is amply present.

13. Cointegration tests confirm that the market rates can to a large degree be explained by the policy rates. For the series which are I(1), this indicates that there is a high degree of co-movement between policy and market rates. The one pair of series that fails the cointegration test consists of the Hungarian short-term market rate and the Hungarian policy rate. This is presumably due to the short series being tested. From the above, the general conclusion is that the policy rate is a highly relevant explanatory variable for the market lending rate in the long run, as can be expected in a market economy. This allows the estimation of the ECM specifications.

14. The estimation results for the basic ECM for each country, as specified in equation (2), are in Table 2. The fit of the estimated equations, as indicated by the R2, is low for all of the equations. At the same time, the Durbin-Watson (1950, 1951) test statistic indicates little serial autocorrelation in the residuals. Both effects are the normal consequences of estimating a model in first differences. The main parameter of interest in the ECM is the estimate c(3) of the coefficient γ3, which indicates the speed of adaptation of the short-term dynamics to the long-run equilibrium equation. This coefficient estimate thus is a measure of the speed of the pass-through of the policy interest rate to the market rates, and hence of the effectiveness of the interest rate channel. Since the coefficient indicates adaptation to the long-run equilibrium, it is expected to be negative.

15. For the series on rates on outstanding loans, the hypothesis that the interest rate pass-through is low in Romania compared to other transition countries, is contradicted. For most countries in the sample, the estimated adaptation coefficient c(3) is negative and significantly different from 0 at the 5 percent uncertainty level. However, in the case of Slovenia, the coefficient estimates are significantly different from 0 only at the 8 percent (long-term rates) or 11 percent (short-term rates) uncertainty levels. In the sample, the coefficient estimates range from -0.08 to -0.39, with almost all estimates being in the range -0.09 to -0.18. The coefficient estimates for Romania, at -0.14 and -0.15 for the short- and long-term rate respectively, are not substantially different from the estimates for the other transition countries in the sample. Statistical testing indicates that the adaptation coefficient for the short-term rate is significantly larger than -0.08 (the lower bound of the estimates for the other countries) at the 5 percent uncertainty level, while the same holds for the long-term coefficient estimates, but only at a 12 percent uncertainty level.

E. Results on Deposit, Newly Issued Loan Rates, and Panel Estimations

16. Estimation results for deposit rate data also reject the hypothesis that the passthrough in Romania is weaker than in other countries. The estimation results for the long-run equilibrium equation for the deposit rates are in Table 3. Most series are cointegrated, indicating that estimation through ECM methodology is warranted. The results for the ECMs for deposit rates in the individual countries are in Table 4. All the estimates of the adaptation coefficients are negative and in most cases they are significantly different from 0, the exceptions being the estimates in the long-term rate equation for the Slovak Republic, and in the short-term rate equation for Poland. The other estimates are in the range -0.13 to -0.60, with the estimate for the long-term rate equation for Romania being -0.24. Once again, statistical testing shows that Romania does not stand out as exhibiting an exceptionally slow speed of adaptation.

17. Estimation results on series for newly issued loans suggest that the pass-through is fast and almost one-to-one. However, data on newly issued loans are not available for most countries in the sample. Time series comprising more than a year are only available for the Czech and Slovak Republics, while for Romania, time series spanning just 10 months are available. The series for Romania are too short even to perform unit root tests and hence are not suited for analysis in an ECM framework. The estimation results are shown in Tables 5 and 6. They confirm that monetary policy transmission from policy rates to market rates is generally fast and almost one-to-one. In the long-term equilibrium equation, all estimates of the policy rate coefficients are highly significant, with estimated values between 0.83 and 1.21, i.e. close to 1. While the Czech long-rate coefficient estimate at 0.83 is still significantly different from 1 at the 5 percent uncertainty level, the Czech short-rate coefficient and both the long- and short-rate Slovak coefficients are not statistically different from 1.

18. Pooling the data series in a panel regression yields inconclusive results. The results of a fixed effects panel estimation with a common coefficient on the policy rate confirm the policy rate as a highly significant variable for the market rates, with values of the t-statistic of 29.0 and 36.5 for the short- and long-term equation respectively. Further estimation in an ECM framework, using the residuals from the panel regression for the long-run equilibrium equation, does not yield any conclusive results. The cause presumably lies in the fact that significant changes in monetary policies in the different countries in the sample occurred at very different points in time. Hence, the residuals of the long-run equilibrium relation look very different when this relationship is estimated in a panel than when estimated for the countries individually.

F. Results: Time Consistency

19. Estimation results for Romania clear differ when different time periods are taken into account (Table 7). To see if the above results are constant over time or whether the market evolved over time, the data series for Romania are split in two, taking as the break point the first month in which the policy interest rate was below 40 percent. The two samples are October 1999- June 2001 and June 2001-January 2004. Estimation results for the different samples clearly differ, as seen in Table 7.

20. In the earlier period the policy rate does not significantly influence the market interest rates. In addition, no cointegration between market and policy rates is found, which also prevents a well-founded interpretation of the estimation results of the ECMs.

21. In sharp contrast, in the later period, the policy rate is highly significant for the market rates and the series are cointegrated. In addition, the coefficients for the policy rates are considerably higher than the estimates for the full sample, a difference which is statistically significant at the 5 and 10 percent level for the short- and long term rate series respectively. These higher estimates indicate that the Romanian banking market has developed towards a more complete market while banking competition increased.

22. Moreover, the interest rate pass-through in Romania has increased over time. The estimates of the adaptation coefficients in the ECMs indicate a significantly swifter adaptation of short-term dynamics to the long-run equilibrium than in the regressions for the full sample. In other words, the interest rate channel of monetary policy may have become more effective over time.

G. Conclusions

23. Claims that the particular features of Romania’s monetary policy regime result in a lower effectiveness of its interest rate instrument are contradicted by the results of this study, which can be summarized as follows:

  • The estimates of interest rate pass-through from policy interest rates to rates on the outstanding volume of loans and deposits in Romania are in line with coefficient estimates for other transition economies in the region.

  • Results for data series on newly issued loans suggest that, in some of the transition countries in the sample, market rates for new loans react to policy rate changes quite fast. For Romania, however, the time series span too short a period.

  • Panel data regressions are inconclusive, likely due to differences in the timing of significant changes in monetary policy in the different countries in the sample. Hence, fitting the same long-run relationship on all countries in the sample yields systematic distortions in the residuals for the individual countries.

24. Moreover, studying the Romanian loans market for different time periods strongly suggests that the interest rate pass-through from policy to market rates has become more pronounced over time. It also suggests that the Romanian banking market was further developed in the later years compared to the period before mid-2001, and became more competitive, with less market power for individual banks.

APPENDIX TABLES

H. Appendix: Estimation Tables

Table 1.

Country Long-Term Equations - Loan Rates

Country_Rate, t = c(1) + c(2) * Country_Policy_Rate, t

article image

Using the standard Johansen Cointegration Test

Note: all series are I(1) at the 5 percent uncertainty level, except Rom_St_Out, which is I(2) and SVN_St_Out and SVN_Pol, which are I(0).
Table 2.

Country ECM Estimation Results - Loan Rates

D(Country_Rate), t = c(1) + c(2) * D(Country_Rate), t-1 + c(3) * L-T-Eq_Resi

article image
Table 3.

Country Long-Term Equations - Deposit Rates

Country_Rate, t = c(1) + c(2) * Country_Policy_Rate, t

article image

Using the standard Johansen Cointegration Test

Note: all series are I(1) at the 5 percent uncertainty level, except SVN_Dep_St_Out and SVN_Pol, which are I(0).
Table 4.

Country ECM Estimation Results - Deposit Rates

D(Country_Rate), t = c(1) + c(2) * D(Country_Rate), t-1 + c(3) * L-T-Eq_Resi

article image
Table 5.

Country Long-Term Equations - Rates on Newly Issued Loan

Country_Rate, t = c(1) + c(2) * Country_Policy_Rate, t

article image

Using the standard Johansen Cointegration Test

These test results should be treated with caution, as no unit root tests could be performed on the series.

Note: all series for the Czech Republic and Slovakia are I(1) at the 5 percent uncertainty level, while the series for Romania are too short to perfrom unit root tests.
Table 6.

Country ECM Estimation Results - Rates on Newly Issued Loans

D(Country_Rate), t = c(1) + c(2) * D(Country_Rate), t-1 + c(3) * L-T-Eq_Resid

article image

These test results should be treated with caution, as no unit root tests could be performed on the series.

Table 7.

Romania: Estimation Results for Different Samples - Loan Rates

article image

Using the standard Johansen Cointegration Test

References

  • Anthoni, D., I. Udrea, and H. Braun, (2003), “Monetary Policy Transmission in Romania,”, National Bank of Romania Occasional Paper 3. http://www.nbr.ro.

    • Search Google Scholar
    • Export Citation

  • De Bondt, G., (2002), “Retail Bank Interest Rate Pass-Through: New Evidence at the Euro Area Level,” ECB Working Paper 136.

  • Borio, C.E.V., (1997), “The Implementation of Monetary Policy in Industrial Countries: A Survey,” BIS Economic Papers 47.

  • Carare, A., A. Schaechter, M.R. Stone, and M. Zelmer, (2002), “Establishing Initial Conditions in Support of Inflation Targeting,” IMF Working Paper 02/102.

    • Search Google Scholar
    • Export Citation

  • Cottarelli, C., and A. Kourelis, (1994), “Financial structure, Bank Lending Rates, and the Transmission Mechanism of Monetary Policy,” IMF Staff Papers, vol. 41, pp. 587–623.

    • Search Google Scholar
    • Export Citation

  • Dickey, D., and W. Fuller, (1981), “Likelihood Ratio Tests for Autoregressive Time Series with a Unit Root,” Econometrica, vol. 49, pp. 1057–1072.

    • Search Google Scholar
    • Export Citation

  • Durbin, J., and G. Watson, (1950), “Testing for Serial Correlation in Least Squares Regression – I,” Biometrika, vol. 37, pp. 409–428.

    • Search Google Scholar
    • Export Citation

  • Durbin, J., and G. Watson, (1951), “Testing for Serial Correlation in Least Squares Regression – II,” Biometrika, vol. 38, pp. 159–178.

    • Search Google Scholar
    • Export Citation

  • Engle, R., and C. Granger, (1987), “Co-integration and Error Correction: Representation, Estimation and Testing,” Econometrica, vol. 35, pp. 251–276.

    • Search Google Scholar
    • Export Citation

  • Johansen, S., (1988), “Statistical Analysis of Cointegrating Vectors,” Journal of Economic Dynamics and Control, vol. 12, pp. 231–254.

    • Search Google Scholar
    • Export Citation

  • Johansen, S., (1991), “Estimation and Hypothesis Testing of Cointegrating Vectors in Gaussian Vector Autoregressive Models,” Econometrica, vol. 59, pp. 1551–1580.

    • Search Google Scholar
    • Export Citation

  • Schaechter, A., M.R. Stone, and M. Zelmer, (2000), “Adoption of Inflation Targeting: Practical Issues for Emerging Market Countries,” IMF Occasional Paper 202.

    • Search Google Scholar
    • Export Citation

  • Svensson, L., (1998), “Open-Economy Inflation Targeting,” NBER Working Paper 6545.

STATISTICAL APPENDIX

Table 1.

Romania: GDP by Origin, 1993-2003

article image
Source: National Institute of Statistics. ESA 79 methodology in 1993-97, ESA 95 methodology in 1998-2003.

Semifinal data.

Provisional data.

Including electric and thermal energy, gas and water.

Services including financial intermediation services indirectly measured.

Net taxes.

Table 2.

Romania: GDP by Expenditure, 1993-2003

(In percent)

article image
Source: National Institute of Statistics. ESA 79 methodology for 1993-98, ESA 95 methodology for 1999-2003.For shares of GDP, ESA 79 methodology for 1993-97, ESA 95 methodology for 1998-2003.

Semifinal data.

Provisional data.

Table 3.

Romania: Investment by Sector, 1993-2003

(In billions of lei at current prices)

article image
Source: National Institute of Statistics.
Table 4.

Romania: Saving-Investment Balance, 1993-2003

(Current prices)

article image
Sources: National Institute of Statistics and National Commission for Economic Forecasting.

Semidefinitive data.

Provisional data.

Preliminary data of National Commission for Economic Forecasting.