Abstract
This Selected Issues paper and Statistical Appendix provides an estimate of the long-term path of Romania’s real exchange rate. The paper describes the recent history of the real exchange rate and its main determinants. A model is developed of a time-varying long-term exchange rate path, mainly driven by the terms of trade and net foreign direct investment. This long-term path is then used to assess the developments in Romania’s actual exchange rate. The paper also examines Romania’s growth potential in the medium term.
I. An Assessment of the Real Exchange Rate in Romania1
A. Introduction
1. The current monetary policy in Romania is focused on promoting gradual disinflation without putting external competitiveness at risk. In conducting monetary policy, the National Bank of Romania (NBR) primarily employs the exchange rate as a nominal anchor. In choosing the exchange rate path, the NBR seeks to strike a balance between taming inflation and avoiding unwarranted real appreciation, which might arise if other policies are not supportive. If the real exchange rate overshoots its long-run level, external competitiveness is negatively affected and the exchange rate becomes unsustainable. Therefore, assessing the real exchange rate, and in particular, trying to determine if the actual rate is over- or undervalued relative to its long-run level, is a crucial issue for conducting monetary policy in Romania.
2. This chapter provides an estimate of the long-run path of Romania’s real exchange rate. After briefly reviewing the literature, the chapter describes the recent history of the real exchange rate and its main determinants. A model is then developed of a time-varying long-run exchange rate path, mainly driven by the terms of trade and net foreign direct investment. This long-run path is then used to assess the developments in Romania’s actual exchange rate.
B. Background
3. Following the classic example of Balassa-Samuelson,2 several theories suggest that the long-run real exchange rate is time-varying and thus provide a rationale for deviations from purchasing power parity (PPP), according to which the long-run value of the real exchange rate is constant. A common procedure to estimate the long-run real exchange rate path applies the cointegration framework, which is used to test whether a long-run relationship between the real exchange rate and various economic fundamentals exists. If such a relationship can be established, it defines the long-run real exchange rate as a function of the identified explanatory variables.3
4. During the past decade, several studies have analyzed real exchange rates in transition countries. For a broad sample of transition countries, including Romania, Halpern and Wyplosz (1997) note that the real exchange rate exhibits strikingly similar features across these countries during the early 1990’s, irrespective of the exchange rate regime. Following an initial sharp depreciation at the beginning of economic transition, the real exchange rate trends upwards surpassing its initial value in most cases. Halpem and Wyplosz (1997), as well as Krajnyak and Zettelmeyer (1998), use U.S. dollar wages as a measure for the real exchange rate. Both studies estimate equilibrium dollar wages that suggest a pronounced undervaluation of domestic currencies in the beginning of the transition period, with the gap between the actual and the equilibrium exchange rate subsequently gradually declining in most of these countries, but not fully closing by 1995-96.4 In addition, the estimated path for equilibrium dollar wages trends upward in these studies. The observed increase in actual dollar wages therefore reflects both a recovery from initial undervaluation and an upward equilibrium trend. However, De Broeck and Slok (2001) argue that underlying productivity gains in the tradable sector have been the major source of real exchange rate movements in several central and eastern European and Baltic countries. In contrast, Barlow and Radulescu (2002) conclude that mean reversion following the sharp depreciation of the Romanian real exchange rate explains most of the subsequent appreciation.
5. This paper argues that the persistent real exchange rate appreciation in Romania since the mid-1990s has been caused by changes in underlying real factors, whereas part of the real exchange rate appreciation in the early 1990s could reflect initial overshooting at the beginning of the transition process.
C. Romania’s Real Exchange Rate and its Main Determinants, 1991–2001
6.Figure 1 shows the evolution of the annual CPI-based real effective exchange rate in Romania during 1991-2001.5 After a sharp depreciation in 1991/1992, the real exchange rate recovered in 1992/1993 correcting for a likely initial overshooting. Subsequently, the real exchange rate appreciated by about 40 percent until 2001.
Romania: The Real Effective Exchange Rate, 1991-2001.
Citation: IMF Staff Country Reports 2003, 012; 10.5089/9781451832785.002.A001
Source: IMF International Financial Statistics; Logarithms of CPI-based real exchange rate.Romania: The Real Effective Exchange Rate, 1991-2001.
Citation: IMF Staff Country Reports 2003, 012; 10.5089/9781451832785.002.A001
Source: IMF International Financial Statistics; Logarithms of CPI-based real exchange rate.Romania: The Real Effective Exchange Rate, 1991-2001.
Citation: IMF Staff Country Reports 2003, 012; 10.5089/9781451832785.002.A001
Source: IMF International Financial Statistics; Logarithms of CPI-based real exchange rate.7. In the literature, the following variables have been identified that can influence the real exchange rate in the Iong-run:6 (a) relative income levels; (b) the productivity differential between a country and its main trading partners; (c) net capital inflows; and (d) the terms of trade. Figure 2 shows the evolution of these variables in Romania over the past decade.
Romania: The Main Determinants of the Real Exchange Rate, 1991-2001.
Citation: IMF Staff Country Reports 2003, 012; 10.5089/9781451832785.002.A001
Source: National Institute for Statistics, Romania and IMF International Financial Statistics. The scales are in logarithms. The difference between any teo time points corresponds to the variable’s growth rate.Romania: The Main Determinants of the Real Exchange Rate, 1991-2001.
Citation: IMF Staff Country Reports 2003, 012; 10.5089/9781451832785.002.A001
Source: National Institute for Statistics, Romania and IMF International Financial Statistics. The scales are in logarithms. The difference between any teo time points corresponds to the variable’s growth rate.Romania: The Main Determinants of the Real Exchange Rate, 1991-2001.
Citation: IMF Staff Country Reports 2003, 012; 10.5089/9781451832785.002.A001
Source: National Institute for Statistics, Romania and IMF International Financial Statistics. The scales are in logarithms. The difference between any teo time points corresponds to the variable’s growth rate.8. In case of imperfect capital markets (see De Gregorio and Wolf (1994)) or nonhomothetic preferences for tradables and nontradables (see Bergstrand (1991)), an increase in relative income can raise the relative demand for nontradables and result in real appreciation. An increase in relative income can also bid up the relative price of fixed, nontradable assets such as land.
9. The Balassa-Samuelson hypothesis states that if a country’s productivity gains in the tradables sector outpace those in the nontradables sector relative to its trading partners, the relative price increase in nontradables would give rise to real appreciation. Although this effect would also be associated with a rise in the country’s relative income, this paper investigates the Balassa-Samuelson effect separately by directly employing the productivity differential between Romania and its main trading partners in the tradables sector.
10. Other studies, including Halpern and Wyplosz (1997), have suggested that net capital flows, in particular sustainable or long-run capital flows, tend to produce real appreciation.7 Following Paiva (2001), sustainable or long-run capital flows are proxied by using net foreign direct investment flows.
11. Several studies, including De Gregorio and Wolf (1994), have found a strong positive effect of the terms of trade on the real exchange rate.8 In transition countries the sluggish integration into Western markets could have gradually improved the terms of trade during the 1990’s (see Halpern and Wyplosz (1997)). Traded goods produced by transition countries were initially of low quality and poorly marketed. An improvement in quality and marketing procedures for exports would tend to drive upward the terms of trade.
D. Empirical Methodology and Data Analysis
12. According to Lane and Milesi-Ferretti (2002), the estimation of a long-run exchange rate path includes the following steps: (a) testing for the presence of unit roots in the data series; (b) testing for the existence of a cointegrating relationship; and (c) estimating the cointegrating relationship. The data are annual over the period 1991-2001. The real effective exchange rate is CPI-based, with weights set according to Romania’s trading pattern in 1995. Relative GDP per capita is measured as Romania’s real GDP per capita in PPP terms relative to the trade weighted real GDP per capita of its main trading partners9. Total productivity in the tradables sector is proxied by using labor productivity because reliable capital stock data, as well as capital utilization rates, are not available for Romania. Labor productivity for the industrial sector over the past decade is provided by the National Institute for Statistics in Romania. The productivity differential is then calculated as the difference between the Romanian productivity level and the trade weighed average of its main advanced trading partners productivity levels. Sustainable net capital flows are proxied by the net foreign direct investment to GDP ratio. The terms of trade are defined as the ratio of export unit values to import unit values as provided by Romania’s National Institute for Statistics. All variables are transformed into logarithms. Given the relatively short time period and a limited number of observations, the results of this econometric exercise should be treated cautiously.
13.Table 1 presents the unit root test results for the augmented Dickey-Fuller and Phillips-Perron tests. The null of nonstationarity in levels cannot be rejected for any of the variables at the 10 percent significance level. Although the individual series are nonstationary, a linear relationship may still exist among some of them such that the residual is stationary. These variables are then said to be cointegrated. Given the short time period, cointegration is only tested for all possible pairs of explanatory variables. Table 2 reports the results of such cointegration tests applying the Johansen procedure. The Johansen test rejects the null of no cointegration in favor of a single cointegration vector in several cases, including specifications that allow for a deterministic trend in the cointegrating equation.10 For all these cases in which the presence of a cointegrating relationship has been established, a long-run equation is estimated using Phillips and Hansen’s (1990) Fully Modified OLS estimates.11
Unit Root Tests
Unit Root Tests
| ADF | Phillips-Perron | |
|---|---|---|
| REER | -1.50 | -1.85 |
| (a) Relative GDP per capita | -1.94 | -1.88 |
| (b) Productivity differential | -2.39 | -1.21 |
| (c) NFDI to GDP ratio | -1.48 | -2.24 |
| (d) Terms of trade | -2.06 | -1.36 |
Unit Root Tests
| ADF | Phillips-Perron | |
|---|---|---|
| REER | -1.50 | -1.85 |
| (a) Relative GDP per capita | -1.94 | -1.88 |
| (b) Productivity differential | -2.39 | -1.21 |
| (c) NFDI to GDP ratio | -1.48 | -2.24 |
| (d) Terms of trade | -2.06 | -1.36 |
Cointegration Tests
Cointegration Tests
| Johansen (1) | Johansen (2) | |
|---|---|---|
| REER and (a) and (b) | 18.02 | 40.26 |
| REER and (a) and (c) | 51.83** | 57.53** |
| REER and (a) and (d) | 28.20 | 41.21 |
| REER and (b) and (c) | 34.74* | 44.91* |
| REER and (b) and (d) | 24.78 | 50.18** |
| REER and (c) and (d) | 41.39** | 45.48** |
Cointegration Tests
| Johansen (1) | Johansen (2) | |
|---|---|---|
| REER and (a) and (b) | 18.02 | 40.26 |
| REER and (a) and (c) | 51.83** | 57.53** |
| REER and (a) and (d) | 28.20 | 41.21 |
| REER and (b) and (c) | 34.74* | 44.91* |
| REER and (b) and (d) | 24.78 | 50.18** |
| REER and (c) and (d) | 41.39** | 45.48** |
14. For all three specifications that included the productivity differential in the cointegrating relationship, FMOLS estimation did not produce a significant parameter estimate for the productivity differential. These results indicate that the Balassa-Samuelson effect fails to explain much of the real exchange rate movements in Romania over the past decade.12 Assuming a declining positive time trend resulted in a significant estimate for relative GDP per capita. However, the parameter estimate was strongly negative and net foreign investment turned out to be insignificant in that case. Only in the case when net capital flows and the terms of trade were employed as explanatory variables and no time trend was included, the FMOLS procedure yielded significant parameter estimates for both explanatory variables. These results are reported in Table 3.13 The terms of trade exert a strong positive effect on the real exchange rate. Net capital flows also tend to push upward the real exchange rate but the quantitative effect is small.14
Estimates of The Long-run Relationship
Estimates of The Long-run Relationship
| NFDI | Terms of trade | Constant | |
|---|---|---|---|
| Phillip-Hansen | 0.13 | 1.22 | -0.27 |
| estimates | |||
| Standard error | (0.02) | (0.22) | (0.45) |
| P-value | 0.000 | 0.001 | 0.575 |
Estimates of The Long-run Relationship
| NFDI | Terms of trade | Constant | |
|---|---|---|---|
| Phillip-Hansen | 0.13 | 1.22 | -0.27 |
| estimates | |||
| Standard error | (0.02) | (0.22) | (0.45) |
| P-value | 0.000 | 0.001 | 0.575 |
15. Figure 3 shows the actual and fitted values according to the estimation results shown in Table 3, as well as the residuals. Fitted values for the real exchange rate represent the long-run real exchange rate under the hypothesis that the terms of trade and net foreign investment have been the main real variables affecting the real exchange rate path over the past decade. After deviating substantially from its long-run path in both directions in the early 1990s, the real exchange rate became strongly overvalued in 1998, the year in which Romania’s financial crisis culminated. The sharp devaluation of the nominal exchange rate that the NBR engineered subsequently led to a somewhat undervalued real exchange rate in 1999. A depreciated real exchange rate in 1999 and the modest appreciation during 2000-01, consistent with the long-run exchange rate trajectory, most likely contributed to Romania’s strong export performance during the past two years.
Romania: Actual and Fitted Values for the Real Exchange Rate and Residuals, 1991-2001
Citation: IMF Staff Country Reports 2003, 012; 10.5089/9781451832785.002.A001
Romania: Actual and Fitted Values for the Real Exchange Rate and Residuals, 1991-2001
Citation: IMF Staff Country Reports 2003, 012; 10.5089/9781451832785.002.A001
Romania: Actual and Fitted Values for the Real Exchange Rate and Residuals, 1991-2001
Citation: IMF Staff Country Reports 2003, 012; 10.5089/9781451832785.002.A001
16. Regarding expectations about the future long-run path of the real exchange rate, the above findings have several important implications. The positive trend of the terms of trade is likely to fade out in the near future as Romanian export companies become established players in foreign markets and, therefore, would not be expected to keep pushing the long-run real exchange rate upward. Given the relatively weak impact of net capital flows on the long-run real exchange rate, the prospects for further sustainable appreciation coming from this source would also be weak. Against this background, any sizable real appreciation of the actual real exchange rate in the years to come could likely outpace its long-run path. The above results include that a significant statistical relationship between the long-run real exchange rate and the productivity differential or relative GDP could not be found for the past decade. However, these findings do not rule out the possibility that those factors gain a more prominent role in affecting the long-run real exchange rate in the future. In addition, the sizable undervaluation of the Romanian real exchange rate according to the absolute purchasing power parity hypothesis may also contribute to further real appreciation.
E. Conclusion
17. This chapter provided an assessment of the real exchange rate in Romania over the past decade. The investigation started with the assumption that real factors have been driving the sustained real appreciation over the past decade. An econometric analysis led to the conclusion that the time-varying path for the long-run real exchange rate had been mainly determined by the terms of trade and net capital flows, both pushing the long-run exchange rate path upward. The estimated long-run path indicated a significant overvaluation of the’ real exchange rate in 1998 which was reversed in 1999 and set the basis for the strong export performance in 2000 and 2001. With structural improvement in the terms of trade likely to diminish and the relatively weak impact of capital inflows, the scope for further real appreciation seems limited. However, in view of the relatively short time period on which the analysis was based on, it cannot be ruled out that other real effects such as the Balassa-Samuelson effect could assume a more important role in determining the real exchange rate in Romania and could possibly allow for further significant and sustainable real appreciation.
References
Balassa, B., 1964, “The purchasing power parity Doctrine: A Reappraisal”, Journal of political Economy, 72, 584- 96.
Barlow, D. and R. Raduleseu, 2002, “Purchasing Power Parity in the Transition: the Case of the Romanian Leu Against the Dollar”, Post-Communist Economics, 14, 123- 135
Bergstrand, J., 1991, “Structural Determinants of Real Exchange Rates and National Price Levels: Some Empirical Evidence’, American Economic Review, 81, 325- 34.
Coorey, S., M. Mecagni and E. Offerdahl, 1996, “Disinflation in Transition Economies: The Role of Relative Price Adjustments”, IMF Working Paper 96/138, (Washington: International Monetary Fund).
De Broeck, M., and T. Slok, 2001, “Interpreting Real Exchange Rate Movements in Transition Countries”, IMF Working Paper 01/56, (Washington: International Monetary Fund).
De Gregorio, J. and H. C. Wolf, 1994, “Terms of Trade, Productivity, and the Real Exchange Rate”, NBER Working Paper, No. 4807, (Cambridge Massachusetts: NBER).
Desruelle, D. and A. Zanello, 1997, “A Primer on the IMF’s Information Notice System”, IMF Working Paper 97/71, (Washington: International Monetary Fund).
Edwards, S., 1989, Real Exchange Rates, Devaluation and Adjustment: Exchange Rate Policy in Developing Countries. (Cambridge Massachusetts: MIT Press).
Halpern, L. and C. Wyplosz, 1997, “Equilibrium Exchange Rates in Transition Economies”, IMF Staff Papers, 44, 430- 61.
Isard, P., H. Faruqee, G. R. Kincaid and M. Fetherston, 2001, Methodology for Current Account and Exchange Rate Assessments, IMF Occasional Paper No. 209 (Washington: International Monetary Fund).
Krajnyak, K. and J. Zettelmeyer, 1998, “Competitiveness in Transition Economies: What Scope for Real Appreciation?”, IMF Staff Papers, 45, 309- 62.
Lane, P. R. and G. M. Milesi-Ferretti, 2002, “Long-run Determinants of the Irish Real Exchange Rate”, Applied Economics, 34, 549- 553.
Paiva, C, 2001, “Competitiveness and the Equilibrium Exchange Rate in Costa Rica”, IMF Working Paper 01/23, (Washington: International Monetary Fund).
Phillips, P. C. B. and B. E. Hansen, 1990, “Statistical Inference in Instrumental Variables Regression with 1(1) Process”, Review of Economic Studies, 57, 99- 125.
Rogoff, K., 1996, “The Purchasing Power Parity Puzzle”, Journal of Economic Literature, 34, 647- 668.
Samuelson, P. A., 1964, “Theoretical Notes on Trade Problems”, Review of Economics and Statistics, 46, 145- 54.
Prepared by Thomas Harjes.
Balassa (1964) and Samuelson (1964) argued that if a country’s productivity in the tradables sector increases relative to its trading partners’ (and relative to its own nontradables sector), its real exchange rate would appreciate.
Many studies refer to such an empirical long-run relationship as defining the equilibrium real exchange rate. In that setting, one option is to define the equilibrium exchange rate as the exchange rate consistent with both internal and external equilibrium (Edwards (1989)). However, as Isard and others (2001) point out, the derived estimates of equilibrium exchange rates are then conditional on assumptions about the equilibrium values of the explanatory variables which are difficult to establish for Romania over the past decade.
In the case or Romania, the degree of potential “undervaluation” of the real exchange rate is large when measured as U.S. dollar wages. Halpern and Wyplosz (1997) estimate that in 1996, Romania’s actual dollar wages were at about 25-35 percent of their equilibrium level.
Other variables include real public spending, the degree of openness and net foreign assets.
However, the accumulation of net foreign liabilities would require Romania to run a trade surplus in the future and may then require a real depreciation of the exchange rate.
There are various channels through which the terms of trade can affect the real exchange rate and in general, the overall effect is ambiguous. De Gregorio and Wolf (1994) show that in a small open economy model the terms of trade positively affect the real exchange rate if imperfect capital mobility is introduced.
Romania’s main advanced trading partners include: Austria, France, Germany, Greece, Italy, the Netherlands, the U.K. and the U.S.A. PPP conversion factors are provided by the World Bank.
As noted in Coorey, Mecagni and Offerdahl (1996), transition countries inherited a set of natural resource prices below world market prices and kept administered prices including public utility prices considerably below cost recovery levels in the beginning of the transition period. When these prices are raised toward world market prices and cost recovery levels, the measured real exchange rate appreciates. This effect can be captured by allowing for a deterministic time trend in the cointegrating equation.
If the regressors and the residuals are correlated, the t-statistics of standard OLS estimators are invalid. The Phillips-Hansen FMOLS estimation takes account of these possible correlations in a semi-parametric manner. In all cases in which the cointegrating equation includes a trend, a linear and alternatively a “root” trend has been included in the FMOLS estimation.
Figure 1 and Figure 2 show that both the real exchange rate and the productivity differential trend upward over the past decade. However, in some years and especially during the 1997-99 crisis the variables move in opposite directions and weaken the case for the productivity differential being a significant contemporaneous explanatory variable. The limited number of observations do not allow for a more elaborate model structure, including lagged variables or trying to control for the above event, that could yield a different outcome. Simply using one-period lags for the productivity differential did not change the outcome.
Standard OLS estimation led to the same results regarding the significance of the various explanatory variables.
Net capital flows were part of all significant cointegrating equations. However, they alone do not drive the results: regressing the real exchange rate on net capital flows only or net capital flows and a time trend yielded a poor fit and insignificant parameter estimates.
