This Selected Issues paper analyzes the developments and determinants of inflation in Romania, and reviews the salient trends in public finance. The study describes the monetary policy issues and the improvements required to clean up the financial sector. The paper chronicles the balance-of-payments crisis in 1999, the external viability trends, reviews the economic and financial implications, and assesses Romania's compliance with EU economic criteria. The paper also provides a statistical appendix for the country.

Abstract

This Selected Issues paper analyzes the developments and determinants of inflation in Romania, and reviews the salient trends in public finance. The study describes the monetary policy issues and the improvements required to clean up the financial sector. The paper chronicles the balance-of-payments crisis in 1999, the external viability trends, reviews the economic and financial implications, and assesses Romania's compliance with EU economic criteria. The paper also provides a statistical appendix for the country.

I. Inflation in Romania—Developments and Determinants1

A. Overview

1. Inflation in Romania has been high and variable through the past decade, owing fundamentally to stop-go stabilization efforts and widespread financial indiscipline. This financial indiscipline has taken several forms over time, including large fiscal and quasi-fiscal deficits, accumulation of arrears, and outbreaks of wage growth well in excess of productivity.

2. Econometric evidence highlights the role of unit labor costs, and to a lesser extent the exchange rate, in driving inflation. Unit labor costs have been the leading proximate determinant of inflation, with deeply rooted financial indiscipline at the enterprise level being largely reflected in higher wages than justified by productivity, or than could even be paid in many enterprises in the absence of soft budget constraints. The influence of the exchange rate on inflation has also become increasingly clear over the past few years, in the wake of the full liberalization of the foreign exchange market.

3. The role of money and credit growth in causing inflation has also been important, though harder to demonstrate empirically. Episodes of excessive money growth led to a buildup of inflationary pressure, but the actual path of inflation was determined largely by policy decisions regarding the timing and magnitude of price liberalizations and exchange rate adjustments. The high rates of money growth in the mid-1990s are still the most plausible explanation for the magnitude of the inflationary spike following the last round of major price liberalization in 1997.

4. This chapter reviews recent developments in inflation in Romania, and analyzes its key determinants. Part B provides background on aggregate and sectoral price developments over the past decade, and the process of price liberalization. Part C reviews developments in several variables commonly identified in the literature as sources of price pressures, including wages, the exchange rate and monetary aggregates. Part D examines empirically the relationships among prices and wages, money and the exchange rate through the 1990s using vector autoregression (VAR) models. Part E concludes.

B. Price Developments

Price Outcomes

5. Inflation has been high and variable through the past decade. On a 12-month basis, inflation reached some 200-300 percent at the start of the transition. Inflation eased steadily between mid-1993 and mid-1995, reaching a low of 25 percent, but again accelerated from second half of 1995. The dramatic surge in early 1997 was associated with the liberalization of agricultural and energy prices;1 12-month inflation peaked just below 180 percent in mid-1997 and remained very high until 1998, but monthly rates fell swiftly after the initial surge.

6. Although down from its high levels following the last round of price liberalization in 1997, inflation has remained volatile. Inflation slowed in 1998, with the 12-month rate bottoming out at 33 percent in February 1999, in response to tight monetary policy and a slower rate of depreciation of the leu. However, against a background of a large fiscal deficit and continuing rapid wage growth, this slowdown came at the cost of a large real appreciation and severe loss of competitiveness. The subsequent large corrective depreciation contributed to a renewed pickup in inflation through 1999, and consumer prices increased by 57 percent in the year to January 2000.

7. Aggregate prices are particularly strongly influenced by food prices, which account for around half of the CPI basket. The prices of food, non-food goods, and services increased at broadly similar rates during the mid-1990s, but have grown at increasingly divergent paces over the past couple of years (Figure I.1b). Services prices rose especially rapidly in 1999, in part reflecting the effects of large increases in administered prices (see below). In 2000, drought has caused a sharp rise in food prices and frustrated progress in disinflation. Notwithstanding the high weight of food in the CPI basket, most other measures of inflation, such as producer prices and the household consumption and GDP deflators, have mostly moved broadly in line with the CPI (Table I.1). A notable exception is 1998, when consumer prices increased by 59 percent but producer prices rose by only 33 percent (Figure I.1a).2

Figure I.la.
Figure I.la.

Consumer and Producer Prices

Twelve-month-ended percentage change

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

* From 1998, refers only to production for the domestic market.
Figure I.lb.
Figure I.lb.

Consumer Prices

Twelve-month-ended percentage change

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

Table I.1.

Measures of Inflation, 1994-99

(Percentage change, period average)

article image
Source: National Commission for Statistics.

Price liberalization

8. Price liberalization in Romania was fitful and protracted, with the last major round of liberalization delayed until 1997. Moreover, estimates vary on the degree to which prices remained controlled during the mid-1990s; in many cases, consumer prices were liberalized but the raw and basic material prices were not; in others, “liberalized” prices were heavily distorted by subsidies, especially in the agricultural sector.

9. Initial progress in price liberalization seemed encouraging. Demekas and Khan (1991) reported that most prices were liberalized in three rounds, in November 1990, in April 1991, and in July 1991, after which the authorities claimed around 80 percent of consumer prices were market-determined. Price controls and subsidies for most other consumer goods were supposedly eliminated in 1993.

10. Nevertheless, formal and informal price controls persisted, or were reintroduced, during the mid-1990s; controls on food prices were especially pervasive. The OECD (1993) noted that prices on many consumer items, notably in state-owned retail stores, were still not market-determined, instead being subject to supervision based on strict mark-up limits. More formally, Government Decision 45/1994 declared a wide range of items as being of “national importance” and subject to review (and influence) by the Competition Office.3 The IMF (1997) reports that, in addition to an array of producer price controls, the government had maintained direct wholesale and retail price ceilings on a number of sensitive food items,4 which accounted for 28 percent in the total consumption basket. With energy, utility, transport and telecommunications prices also administered, this implied that nearly 40 percent of the consumer price basket was still controlled as at end-1996.

11. Most prices still subject to control were liberalized in early 1997. Agricultural prices were liberalized in February 1997; with the trade regime also substantially liberalized shortly afterwards, agricultural prices are now market-determined. In the industrial sector, most administered producer and retailer prices were liberalized by March 1997, although price controls were retained for a short list of goods supplied in monopoly markets, notably energy; these prices have subsequently been adjusted (in most cases by the Competition Office) in line with movements in the exchange rate and/or consumer price index.5 Fuel prices were deregulated in September 1998. Administered and regulated prices now account for about 14 percent of the CPI basket.6

12. Administered prices have grown sharply since 1997 (Table I.2). Large increases in electricity and heating prices reflect the phasing out of the dual pricing system, where low household prices were cross-subsidized by higher prices for other economic agents. Domestic thermal energy prices were increased sharply in 1999 to bring them closer to world prices.

Table I.2.

Developments in Administered Prices, 1997-99

Percentage change, end-period

article image
Source: National Commission for Statistics; and staff estimates.

C. Determinants of Inflation

13. Theoretical considerations suggest several variables to be included in an empirical study of inflation. Under certain stringent assumptions—notably, perfect wage and price flexibility, and a stable equilibrium real exchange rate—conventional economic theory suggests that money alone can explain inflation. However, relaxing each of these rather stringent assumptions introduces new potential explanators of inflation. Relaxing the assumptions of wage and price flexibility implies roles for unit labor costs and pricing policies respectively in explaining inflation; similarly, instability in the real exchange rate implies the exchange rate is also potentially relevant in explaining inflation.

14. Research on inflation in transition economies has highlighted both the role of traditional cost-push and demand-pull factors in generating and sustaining inflation, and also the effect of relative price adjustment in retarding disinflation.7 Each of these factors appears to have been present in Romania. Cost-push pressures have resulted from episodes of wage growth well in excess of productivity—owing fundamentally to a lack of financial discipline—as well as real depreciations of the leu. Demand-pull factors have included monetary accommodation of fiscal and/or quasi-fiscal deficits, again reflecting financial indiscipline and pervasive soft budget constraints at the enterprise level. Relative price adjustment has been especially protracted in Romania owing to the piecemeal and occasionally reversed process of price liberalization.

15. This section reviews the behavior of key variables associated with inflation in other transition economies: money and credit, wages and the exchange rate.

Money and credit

16. Figure I.2 shows the growth of M2, M2 including foreign currency deposits (M2X), and domestic credit. Growth in the monetary aggregates surged on a number of occasions in the 1990s when the National Bank of Romania (NBR) was forced to accommodate large fiscal and quasi-fiscal deficits, of loss-making state-owned enterprises in general and of the agriculture sector in particular. The most egregious case was the surge in money growth which peaked in late 1994, when the authorities launched a very large program of subsidized agricultural financing.8

Figure I.2.
Figure I.2.

Money, Credit, and Prices

Twelve-month-ended percentage change

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

* M2 plus foreign currency deposits.

17. Figure I.2 suggests that the relationship between money and prices was rather loose for much of the 1990s, though a positive correlation has become somewhat clearer over the past two years. Especially noteworthy is the low rate of inflation relative to money in the mid-1990s, and the much higher inflation rate in 1997. This would be consistent with the existence of a monetary overhang arising from the variety of price controls in effect through the mid-1990s, and which was apparently run down following the foreign exchange and price liberalizations in 1997.

18. Money growth may not correlate highly with inflation because controls on price and exchange rate movements had severed the links through which inflationary effects may pass. Episodes of excessive money growth over the past decade led to a buildup of inflationary pressure, but the actual path of inflation was determined largely by policy decisions regarding the timing and magnitude of price liberalizations and exchange rate adjustments.

19. The high inflation rates relative to money growth in the early 1990s imply significant demonetization of the Romanian economy since the start of the transition (Figure I.3). The increasing share of foreign currency deposits in the monetary aggregates, though significant in itself, has only modestly countered the overall trend of demonetization.

Figure I.3.
Figure I.3.

Demonetization

Percent of GDP

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

20. Some remonetization took place between 1994 and 1996, but with several undesirable features which rendered it unsustainable. The OECD (1998) observes that the remonetization process funneled subsidies to loss-making sectors of the economy, in particular the agriculture and energy sectors, using NBR credits, which undermined the attempts of the central bank attempts to reduce inflation.9 This policy was completely opposed to the authorities’ stated intention of allowing market forces to determine the sectoral allocation of credit.

Soft Budget Constraints and Inter-Enterprise Arrears

Soft budget constraints have taken a number of forms in Romania. One important form has been the soft credits extended to the agriculture sector, including NBR credits, especially in the mid-1990s. Another important manifestation of financial indiscipline, which has yet to be adequately addressed, has been inter-enterprise arrears. Inter-enterprise arrears were equivalent to 42 percent of GDP as at end-1999, and apart from a fall in 1997, have risen steadily each year from around 20 percent of GDP as at end-1994.

Some level of arrears might be inevitable in a demonetized economy such as Romania (Figure I.3), especially in the aftermath of demonetization in the early 1990s. Clearly there is a failure of financial intermediation in an economy in which credit is less than 20 percent of GDP, and the development of inter-enterprise credits is a natural response. Consequently, the question of when an inter-enterprise credit becomes an arrear is an important reporting issue, and the extent to which the arrears data include normal trade credits as well as overdue payments is not fully clear. Nevertheless the continuing increases in arrears since the mid-1990s reflect deeper financial indiscipline.

This problem has been particularly acute in the case of the major utilities – both in terms of the utilities’ tax arrears to the government, and the arrears to the utilities of other enterprises such as nationally owned loss-making companies, and local utilities. Despite the scale of the utilities’ losses and arrears, wages at the utilities remain some of the highest in Romania; in the December quarter of 1999, average wages at the three major utilities were 2.2 times the economy-wide average wage.

Soft budget constraints and weak corporate governance in the state sector have allowed episodes of faster wage growth, notably in 1995,1996, and 1998. Consequently, wages in state-owned utilities and many loss-making companies are among the highest in the country. In the December quarter of 1999, average wages at the ten largest regies autonomes exceeded the economy-wide average wage by more than 60 percent.10 In turn, this complicates the process of restructuring. Workers with high-wage, low-productivity jobs – such as those in the mining and energy sectors – clearly face high opportunity costs from moving to higher productivity but lower wage jobs, and have strong incentives to resist restructuring.

Wages

21. Wage growth contributes to inflation in several ways. In general, wage increases which exceed increases in productivity generate inflationary pressures; large wage increases in response to an initial inflation shock contribute to inflationary inertia and hence sustain inflationary pressures. Additional effects operate when budget constraints are soft. Sahay and Vegh (1996) note the role of wage bill increases in inducing monetary expansion via the expansion of credit to state enterprises and to the government.

22. Figure I.4 shows growth in nominal average wages, and smoothed growth in nominal unit labor costs in industry. The extremely close correlation between growth in unit labor costs and consumer prices is striking, especially before 1997.

Figure I.4.
Figure I.4.

Wages and Prices

Twelve-month-ended percentage change

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

* Three-month moving average.** Economy-wide from January 1994; industry to December 1993.

23. Figure I.5 highlights the volatility in real wages. Real wage growth was particularly strongly in 1995 and 1996, with high wage increases recorded in the state-owned régies autonomes (RAs) and commercial companies, despite the large losses recorded in these sectors. These wage increases were financed in part by a large accumulation of arrears - a non-monetary variant of the Sahay-Végh credit expansion (see Box I.2). Wages again grew strongly in 1998, led by increases in the budgetary sector and RAs. Although wage growth in 1998 would appear only modest if deflated by the CPI, U.S. dollar wages increased by over 30 percent, and wages deflated by the PPI rose by around 20 percent—against a background of a second year of deep recession.

Figure I.5.
Figure I.5.

Real Wages

Twelve-month-ended percentage change

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

24. In between periods of rapid growth, real wages fell significantly in 1994, even more sharply in 1997 (owing to the unexpectedly high inflation outcome), and more modestly in 1999-2000. Notably, each of these falls in real wages has presaged a period of significant progress in disinflation.

Wage Policies

Policy efforts to contain wages have encountered difficulties. In the early 1990s, the authorities experimented with punitive tax penalties on wage bills in excess of a reference level; non-wage remuneration tended to increase in response. Tax-based incomes policies were abandoned in 1995, and no official wage policy was in place in 1996 (Oprescu, 2000).

In conjunction with the 1997 Stand-By Arrangement with the IMF, the authorities agreed to limit the growth of the average wage for 1997 compared with the average wage for the fourth quarter of 1996 to 75 percent of consumer price inflation over the same period, with the policy applying to the state sector, including the budgetary sector, “régies autonomes” and national companies, and loss-making commercial companies retained by the State Ownership Fund. The policy began to weaken by August 1997, with the ceilings not observed in the budgetary sector or by the régies autonomes, and wage growth picked up in 1998.

Wage policy was tightened in 1999 as one of the conditions of the 1999 (and now extended) Stand-By Arrangement, with a degree of success. The authorities undertook to limit the increase in the wage bill for the state budget sector to 28 percent in nominal terms over the whole of 1999, implying a 9 percent real decline based on then-projected inflation. The target for end-December 1999 was breached by nearly 5 percent, in part as a result of the authorities’ decision to increase the defense/security sector wage bill by 80 percent. For the rest of the state sector, nominal wage bills in 1999 were restricted to four times their level in the December quarter of 1998. In the case of the RAs and national companies, this policy delivered a 20 percent reduction in these companies’ overall real wage bill.

Under the extended Stand-by Arrangement, the authorities undertook to limit the increase in the nominal wage bill of the state sector to 40 percent in 2000, implying an increase of 1 percent in real terms on the basis of the originally targeted rate of inflation. Within this overall target, the real wage bill for the budgetary sector was envisaged to rise by 12 percent and that for the rest of the state sector to decline by 10 percent. The authorities justified this differentiated treatment of employees within the state sector on equity as well as efficiency grounds, with wages in the utilities among the highest in the country, and despite cuts last year, still almost twice as high as those in the budgetary sector in early 2000. Recent slippages in wage policy imply that the state sector wage bill is now likely to increase by 58 percent in 2000.

Exchange rate

25. Figure I.6 shows the evolution of the leu / U.S. dollar exchange rate, a trade-weighted nominal exchange rate (70 percent deutsche mark/euro, 30 percent U.S. dollar), and consumer prices. Figure I.6 suggests that consumer prices are highly responsive to movements in the exchange rate, although with some lag.11 The large depreciation through late 1998 and early 1999 appears to have contributed strongly to the more gradual pickup in inflation through 1999.

Figure I.6.
Figure I.6.

Exchange Rate and Prices

Twelve-month-ended percentage change

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

* 70 percent DM / euro, 30 percent USD.

26. The erratic behavior of the exchange rate reflects a series of policy reversals, especially in the early to mid-1990s. The IMF (1997) reports that notwithstanding frequent commitments to a flexible exchange rate, the authorities repeatedly intervened to maintain the exchange rate at overvalued levels, effectively as a subsidy to energy-intensive enterprises, only to be forced into allowing periodic depreciations as reserves ran low. The foreign exchange market was subject to particularly severe distortions in 1996; in response to a sharp depreciation of the leu in early 1996, the authorities withdrew the licenses of all foreign exchange dealers except for four state-owned banks. Attempts to set the exchange rate by administrative means were abandoned following the 1996 election, and the exchange rate is now a managed float.

27. Figure I.7 highlights the considerable volatility of Romania’s real effective exchange rate.12 Coorey et al. (1998) note that a real appreciation can have different implications for inflation depending on the nominal exchange rate regime: real appreciation associated with a stable nominal exchange rate generally implies capital inflows, monetary expansion and higher inflation; but when the nominal exchange rate is flexible, real appreciation is generally associated with nominal appreciation—or in the case of Romania, slower nominal depreciation—implying downward pressure on inflation.

Figure I.7.
Figure I.7.

Measures of the Real Exchange Rate

January 1995 = 100

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

* From 1998, PP1 refers only to production for the domestic market.** Three-month centered moving average.

D. Empirical Analysis

28. While casual inspection yields considerable information on the relationships between inflation and its determinants, econometric analysis offers additional insights. The analysis uses vector autogression (VAR) techniques, which are appealing in this context because they require no a priori assumptions about the exogeneity of the policy and other variables— exogeneity is instead tested below—and they avoid problems of simultaneity bias, given the potential for contemporaneous relationships among the variables. The analysis uses the variables described in section C, as well as an activity variable to complete the system:

  • Consumer price index—CPI

  • Industrial production—IP (proxy for activity)

  • Unit labor costs in industry—ULC

  • Nominal exchange rate—NER (lei /U.S. dollar), NTWI (weighted average of lei against U.S. dollar and DM/euro)

  • Monetary aggregates—M2, M2X (including foreign currency deposits), and CRED (domestic credit)

All variables have been logged and run from January 1991 to March 2000.13

29. The first step in the analysis is to determine the order of integration of the variables, in order to avoid misspecifying the model. Results from a variety of unit root tests (Appendix I) indicate on balance that each of the variables is I(1), i.e. integrated of order 1.14

30. The non-stationarity of the data motivates the use of the multivariate Johansen procedure to detect the presence of long-run stationary (“cointegrating”) relationships among the non-stationary variables. An advantage of the Johansen procedure is that it also allows the researcher to investigate the speed of adjustment to long-run equilibrium, and so to test for (weak) exogeneity of the explanatory variables (if the speed of adjustment of a variable is not significantly different from zero, the variable is weakly exogenous).15 The procedure is briefly explained in Appendix II.

31. Tests for cointegration were performed using unrestricted 5-variable VARs, using the various measures of exchange rates and monetary aggregates, with four lags and eleven centered seasonal dummies. The test results shown in Appendix II provide evidence for the existence of one cointegrating relationship across a range of specifications.

32. Each VAR was estimated with the constraint of one cointegrating relationship to give estimates of the long-run relationships, which were then tested for significance, or “exclusion” from the long-run relationship. In most cases, the CPI and ULC were found to be most strongly significant; also, weak exogeneity was usually rejected for the CPI, implying that the CPI adjusts to shocks to the rest of the system. The exchange rate was found to be significant when a structural dummy in 1997 was included. The monetary aggregates (and unsurprisingly, activity) were not found to be significant; this is consistent with the interpretation that the linkages between prices and monetary aggregates have been weakened by periods of monetary overhang and price controls, as well as unstable money demand.

33. To narrow the focus on the relationships between the CPI, ULC and exchange rate, three-variable VARs were estimated. The results are sensitive to the inclusion of the dummy in 1997, but are intuitively plausible when the dummy is included. All three variables are significant and correctly signed; exogeneity cannot be rejected for the exchange rate and unit labor costs, but is rejected for the CPI. Note that the finding of long-run exogeneity for unit labor costs does not imply that wages do not react to inflation in the short run; but it does imply that in the long run, wages are determined by real instead of nominal factors. This model was reestimated holding the exchange rate and unit labor costs exogenous, to yield the following long-run vector:

LCPI = 0.156 LNTWI + 0.846 LULC

34. These parameters appear plausible, but should still be treated with some caution. Figure I.8 shows that the model does quite a reasonable job of explaining inflation (R2=0.672). However, tests on the residual properties and the structural stability of the model (Figure I.9) point to problems with the parameters of the model in early 1997. This is not very surprising given the extent of structural changes at this time, including the liberalization of prices and the exchange rate. Unfortunately, the sample period since 1997 is still very short, and does not yet lend itself to reliable modeling.

Figure I.8.
Figure I.8.

Romania: Estimates of CPI Determinants, 1991-99

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

Figure I.9.
Figure I.9.

Romania: Residual Properties of the CPI Estimate, 1997-99

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

35. Because of the susceptibility of the levels data to structural breaks, it is also useful to examine the differenced variables as well, to shed more light on the shorter-run dynamics.16 Results of Granger causality among the variables are reported in Table I.3.

Table I.3:

Granger Causality Tests

article image
article image
Standard F-tests; (*), (**), (***) indicate rejection of the null hypothesis at significance levels of 1, 5, and 10 percent, respectively. The null hypothesis is “no Granger causality”.

36. Several results are particularly noteworthy, and hold up across most lags. First, there is strong evidence that unit labor costs growth and depreciation of the exchange rate both Granger cause inflation; also, while lei-only M2 does not appear to Granger cause inflation, there is strong evidence that broader monetary aggregates including foreign currency deposits do Granger cause inflation. Second, there is strong evidence that inflation Granger causes unit labor costs, confirming the bidirectional causality between wages and prices at least in the short run. Finally, there is also some evidence that unit labor costs Granger cause M2 and M2X, consistent with the hypothesis that money growth has accommodated growth in wages and prices.

37. Variance decompositions and impulse response functions have been obtained from a 4-lag unrestricted VAR. The VAR is identified using the Choleski decomposition, which implies—unlike in the Johansen procedure—that the ordering of the variables can affect the results significantly.17 The ordering shown here assumes that movements in the exchange rate and unit labor costs feed into inflation, which is then accommodated by money.

38. Table I.4 shows that substantial proportions of the forecast error variation in inflation can be attributed to innovations in unit labor costs, and that the reverse is also true, as expected on the basis of the Granger causality tests. Money also explains a significant proportion of the forecast error variance in inflation. These results appear relatively robust to the ordering of the VAR. However, results for the exchange rate do appear sensitive to the VAR ordering; the result that much of the forecast error variance in M2X is explained by the exchange rate but not vice versa is reversed when the VAR is reordered.

Table I.4:

Variance Decompositions

article image

39. Figure I.10 shows impulse response functions (IRFs) to a one standard deviation structural shock, with bootstrapped standard error bands.18 Again, the clearest results appear to be for inflation and unit labor costs. Inflation responds to unit labor costs within one month, with the response persisting for over a year. Unit labor costs tend to respond sharply to innovations in inflation within about two months. Inflation also appears to respond to innovations in money with a lag of about two months, and to innovations in the exchange rate in about three months, though these responses decay more quickly than responses to unit labor costs.

Figure I.10.
Figure I.10.
Figure I.10.

Impulse Response Functions

One standard-deviation structural shock, ± 2 S.E.

Citation: IMF Staff Country Reports 2001, 016; 10.5089/9781451832716.002.A001

E. Conclusions

40. Inflation in Romania has reflected a number of causes, the most fundamental of which have been lax macroeconomic policies and widespread financial indiscipline. Inflation has been generated by monetary accommodation of large fiscal and quasi-fiscal deficits, and by rapid wage growth unsupported by productivity and financed by arrears; and it has been sustained by inflation inertia and protracted relative price adjustment.

41. The econometric evidence points to the role of unit labor costs in driving inflation. Reading through the statistical noise, unit labor costs emerge as the most plausible explanator of inflation in the long-run, at least in a proximate sense. It is clear that wages also respond to inflation in the shorter run, highlighting the role of inertia in sustaining inflation. But it is also true that in the long run, real wages are determined, whether efficiently or inefficiently, by real factors. The combination of wage levels well above those justified by productivity (even if those wages are low in absolute terms), especially in conjunction with widespread overstaffing—and more fundamentally, the failure to enforce financial discipline—has clearly done much to fuel inflation.

42. There is also evidence of the increasing importance of the exchange rate for inflation. Though somewhat sensitive to specification, the econometric evidence generally confirms the importance of the exchange rate for inflation—as does, perhaps even more persuasively, the pickup in inflation through 1999. This relationship should emerge even more clearly over the next few years, as more data becomes available for the period following the liberalizations of the exchange rate and of prices, and the Romanian economy continues to open up.

43. The correlation between money growth and inflation is less visible, owing to price controls and fluctuations in the real exchange rate. There is little statistical evidence of a stable relationship between money growth and inflation. However, a plausible reason for this is the monetary overhang present at the start of the 1990s, and which built up again in the mid-1990s ahead of the last round of price liberalization; another plausible reason is the fluctuation of the real exchange rate. With the major price distortions now eliminated, the increasing correlation between money growth and inflation over the past few years suggests that this traditional relationship is now be reasserting itself.

44. Romania’s experience highlights the need to unburden monetary policy as a prerequisite for a sustained reduction in inflation. Although inflation is ultimately a monetary phenomenon, pressures for accommodating fiscal and quasi-fiscal deficits proved irresistible on several occasions through the past decade. If Romania is to make the progress in disinflation envisaged in its medium-term strategy, monetary policy needs greater support from incomes and fiscal policies, and greater progress in enforcing financial discipline.

Appendix I: Unit Root Tests

Appendix II. The Johansen Procedure

This study uses the Johansen and Juselius (1990, 1992) procedure to test for cointegrating vectors in multivariate models.

The procedure is based on the following p-dimensional VAR with k lags:

Xt=A1Xt-1++AkXt-k+μ+ΨDt+εt,t=1,,T(AII.1)

where Xt is the p × 1 vector of variables of interest, Dt is an (optional) matrix of centered seasonal dummies and εt, is a vector of Gaussian i.i.d. errors. This levels model may be rewritten in error-correction form:

ΔXt=Γ1ΔXt-1++Γk-1ΔXt-k+1+ΠXt-k+μ+ΨDt+εt,t=1,,T(AII.2)

where the Γ and Π matrices are given by:

Γi=-[I-i=1k-1πi](AII.3)

and

Π=-[I-i=1kπi](AII.4)
A01app02

Under the hypothesis of cointegration, the p × p matrix Π contains information about the long-run relationships among the variables in Xt. This hypothesis will depend on the rank r of Π. There are three possibilities to consider:

  1. r may be full, i.e. the rank of Π is equal to p. In this case, all the variables of Xt are stationary, and a standard VAR should be estimated in levels. In general, this will not occur when one or more of the variables is I(1).

  2. r is zero, i.e. Π is a null matrix. In this case, no long-run relationships exist among the variables, and the VAR must be estimated in differences.

  3. r is between 0 and p, and represents the number of cointegrating vectors among the variables in Xt. This implies that there exist p × r matrices α and β such that Π = αβ’, where β is a matrix of r cointegrating vectors, and α is a matrix of adjustment coefficients. Even though the elements of Xt may be non-stationary, the cointegrating vectors represent linear combinations of these elements which are stationary, i.e. β’Xt is stationary.

Johansen and Juselius have devised two likelihood ratio tests to determine the rank of Π. The first test is the trace test, which tests the hypothesis that rp against the general alternative of stationarity. The trace statistic is given by:

Trace=-TΣi-r+1pln(1-λ^i)(AII.5)

The second test is the maximal eigenvalue test, which tests the null hypothesis of at most r cointegrating vectors against the alternative of r + 1 cointegrating vectors. The maximal eigenvalue test statistic is given by:

λmax=-Tln(1-λ^r+1)(AII.6)

Critical values for these tests have been generated by Osterwald-Lenum (1992); updated critical values for the trace tests appear in Hansen and Juselius (1995).

If cointegration is found and the cointegrating vectors β are estimated, the significance of the coefficients of β may then be tested by additional likelihood ratio tests. Not all of the variables in the model need be important in the long-run relationships, and so each variable may be tested for exclusion. The test takes the form of the restriction βi = 0 for the ith variable. The test statistic is defined as:

LR=Σi=1rln(1-λ˜i)/(1-λ^i)(AII.7)

where r is the number of cointegrating vectors, and λ˜i

article image
and λ^i
article image
are the eigenvalues from the restricted and unrestricted estimates of β respectively. The test statistic is distributed as χ2 with the degrees of freedom equal to the number of restrictions.

The α matrix contains the “adjustment” vectors, which describe the speed with which the dependent variables adjust to long-run equilibrium. The variables of the model may then be tested for exogeneity, by testing the restriction αj = 0 for the jth variable, in a test procedure similar to the exclusion test. If this restriction cannot be rejected for a given variable, then that variable will be weakly exogenous to the long-run relationship, since it will not adjust to shocks to other variables.

Tests for cointegration were performed using unrestricted 5-variable VARs, using the various measures of exchange rates and monetary aggregates, with four lags and eleven centered seasonal dummies. The tests were performed with and without a trend in the cointegrating space. Tests were also performed including a structural break in February 1997.19 The maximum eigenvalue statistic and the trace statistic provide evidence for the existence of one cointegrating relationship across a range of specifications (Tables AII.1 and AII.2).

Table AII.1.

Tests for Cointegration, Unrestricted Model

article image
(*), (**), (***) indicate rejection of the null hypothesis at significance levels of 1, 5, and 10 percent, respectively.
Table AII.2

Tests for Cointegration, Unrestricted Model, Break in February 1997

article image
(*) (**) (***) indicate rejection of the null hypothesis at significance levels of 1, 5, and 10 percent, respectively.

Each VAR was estimated with the constraint of one cointegrating relationship to give estimates of the long-run relationships (the coefficients of the “β vector”), which were then tested for significance, or “exclusion” from the long-run relationship (Tables AII.3 and AII.4). In most cases, exclusion hypotheses were generally rejected most strongly for the CPI and ULC; also, weak exogeneity was usually rejected for the CPI. In general, exclusion could not be rejected for the exchange rate in the absence of a structural dummy in 1997, but was rejected when the dummy was included. Exclusion could not be systematically rejected for the monetary aggregates or activity.

Table AII.3

Estimates and Tests of the Long-Run Relationships, Unrestricted Model

article image
article image
Table AII.4

Estimates and Tests of the Long-Run Relationships, Unrestricted Model, Break in February 1997

article image
article image

To narrow the focus on the relationships between the CPI, ULC and exchange rate, three-variable VARs were estimated:

article image
(*) and (**) indicate rejection of the null hypothesis at significance levels of 1 and 5 percent, respectively.

The results are sensitive to the inclusion of the dummy. Without the dummy (model A), exclusion cannot be rejected for the exchange rate, and exogeneity cannot be rejected for the CPI—in other words, the CPI helps explain unit labor costs but is not itself explained.

However, including the dummy (model B) yields intuitively plausible results. Rejection can be excluded for all three variables, which are correctly signed; exogeneity cannot be rejected for the exchange rate and unit labor costs, but is rejected for the CPI. Note that long-run exogeneity for unit labor costs does not imply that wages do not react to inflation in the short run; but it does imply that in the long run, wages are determined by real instead of nominal factors.

This model was reestimated holding the exchange rate and unit labor costs exogenous, to yield the following long-run vector:

LCPI = 0.156 LNTWI + 0.846 LULC

These parameters appear plausible, and Figure I.8 shows that the model does quite a reasonable job of explaining inflation (R2=0.672). However, there is some evidence of nonnormal residuals, as well as of first-order autocorrelation. Moreover, recursively estimated one-step ahead prediction tests and tests for constancy of the long-run parameters (Figure I.9) point to problems with the parameters of the model in early 1997, probably reflecting the price liberalization episode in early 1997.

References

  • Coorey, Sharmini, Mauro Mecagni, and Erik Offerdal, 1998, “Disinflation in Transition Economies: The Role of Relative Price Adjustment”, in Moderate Inflation: The Experience of Transition Economies, ed. by Carlo Cottarelli GyOrgy Szapary (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Demekas, Dimitri G. and Mohsin S. Khan, 1991, The Romanian Economic Reform Program, IMF Occasional Paper No. 89 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Ericsson, Neil R., 1992, “Cointegration, Exogeneity and Policy Analysis: An Overview”, Journal of Policy Modeling, Vol. 14, No. 3, pp. 251-80.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hansen, Henrik and Katarina Juselius, 1995, CATS in RATS: Cointegration Analysis of Time Series (Estima).

  • International Monetary Fund, 1997, Romania—Recent Economic Developments, IMF Staff Country Report No. 97/46, June.

  • Johansen, Søren and Katarina Juselius, 1990, “Maximum Likelihood Estimation and Inference on Cointegration - with Applications to the Demand for Money”, Oxford Bulletin of Economics and Statistics, Vol. 52, No. 2, pp. 169-210.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Johansen, Søren and Katarina Juselius, 1992, “Testing Structural Hypotheses in a Multivariate Cointegration Analysis of the PPP and UIP for the UK”, Journal of Econometrics, Vol. 53, pp. 211-44.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Oprescu, Gheorghe, 2000, “The Labor Market in Romania”, in Economic Transition in Romania: Past, Present and Future, ed. by Christof Rühl and Daniel Diani (Bucharest: World Bank and Romanian Center for Economic Policies).

    • Search Google Scholar
    • Export Citation
  • Organization for Economic Co-operation and Development, 1993, Romania: An Economic Assessment (Paris: OECD).

  • Organization for Economic Co-operation and Development, 1998, OECD Economic Surveys, 1997-1998: Romania (Paris: OECD).

  • Organization for Economic Co-operation and Development, 2000, Labor Market and Social Policies in Romania (Paris: OECD).

  • Osterwald-Lenum, Michael, 1992, “The Asymptotic Distribution of the Maximum Likelihood Cointegration Rank Test Statistics”, Oxford Bulletin of Economics and Statistics, Vol. 54, No. 3, pp. 461-71.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ross, Kevin, 1998, “Post Stabilization Inflation Dynamics in Slovenia”, IMF Working Paper 98/27 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Sahay, Ratna and Carlos A. Végh, 1996, “Inflation and Stabilization in Transition Economies: an Analytical Interpretation of the Evidence”, Journal of Policy Reform, Vol. 1, No. 1, pp. 75-108.

    • Crossref