NAIRU Estimation in Romania (including a comparison with other transition countries) Student: Otilia Iulia Ciotau Supervisor: Professor Moisa Altar THE ACADEMY OF ECONOMIC STUDIES DOCTORAL SCHOOL OF FINANCE AND BANKING BUCHAREST,2004

Doctoral School of Finance and Banking – June, 2004 Contents  The paper’s incentives  Features of unemployment rate in Romania  Estimation methods  Comparison of results  Concluding remarks

Doctoral School of Finance and Banking – June, 2004 Natural Rate and NAIRU Is there any difference?  Natural rate of unemployment - Friedman (1968), Phelps (1968 ): the level of unemployment to which the economy would converge in the long run in the absence of structural changes to the labor market;  NAIRU (Non-Accelerating Inflation Rate of Unemployment) - Modigliani and Papademos (1975): the rate of unemployment at which there is no tendency for inflation to increase or decrease

Doctoral School of Finance and Banking – June, 2004 Are NAIRU estimates useful?  “I have become convinced that the NAIRU is a useful analytic concept. It is useful as a theory to understand the causes of inflation. It is useful as an empirical basis for predicting changes in the inflation rate. And, it is useful as a general guideline for thinking about macroeconomic policy.” Stiglitz, J., Reflections on the Natural Rate Hypothesis

Doctoral School of Finance and Banking – June, 2004 Features of Unemployment Rate in Romania  The labor market have been strongly affected by the adjustment process from centrally planned to market- oriented economies;  Mass lay-offs;  Issues about underestimation of unemployment rate (masked unemployment, methodology);  Labor force working in informal economy;  Active measures for unemployment mitigation (Law no76/2002).

Doctoral School of Finance and Banking – June, 2004 Unemployment Rate in Romania (1994:1 – 2004:1)

Doctoral School of Finance and Banking – June, 2004 Estimation methods  Statistical methods Hodrick-Prescott Filter Univariate UC Bivariate UC (Okun’s approach) Multivariate UC  Reduced-form methods Phillips curve with constant NAIRU Elmeskov method Phillips curve with time-varying NAIRU

Doctoral School of Finance and Banking – June, 2004 Hodrick-Prescott ( =1600)

Doctoral School of Finance and Banking – June, 2004 Univariate UC for Romania  Fitted model: - is generated by the stochastic process: k t and k t * are uncorrelated w.n. with the same variance. - and its reduced form is a restricted ARMA(2,1):

Doctoral School of Finance and Banking – June, 2004 Seasonal component and intervention variable  The seasonal pattern is the sum of [s/2] (two for quarterly data) cyclical components, with frequencies : - same variance to each harmonic. is a pulse intervention variable:

Doctoral School of Finance and Banking – June, 2004 The maximum likelihood estimates are: 95% confidence interval for NAIRU: (2003:2) % (2003:3) % (2003:4) % (2004:1) % Back

Doctoral School of Finance and Banking – June, 2004 Estimated parameters for the cycle: Period: ( 'years') Amplitude: Rho: Variance:

Doctoral School of Finance and Banking – June, 2004 Unemployment Rate Forecast (2004:2) % (2004:3) % (2004:4) % (2005:1) % 95% confidence interval for unemployment rate forecast:

Doctoral School of Finance and Banking – June, 2004 Univariate UC for Czech R. and Lithuania  Fitted model:  Intervention variables: Irr & Irr for Czech R.

Doctoral School of Finance and Banking – June, 2004 NAIRU (UC-1 trend) Czech R.

Doctoral School of Finance and Banking – June, 2004 UC-1 slope for Czech R.

Doctoral School of Finance and Banking – June, 2004 Unemployment gap Czech R.

Doctoral School of Finance and Banking – June, 2004 Bivariate UC: unemployment rate and real GDP ( 1994:1-2003:3)  Okun’s law  SUTSE ( Seemingly Unrelated Time Series Equations ): Intervention variable: For unemployment series: irr 2002:1; For GDP: level 1997:1.

Doctoral School of Finance and Banking – June, 2004 Common cycles Estimated parameters for the cycle: Period: ( 'years'); Amplitude unemployment gap : ; Amplitude GDPgap: ; Rho:

Doctoral School of Finance and Banking – June, 2004 NAIRU (trend UC-2) and unemployment gap (cycle UC-2) 95% confidence interval for NAIRU: (2003:1) % (2003:2) % (2003:3) % UC-1 NAIRU

Doctoral School of Finance and Banking – June, 2004 Potential Output (trend UC-2) and Output Gap

Doctoral School of Finance and Banking – June, 2004 Unemployment Rates in Transition Economies

Doctoral School of Finance and Banking – June, 2004 Multivariate framework  SUTSE model for six countries: Series are linked via the off diagonal elements in and ; This approach allows for detection of common features (Engle and Kozicki 1993): like trend, cycle, seasonal. Estimated parameters for the similar cycle: Rho = 0.96 Period = ( ‘years’)

Doctoral School of Finance and Banking – June, 2004 Correlation between cyclical components  Czech R.  Hungary0.983  Lithuania  Polonia  Slovakia  Romania

Doctoral School of Finance and Banking – June, 2004 Short-run commovements between unemployment rate in Czech R. and Hungary

Doctoral School of Finance and Banking – June, 2004 Correlation between seasonal components  Czech R.  Hungary  Lithuania  Polonia  Slovakia  Romania

Doctoral School of Finance and Banking – June, 2004 Seasonal comovements between unemployment rate in Poland and Romania

Doctoral School of Finance and Banking – June, 2004 Seasonal components in unemployment rate: Romania

Doctoral School of Finance and Banking – June, 2004 Seasonal components in unemployment rate: Poland

Doctoral School of Finance and Banking – June, 2004 Seasonal comovements between unemployment rate in Hungary and Slovakia

Doctoral School of Finance and Banking – June, 2004 Seasonal components in unemployment rate: Hungary

Doctoral School of Finance and Banking – June, 2004 Seasonal components in unemployment rate: Slovakia

Doctoral School of Finance and Banking – June, 2004 NAIRU (UC-2 trend) and unemployment gap in Romania Amplitude:

Doctoral School of Finance and Banking – June, 2004 NAIRU (UC-2 trend) and unemployment gap in Czech R. Amplitude:

Doctoral School of Finance and Banking – June, 2004 NAIRU (UC-2 trend) and unemployment gap in Lithuania Amplitude:

Doctoral School of Finance and Banking – June, 2004 NAIRU (UC-2 trend) and unemployment gap in Poland Amplitude:

Doctoral School of Finance and Banking – June, 2004 NAIRU (UC-2 trend) and unemployment gap in Slovakia Amplitude:

Doctoral School of Finance and Banking – June, 2004 NAIRU (UC-2 trend) and unemployment gap in Hungary Amplitude:

Doctoral School of Finance and Banking – June, 2004 Testing for hysteresis  ADF, PP: we cannot reject the unit root hypothesis for any of the six series;  Zivot and Andrews (1992) : unit root test with structural break endogenously determined (prg. EViews)

Doctoral School of Finance and Banking – June, 2004 Zivot, Andrews test results Country AIC Model A AIC Model B AIC Model C AIC Model D Best model Estimated for best model Tmin Unit root test outcome Czech R C Not significant at 10% Hungary B Significant at 5% Poland B Not significant at 10% Slovakia D Significant at 1% Lithuania C Not significant at 10% Romania C Significant at 1%

Doctoral School of Finance and Banking – June, 2004 Breakpoints endogenously determined by ZA test CountryBreakpointSignificance Czech Republic 1998 q2 Not significant at 10% Hungary 2001 q2 Significant at 5% Poland 1998 q2 Not significant at 10% Slovakia 1998 q4 Significant at 1% Lithuania 2003 q3 Not significant at 10% Romania 2001 q4 Significant at 1%

Doctoral School of Finance and Banking – June, 2004 Reduced-form methods  “Triangle model of inflation” (Gordon) where Estimation of a constant NAIRU requires the introduction of a constant in (1): For a time-varying NAIRU we use (1) as the measurement equation for a state space representation estimated with Kalman filter.

Doctoral School of Finance and Banking – June, 2004 Constant NAIRU (u* = 6.98%) Method: Least Squares Sample: 1994:1 2004:1 VariableCoefficientStd. Errort-StatisticProb. C DINF(-2) SOM(-1) DSOM CFE OILM(-1) REER R-squared Adjusted R-squared

Doctoral School of Finance and Banking – June, 2004 Elmeskov Method  simplified „accelerationist” version of Phillips curve:  An estimate of is obtained for any two consecutive periods as which is substituted in (1) to give:

Doctoral School of Finance and Banking – June, 2004 Elmeskov Method

Doctoral School of Finance and Banking – June, 2004 Time-varying NAIRU  The basic inflation equation:  is supplemented by a second equation that explicitly allows the NAIRU to vary with time:  The method of estimation is Kalman filter with a standard deviation of 0.2 for the state variable as a “smoothing prior” (Gordon 1997).

Doctoral School of Finance and Banking – June, 2004 Time-varying NAIRU

Doctoral School of Finance and Banking – June, 2004 Comparison of results HPUnivar.UCBivar.UCMultivar.UCRecursiveElmeskovKalman1Kalman

Doctoral School of Finance and Banking – June, 2004 Conclusion  The Romanian NAIRU is lower than in the other countries studied and also rather small comparing to Europe;  NAIRU in Romania is smooth comparing to the other five countries;  Uncertainty of the results

Doctoral School of Finance and Banking – June, 2004 Further direction for research  Estimating NAIRU based on unemployment rate calculated according to international accepted standard  Using the series from claimant count just for improving the accuracy in a bivariate UC model; Harvey and Chung(2000), Estimating the underlying change in unemplyment in the Uk

Doctoral School of Finance and Banking – June, 2004 Predictive-testing (Romania UC-1)

Doctoral School of Finance and Banking – June, 2004 Predictive-testing (Romania UC-1)

Doctoral School of Finance and Banking – June, 2004 Auxiliary observation residuals (Romania UC-1)

Doctoral School of Finance and Banking – June, 2004 CUSUM test UC-1

Doctoral School of Finance and Banking – June, 2004 Bivariate UC – better fit for unemployment than in univariate case

Doctoral School of Finance and Banking – June, 2004 Predictive testing: bivariate GDP

Doctoral School of Finance and Banking – June, 2004 Forecast for GDP and unemployment rate

Doctoral School of Finance and Banking – June, 2004 Predictive testing for multivariate UC-1

Doctoral School of Finance and Banking – June, 2004 Forecast multivariate UC