Academy of Economic Studies Bucharest Doctoral School of Finance and Banking DISSERTATION PAPER OUTPUT AND UNEMPLOYMENT DYNAMICS IN ROMANIA Student: SINCA FLORIN EUGEN Supervisor: Professor MOISA ALTAR Presentation content 1. Introduction 2. The importance of output and unemployment dynamics in Romania 3. Theoretical connection between output and unemployment 4. Unit root tests 5. Estimation of ARIMA (p,1,q) models for output 6. Estimation of ARIMA (p,1,q) models for unemployment rate 7. Bivariate analysis Granger causality tests 8. A VAR analysis of the joint evolution of output and unemployment 9. Estimating Okuns coefficient using dynamic OLS 10. Conclusions 2 1. Introduction Output and unemployment dynamics is a subject of intense macroeconomic importance. It has been analyzed both in univariate in bivariate models: Campbell and Mankiw (1987) Blanchard and Quah (1989) Evans (1989) Weber (1995) Leon-Ledesma and McAdam (2003) 3 2. The importance of output and unemployment dynamics in Romania as a present candidate and future member of the European Union Romania must undertake labour market reforms unemployment rate (7.2 % in 2003) is currently lower than EU average, but long unemployment duration (24.1 months during the

third quarter of 2003) is a major problem and creates the image of a stagnant pool for Romanian unemployment Romania has to pursue high and constant economic growth rates and in this context shocks may affect both unemployment and output it is important to maintain an economic growth of 5 % in 2004 and 2005 without deteriorating labour market equilibriums 4 3. Theoretical connection between output and unemployment Following Blanchard and Fisher (1989), the relation between output and labour is given by the production function: Y sF L F' 0 (1) F" 0 (2) equation (1) gives the straight relation between output and labour and not between output and unemployment labour force is represented by the total number of employees in the economy and in some cases the connection between the number of employees and the unemployment rate is weak if we consider the existence of a negative relation between the total number of employees and the unemployment rate, then there is a negative relation between output and unemployment rate 5 This negative relation between cyclical fluctuations in output and the level or the change of unemployment rate is known as Okuns Law Okuns Law says that an increase of 3 % in output over its normal growth rate over a year leads to an increase of 2 % in employment and a decrease of 1 % in the unemployment rate 4. Unit root tests 160

13 150 12 140 11 130 10 120 9 110 8 7 100 6 90 93 94 95 96 97 98 99 00 01 02

03 93 94 95 96 output 97 98 99 00 01 02 03 unemployment rate The variables of this econometric analysis are: industrial production index, denoted as output (IPI, monthly observations 1993:01 2003:12, seasonally adjusted) unemployment rate (UNR, monthly observations 1993:01 2003:12, seasonally adjusted) 7 p (3) yt a0 yt 1 i yt i t i 1 RESULTS OF THE ADF TEST Macroeconomic variable p

a0 +1 Conclusions Industrial production index (IPI), full sample 3 6.550 (1.696) -0.053 (-1.694) 0.947 I (1) Unemployment rate (UNR), full sample 5 0.419 (1.985) -0.045 (-2.067) 0.955 I (1) Unemployment rate (UNR), 1993:01 2001:12 12 0.552 (3.308) -0.058

(-3.368) 0.942 I (0) RESULTS OF THE KPSS TEST Macroeconomic variable KPSS statistic Conclusions Industrial production index (IPI), full sample 0.237 I (0) Unemployment rate (UNR), full sample 0.113 I (0) Unemployment rate (UNR), 1993:01 2001:12 0.150 I (0) 8 5. Estimation of ARIMA (p,1,q) models for output L IPI t L t (4) 1 IPI t L L t A L t 1 1

1 IPI 1 L L L t 1 L A L t B L t (5) (6) i Bi A j (7) j 0 The long-run impulse response represents the response of IPI t+i to an innovation at time t, for large i and is given by: (8) A1 9 The estimated ARMA (2,2) model for the first difference of output, having standard errors in parentheses, is: IPI t 0.636IPI t 1 0.760IPI t 2 t 0.341 t 1 0.705 t 2 0.127 0.099 0.136 (9) 0.107 Inverted AR roots are 0.32 0.81i Inverted MA roots are 0.17 0.82i Akaike information criterion is 5.809 Schwarz criterion is 5.898 Adjusted R-squared is 0.163 Ljung-Box Q-statistics are: Q(8)=1.938 (0.747) ; Q(16)=18.310 (0.107) ; Q(34)= 40.836 (0.090) Breusch-Godfrey Serial Correlation LM Test for 10 lags is 11.640 (0.309) The long-run impulse response is 0.854 implying that an

innovation of 1 percent in output leads to a revision of the forecast in the long-run by an amount less than 1 percent. 10 6. Estimation of ARIMA (p,1,q) models for unemployment rate L UNRt c0 xt L t (10) xt is a dummy variable that takes value 1 in 2002:01 and 0 in all other periods The estimated model for the first difference of unemployment rate, having standard errors in parentheses, is AR (1): UNRt 2.685 xt 0.550UNRt 1 (11) Inverted AR 0root 0.074 .204 is 0.55 Akaike information criterion is 0.059 and Schwarz criterion is 0.015 Adjusted R-squared is 0.620 Ljung-Box Q -statistics are: Q(8)=2.786 (0.904) ; Q(16)=15.801 (0.395) ; Q(34)= 30.553 (0.590) Breusch-Godfrey Serial Correlation LM Test for 10 lags is 5.226 (0.875) 11 The sharp increase of the unemployment rate at the beginning of 2002 had only a temporary effect unemployment increase in January 2002 and the subsequent effects according to the estimated AR (1) model Jan Feb Mar Apr May June July

Aug Sept Oct Nov Dec 2.685 1.476 0.812 0.446 0.245 0.135 0.074 0.040 0.022 0.012 0.006 0.003 12 7. Bivariate analysis Granger causality tests Granger causality tests between output growth and unemployment growth, full sample 1993:01 2003:12 Null hypotheses Lags number 1 2

3 8 12 F-stat Prob F-stat Prob F-stat Prob F-stat Prob F-stat Prob UNR does UNR does not Granger cause UNR does IPI 0.124 0.725 1.813 0.167 1.106 0.349 0.678 0.709 0.479

0.922 UNR does IPI does not Granger cause UNR does UNR 0.889 0.347 0.662 0.517 1.204 0.311 1.525 0.157 1.115 0.356 No causality between output and unemployment for the full sample No causality between unemployment and output for the full sample 13 Granger causality tests between output growth and unemployment growth, shorter sample 1993:01 2001:12 Null hypotheses Lags number 1 2 3 8 12

F-stat Prob F-stat Prob F-stat Prob F-stat Prob F-stat Prob UNR does UNR does not Granger cause UNR does IPI 1.982 0.162 2.960 0.056 1.912 0.132 1.117 0.360 1.492 0.147 UNR does IPI does not Granger cause UNR does UNR

0.210 0.647 0.200 0.818 0.406 0.748 1.164 0.330 1.466 0.158 There is a weak causality between unemployment and output with two lags. 14 8. A VAR analysis of the joint evolution of output and unemployment Impulse response functions Response of output growth to output growth Response of output growth to unemployment growth 5 1.0 4 0.5 3 0.0 2 1

-0.5 0 -1.0 -1 -1.5 -2 -3 -2.0 1 2 3 4 5 6 7 8 9 10 11 12 1 Response of unemployment growth to output growth 3 4 5

6 7 8 9 10 11 12 Response of unemployment growth to unemployment growth .10 .4 .05 .3 .00 .2 -.05 .1 -.10 .0 -.15 -.1 -.20 2 -.2 1

2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 15

Variance decomposition for output growth Period Output growth Unemployment growth 1 100.000 0.000 2 99.995 0.005 4 97.648 2.352 6 97.611 2.389 8 97.597 2.403 10 97.595 2.405 12 97.594 2.406

14 97.593 2.407 Variance decomposition for unemployment growth Period Output growth Unemployment growth 1 0.106 99.894 2 2.537 97.463 4 4.150 95.850 6 4.227 95.773 8 4.228 95.772 10 4.230

95.770 12 4.230 95.770 14 4.231 95.769 16 9. Estimating Okuns coefficient using dynamic OLS Output gap: y tc y t y tn (12) Cyclical unemployment rate: U tc U t U tn (13) An autoregressive-distributed lag model is estimated for the cyclical unemployment rate: U c t k iU c t i i 1 k

i y tc i t (14) i 1 The impact of a change in output gap on cyclical unemployment rate in the long-run is given by: k a LR i (15) i 1 1 k i 1 i 17 Potential output and natural unemployment rate are determined by Hodrick Prescott filter, using the smoothing parameters: 14,400 ; 10,000 and 40,000. Okuns coefficient is estimated for the full sample 1993:01 2003:12 and also for the shorter sample 1993:01 2001:12. There are no important differences between these estimates and the value of Okuns coefficient lies between 0.1239 and 0.0943. 18 10. Conclusions considering the results of the unit root tests and the estimated

AR (1) model, the hysteresis hypothesis is rejected for the unemployment rate an output innovation of 1 percent leads to a revision of output forecast in the long-run of 0.854 percents there is a weak Granger causality between unemployment growth and output growth with a lag of two months the results of the estimated VAR illustrate a negative relation between output and unemployment attributed to Okuns law according to VAR results, output presents a higher degree of persistence than unemployment; after ten months 1 % of the output initial innovation is still present in output, but only 0.62 % of the unemployment initial innovation is still present in unemployment variance decomposition for the estimated VAR shows that output explains a larger part of unemployment variance, while unemployment explains only a small part of output variation19 References Apel, Mikael; Hansen, Jan and Hans Lindberg (1996) Potential Output and Output Gap ; Sveriges Riksbank Economic Review 3, 24-35. Blanchard, Olivier and Lawrence Summers (1986) Fiscal Increasing Returns, Hysteresis, Real Wages and Unemployment ; NBER Working Paper No 2034. Blanchard, Olivier and Stanley Fisher (1989) Lectures on Macroeconomics ; MIT Press, Cambridge. Blanchard, Olivier and Danny Quah (1989) The Dynamic Effects of Aggregate Demand and Supply Disturbances; The American Economic Review 79, 4 (Sep), 655-673. Blanchard, Olivier (1991) Wage Bargaining and Unemployment Persistence ; Journal of Money, Credit and Banking 23, 3 (Aug), 277-292. Blanchard, Olivier and Pedro Portugal (2000) What Hides Behind an Unemployment Rate: Comparing Portuguese and U.S. Labour Markets. Blanchard, Olivier (2004) Explaining European Unemployment. Brunello, Giorgio (1990) Hysteresis and the Japanese Unemployment Problem: A Preliminary Investigation ; Oxford Economic Papers 42, 3 (Jul), 483-500. 20 Campbell, John Y. and N. Gregory Mankiw (1987a) Are Output Fluctuations Transitory ? ; Quarterly Journal of Economics 102, 4 (Nov), 857-880. Campbell, John Y. and N. Gregory Mankiw (1987b) Permanent and Transitory Components in Macroeconomic Fluctuations ; The American Economic Review (May), 111-117. Campbell, John Y. and Pierre Perron (1990) Pitfalls and Opportunities: What Macroeconomists Should Know About Unit Roots ; NBER Technical Working Paper No 100. Cerra, Valerie and Sweta Chaman Saxena (2000) Alternative Methods of Estimating Potential

Output and the Output Gap: An Application to Sweden ; IMF Working Paper. Cheung, Yin-Wong and Menzie David Chinn (1996) Deterministic, Stochastic, and Segmented Trends in Aggregate Output: A Cross-Country Analysis ; Oxford Economic Papers 48, 1 (Jan), 134162. Coe, David T. and John McDermott (1995) Does the Gap Model Work in Asia ? ; IMF Staff Papers 44, 1 (March), 59-80. Dowrick, Steve (2003) Ideas and Education: Level or Growth Effects ? ; NBER Working Paper No 9709. Enders, Walter (1995) Applied Econometric Time Series ; John Wiley & Sons, Inc. Evans, George W. (1989) Output and Unemployment Dynamics in the United States: 1950-1985 ; Journal of Applied Econometrics 4, 3 (Jul.-Sep), 213-237. European Commission (2003) 2003 Regular Report on Romanias Progress Towards Accession. Friedman, Benjamin M. and Michael L. Wachter (1974) Unemployment: Okuns Law, Labour Force, and Productivity ; Review of Economics and Statistics 56, 2 (May), 167-176. 21 Gordon, Robert J. and Peter K. Clark (1984) Unemployment and Potential Output in the 1980s ; Brookings Papers on Economic Activity 2, 537 568. Gordon, Robert J. (1995) Is There a Tradeoff Between Unemployment and Productivity Growth ? ; NBER Working Paper No 5081. Hamilton, James D. (1994) Time Series Analysis ; Princeton University Press, Princeton, New Jersey. Leon-Ledesma, Miguel and Peter McAdam (2003) Unemployment, Hysteresis and Transition ; European Central Bank Working Paper No 234. Murphy, Kevin M. and Robert Topel (1997) Unemployment and Nonemployment; The American Economic Review 87, 2, (May), 295-300. Pindyck, Robert S. and Daniel L. Rubinfeld (1998) Econometric Models and Economic Forecasts ; Irwin / McGraw-Hill. Proietti, Tommaso; Musso, Alberto and Thomas Westermann (2002) Estimating Potential Output and the Output Gap for the Euro Area: a Model-Based Production Function Approach ; European University Institute Working Paper ECO No 2002/9. Rudebusch, Glenn D. (1993) The Uncertain Unit Root in Real GDP ; The American Economic Review 83, 1 (March), 264-272. Strazicich, Mark C.; Tieslau, Margie and Junsoo Lee (2001) Hysteresis in Unemployment ? Evidence from Panel Unit Root Tests with Structural Change. Weber, Christian E. (1995) Cyclical Output, Cyclical Unemployment, and Okuns Coefficient: A New Approach ; Journal of Applied Econometrics 10, 4 (Oct-Dec), 433-445. 22