TESTING OF THE OKUN S LAW IN ROMANIA

Professor Marin DINU, PhD Leturer Marius-Corneliu MAINAŞ, PhD Email: marinasmarius@yahoo.fr Assoiate Professor Cristian SOCOL, PhD Leturer Aura-Gabriela SOCOL, PhD TESTING OF THE OKUN S LAW IN OMANIA Abstrat. The objetive of this study is to estimate the Okun s oeffiient, and to hek the validity of Okun s law in omania during the period 1999-28. For this purpose, we used data series of the unemployment rate, alulated aording to Eurostat methodology and industrial prodution index, used as a proxy for the hange in real GDP. Firstly, we have presented theoretial arguments of the asymmetri relationship between unemployment and eonomi growth. Seondly, we tested the hypothesis of unemployment hysteresis in omania. Finally, we estimated Okun`s oeffiient based on the ADL (autoregressive distributed lag) model, whih we introdued two dummy variables to apture the impat of hysteresis in unemployment. The estimated value was -.26, a value greater than previous estimates for the period 1992-24 (-.17), but lose to the oeffiient estimated by Okun (-.3). Keywords: hysteresis in unemployment; Okun law; output-gap; stationarity; ADL model. JEL Classifiation: E24, E27, J8 Aording to Okun's law, a redution of domesti prodution below its potential auses an inrease in unemployment over the natural rate of unemployment, its hanging being less than the prodution. Due to speifi rigidities of the labour market and good market (suh as menu osts, effiieny wages, minimum wages) pries and wages do not respond simultaneously to hange of aggregate demand. Consequently, an inrease of demand will indue an inrease in prodution in the short term, whih will be refleted in reduing unemployment. Testing Okun law is sensitive to the nature and intensity of the eonomy s business yle phase. Thus, in periods of eonomi reession, the unemployment rate will reat to a greater extent to the variation in prodution than in periods of inflationary gap when the unemployment rate remains around the natural unemployment rate. The struture of this study is onsistent with objetive namely testing Okun law in omania. Thus, in the first setion we have presented a series of theoretial arguments of the asymmetri relationship between unemployment and eonomi growth. In the next setion we tested the hypothesis of unemployment hysteresis in omania. Its presene is a sign of manifestation of strutural breaks that prevent a return to past values of unemployment. In the third

Marin Dinu, Marius Marinas, Cristian Sool, Gabriela Sool setion I estimated Okun oeffiient based on the ADL model, whih we introdued two dummy variables to apture the impat of hysteresis. The relationship between prodution and unemployment To illustrate the asymmetri response of unemployment to hange of prodution we used the model proposed by Shapiro and Stiglitz (1984). We realised the deomposition of output (Y), aording to the following equation: Y ' ' Y L Y = f ( Y / L, L ); Y Y > ; Y (1) = / L L > L where L is the employed population, and Y/L labour produtivity (affeted only by the workers). The two variables are in a positive relationship with the output and its variation is approximately equal to the amount of labour produtivity and hanges employment. A derease in employment will involve eteris paribus, an inrease of unemployment rate (U ) and onsequently an inrease in labour produtivity. Depending on the positive relationship between unemployment and produtivity, the previous equation an be rewritten in terms of unemployment rate, as follows: dy du dy dy / L = + dy / L du Sine the first derivative of employment with respet to the unemployment rate is negative (inverse relationship of these variables) and the produtivity is positive, then the first derivative of output is influened by the ratio between these two terms: - if dy dy / L dy dl, then will be a positive relation between dy / L du dl du > output and unemployment rate, at low levels of unemployment (area I, in figure 1); - if dy dy / L dy dl, then will be a negative relation between dy / L du < dl dy dl output and unemployment rate (area II, in figure 1). Y du dl du (2) I II U * Figure1. The relationship between output and unemployment U

Testing of the Okun s Law in omania Okun's law is validated in the seond area of the figure, in the event of negative shoks to aggregate demand or aggregate supply. The first area orresponds to strutural hanges of the eonomy that generate both short-term output and unemployment (strutural) inreases, meaning that law Okun is not heked. Another possible explanation of the inompatibilities inverse relationship between prodution and unemployment is the result of the unemployment rate dependene of other fators suh as degree of mobility of labour, demand for foreign labour fore, labour law and the shadow eonomy. For example, the emergene of a reession in an eonomy (the relative output derease) may be aompanied by the migration of unemployed (thus lowering the internal rate of unemployment) to other eonomies haraterized by labour fore defiit. Testing of unemployment hysteresis in omania (1999-28) Unemployment hysteresis term was introdued by Phelps (1972) with the signifiane of unemployment persistene at a given level after transitory shoks. Persistene of unemployment (or almost hysteresis) means a very low adjusting speed to the steady-state level. Aording to Lindbek and Snower (1986) the hysteresis phenomenon must be interpreted aording to the model oordinates of insiders-outsiders, so unions maintain wages at a level whih generates inrease of unemployment. omer (21) onsiders that there are two other auses of hysteresis. The first onerns the redution of qualifiations for the unemployed, whih will make harder finding a job when aggregate demand will rise. Under the seond, people in long-run unemployment surrenders to a large proportion of the job searh. It follows that the validation of unemployment hysteresis hypothesis is a signal to enhane the ative labour market poliies. The absene of hysteresis in unemployment is the onsequene of stationarity. If unemployment rate is a stationary proess at the initial level, the effets of various shoks will always be transitory and the unemployment rate will tend to the natural rate (assumed to be idential to the NAIU). If the unemployment rate is non-stationary (beomes stationary by integrating with the first differene), then the shoks that affet this series will have permanent effets on it, ausing its displaement from a point of equilibrium to another. Thus, ylial flutuations will exert an influene on long-run unemployment, the eonomy had the apaity to shok eliminate. From an eonometri viewpoint testing the hysteresis hypothesis involves the use of unit root tests. Studies that have estimated the unemployment hysteresis are sensitive to both the eonomy and analyzed hosen as a referene. Using data from the period 1853-1984 for Frane, Germany, UK and USA, Blanhard (1986) onluded that only the U.S. were haraterized by a stationarity of unemployment (lak of hysteresis in unemployment). oed (1996) validated the results to the U.S. for the period 197-1995 and estimated the presene of hysteresis in 16 OECD ountries. León-Ledesma (22) used a panel methodology and identified the existene of hysteresis in the European Union ountries and its absene in the U.S. León-Ledesma (24) estimated that the Central and Eastern European ountries

Marin Dinu, Marius Marinas, Cristian Sool, Gabriela Sool are haraterised by an unemployment hysteresis in the absene of dummy variables that highlight ertain strutural breaks. In order to test the hypothesis of unemployment hysteresis in omania we used monthly data rate of unemployment for the period 1999:1-28:6, the soure of data being Eurostat. Sine eonomi ativities are seasonal, we used Census X12 proedure to eliminate the influene of seasonality (Figure 2). 9.5 9. 8.5 8. 7.5 7. 6.5 6. 5.5 99 1 2 3 4 5 6 7 Note: Unemployment rate is represented by Y axis and is expressed as a perentage of the labour fore. Soure: Eurostat Figure 2. The unemployment rate in omania (without seasonality) Graphi representation shows that the unemployment rate reahing a maximum of 9.17% in January 22 and a minimum of 5.73% in May of 28. Also, there is the possible break of the strutural data series in January of 22, 24, 26, the orresponding halting the downward trend of unemployment. Sine August 26 there is a desending trajetory of the variable, in the emphasis on labour fore defiit in the omanian eonomy. From a statistial viewpoint, the harateristis of a normal distribution of data series are refleted by the oeffiient of for the asymmetry (skewness) and 3 for the flatted (kurtosis). If unemployment kurtosis value is lose to 3, so the probability of extreme values and persistent unemployment is about the same as for a normally distributed series. Series data shows a positive asymmetry (skewness =.33) indiating a slight upward trend in unemployment over the interval examined (Figure 3). 12 1 8 6 4 2 6. 6.5 7. 7.5 8. 8.5 9. Note: the values of the statistial tests are:.33717 for Skewness, 2.92914 for Kurtosis and 2.23952 for Jarque-Bera test. The probability value is.332214, higher than 5% signifiane level. Thus it aept null hypothesis of normal distribution of residual. Figure 3. Statistial distribution of unemployment rate

Testing of the Okun s Law in omania Jarque-Bera test indiates whether a series is normally distributed. It measures the differene between the oeffiient of asymmetry and the flatted of distribution when analyzed in omparison with the normal. Sine the assoiated probability is 33.22%, then aept the null hypothesis of a normally distributed series of unemployment rate. To test the hypothesis of hysteresis in unemployment rate we used stationary Augmented Dikey-Fuller (ADF) and Phillips-Perron (PP) tests. If you aept null hypothesis (H) that the series has a unit root and the variable is not stationary. A stationary time series is one whose statistial properties suh as mean, variane, autoorrelation, et. are all onstant over time. Also, a stationary series tends revert to the mean value and flutuate around it (has a finite variation). In ontrast, a nonstationary series has a different mean at different moments in time. In aordane with the ADF test with onstant, the series is non-stationary baseline to any of the thresholds of signifiane of 1%, 5% or 1%. The t statisti value is smaller in absolute value ompared to ritial values and assoiated probability is 248, whih suggests aepting null hypothesis (H) of non-stationary. The same result is obtained by using Phillips-Perron test with onstant probability is also higher than referene values of 1%, 5% and 1% (Table 1). Table 1. Testing the unemployment rate stationarity Critial values t-statisti ADF() -2.15939 PP() -2.163649 ADF(1) -11.25261 PP(1) -11.25261 1% -3.489117-3.489117-2.585773-2.585773 5% -2.88719-2.88719-1.943714-1.943714 1% -2.58525-2.58525-1.614834-1.614834 Note: The two tests examine the null hypothesis of non-stationarity. The unemployment rate is non stationary at level, but beome stationary after first differening at 1% level of signifiane (alulated values are less than ritial values). The estimation proedures of stationarity were arried out in Eviews 6. Applying the test for first differene, unemployment rate shows values signifiantly higher than the ritial values for ADF and Phillips-Perron (without trend and onstant), the latter taking into aount strutural breaks. It follows that the variable beomes stationary in first differene, ie is integrated at first order. Thus, between 1999-28 years the monthly rate of unemployment in omania registered hysteresis, shoks that have generated the variation being followed by a return to the previous variable. Taking into aount observations from Figure 2 we tested stationarity of unemployment rate series whereby exluded strutural break in January 22. In these irumstanes, the ADF and PP tests (with onstant) highlights data series stationarity at the level, whih ontradits the hysteresis hypothesis previously estimated (Table 2).

Marin Dinu, Marius Marinas, Cristian Sool, Gabriela Sool Table 2. The lak of hysteresis in unemployment rate Critial values t-statisti ADF() -4.37371 PP() -7.275489 1% -3.489659-3.489117 5% -2.887425-2.88719 1% -2.58651-2.58525 Note: The unemployment rate is stationary at level; the alulated values in Eviews 6 are less than ritial values. It follows that of unemployment rate hysteresis is onsequene of strutural break in data series. This was generated by applying from 1 January 22 a law on minimum inome. Under it, the persons entitled to reeive soial assistane should obtain proof of inome below the guaranteed minimum wage. In these irumstanes, many people have registered at National Ageny of Employment, although not in the job searh. Consequently, the unemployment rate was artifiially inreased from Deember 21, whih is loated at 6.51%. Thus onfirming results obtained by León-Ledesma (24) for a panel that has been inluded and omania. During the transition period have been both strutural hanges in these eonomies and hanges in labour law and soial protetion, whih were refleted in the emergene of hysteresis in unemployment. Its existene implies the use of dummy variables to estimate the Okun oeffiient in the ase of omania. Theoretial framework and the empirial evidene on Okun s law Okun (1962) estimated that an inrease in real GDP by 1 perentage point from GDP potential auses a derease by about.3 perentage points of unemployment below its natural rate other words, keeping unemployment at its natural level implies a level of output equal to that potential, aording to the following equation: ( U * t U ) = β ( Y Y ) + ε * t t t t, (3) where: Y t urrent level of GDP; Y * t - potential GDP; U t urrent level of unemployment rate; U * t - natural rate of unemployment (NAIU); β Okun oeffiient; ε t residual Okun oeffiient value (β) is mainly influened by three fators: employment laws. Firms hange their employment levels during the ourse of business and depending on the labour law relating to hiring and firing. Thereby the Okun oeffiient values will be lower in eonomies haraterized by higher level of employment protetion. the degree of labour market flexibility. In eonomies haraterized by a redued flexibility of the labour market (ie diminishing aommodation of firms and working age population to the eonomi hanges) will ontinue a high level of unemployment in periods of reession and job vaanies in periods of expansion. Therefore, the Okun oeffiient will be lower.

Testing of the Okun s Law in omania business yle phase and its intensity. Beause is not linear relation between hanging prodution and the employment, it is expeted that during of overheating unemployment is less responsive to hanges in prodution. In this ase the oeffiient β will have a lower value. In ontrast, during periods of prolonged reession, the unemployment rate will reat more to output derease. β oeffiient will have a value greater than the first ase. Studies that have tested the Okun oeffiient are sensitive to the nature of the output gap reorded and to the duration of expansion and eonomi reession. Lee (2) established that the relationship between unemployment and growth is not stable over time but onluded that the impat of output growth on employment is still valid. He also estimated that strutural breaks of the Okun law model is aused by redution in produtivity and restruturing ompanies. Sıgner (22) has analyzed the Okun oeffiients for the 15 OECD ountries during 196-1999 years and has estimated the.12 absolute value in Japan and.82 in the ase of the Netherlands. Coeffiients were not statistially signifiant for four of the 15 eonomies examined: Austria, Italy, Belgium and Switzerland. Gabrish (26) has tested the relationship between unemployment and eonomi growth between 1992 and 24 for eight of the new EU ountries and identified the existene of a stable relationship between these variables only in the last stage of transition. Okun oeffiients had values between.85 for Hungary and Latvia 2.3 ase in the interval 1998:1-24:4. In Poland and Slovakia the Okun oeffiient was not different from zero, one explanation being that the existene of a high natural rate of unemployment. Empirial investigation of Okun oeffiient in omania (1999-28) In this study we estimate Okun oeffiient by ordinary least squares (OLS) proedure using Eviews 6 software. To test this oeffiient in the ase of omania we used equation (3) relating to the ylial values of output and unemployment. As a proxy for the output hanges we used index of industrial prodution, beause this variable influenes the ylial evolution of GDP and employment level of the omanian eonomy. For example, industrial prodution dereases in periods of reession, whih redue the employment and the aggregate demand in the whole eonomy. Therefore it will shrink the ontribution of servies to GDP and the deline of domesti output. The soures of monthly data series on unemployment and industrial prodution index are Eurostat database and the Tempo database of National Institute of Statistis form omania. To assure omparability of dates, industrial prodution index was alulated using the fixed base year 2. The period analysis is 1999:1 and 28:6 years i.e. 112 observations. To remove seasonality of these variables we used the proedure Census X-12 in Eviews 6 software. We obtained the ylial values of these variables by removing trend omponents of series from data series. The trend where extrated using Hodrik-Presott filter with lambda parameter λ = 144, that orresponds to monthly data series.

Marin Dinu, Marius Marinas, Cristian Sool, Gabriela Sool The trend of industrial prodution index has been limbing in the analysis period, registering an inrease of approximately 5% ompared to 2. This trend reflets the inreased prodution potential of omanian industry, as a result of intensifiation of ompetitive pressure that aompanied the development of trade relations with developed eonomies. Trend unemployment rate (orresponding natural rate of unemployment, NAIU) did not reat immediately to developments in industrial prodution. Thus, it inreased from 7% in 1999 to 7.7% in 23:2, after whih it dereased to a level of 5.81% at the end of the period analyzed. An industrial prodution index below its trend desribes a negative output-gap and one higher than its trend desribes a positive output-gap. As regards the ylial unemployment that take negative values when the urrent rate of unemployment is lower than the natural and positive otherwise (Figure 4). 8 4-4 -8-12 99 1 2 3 4 5 6 7 Cylial unemployment rate Output-gap Figure 4. The relation between output-gap and the ylial rate of unemployment Note: On the Y axis are the hanges (in perents) of the output-gap and ylial unemployment rate. Soure: Eurostat and National Institute of Statistis (omania) Estimation method Aording orrelation between unemployment and output desribed by Okun in equation (3), the existene of a positive output gap generates a redution in unemployment below its potential, ie a negative ylial unemployment. This orrelation is validated between 21 and 22 years and a time shift some of the other years. Thus, a positive output gap at the end of 22 refleted a redution in unemployment in 23, and the negative gap at the end of 23 the unemployment rate inreased from trend in 24. Aording to Figure 4, the response of ylial unemployment to negative output-gap has been relatively low sine 27, as inreased tendeny to overheating of the eonomy has generated a weak ompensation of the defiit of labour ourred. Testing equation (3) in omania we have obtained an insignifiant regression from a statistial viewpoint, whih proves that the ylial unemployment rate does not vary depending on the urrent output gap, but in light of the delayed operation.

Testing of the Okun s Law in omania Therefore we estimated Okun oeffiient for omania, using the the dynami model ADL (autoregressive distributed lag), whih takes the following form: U t m n p β j U + γ t i i t i + j= 1 i= l= 1 = α + δ l Dummy l + ε t (4) where, U ylial rate of unemployment; Y ylial industrial prodution index (output-gap) The dependent variable of the above equation is influened both by the lagged rate of ylial unemployment, output-gap and dummy variables, whih will be introdued to reflet strutural breaks of unemployment rate. We estimated longrun equilibrium of the Okun s oeffiient by equalisation atual values with lagged variable. * U, t = U, t 1 =... = U, t m = U (5) U Y = Aording with equations (5) and (6), equation (4) is rewriting: t = Yt 1 =... = Yt n Y (6) Aording to the above relationship, long-run response of unemployment ylial to output-gap levels will be alulated as follows: β * * (1 i = = m 1 j = 1 Empirial results n m j= 1 γ β ) = i β j j n i= γ ( Y i t i ) Based on the Akaike information riterion (AIC) and Shwarz information riterion (SIC) 3 we hosed the ADL (1,3), to whih we added two dummy variables relating to January of 22, and 24, to haraterize the strutural breaks of unemployment rate. The oeffiients from the OLS estimation for equation (4) are reported in below equation: U =.924 U ( 1).1 * Y +.9 * Y ( 1) +.4 * Y ( 2 ) (9).19 ( 3) + 2.61 Dummy + 1.38 Dummy.3 22 24 The model is valid from a statistial viewpoint, beause: the value of explanatory regression is high adjusted 2 is.88 (as a result of the inlusion of lagged temporary variables also); the errors of regression are not auto-orrelated; Durbin-Watson test is 1.75 and the probability assoiated of Breush-Godfrey test is 61% (with 3 lags); the riteria for hoie of model has negative values and minimum (ompared to other simulations performed): AIC = -.44, and SIC = -.25; (7) (8)

Marin Dinu, Marius Marinas, Cristian Sool, Gabriela Sool there is not the phenomenon of errors heteroskedastiity, the probability assoiated with this test is 42%, a value higher than the aeptane of null hypothesis; regression oeffiients are stable, whih refleted both the CUSUM test and CUSUM of squares ( Figures 5). 3 1.4 2 1.2 1 1..8.6-1.4-2.2. -3 24 25 26 27 -.2 24 25 26 27 CUSUM 5% Signifiane CUSUM of Squares 5% Signifiane Fig. 5. The CUSUM and the CUSUM of Squares tests of oeffiients stability Note: The CUSUM test takes the umulative sum of reursive residuals and plots its value against the upper and lower bounds of the 95% onfidene interval at eah point. The CUSUM of squares statisti is a umulative sum of squares residuals, expressed as a fration of these squared residuals summed over all observations. The test is plotted with 5% onfidene bounds. As CUSUM test, we an find the parameter instability when the umulative sum of squares goes outside the area between the two ritial lines. The insignifiant oeffiients of regression (9) at the 1% signifiane level are written with asterisk. Aording to it, negative relationship between short-term output-gap and ylial unemployment rate is statistially valid if the independent variable is delayed with three lags. Suh inrease of one perentage point of the output gap will be refleted in a redution in ylial unemployment of more than three months to.2 perentage points. Also, 92.4% inrease in unemployment in the previous month was found in the urrent rate of unemployment, whih is evidene of persistent unemployment in the omanian eonomy. The variable dummy 22 has generated an inrease in ylial unemployment to 2.61 perentage points, and dummy 24 an inrease with 1.38 perentage points, whih ontributed to the inrease to a signifiant extent of unemployment in those years. For example, inreased unemployment in early 22 was refleted by an inrease of over 1 perentage point in the year (figure 6).

Testing of the Okun s Law in omania 3 2.5 2 (%) 1.5 1.5 I II III IV V VI VII VIII IX X XI XII the months of the 22 year Figure 6. The impat of unemployment rate inrease in January 22 Note: On the Y axis is the inrease of unemployment rate (in perentage points) brought about by strutural break from January 22. To alulate the oeffiient Okun on the basis of (7) and (8) equations we hold the signifiant regression oeffiients and the value obtained was -.26. Consequently, an inrease of industrial prodution over its trend by 1 perentage point will ause a long-run redution of.26 perentage points below the rate of natural rate of unemployment (the trend of unemployment rate). The result for Okun oeffiient in the ase of omania is relatively lose to the standard (about -.3) and higher than estimated Caraiani (26) for the period 1992-24 (-.17). It follows that the sensitivity of the unemployment rate gap to hange in domesti prodution inreased as a result of further growth in the period 25-28. Improving flexibility of labour fore, migration of the working-age population and rise of investments may be other explanations of the Okun oeffiient inrease (absolute value). If the initial analysis would exlude the last 14 months, the range has been the tendeny of the eonomy overheating, then the absolute value of estimated Okun oeffiient is higher (.35). Therefore, an inflationary gap auses a weak reation of employment and unemployment rate will reat less to prodution hanges. Testing for asymmetry in Okun oeffiient in omania The asymmetry of unemployment rate to hanges in prodution has been validated in the United States by othman (1991), Brunner (1997) and Silvapulle (24). Aording to their ylial unemployment rate is more sensitive to negative output-gap than in the positive output-gap (ie inflationary gap). To test the asymmetrial reation of unemployment to intensity and duration of output-gap we

Marin Dinu, Marius Marinas, Cristian Sool, Gabriela Sool used the following methodology used by Lee (2). First we divided the outputgap data series in two sub-periods, an inflationary output-gap gap (DI) and the other restritive output-gap (D), using the below variables. I U DI t 1, ify =, ify > < ; I D Equation (4) is rewriting: 1, ify =, ify < > m n n p β j U + γ i I t i DI t i + γ i I D t i + j= 1 i= i= 1 l= 1 = α + δ l Dummy l + ε t Depending on the AIC and SIC riteria we seleted three lags for the restritive output-gap and no lag for inflationary output-gap. We also have used the unemployment rate lagged by one month and two dummy variables used in previous regression and the model thus obtained satisfies the onditions of validity and stability oeffiients in equation (12) by OLS proedure 3. (1) (11) U.16 2.63 =.929 * I D Dummy U ( 1).14 ( 2 ).26 * I I D D 22 + 1.38 Dummy 24 + ( 3 ) +.6.12 * I.12 D * I DI ( 1) + + (12) The insignifiant oeffiients of regression at the 5% signifiane level are written with asterisk. Cylial unemployment rate is sensitive to the restritive output-gap lagged by three months, while the response to inflationary output gap is zero with a probability of 25%. Thus output derease by 1 perentage point below potential was refleted by a gap of three months the unemployment rate rising to.26 points. Calulating the long-run Okun s oeffiient (based on the equation 8) we obtained a value of -.37, a value higher (as absolute value) than initial value estimate in equation (9) (-.26). It thus onfirms the results of other studies that the unemployment rate tends to hange more in the event of a restritive output-gap. It thus validates the asymmetri Okun's oeffiient in the ase of the omanian eonomy. Conlusion In order to estimate the Okun oeffiient for the omanian eonomy we onstrut the variables of the model, namely the industrial prodution yle (proxy for output-gap) and the unemployment yle, and then desribe their dynami. Using an ADL model in whih we inluded two dummy variables were surprised that the impat event of rupture of the strutural unemployment rate, we obtained a -.26 value for long-run version of Okun s oeffiient. This value is lose to the standard (approximately -.3), but higher than estimated in 26 for the period 1992-24 (-.17). Inreasing (in absolute terms) of the Okun oeffiient is the effet of further eonomi growth in omania and redued labour market rigidities. It also reorded the highest values if we exluded from initial estimation the period haraterized by a labour market defiit. In addition, this oeffiient has reated to

Testing of the Okun s Law in omania a greater extent restritive output-gap, onfirming the results of other estimates. Even in this study we have analyzed the relationship between unemployment and industrial prodution during the eonomi expansion, though the results obtained are also useful for the period of eonomi risis. omania's entry into reession sine the third quarter of 28 generated a rapid inrease in the unemployment rate up to a maximum of 8.4% in Marh 21. Therefore, the Okun oeffiient will be higher than the value estimated in this study, showing asymmetri reation of unemployment to eonomi hange. Notes 1. This paper represents a partial dissemination of the postdotoral researh projet CNCSIS, HUMAN ESOUCES type, Maroeonomi modeling of the relationships between the asymmetri shoks, onvergene of business yles and mehanisms of adjustment in the ontext of omania's adhesion to the euro area, No 78/3.8.21, Projet Manager Marinaş Marius- Corneliu. 2. AIC and SIC are methods of omparing alternative speifiations by adjusting the sum of squared residuals (SSE) for the sample size (n) and the number of oeffiients in the model (K). AIC = Log(ESS/n) + 2(K) / n; SIC = Log(ESS/n) + Log(n)(K) / n. To use AIC and SC, estimate two or more models and alulate AIC and SC for eah equation. The lower AIC and SC are the better speifiation. 3. AIC and SIC values are -.43, respetively -.24 and adjusted -squared is.88. The oeffiients of the linear regression model are stable over time aording with stability tests of the oeffiients (CUSUM and CUSUM of Squares). The detet the autoorrelation in the residuals from regression, we have used the Breush Godfrey serial orrelation LM test. The null hypotesis of the test is H: no autoorrelation and it is aepted beause the test probability is 27.3% higher than 5% ritial value. To test for heteroskedastiity we have used White Test; its null hypothesis H: errors are homoskedasti. The probability is 94.5% higher than ritial values thus it rejets presene of heteroskedastiity.

Marin Dinu, Marius Marinas, Cristian Sool, Gabriela Sool EFEENCES [1] Altig, D., Fitzgerald, T., upert, T.(22), Okun's Law evisited: Should We Worry about Low Unemployment?. Publi Administration and Publi Poliy, 97, 135 142; [2] Blanhard, O.J., Summers, L.H. (1986), Hysteresis and the European Unemployment Problem. NBE Maroeonomis Annual 1, 15-78; [3] Brunner, A. D.(1997), On the Dynami Properties of Asymmetri Models of eal GNP. eview of Eonomis and Statistis, 79, 321-326; [4] Caraiani, P. (26), Alternative Methods of Estimating the Okun Coeffiient. Appliations for omania. Journal for Eonomi Foreasting, Institute for Eonomi Foreasting, vol. 3(4), 82-89; [5] Dreger, C. ; Marellino, M. (27), A Maroeonomi Model for the Euro Eonomy. Journal of Poliy Modeling, 29(1), 1-13; [6] Freeman, D. (2), A egional Test of Okun s Law. International Advanes in Eonomi esearh, 6, 557 57; [7] Gabrish, H.; Busher, H. (26), The elationship between Unemployment and Output in Post-ommunist Countries. Post-Communist Eonomies, Vol.18, No. 3, 261-276; [8] Gordon,.J.(1984), Unemployment and Potential Output in the 198 s, Brookings Papers on Eonomi Ativity, Vol. 2, pp. 537-564; [9] Izyumoy, A.;Vahaly. J. (22), The Unemployment-output Tradeoff in Transition Eonomies: Does Okun s Law Apply? Eonomis of Planning, Vol.35, 317-331; [1] Lee, J. (2), The obustness of Okun s Law: Evidene from OECD Countries. Journal of Maroeonomis 22 (2), 331 356; [11] León-Ledesma, M. A.; MAdam, P. (24), Unemployment, Hysteresis and Transition. Sottish Journal of Politial Eonomy, 51(3), 277-41; [12] León-Ledesma, M. A. (22), Unemployment Hysteresis in the US States and the EU: A Panel Approah. Bulletin of Eonomi esearh, 54, 95-13; [13] Lindbek, A.; Snower, D. J. (1986), Effiieny Wages versus Insiders and Outsiders. European Eonomi eview, 31, 47-416; [14] Malley,J.; Molana, H. (28), Output, Unemployment and Okun's Law: Some Evidene from the G7.Elsevier, Eonomis Letters 11, 113 115; [15] Moosa, I. (1997), A Cross ountry Comparison of Okun s Coeffiient. Journal of Comparative Eonomis, 24, 335 356; [16] Murphy, K. M.; Topel,. (1997), Unemployment and Nonemployment. The Amerian Eonomi eview, 87, 295-3; [17] Okun, A.M. (1962),Potential GNP: Its Measurement and Signifiane. Amerian Statistial Assoiation, Proeedings of the Business and Eonomis Setion, 98 13;

Testing of the Okun s Law in omania [18] Phelps, E. S. (1972), Inflation Poliy and Unemployment Theory. New York: W. W. Norton and Co.; [19] Prahowny, M. (1993), Okun s Law: Theoretial Foundations and evised Estimates. The eview of Eonomis and Statistis, 75, 331 336; [2] øed, K. (1996), Unemployment Hysteresis Maro Evidene from 16 OECD Countries. Empirial Eonomis, 21(4), 589-6; [21] omer, D. (21), Advaned Maroeonomis. 2 nd ed. New York: MGraw-Hill/Irwin; [22] othman, Ph. (1991), Further Evidene on the Asymmetri Behaviour of Unemployment ates over the Business Cyle. Journal of Maroeonomis, 13(2), 291-298; [23] Shnabel, G. (22), Output Trends and Okun s Law. BIS Working papers, Monetary and Eonomi Department, 111, 1-34; [24] Shapiro, C.; Stiglitz, J.E. (1984), Equilibrium Unemployment as a Worker Disipline Devie. Amerian Eonomi eview, 74, 433 444; [25] Silvapulle, P.; Moosa, I. A.; Silvapulle, M. (24), Asymmetry in Okun s Law. Canadian Journal of Eonomis, 37 (2), 353-374; [26] Sögner, L.; Stiassny, A. (22), An Analysis on the Strutural Stability of Okun s Law A Cross-ountry Study. Applied Eonomis, 14, 1775-1787; [27] Solow,. M. (2), Unemployment in the United States and in Europe: A Contrast and the easons. CESifo Working Paper, 46 (1), pp. 1-13; [28] Villaverde, J.; Maza A. (28), The obustness of Okun sllaw in Spain, 198 24 egional Evidene. Journal of Poliy Modeling 31, 289 297.