Okun's law in Austria

Christl, Michael; Köppl-Turyna, Monika; Kucsera, Dénes Article Okun's law in Austria DANUBE: Law, Economics and Social Issues Review Provided in Cooperation with: European Association Comenius (EACO), Brno Suggested Citation: Christl, Michael; Köppl-Turyna, Monika; Kucsera, Dénes (2017) : Okun's law in Austria, DANUBE: Law, Economics and Social Issues Review, ISSN 1804-8285, De Gruyter, Warsaw, Vol. 8, Iss. 2, pp. 97-110, https://doi.org/10.1515/danb-2017-0007 This Version is available at: https://hdl.handle.net/10419/184526 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0 DANUBE: Law and Economics Review, 8 (2), 97–110 DOI: 10.1515/danb-2017-0007 97 OKUN’SLAW IN AUSTRIA Michael Christl1, Monika Köppl-Turyna2, Dénes Kucsera3 Abstract We estimate the classic and the dynamic variant of Okun’s law for the Austrian labor market. We find that, for recent periods, the growth rate necessary to stabilize the unemployment rate equals 2.8 percent. Moreover, we find that the rate has been growing in recent quarters due to the increasing labor force size and the effects of the crisis. The latest prediction of the employment threshold lies above 3 percent, much above forecasted GDP growth up to 2017. Keywords Unemployment, Growth, Okun’s Law, Austria I. Introduction When the financial crisis hit the Austrian economy in 2009, a huge decline of almost 4 percent in real GDP was observed. Most people expected an increase in the unemployment rate; however, this expectation did not come true. At least until 2012, the unemployment rate stayed on quite a stable level, while GDP growth remained low for several years after the crisis. This brings into question whether the relationship between the unemployment rate and GDP growth – also known as Okun’s law – has changed over time. Okun’s law goes back to Okun (1962) and describes a long-run relationship between output and the unemployment rate. In its simplest form, it is often stated as a three-to- -one relation, i.e. a decrease of one percent in real output increases unemployment by 0.3 percent. Hence, the law determines how much growth in the economy is necessary to stabilize the unemployment rate. If growth lies above this threshold, the unemployment rate decreases and vice versa. The original findings of Okun have been criticized in the literature following his publication. Dornbusch and Fischer (1988) find that estimates of potential output and the link between unemployment and the GDP gap change over time 1Agenda Austria, Schottengasse 1/3, 1010 Vienna. E-mail: [email protected]. 2Agenda Austria, Schottengasse 1/3, 1010 Vienna. E-mail: [email protected]. 3Agenda Austria, Schottengasse 1/3, 1010 Vienna. E-mail: [email protected]. 98 Michael Christl, Monika Köppl-Turyna, Dénes Kucsera: Okun’s Law in Austria and cannot be taken as a constant rule. Cuaresma (2003) shows that the contemporaneous effect of growth on unemployment is not symmetric and significantly higher in recessions. Empirical estimates indicate that a two to three percent GDP growth above natural or average GDP growth causes unemployment to decrease by one percentage point and vice versa (see Blanchard and Fischer (1989) or Romer (1996)). Mankiw (1994) argues that Okun’s rule of thumb is closer to two than to three percentage points. Sögner (2001) analyzes the dependence of unemployment growth on annual real GDP growth for Austria. He finds neither structural breaks nor outliers in Okun’s law in a sample of quarterly Austrian data covering the period 1977 to 1995. Moreover, Sögner (2001) found that neither changes in the political system, governmental employment programs nor migration have caused a systematic change in Austria’s Okun’s law. Hence, he estimates Okun’s law at a constant level of 4.16 Percent in the period between 1977 to 1995. Sögner and Stiassny (2002) again analyzed Okun’s law for 15 OECD countries for the time period between 1964 and 2000 and found structural breaks in the Okun relation for some OECD countries, but not for Austria. Discussion regarding the stability of Okun’s law came up in many countries after the global financial crises. In the US, for instance, unemployment declined considerably despite weak growth. Several papers address this question – Meyer, Tasci et al. (2012) and (Owyang, Sekhposyan et al., 2012) for the US, Beaton (2010) for the US and Canada, Österholm (2016) for Sweden and Ball, Leigh and Loungani (2013) for 20 advanced economies. In a recent paper, Christl, Köppl-Turyna and Kucsera (2016) showed that recent developments in the Austrian labor market can be attributed to an undergoing structural change in some of the most important sectors of the Austrian economy. Such changes might indeed influence Okun’s law and thus questions the stability of Okun’s law that has been reported for the period before the year 2000 by Sögner (2001) and Sögner and Stiassny (2002). This paper contributes to the discussion on Okun’s law’s stability by analyzing the development of Okun’s law in Austria for the time period from 1970 to 2015. Our results suggest that it is not stable over time. The estimated models show that, after entry into the European Union, the employment threshold decreased by almost one percentage point. Even though the employment threshold was decreasing, we again found an upward trend in recent years. This contrasts with findings for other European countries, such as Sweden, where Österholm (2016) showed that the employment threshold decreased continuously over the period from 2004 to 2014. In section II, we describe the data and in section III we briefly present the empirical model. Section IV presents the empirical findings and robustness analyses. Finally, Section V concludes the paper. II. The Data In the following analysis, we use both yearly and quarterly data. As yearly data are available for a longer period, the main part is based on these data. Nevertheless, in order to provide more accurate results, and also as a robustness check, the analysis is repeated using quarterly data. This section describes both datasets. DANUBE: Law and Economics Review, 8 (2), 97–110 DOI: 10.1515/danb-2017-0007 99 The yearly data on unemployment is taken from Public Employment Service Austria (AMS), which counts all registered unemployed persons. Real GDP growth is taken from the database of the Austrian Federal Economic Chamber. Both time series are available from 1950 until 2014. The quarterly real GDP growth and unemployment data are taken from the OECD database. Similarly to the yearly data, the second series count all registered unemployed persons. Unlike the yearly data, however, the quarterly time series are available only for a shorter period and cover the time span from Q1 1988 until Q1 2015. Both series are seasonally adjusted. A visual inspection of the data in use reveals that the unemployment rate might be stationary, whereas the GDP index has been showing an upward trend. Yet a simple visual inspection cannot allow us to easily determine the stationarity of the unemployment series or whether the trend observed for GDP is deterministic or stochastic. To test for the stationarity of the time series, we use the ADF test and the PP test. The test results for the variables of interest are presented in Table 1. Both tests suggest that the level data has a unit root, also when considering the possibility of a deterministic trend in the real GDP index. Table 1: Unit root tests (p-values) U log(rGDP) ADF PP ADF PP Yearly data level 0.04 0.72 0.46 0.86 level, trend 0.30 0.29 dif 0.03 0.01 0.01 0.01 Quarterly data level 0.29 0.05 0.79 0.96 level, trend – – – – dif 0.00 0.00 0.00 0.00 Taking first differences solves the problem of non-stationarity. We still need to check whether the data is not cointegrated. Using information criteria, we can establish that one lag of the differenced data should be used to test for cointegration. The results of the Johansen test are presented in Table 2. Table 2: Cointegration test Max rank Parms Eigenvalue Trace Crit. value Yearly data 0 2 10.38 15.41 1 5 0.14 0.55 3.76 2 6 0.01 Quarterly data 0 6 14.94 15.41 1 9 0.09 5.09 3.76 2 10 0.05 100 Michael Christl, Monika Köppl-Turyna, Dénes Kucsera: Okun’s Law in Austria We cannot reject the null hypothesis that there is no cointegrating relationship between GDP and the unemployment rate, therefore the OLS method can be applied, since the differenced series of the unemployment rate and real GDP growth are stationary. Figure 1 presents the development of the first differences of the unemployment rate and GDP growth over time. Figure 1: Unemployment rate (first difference) and real GDP growth -.02 -.01 0 .01 .02 d(unemployment rate) -.15 -.1 -.05 0 .05 .1 .15 GDP growth 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Year GDP growth d(unemployment rate) Source: Public Employment Service Austria (AMS) and Austrian Federal Economic Chamber (WKO) III. The empirical model Traditionally,Okun’s lawstatestheempirical relationshipbetweenchanges inthe unemployment rate ∆utand changes in output ∆yt. The “classical version” of Okun’s law can be written as: ∆ut=β0+β1∆yt+t,(1) where β1is the so-called “Okun’s coefficient”. One would expect Okun’s coefficient to be negative, so that output growth is associated with a falling unemployment rate and negative output growth is associated with a rising unemployment rate. The ratio −β0/β1 measures the rate of output growth consistent with a stable unemployment rate, or, in other DANUBE: Law and Economics Review, 8 (2), 97–110 DOI: 10.1515/danb-2017-0007 101 words, how quickly the economy would typically need to grow to maintain a given level of unemployment, the so-called “employment threshold”. To control for the lagged reaction on the labor market, the lags of the variables are included in the model, which leads to the “dynamic version” of Okun’s law. Changes in Okun’s law (Okun’s coefficient) are affected by the business cycle, and also by the variation in the timing and relation between growth and unemployment rates (see, e.g. Knotek II, 2007). These facts suggest that the dynamic version of Okun’s law should be the preferred model. Additionally, Meyer et al. (2012) suggest allowing for non-linear changes in the relationship between growth and changes in unemployment, which might solve the instability problem of the classic Okun’s law. ∆ut=β0+β1∆yt+β2∆yt−1+β3∆yt−2+γ1∆ut−1+γ2∆ut−2+t(2) Another reason for including past changes in the unemployment rate as explanatory variables in the dynamic version of Okun’s law is to eliminate serial correlation in the error terms, which might constitute a problem when regressing the difference version of Okun’s law.4 In this paper, we examine both versions of Okun’s law for Austria to see whether the results for Okun’s coefficient as well as for the employment threshold are robust across the different specifications. IV. Results Okun’s law – the classical model The linear relationship between changes in the unemployment rate and real GDP growth, as stated by Okun’s law (Equation 1) will be tested for statistical significance in this section. As expected, Okun’s coefficient estimated for the period between 1970 and 2014 is negative and highly significant (see Table 3, Column 2). The rate of output growth that is consistent with a stable unemployment rate − β0 β1equals 3.62 percent, with a 95% confidence interval between 2.97 and 4.27 percent. Subsequently, we split the sample into two periods, one for the period within the European Union (from 1995 onwards) and one before entering the European Union (Table 3, Columns 3 and 4). Okun’s coefficient is again negative and highly significant. The employment threshold − β0 β1in the time period between 1995 and 2014 is distinctly lower, at 2.78 percent, and distinctly higher for the period between 1970 and 1995, at 4.42 percent. 4An alternative specification of Okun’s law is called the “gap version”. This states the relation between the gap of the actual unemployment rate utand the natural rate of unemployment ut∗and the output gap the difference between the actual output ytand the potential output yt∗. (ut−ut∗) = β0+β1(yt−yt∗) + t.(3) As the interpretation of the coefficients differs strongly from the interpretation of the traditional and dynamic equation, we focus only on these two equations. 102 Michael Christl, Monika Köppl-Turyna, Dénes Kucsera: Okun’s Law in Austria Table 3: Regression results for the classical formulation of Okun’s law Dependent variable: ∆U 1970–2014 1970–1995 1995–2014 ∆GDP −0.178∗∗∗ −0.194∗∗∗ −0.175∗∗ (0.025) (0.026) (0.099) Constant 0.006∗∗∗ 0.008∗∗∗ 0.005∗∗ (0.001) (0.001) (0.002) Observations 64 45 20 Adjusted R20.446 0.559 0.100 Residual Std. Error 0.005 (df = 62) 0.004 (df = 43) 0.005 (df = 18) F Statistic 51.636∗∗∗ (df = 1; 62) 56.784∗∗∗ (df = 1; 43) 3.120∗∗∗ (df = 1; 18) D-W stat. 1.84 1.69 2.42 Breusch-Godfrey p-val 0.53 0.36 0.30 Employment threshold 0.0362 0.0442 0.0278 LL 95% 0.0297 0.0332 0.0140 UL 95% 0.0427 0.0469 0.0415 Note: ∗p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01 Figure 2: Rolling regression – classical formulation of Okun’s law (yearly data) Year Employment treshhold 1970 1980 1990 2000 2010 2.5 3.0 3.5 4.0 4.5 5.0 5.5 Note: The figure presents the real output growth necessary to stabilize the unemployment rate that results from the rolling regression (with the length of 20 years) of classical formulation of Okun’s law for yearly data. The sample starts with 1950 to 1970 and ends with the period 1994 to 2004. DANUBE: Law and Economics Review, 8 (2), 97–110 DOI: 10.1515/danb-2017-0007 103 To test the stability of these results over time, we use a rolling regression method for different time periods with a length of 20 years. The sample starts with the period 1950 to 1970 and ends with the period 1994 to 2014. Figure 2 shows the development of the rate of real output growth that is needed for a stable unemployment rate. We can see that the employment threshold changes significantly over time. While in the 1970s it was close to 3.5 percent real GDP growth, it rose to 5.5 percent in the 1990s. Since entering the European Union, the threshold continuously has decreased to a level of 2.7 percent nowadays. Nevertheless, at the end of the examined period, and thus for the period comprising the recent financial crisis, the employment threshold has again started to increase. Okun’s law – dynamic model Thedynamicmodeladdsmorelags of real GDP and of the differences in the unemployment rate. The model specification is highlighted in Equation 2. The number of optimal lags in the model indicated by the Akaike’s information criterion is two. Adding the lag structure to the model still shows statistically significant coefficients for Okun’s law (Table 4, Column 2). Table 4: Regression results for the dynamic formulation of the Okun’s law before and after entering the EU Dependent variable: ∆U 1970–2014 1970–1995 1995–2014 ∆GDP −0.118∗∗∗ −0.139∗∗∗ −0.002 (0.026) (0.025) (0.081) ∆GDPt−1−0.076∗∗ −0.040 −0.418∗∗∗ (0.032) (0.034) (0.084) ∆GDPt−20.015 −0.009 −0.056 (0.032) (0.035) (0.110) ∆Ut−10.138 0.230∗ −0.170 (0.110) (0.128) (0.200) ∆Ut−1−0.248∗∗ −0.291∗∗ −0.613∗∗∗ (0.114) (0.131) (0.186) Constant 0.006∗∗∗ 0.007∗∗∗ 0.012∗∗∗ (0.001) (0.002) (0.004) Observations 62 43 18 Adjusted R20.589 0.664 0.640 Residual Std. Error 0.004 (df = 56) 0.004 (df = 37) 0.003 (df = 12) F Statistic 16.053∗∗∗ (df = 5; 56) 17.613∗∗∗ (df = 5; 37) 7.034∗∗∗ (df = 5; 12) D-W stat. 2.27 2.36 1.81 Breusch-Godfrey p-value 0.06 0.04 0.31 Continued on next page 104 Michael Christl, Monika Köppl-Turyna, Dénes Kucsera: Okun’s Law in Austria Dependent variable: ∆U 1970–2014 1970–1995 1995–2014 Employment threshold 0.0350 0.0384 0.0257 LL 95% 0.0294 0.0326 0.0228 UL 95% 0.0405 0.0440 0.0295 Note: ∗p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01 In the dynamic version of Okun’s law, the rate of output growth consistent with a stable unemployment rate − β0 β1+β2+β3is 3.50 percent in the long-run relation between 1970 and 2014, with a 95% confidence interval of 2.94 percent to 4.05 percent. Dividing the sample again into two periods, before and after entering the EU, we can again observe that the estimation results are not constant over time (Table 4, Columns 3 and 4). While in the period between 1970 and 1995 reactions of the labor market to changes in GDP growth have taken place almost without any delay (β1is highly significant, while the lags show no significance), in the period between 1995 and 2014, the labor market seems to react a year later to changes in output (β2is highly significant). The employment threshold rises to 3.84 percent for the period 1970 to 1995, while it is generally lower, at the level of 2.57 percent, for the period within the European Union. Figure 3: Rolling regression – dynamic formulation of the Okun’s law (yearly data) Year Employment treshhold 1970 1980 1990 2000 2010 2.5 3.0 3.5 4.0 Note: The figure presents the real output growth necessary to stabilize the unemployment rate that results from the rolling regression (with the length of 20 years) of the dynamic formulation of Okun’s law for yearly data. The sample starts with 1950 to 1970 and ends with the period 1994 to 2014.