MPRA Munich Personal RePEc Archive Unemployment and Labour Force Participation in Italy Francesco Nemore Università degli studi di Bari Aldo Moro 8 March 2018 Online at https://mpra.ub.uni-muenchen.de/85067/ MPRA Paper No. 85067, posted 9 March 2018 09:08 UTC
Unemployment and Labour Force Participation in Italy Francesco Nemore Università degli Studi di Bari "Aldo Moro" Dipartimento di Economia, Management e Diritto dell'impresa Largo Abbazia Santa Scolastica, 53 70124 Bari (Italy) francesco.nemore@outlook.it Abstract: This paper provides an empirical investigation of the long-run relationship between unemployment and labour force participation rate in Italy. I test the unemployment invariance hypothesis using a macroeconomic approach and resorting to cointegration techniques. After testing for the non-stationarity of each time series, cointegration analysis results strongly suggest a clear long-run relationship between the two variables independently on gender aggregation level. In order to provide robustness to my results I replicate the analysis by age group aggregation level showing a strong discouraged-worker effect for the middle and old workers group. Keywords: Cointegration, Discouraged worker JEL Classification: E24, J60
Introduction The labour force participation and unemployment rates are commonly used indicators to determine the labour supply and the labour availability that readily can be used in the production process. There has always been ample debate among researchers and scholars looking for a precise definition of the employment status like unemployed, employed or inactive. These definitions can be based on different criteria and usually they can differ greatly across countries. During the eighties the International Labour Office (ILO) has undertaken a work of standardization of these diverse definitions about working conditions to ensure international comparability. These criteria have gradually been adjusted after the coordination action carried out by Eurostat. Yet still now substantial critics seem emerging in the distinction between unemployment and non-participation in the labour force. Economists usually assume that the low unemployment rate in some countries is linked to a temporary excess of labour supply where people are temporarily channelled into not-activity rather than considered unemployed. However, the empirical literature found statistically significant discourage-worker effect (Murphy and Topel, 1997). This would lead individuals to end an active job research because of the perceived low probability of finding vacancies. When the cyclical nature of the economy can affect such perceptions it is likely that the unemployment labour force participation is guided by a more or less significant causality (Elmeskov and Pichelmann, 1993). 2
The labour market in Italy has been subject to several reforms in recent years. After the reform of the Monti s government and subsequent amendments and additions made by the Letta s executive, the prime minister Matteo Renzi has outlined a new program of reforms designed to affect the labour market and welfare through the Jobs Act. The Italian labour market was always characterized by a high degree of structural rigidity. The system centred on the Statute of workers favoured a rigid protection of existing jobs through the imposition of high firing costs and the institution of wage supplementation that allowed maintaining an umbilical cord between the unemployed workers and companies of origin. In comparison with its European partners, the Italian labour market continues to be distinguished by its small size: the values of the activity and employment rates are comparatively low, while the unemployment rate is high. Because of such substantially typical characters of the Italian labour market and the persistency effects of the economic crisis in 2007-2008 the theoretical hypothesis of independence between unemployment rate and labour force participation (Layard et al., 1991) could not be verified. The aim of this contribution is to explore the long-run relationship between unemployment and labour force participation in Italy. In the wake of previous studies (Osterholm, 2009 and Emerson, 2011) I proceed to test the unemployment invariance hypothesis using a macroeconomic approach and resorting to cointegration techniques. In the first part the relationship between unemployment and labour force participation is analysed using data aggregation by gender. In order to investigate the robustness of my results I then proceed to repeat the analysis by age group independently on the gender aggregation. 3
Data and empirical analysis I use seasonally adjusted monthly data on unemployment and labour force participation rate for workers aged 16-64 in Italy over the sample period from January 1998 to December 2014. The data come from the Labour Force Survey (LFS) available on the Eurostat databases. Figure 1 shows the time series of the labour-force participation rate and the unemployment rate in the sample period. The overall unemployment rate dropped significantly since 1998, reaching a minimum of around 6% in 2007. As a result of the financial and economic crisis it starts a growing trend with a peak of 12% in 2014. Same trends are found for the unemployment rates of males and females. Differently the labour force participation is strictly increasing until 2004 where it seems getting stability at around 63%. The recession of 2008 determines a reduction in the labour participation rate by about a percentage point until 2012 where the trend starts to increase with a maximum of 64.3% in 2014. Whereas the female labour force participation follow the overall trend the male participation rate has increased gradually until 2005 and then declined until 2011 but with many intermediate fluctuations. Firstly I proceed to analyse the time series properties using two different unit-root tests: the Augmented Dickey Fuller test with GLS detrending (Elliot et al., 1996) and the KPSS test (Kwiatkowski et al., 1992). The Dickey-Fuller t test for a unit root is a test in which the series have been transformed by a generalized least-squares regression. The KPSS test differs from the commonly used unit root tests (such as 4
Dickey Fuller test) by having a null hypothesis of stationarity. The tests may be conducted under the null of either level stationarity or trend stationarity. Inference from this test is complementary to that deriving from tests based on the Dickey-Fuller distribution. The KPSS test is often used in conjunction with other ordinary tests to investigate the possibility that a series is fractionally integrated. The results of the two-unit root tests are showed in Table1. Independently on gender aggregation level both tests confirm that unemployment (u t ) and labour force participation rate (p t ) are not stationary at 1% significance level. Having showed that all time series variables are unit-root processes, I proceed to check if there is a longrun relationship between them through a cointegration test. If variables are cointegrated unemployment and labour force participation rates are related in the long-run. The opposite will result if they are not cointegrated. In the first case a relevant economic intuition can be inferred: the theoretically unemployment invariance hypothesis would not be confirmed in an empirical setting. I test the cointegration applying Johansen's methodology (Johansen, 1988, 1991). The model is a finite-order VAR designed as:! x! = c + A! x!!! + e!!!! where x! = p!, u! is a vector of non-stationary variables containing the labour force participation rate, p!, and the unemployment rate, u!, A! is a 2x2 matrix of parameters, and e! is a 2x1 vector of residuals. Rewriting this unrestricted VAR we get: 5
!!! x! = c + Γ! x!!! + Πx!!! + e!!!! where Γ! =!!!!!!! A! I and Π =!!! A! I. The unemployment rate, u!, and the labour force participation rate, p!, are cointegrated if and only if the coefficients of Π matrix has rank equal to one. In this case Π is decomposed as Π = αβ, where α and β are 2x1 vectors, α contains the adjustment parameters in the vector error correction model (VEC), and β contains the cointegration vector. The Johansen s approach produces two statistics: the Johansen s trace, J!"#$%, and the Johansen s maximum eigenvalue, J!"#, that allow to check the presence of cointegrating vectors in each relationship. The selection of the number of lags for each cointegration analysis is conducted through the Akaike information criterion, the Hannan-Quinn criterion and the Schwarz information criterion. Table 2 reports the results of cointegration analysis. I show that only one cointegrated vector is supported in all three cases, though at different significance level. These results would confirm that unemployment and labour force participation are related in the long-run thus violating the unemployment invariance hypothesis. To check what kind of relationship exists between the two variables, I proceed to estimate the coefficients of each cointegrated vector. Table 3 shows that no estimated coefficient is statistically significant considering the standard errors. Such results would presume the absence of any discourage-worker effect for Italy differently from work Osterholm (2010) and Emerson (2011), where this effect is statistically proven for Sweden and USA. Assuming that the age drives worker s motivation in any active search for vacancies, I proceed to repeat my analysis by considering a different level 6
of data aggregation. The aggregation by age distinguishes unemployment and labour force participation among two age groups: 15-24 (young) and 25-64 (middle and old). Table 4 shows the unit root test results. Again, the non-stationarity of each time series is found for both age groups. Now, the existence of a long-run relationship between unemployment and labour force participation rate can be proven running the usual cointegration test. Table 5 presents cointegration analysis results. The existence of a single cointegrated vector in both age groups confirms the hypothesis of a long-run relationship between the two cointegrated variables. Finally, Table 6 reports the estimated coefficients. Unlike the previous cointegration framework, the new coefficients allow for a direct interpretation: for middle and older workers a lower labour participation rate is always associated with a higher unemployment rate. Therefore the presence of the aforementioned discouragedworker effect is proved. Concluding Remarks This empirical contribution attempts to provide a testing of the unemployment invariance hypothesis for the Italian labour market. By using cointegration tests I demonstrate a long-run relationship between unemployment and labour force participation rates regardless gender aggregation level. In order to sustain my results, I replicate the analysis with age group level data. Robustness checks presents clear evidence of the supposed long-run relationship and support a discouraged-worker effect for the middle and old workers age group. These results contradict the 7
unemployment invariance hypothesis and would call policy makers to draw the necessary implications when labour market policies need to be revisited. REFERENCES: Elliott, G., Rothenberg, T. J., & Stock, J. H. (1992). Efficient tests for an autoregressive unit root. Econometrica, 64, 813-836. Elmeskov, J., & Pichelmann, K. (1993). Unemployment and Labour Force Participation. Economic Department Working Paper, No. 130. OCSE. Emerson, J. (2011). Unemployment and labor force participation in the United States. Economics Letters, 111(3), 203-206. Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of economic dynamics and control, 12(2-3), 231-254. Johansen, S. (1991). Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models. Econometrica: Journal of the Econometric Society, 1551-1580. Kwiatkowski, D., Phillips, P. C., Schmidt, P., & Shin, Y. (1992). Testing the null hypothesis of stationarity against the alternative of a unit root: How sure are we that economic time series have a unit root?. Journal of econometrics, 54(1-3), 159-178. Layard, P. R. G., Nickell, S. J., & Jackman, R. (2005). Unemployment: macroeconomic performance and the labour market. Oxford University Press. Murphy, K. M., & Topel, R. (1997). Unemployment and nonemployment. The American Economic Review, 87(2), 295-300. Österholm, P. (2010). Unemployment and labour-force participation in Sweden. Economics Letters, 106(3), 205-208. 8
Total unemployment rate Total labour-force participation rate tot_unemp 6 8 10 12 14 1998m1 2000m1 2002m1 2004m1 2006m1 2008m1 2010m1 2012m1 2014m1 date tot_lfpr 58 60 62 64 1998m1 2000m1 2002m1 2004m1 2006m1 2008m1 2010m1 2012m1 2014m1 date Male unemployment rate Male labour-force participation rate unemp_men 4 6 8 10 12 1998m1 2000m1 2002m1 2004m1 2006m1 2008m1 2010m1 2012m1 2014m1 date lfpr_men 72.5 73 73.5 74 74.5 75 1998m1 2000m1 2002m1 2004m1 2006m1 2008m1 2010m1 2012m1 2014m1 date Female unemployment rate Female labour-force participation rate unemp_women 8 10 12 14 16 1998m1 2000m1 2002m1 2004m1 2006m1 2008m1 2010m1 2012m1 2014m1 date lfpr_women 44 46 48 50 52 54 1998m1 2000m1 2002m1 2004m1 2006m1 2008m1 2010m1 2012m1 2014m1 date Fig. 1. Unemployment and labour-force participation rate in Italy, January 1998 to December 2014. 9
Table 1: Univariate unit-root tests on individual series All Male Female p t u t p t u t p t u t ADF-GLS -1.231-1.090-1.014-1.325-1.588-1.020 KPSS 0.244*** 0.345*** 0.257*** 0.344*** 0.221*** 0.347*** Note: ADF-GLS is the test-statistic from the augmented Dickey-Fuller test with GLS detrending, where lag length is chosen based on the Schwarz information criterion. KPSS is the test statistic from Kwiatowski, Phillips, Schmidt and Shin test. *** indicate significance at 1% level, Sample covers January 1998 to December 2014. Table 2: Cointegration tests All Male Female J trace J max J trace J max J trace J max H 0 : r = 0 13.831* 13.422* 14.212* 13.997* 17.705** 17.563** H 0 : r = 1 0.409 0.409 0.215 0.215 0.141 0.141 Note: Lag length of the VAR is selected using the Akaike information criterion, the Hannan-Quinn criterion and the Schwarz information criterion. * and ** indicate significance at the 10% and 5% levels, respectively. Sample covers January 1998 to December 2014. Table 3: Estimated cointegrating vector All Male Female p t-1 1.00 1.00 1.00 u t-1-0.005 0.063-0.179 (0.189) (0.074) (0.395) constant -62.118-74.640-49.530 Note: Standard errors in parenthesis. Sample covers January 1998 to December 2014. 10
Table 4: Univariate unit-root tests on individual series Age-group 15-24 25-64 p t u t p t u t ADF-GLS -2.142-1.436-0.737-0.823 KPSS 0.226* 0.342*** 0.341*** 0.344*** Note: ADF-GLS is the test-statistic from the augmented Dickey-Fuller test with GLS detrending, where lag length is chosen based on the Schwarz information criterion. KPSS is the test statistic from Kwiatowski, Phillips, Schmidt and Shin test. * and *** indicate significance at 10% and 1% level, respectively. Sample covers January 1998 to December 2014. Table 5: Cointegration tests Age-group 15-24 25-64 J trace J max J trace J max H 0 : r = 0 13.057* 12.941* 18.862** 18.411** H 0 : r = 1 0.116 0.116 0.451 0.451 Note: Lag length of the VAR is selected using the Akaike information criterion, the Hannan-Quinn criterion and the Schwarz information criterion. * and ** indicate significance at the 10% and 5% levels, respectively. Sample covers January 1998 to December 2014. Table 6: Estimated cointegrating vector Age-group 15-24 25-64 p t-1 1.00 1.00 u t-1 0.160 0.507 (0.207) (0.197) constant -38.708-80.454 Note: Standard errors in parenthesis. Sample covers January 1998 to December 2014. 11
12