Regional and Sectoral Evidence of the Macroeconomic Effects of Labor Reallocation: A Panel Data Analysis

Bakas, Dimitrios; Panagiotidis, Theodore; Pelloni, Gianluigi Working Paper Regional and Sectoral Evidence of the Macroeconomic Effects of Labor Reallocation: A Panel Data Analysis Quaderni - Working Paper DSE, No. 902 Provided in Cooperation with: University of Bologna, Department of Economics Suggested Citation: Bakas, Dimitrios; Panagiotidis, Theodore; Pelloni, Gianluigi (2013) : Regional and Sectoral Evidence of the Macroeconomic Effects of Labor Reallocation: A Panel Data Analysis, Quaderni - Working Paper DSE, No. 902, Alma Mater Studiorum - Università di Bologna, Dipartimento di Scienze Economiche (DSE), Bologna, https://doi.org/10.6092/unibo/amsacta/3832 This Version is available at: https://hdl.handle.net/10419/159741 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/3.0/ ISSN 2282-6483 Regional and Sectoral Evidence of the Macroeconomic Effects of Labor Reallocation: A Panel Data Analysis Dimitrios Bakas Theodore Panagiotidis Gianluigi Pelloni Quaderni - Working Paper DSE N°902 REGIONAL AND SECTORAL EVIDENCE OF THE MACROECONOMIC EFFECTS OF LABOR REALLOCATION: A PANEL DATA ANALYSIS? DIMITRIOS BAKASa,f †THEODORE PANAGIOTIDISb,f ‡GIANLUIGI PELLONIc,d,e,f †† aDepartment of Economics, University of Athens, Greece bDepartment of Economics, University of Macedonia, Greece cDepartment of Economics, University of Bologna, Italy dDepartment of Economics, Wilfrid Laurier University, Canada eJohns Hopkins University Bologna Center, Italy fRimini Centre for Economic Analysis (RCEA), Italy July 2013 ABSTRACT This paper re-examines Lilien’s sectoral shifts hypothesis for U.S. unemployment. We employ a monthly panel that spans from 1990:01 to 2011:12 for 48 U.S. states. Panel unit root tests that allow for crosssectional dependence reveal the stationarity of unemployment. Within a framework that takes into account dynamics, parameter heterogeneity and cross-sectional dependence in the panel, we show that sectoral reallocation is significant not only at the aggregate level but also at the state level. The magnitude and the statistical significance of the latter as measured by Lilien’s index increases when both heterogeneity and cross-sectional dependence are taken into account. Keywords: Unemployment · Sectoral Shifts · Employment Fluctuations · Dynamic Panel Data · Parameter Heterogeneity ·Cross-Sectional Dependence. JEL Classification: C33 ·E24 ·E32 ·J21 ·R23 ? Acknowledgments: We would like to thank Karim Abadir, Ramazan Gençay, Angelo Melino, Roberto Patuelli, Hashem Pesaran, Vasilis Sarafidis and Thanasis Stengos for their useful comments and suggestions on earlier versions of this work. Also we would like to thank the participants of the 2nd Time Series Worskhop of the Rimini Center for Economic Analysis. All remaining errors are our own. †E-mail: [email protected] (Corresponding Author) ‡E-mail: [email protected] ††E-mail: [email protected] 1 INTRODUCTION The analysis of the macroeconomic effects of labor reallocation has been developed along several dimensions. Earlier analysis have focused on one-dimensional characteristics: sector, plant, dimension, labor turnover, region, real wage, exchange rate and money supply among others, while subsequent work has deepened their analytical frameworks by embodying concurrent disaggregations along multiple dimensions.1This study follows the latter and, by bringing together the sectoral and regional dimensions, follows the path set by the pioneering efforts of Medoff (1983) and Neumann and Topel (1991). In this work, we explore the impact of a purged measure of labor reallocation on unemployment using an extensive panel data for the United States. The novel aspects of this article are linked to recent developments in panel data econometrics concerning dynamics, heterogeneity and cross-sectional dependence which, to the best of our knowledge, have never been applied before in this area of research and certainly not in the context of our model. First, we extend the fixed effects approach, by using the Driscoll and Kraay’s (1998) estimator. Second, in order to obtain consistent estimates in a dynamic panel with substantial heterogeneity across regions, we use both the Pesaran and Smith’s (1995) Mean Group Estimators (MG) and Pesaran et al.’s (1999) Pooled Mean Group estimator (PMG). Third, since estimators assuming cross-sectional independence across regions could be inefficient, we extend the previous heterogeneous slopes estimation procedure by implement the Common Correlated Effects (CCE) estimator, its pooled-CCE (CCEP) and mean group-CCE (CCEMG) extensions proposed by Pesaran (2006), as well as the recently developed Augmented Mean Group (AMG) estimator by Bond and Eberhardt (2009) and Eberhardt and Teal (2010) that accounts for cross-sectional dependence by means of a ‘common dynamic process’. Finally, we have generated one of the largest dataset for such a macroeconometric experiment given that previous studies have suffered from limited degrees of freedom. The remainder of the paper is organized as follows. Section 2 provides a short review of the essential literature background. Section 3 discusses the econometric model and estimation methodology. Section 4 presents the data and provides a preliminary data analysis. Section 5 reports the empirical results. Finally, in Section 6, concluding remarks are provided. 2 LITERATURE BACKGROUND Medoff (1983) is the pioneering work on the effects of the sectoral shifts hypothesis (SSH) on unemployment at a regional level following the seminal paper of Lilien (1982). The former study sheds light upon the differences in labor market imbalance in the North-East-Atlantic U.S. states and South-West-Pacific U.S. states. Using a battery of alternative definitions of the Beveridge curve, the relationship between unemployment and vacancies, and of reduced form equations, Medoff (1983) could relate the outward shift of the U.S. Beveridge curve(s) between the preand post1973 periods and the variable labor conditions across areas in the same period. The emerging evidence in favor of labor market imbalance across time and regions corroborates the possibility that much of that period unemployment was structural. The analytical framework of the paper reflects the state of the art at the time of its writing and has been inevitably superseded by subsequent developments. Neumann and Topel (1984;1991) brings the analysis in the modern era. Their paper studies the 1 determinants of geographic unemployment differentials in the United States for the period 1948-1981. Neumann and Topel (1991) examine an ‘islands model’ featuring independent labor markets characterized by specific industries and labor force. They wish to test whether demand uncertainty and diversification are important determinants of equilibrium unemployment differentials among labor markets. Using pooled time-series-cross-section regional data they propose unemployment as a function of a period effect component common to all markets and three regressors: an estimate of the covariance structure of local labor demands, an index of local sensitivity to industry specific oscillations and a market-specific index of structural change in the sectoral distribution of employment. Neumann and Topel (1991) strategy is to construct a measure of sectoral shocks which can be separated in permanent (related to labor reallocation) and transitory (associated with local cycles and other random events) changes in the sectoral composition of demand. They run alternative specifications of their basic model pooled across states and over time for the selected period using fixed effects estimators as dictated by the prevailing state of the art. The emerging outcomes bear out that demand uncertainty and diversification are important determinants of equilibrium unemployment regional differentials and that regional differences in unemployment are quite large, and remarkably persistent over time. A third result is that permanent sectoral demand shifts are significant determinants of unemployment but their impact is modest relative to typical cyclic fluctuations in unemployment. Thus the regional analysis of Neumann and Topel (1991), contrary to Medoff (1983), cannot corroborate the relevance of sectoral shifts. Two factors may affect negatively their analysis. First, their dispersion index may belittle the role of allocative shocks (Shaw,1989). Second, as they use fixed effect estimators their outcomes could be subject to significant potential bias, occurring when parameter heterogeneity and error cross-section dependence are ignored. Subsequent sectoral shifts analyses using panel data techniques have mostly focused on the sectoral and not the regional dimension, are more concerned with wage differentials than unemployment and have employed fixed effects estimators (Shaw,1989;Keane,1991;Keane and Prasad,1996).2 In the light of the more recent developments in panel data econometrics (dynamics, heterogeneity and cross-sectional dependence) it appears that these previous results need to be extended and revised. To the best of our knowledge only De Serres et al. (2002) attempt an updating of panel data methodology in this area. They study the share of wages in total income for five European countries and the US. Part of the analysis uses pooled time-series-cross-section data and takes into account heterogeneity by means of the Pesaran et al.’s (1999) Pooled Mean Group estimator. The authors conclude that the persistent trend decline in the aggregate real wage share in the 1980’s and 1990’s is not necessarily only the effect of a pervasive wage moderation but could also reflect changes in sectoral composition. However, the methodological improvement of De Serres et al. (2002) is limited and their focus is only on European wages, ignoring US unemployment behavior which has been the focus of most literature. As we have stressed above, all of the existing contributions have not or could not properly handle panels which are long and wide as well as they did not take into account the potential interdependence of the individual units. It is the purpose of this paper to remedy this state of the art by looking at both a more complete dataset on the one hand and accounting for cross-sectional dependence on the other. 2 3 ECONOMETRIC MODEL AND METHODOLOGY Using pooled time-series-cross-section data on state unemployment and employment for the United States, we estimate Lilien’s dynamic reduced equation of the form: Ui,t=µi+φiUi,t−1+βiσi,t+λ0 izt+"i,t, (1) where Ui,tis the unemployment rate for state iat time t;σi,tis a measure of employment cross-sectoral dispersion; the vector ztrepresents a vector of control variables that capture aggregate demand shocks, common to all states, which in our specification is a 2 ×1 vector that includes the measures of expected, ∆Log(Mt), and unexpected, Ht, money growth. Finally, µistands for a set of state-specific fixed effects capturing the influence of unobserved state-specific heterogeneity and "i,tis the error term. Following Lilien (1982), the dispersion proxy for each state iat time tis calculated as the weighted standard deviation of the cross-sectoral employment growth rates using a K-sectoral decomposition as follows: σi,t=  K X j=1 nj,i,t Ni,t (∆ln nj,i,t−∆ln Ni,t)2  1/2 , (2) where nj,i,tis employment in sector jfor state iat time t,Ni,t=PK j=1nj,i,tis aggregate employment at time tfor state i,Kis the number of sectors (with j=1, 2, . . ., Ksectors) in the state iand the term nj,i,t Ni,t is the sector jshare of total regional employment at time t. Because of the problem of ‘observation equivalence’ embedded in the Lilien’s σi,tmeasure (Lilien, 1982;Abraham and Katz,1986; for a full discussion of the issue see Gallipoli and Pelloni,2008), we filter out aggregate effects from the dispersion proxy (σi,t) by decomposing it into an idiosyncratic component and a component measuring the response to aggregate shock. To obtain the ’purged’ measure, we have regressed σi,ton the vector of aggregate variables ˜ zt:3 σi,t=αi+ q X j=0 ϕj˜ zt−j+ui,t. (3) The estimated residual b ui,tfrom Equation 3 stands as the ‘purged’ component of σi,t. This ‘purged’ dispersion index, measuring only the reallocation shocks, is then used in the reduced form unemployment Equation 1. We have included expected and unexpected money growth to capture the potential money surprises of segmented markets models (Lucas,1990;Fuerst,1992).4In our analysis, the measure of unanticipated money growth has been generated by estimating a GARCH (1,1) model for ∆Log(Mt)and interpreting the estimated conditional variance as a parametric proxy of unanticipated money growth.5 In order to analyze the effect of sectoral shifts using panel regressions for the U.S. states we need to consider the issues of dynamics, heterogeneity and cross-sectional dependence that emerges from the specification form of Equation 1. The standard empirical macroeconometric literature suggests using traditional pooled estimators 3 adopted from the microeconometric literature, such as the least square dummy variable estimator allowing for individual fixed effects. The fixed effects (FE) model allows the intercepts to be differ across regions, while all other coefficients forced to be identical, and can be estimated by OLS method using a simple transformation (within estimator). We extend the fixed effects estimator, by using the Driscoll and Kraay’s (1998) extension of nonparametric variance-covariance matrix estimation, which produces heteroskedasticity and autocorrelation consistent standard errors that are robust to the presence of general forms of spatial and cross-sectional dependence. The presence of a lagged dependent variable among the regressors (Ui,t−1) results to a biased OLS fixed effects estimator when T is fixed (Nickell,1981). Therefore, several suggestions have proposed in the literature. Kiviet (1995) proposes a bias corrected fixed effects estimator, while Arellano and Bond (1991) propose a Generalized Method of Moments (GMM) estimation procedure to deal with the issue of lagged dependent variable. Specifically, the difference GMM estimator (AB-GMM) of Arellano and Bond (1991) firstly transforms the model by first differencing to eliminate the individual effects and then uses the GMM framework of Hansen (1982). Following the work of Arellano and Bover (1995), Blundell and Bond (1998) propose an extended system estimator that applies additional moment conditions, the system GMM estimator (BB-GMM). The previous standard pooled estimators assumes slopes homogeneity across regions, and according to the work of Pesaran and Smith (1995), these estimators yield inconsistent estimates in the case of a dynamic panel data model when the slope coefficients differ across regions. Given the existing differences in labor market across the U.S. states, the homogeneity assumption is quite restrictive, and therefore the usage of pooled estimation methods may lead to substantially heterogeneity bias in the estimated parameters of Lilien’s panel version of Equation 1. One way to obtain consistent estimates in dynamic panels with considerable heterogeneity across regions is to use estimators that allows for considerable slope heterogeneity across regions. In fact, Pesaran and Smith (1995) propose the Mean Group Estimators (MG) that consists of estimating separate OLS regressions for each region and then calculating averages of the specific coefficients over groups. Furthermore, Pesaran et al. (1999) suggest an intermediate estimator that imposes long-run slope homogeneity between regions but allows for short-run parameters heterogeneity. The pooled mean group (PMG) estimator involves both pooling and averaging of the individual regression coefficients in order to obtain more efficient estimates that the MG estimators under the assumption of slopes homogeneity. Another important issue, that evolve in the regional panel sectoral shifts analysis of unemployment, is the issue of cross-sectional dependence among states. Interdependence across cross sections is a considerable characteristic in the analysis of macro and regional panel data models, and estimators based on the assumption of cross-sectional independence may prove inefficient or even inconsistent (Sarafidis and Wansbeek,2012). Therefore, we extend the heterogeneous slopes estimation procedure by implementing the Pesaran (2006) Common Correlated Effects (CCE) estimators that account for the presence of unobserved common factors by using cross-section averages of the dependent and independent variables as additional regressors. Moreover, it has been shown that the CCE estimator still provides consistent estimates of the slope coefficients and their SEs under the more general case of multifactor error structure and spatial error correlation (Pesaran and Tosetti,2011). Specifically, we consider the mean group CCE (CCEMG) extension of the estimator proposed in Pesaran and Smith (1995) as well as the pooled 4 CCE (CCEP) version that assumes slopes homogeneity while it allows for different common effects coefficients across i.6Finally, we implement the recently Augmented Mean Group (AMG) estimator proposed by Bond and Eberhardt (2009) and Eberhardt and Teal (2010) that accounts for cross-sectional dependence by means of a ‘common dynamic process’ in the regional regressions. The ‘common dynamic process’ is extracted using year dummy coefficients of a pooled regression in first differences and represents the levels-equivalent mean evolution of unobserved common factors across all regions (Eberhardt and Teal,2010).7 We, therefore, continue our analysis by estimating the Lilien’s dynamic reduced form unemployment relationship for the U.S. states panel, taking into account the issues of dynamics, heterogeneity, nonstationarity and cross-sectional dependence, by using alternative estimation approaches for homogeneous and heterogeneous panel data. 3.1 POOLABILITY TESTS An important issue for the estimation of panel data models is the assumption of common slope coefficients across regions and/or over time, i.e. that βi=βwith i=1,2,..., Nand/or βt=βwith t=1,2,..., T. Following Baltagi (2008), this can be tested by a Chow test, see Chow (1960), that is extended to the case of Nand/or Tlinear regressions. The test for the poolability of the data across regions (time) simply compares the restricted residual sum of squares (RSSr) of the fixed effects model with the unrestricted residual sum of squares (RSSu) obtained by the region-specific (time-specific) OLS regressions. Under the null hypothesis of poolability across regions the F-statistic can be defined as: F=(RSSr−RSSu)/(N−1)K RSSu/N(T−K)∼F((N−1)K,N(T−K)), (4) and under the null hypothesis of poolability over time we have: F=(RSSr−RSSu)/(T−1)K RSSu/T(N−K)∼F((T−1)K,T(N−K)). (5) 3.2 CROSS-SECTIONAL DEPENDENCE TEST In order to determine the existence of cross-sectional dependence among states, we employ the simple test suggested by Pesaran (2004). The Cross-Section Dependence test statistic is based on the average of pair-wise correlation coefficients ( ˆ ρi j) of the OLS residuals, obtained from the individual ADF regressions. The CDS test is given by: CDS =v u u t2T N(N−1)  N−1 X i=1 N X j=i+1 ˆ ρi j . (6) The CDS statistic under the null of cross-independence is distributed as a two-tailed standard normal distribution, i.e. CD ∼N(0,1)for Ti j >3 and sufficient large N.Baltagi et al. (2007) provide evidence that the CDS test can be also employed as a useful diagnostic test for various models of spatial dependence. 5 3.3 PANEL UNIT ROOT TESTS Before proceeding with the panel estimation, we need to examine the order of integration of the series under consideration. We employ the IPS panel unit root test of Im et al. (2003) as well as the CIPS panel test of Pesaran (2007) that takes into account cross sectional dependence among panel members. 3.3.1 PANEL UNIT ROOT TESTS WITHOUT CROSS SECTIONAL DEPENDENCE This panel test is an extension of the univariate ADF regression as follows: ∆yi,t=αi+φiyi,t−1+ pi X j=1 θi,j∆yi,t−j+"i,t, (7) where yi,tstands for each series under consideration for state iat time t. The null hypothesis is that all series contains a unit root, φi=0 for all i(with i=1, 2,..., N), while the alternative hypothesis assumes that some of the Npanel units are stationary with individual specific autoregressive coefficients. Im et al. (2003) propose a test based on the average of the ADF statistics computed for each individual in the panel. Specifically, the IPS statistic is defined as: ¯ tN,T=1 N N X i=1 tiT (pi,θi). (8) Under the assumption of cross-sectional independence, this statistic is shown to converge to a normal distribution. 3.3.2 PANEL UNIT ROOT TESTS WITH CROSS SECTIONAL DEPENDENCE The IPS test that is based on the restrictive assumption that the series are independent across states i, suffers from serious size distortion and restricted power in the presence of cross-sectional dependence (O’Connell,1998) and cross-sectional cointegrating relationships (Banerjee et al.,2004). In order to overcome this, Pesaran (2007) proposes a simple approach to deal with the problem of cross-sectional dependence. A one-factor model is considered with heterogeneous factor loadings for residuals and suggests to augment the standard ADF regression with the cross-section averages of lagged levels and first-differences of the individual series. The regression used for the ith cross-section unit is defined as: ∆yi,t=αi+φiyi,t−1+ci¯ yt−1+ pi X j=0 θi,j∆¯ yt−j+ pi X j=1 θi,j∆yi,t−j+"i,t, (9) where ¯ yt−1=N−1PN i=1yi,t−1and ∆¯ yt=N−1PN i=1yi,t=¯ yt−¯ yt−1. The CIPS test is based on the average of individual cross-sectionally augmented ADF statistics (CADF) as follows: C I PS =1 N N X i=1 ti(N,T). (10) Simulated critical values of CIPS are listed in Pesaran (2007). Baltagi et al. (2007) show that the CIPS test is found to be robust to the presence of other sources of cross-sectional dependence such as the spatial form. 6 nomics Series Working Papers 515, University of Oxford, Department of Economics. (p. 1,5,5) Fuerst, T. S. (1992), “Liquidity, Loanable Funds, and Real Activity,” Journal of Monetary Economics, 29, 3–24. (p. 3) Gallipoli, G. and G. Pelloni (2008), “Aggregate Shocks vs Reallocation Shocks: an Appraisal of the Applied Literature,” Working Paper Series 27-08, Rimini Centre for Economic Analysis. (p. 3,11,11) Hansen, L. P. (1982), “Large Sample Properties of Generalized Method of Moments Estimators,” Econometrica, 50, 1029–54. (p. 4) Im, K. S., M. H. Pesaran, and Y. Shin (2003), “Testing for Unit Roots in Heterogeneous Panels,” Journal of Econometrics, 115, 53–74. (p. 6,6,8) Keane, M. P. (1991), “Sectoral Shift Theories of Unemployment: Evidence from Panel Data,” Institute for empirical macroeconomics discussion paper, no. 28, Federal Reserve Bank of Minneapolis. (p. 2) Keane, M. P. and E. S. Prasad (1996), “The Employment and Wage Effects of Oil Price Changes: A Sectoral Analysis,” The Review of Economics and Statistics, 78, 389–400. (p. 2) Kiviet, J. F. (1995), “On Bias, Inconsistency, and Efficiency of Various Estimators in Dynamic Panel Data Models,” Journal of Econometrics, 68, 53–78. (p. 4) Lilien, D. M. (1982), “Sectoral Shifts and Cyclical Unemployment,” Journal of Political Economy, 90, 777–793. (p. 1,3,3,3,3,4,5,8,8,8,8,9,10,10) Long, J. B. and C. I. Plosser (1987), “Sectoral vs. Aggregate Shocks in the Business Cycle,” American Economic Review, 77, 333–36. (p. 11) Lucas, R. E. (1972), “Expectations and the Neutrality of Money,” Journal of Economic Theory, 4, 103–124. (p. 11) Lucas, R. E. (1973), “Wage Inflation and the Structure of Regional Unemployment: Comment,” Journal of Money, Credit and Banking, 5, 382–84. (p. 11) Lucas, R. E. (1990), “Liquidity and Interest Rates,” Journal of Economic Theory, 50, 237–264. (p. 3) Medoff, J. L. (1983), “U.S. Labor Markets: Imbalance, Wage Growth, and Productivity in the 1970s,” Brookings Papers on Economic Activity, 14, 87–128. (p. 1,1,1,2) Neumann, G. and R. Topel (1984), “Employment Risk, Sectoral Shifts and the Geographic Distribution of Unemployment,” Discussion Paper 84-3, Economic Research Center/NORC. (p. 1) Neumann, G. R. and R. H. Topel (1991), “Employment Risk, Diversification, and Unemployment,” The Quarterly Journal of Economics, 106, 1341–65. (p. 1,1,2,2,2) Nickell, S. J. (1981), “Biases in Dynamic Models with Fixed Effects,” Econometrica, 49, 1417–26. (p. 4) O’Connell, P. G. J. (1998), “The Overvaluation of Purchasing Power Parity,” Journal of International Economics, 44, 1–19. (p. 6) Pelloni, G. and W. Polasek (1999), “Intersectoral Labour Reallocation and Employment Volatility: A Bayesian Analysis using a VAR-GARCH-M model,” Discussion Papers 99/4, Department of Economics, University of York. (p. 11) Pelloni, G. and W. Polasek (2003), “Macroeconomic Effects of Sectoral Shocks in Germany, The U.K. and, The U.S. A VAR-GARCH-M Approach,” Computational Economics, 21, 65–85. (p. 7,11) Pesaran, M. H. (2004), “General Diagnostic Tests for Cross Section Dependence in Panels,” IZA Discussion Papers 1240, Institute for the Study of Labor (IZA). (p. 5) 13 Pesaran, M. H. (2006), “Estimation and Inference in Large Heterogeneous Panels with a Multifactor Error Structure,” Econometrica, 74, 967–1012. (p. 1,4,9,9,19,19,19,19) Pesaran, M. H. (2007), “A Simple Panel Unit Root Test in the Presence of Cross-Section Dependence,” Journal of Applied Econometrics, 22, 265–312. (p. 6,6,6,8) Pesaran, M. H., Y. Shin, and R. P. Smith (1999), “Pooled Mean Group Estimation of Dynamic Heterogeneous Panels,” Journal of the American Statistical Association, 94, 621–634. (p. 1,1,2,2,4,9,19, 19) Pesaran, M. H. and R. Smith (1995), “Estimating Long-Run Relationships from Dynamic Heterogeneous Panels,” Journal of Econometrics, 68, 79–113. (p. 1,1,4,4,4,9,19,19) Pesaran, M. H. and E. Tosetti (2011), “Large Panels with Common Factors and Spatial Correlation,” Journal of Econometrics, 161, 182–202. (p. 4) Sarafidis, V. and T. Wansbeek (2012), “Cross-Sectional Dependence in Panel Data Analysis,” Econometric Reviews, 31, 483–531. (p. 4) Shaw, K. (1989), “Wage Variability in the 1970s: Sectoral Shifts or Cyclical Sensitivity,” The Review of Economics and Statistics, 71, 26–36. (p. 2,2) Wallis, K. F. (1987), “Time Series Analysis of Bounded Economic Variables,” Journal of Time Series Analysis, 8, 115–123. (p. 7) 14 TABLES AND FIGURES PART A: DESCRIPTIVE STATISTICS Table 1: U.S. States and Abbreviations State Abbrev. State Abbrev. Alabama AL Nebraska NE Arizona AZ Nevada NV Arkansas AR New Hampshire NH California CA New Jersey NJ Colorado CO New Mexico NM Connecticut CT New York NY Delaware DE North Carolina NC Florida FL North Dakota ND Georgia GA Ohio OH Idaho ID Oklahoma OK Illinois IL Oregon OR Indiana IN Pennsylvania PA Iowa IA Rhode Island RI Kansas KS South Carolina SC Kentucky KY South Dakota SD Louisiana LA Tennessee TN Maine ME Texas TX Maryland MD Utah UT Massachusetts MA Vermont VT Michigan MI Virginia VA Minnesota MN Washington WA Mississippi MS West Virginia WV Missouri MO Wisconsin WI Montana MT Wyoming WY N=48 T=264 (1990m01 – 2011m12) Obs =12672 15 Table 2: Summary Statistics Mean Std. Dev. Maximum Minimum Skewness Kurtosis Jarque-Bera Prob PANEL (A): Sectoral Variables Sectoral Employment Shares∗ Total 7.410 0.977 9.632 5.274 -0.092 2.348 242.085 0.000 Mining - Logging - Construction 4.521 0.930 6.884 2.484 -0.071 2.664 70.346 0.000 Manufacturing 5.240 1.153 7.593 2.153 -0.499 2.655 589.474 0.000 Trade - Transportations 5.794 0.971 7.986 3.706 -0.089 2.335 250.428 0.000 Information 3.534 1.098 6.395 1.237 0.066 2.422 185.847 0.000 Financial activities 4.531 1.046 6.841 2.056 -0.069 2.503 140.343 0.000 Professional activities 5.121 1.156 7.728 2.138 -0.227 2.524 228.895 0.000 Education - Health 5.299 1.006 7.526 2.528 -0.044 2.504 133.786 0.000 Leisure - Hospitality 5.048 0.946 7.367 3.136 -0.004 2.365 212.668 0.000 Other services 4.149 0.999 6.248 1.874 -0.098 2.218 342.756 0.000 Government 5.646 0.922 7.836 3.741 -0.069 2.465 161.372 0.000 PANEL (B): Macro Variables ULogistic i,t-2.890 0.346 -1.799 -3.842 0.218 2.952 101.900 0.000 ULogarithmic i,t-2.948 0.327 -1.952 -3.863 0.162 2.903 60.216 0.000 σ13 i,t0.006 0.004 0.085 0.001 5.492 69.419 2.38E+06 0.000 σ9 i,t0.005 0.003 0.044 0.000 3.328 26.559 3.15E+05 0.000 σ4 i,t0.003 0.002 0.043 0.000 4.326 43.993 9.23E+05 0.000 σ9pur ged i,t-7.29E-19 0.003 0.039 -0.005 3.388 27.301 3.34E+05 0.000 ∆Log(Mt)0.004 0.009 0.060 0.003 1.938 13.821 1447.777 0.000 Ht1.18E-04 0.000 0.003 0.000 6.127 47.751 2.35E+04 0.000 Notes:∗indicates variables in logarithms. Table 3: Lilien Index (σ9 i,t) AL AZ AR CA CO CT DE FL GA ID IL IN IA KS KY LA ME MD MA MI MN MS MO MT 1990 0.0049 0.0055 0.0060 0.0034 0.0041 0.0061 0.0134 0.0044 0.0046 0.0083 0.0040 0.0051 0.0041 0.0060 0.0041 0.0050 0.0094 0.0065 0.0057 0.0060 0.0036 0.0068 0.0049 0.0091 2000 0.0037 0.0045 0.0042 0.0031 0.0037 0.0031 0.0098 0.0043 0.0052 0.0058 0.0030 0.0039 0.0050 0.0051 0.0050 0.0043 0.0091 0.0058 0.0055 0.0045 0.0035 0.0051 0.0046 0.0053 2011 0.0055 0.0047 0.0065 0.0024 0.0045 0.0047 0.0082 0.0032 0.0047 0.0066 0.0027 0.0042 0.0043 0.0060 0.0041 0.0058 0.0059 0.0058 0.0052 0.0053 0.0056 0.0046 0.0046 0.0070 1990-2011 0.0038 0.0047 0.0042 0.0035 0.0040 0.0044 0.0099 0.0039 0.0040 0.0061 0.0035 0.0040 0.0042 0.0050 0.0043 0.0051 0.0056 0.0041 0.0037 0.0049 0.0038 0.0054 0.0042 0.0062 NE NV NH NJ NM NY NC ND OH OK OR PA RI SC SD TN TX UT VT VA WA WV WI WY 1990 0.0055 0.0075 0.0081 0.0053 0.0053 0.0041 0.0045 0.0054 0.0040 0.0055 0.0059 0.0035 0.0084 0.0055 0.0068 0.0054 0.0036 0.0043 0.0087 0.0046 0.0047 0.0071 0.0035 0.0085 2000 0.0050 0.0062 0.0046 0.0045 0.0047 0.0053 0.0036 0.0062 0.0034 0.0039 0.0034 0.0043 0.0090 0.0046 0.0069 0.0059 0.0023 0.0039 0.0106 0.0052 0.0066 0.0111 0.0034 0.0069 2011 0.0050 0.0054 0.0064 0.0048 0.0072 0.0050 0.0035 0.0072 0.0032 0.0050 0.0041 0.0029 0.0080 0.0042 0.0067 0.0047 0.0029 0.0051 0.0066 0.0040 0.0044 0.0061 0.0051 0.0074 1990-2011 0.0046 0.0062 0.0054 0.0038 0.0056 0.0033 0.0041 0.0061 0.0033 0.0044 0.0039 0.0031 0.0067 0.0045 0.0058 0.0052 0.0027 0.0046 0.0066 0.0037 0.0044 0.0060 0.0036 0.0068 Notes: See Table 1 for U.S. States Abbreviations. 16 Table 4: U.S. Regional Employment Structure AL AZ AR CA CO CT DE FL GA ID IL IN IA KS KY LA ME MD MA MI MN MS MO MT 1990 (%) Mining - Log - Constr 6.38 6.53 4.93 5.45 5.29 3.92 6.23 7.73 5.50 6.70 4.59 4.92 3.82 4.80 6.74 9.54 6.05 7.38 3.41 3.93 4.05 5.08 4.31 6.25 Manufacturing 22.20 11.90 23.70 15.70 11.20 18.60 13.20 9.46 17.50 13.60 17.30 24.00 17.90 16.30 18.30 11.10 17.40 9.14 16.10 21.30 16.00 24.50 16.70 6.57 Trade - Transportations 19.40 20.90 21.10 19.40 20.00 19.60 19.40 22.70 22.30 22.20 21.70 21.10 22.00 21.90 20.70 22.40 21.00 20.40 19.10 19.60 21.00 20.00 21.50 23.70 Information 1.83 2.20 1.92 3.14 3.43 2.60 1.44 2.38 2.85 1.97 2.48 1.78 2.44 2.79 1.81 1.69 1.91 2.18 2.93 1.77 2.54 1.39 2.85 2.16 Financial activities 4.90 6.65 4.34 6.55 6.88 9.47 10.60 7.17 5.32 4.98 7.08 5.17 5.67 5.48 4.35 5.47 4.96 6.32 6.73 4.88 6.05 4.40 5.95 4.65 Professional activities 7.20 9.63 6.22 12.00 11.80 10.40 14.50 5.80 9.50 8.50 10.80 6.20 5.63 7.70 6.70 7.30 6.29 11.90 11.50 10.10 10.00 5.48 9.30 5.24 Education - Health 7.90 9.12 9.88 8.97 8.50 12.20 9.50 11.30 8.20 8.40 10.20 9.90 12.00 9.60 11.50 10.30 12.40 10.40 15.40 10.40 11.30 8.00 10.90 11.80 Leisure - Hospitality 6.51 10.70 7.04 8.84 11.00 6.66 8.00 13.20 8.05 9.53 7.45 8.25 8.27 8.03 8.19 8.00 9.00 8.51 7.94 8.38 8.45 6.00 8.80 11.80 Other services 3.74 3.89 3.61 3.34 3.68 3.54 3.39 4.30 3.06 2.99 3.92 3.84 4.53 3.76 4.22 3.65 3.23 4.43 3.26 3.65 4.27 3.36 3.90 3.89 Government 20.00 18.40 17.20 16.60 18.20 13.00 13.80 15.80 17.80 21.10 14.50 14.90 17.90 19.60 17.50 20.50 17.90 19.40 13.70 16.00 16.30 21.70 15.80 24.00 Total (000s) 1635.7 1483.1 923.8 12500.2 1520.8 1619.6 347.5 5373.3 2991.8 385.3 5287.6 2522.0 1226.3 1091.9 1487.2 1587.6 535.1 2173.3 2988.2 3946.6 2135.9 936.5 2345.1 297.2 NE NV NH NJ NM NY NC ND OH OK OR PA RI SC SD TN TX UT VT VA WA WV WI WY 1990 (%) Mining - Log - Constr 3.98 10.20 4.59 4.16 8.04 4.02 5.59 5.20 4.31 6.68 5.34 4.91 4.13 6.58 4.87 4.61 7.16 5.02 6.05 6.87 6.03 9.87 4.01 14.40 Manufacturing 13.30 3.90 19.40 14.60 6.55 12.00 26.40 5.85 21.70 13.20 16.30 18.40 21.00 22.60 11.60 22.80 13.30 14.30 16.60 13.40 15.70 13.00 22.80 4.60 Trade - Transportations 23.00 17.90 21.50 22.90 20.30 19.10 20.30 24.30 19.70 20.90 21.10 20.20 17.10 19.70 22.30 22.00 22.20 21.20 19.80 19.90 20.20 21.50 20.00 21.00 Information 3.01 1.77 2.07 3.31 1.84 3.48 1.66 2.25 2.07 1.93 2.15 2.11 2.22 1.50 1.90 2.04 2.49 1.85 2.13 2.56 2.31 1.90 1.94 1.85 Financial activities 6.67 5.11 6.49 6.44 4.76 9.49 4.51 4.79 5.22 5.71 5.79 6.28 5.90 4.62 5.92 5.17 6.46 4.81 5.24 4.97 5.28 4.15 5.41 3.99 Professional activities 8.40 9.50 7.20 12.10 9.80 10.50 7.70 4.50 9.30 8.20 8.20 8.80 9.40 8.60 3.86 7.60 9.00 10.30 5.50 12.00 9.20 5.19 6.70 4.47 Education - Health 10.20 5.63 12.10 10.00 8.80 13.10 7.40 13.50 11.10 9.80 10.30 14.30 15.20 6.40 12.80 9.60 9.50 9.30 13.50 8.30 9.60 11.10 10.40 6.50 Leisure - Hospitality 8.19 31.00 8.90 7.42 10.40 6.99 7.70 9.19 8.19 8.02 8.70 7.42 7.90 8.60 10.10 8.10 8.36 8.74 11.20 8.08 8.90 7.59 8.69 12.00 Other services 3.52 2.70 3.48 3.29 3.76 3.39 3.25 4.88 3.64 3.55 3.68 4.06 3.54 3.08 4.82 2.12 3.70 3.53 3.07 4.03 4.18 5.42 5.09 3.37 Government 19.60 12.20 14.30 15.90 25.80 17.90 15.50 25.50 14.80 22.10 18.50 13.60 13.80 18.30 21.90 16.00 17.80 20.80 16.90 20.00 18.60 20.20 15.00 27.80 Total (000s) 730.8 620.9 508.2 3635.5 580.4 8214.6 3125.5 265.8 4882.3 1183.9 1255.6 5173.0 454.1 1541.9 288.5 2196.0 7101.0 723.6 257.7 2894.3 2142.9 630.0 2291.4 198.5 AL AZ AR CA CO CT DE FL GA ID IL IN IA KS KY LA ME MD MA MI MN MS MO MT 2011 (%) Mining - Log - Constr 4.89 5.12 4.98 4.14 6.20 3.19 4.69 4.54 3.98 5.56 3.62 4.49 4.36 4.62 5.04 9.19 4.60 5.74 3.39 3.35 3.64 5.31 4.02 7.22 Manufacturing 12.70 6.20 13.60 8.90 5.70 10.30 6.10 4.28 9.00 9.10 10.10 16.40 14.00 12.00 11.90 7.30 8.50 4.44 7.90 12.90 11.20 12.30 9.30 3.94 Trade - Transportations 19.50 19.60 20.30 18.90 17.80 18.10 18.00 20.60 21.20 20.10 20.10 19.40 20.40 19.00 20.30 19.60 19.80 17.30 17.10 18.30 18.50 19.70 19.30 20.20 Information 1.25 1.52 1.26 3.07 3.17 1.94 1.38 1.85 2.51 1.55 1.77 1.21 1.89 2.08 1.49 1.25 1.38 1.66 2.58 1.36 1.97 1.09 2.20 1.70 Financial activities 4.96 6.90 4.15 5.42 6.37 8.31 10.20 6.64 5.38 4.84 6.39 4.64 6.80 5.47 4.71 4.98 5.34 5.60 6.41 4.91 6.51 4.11 6.01 4.88 Professional activities 11.40 14.30 10.20 15.10 15.10 12.00 13.30 14.50 14.10 12.30 14.60 10.10 8.37 11.20 10.50 10.30 9.61 15.60 14.80 14.00 12.30 8.62 12.50 9.56 Education - Health 11.60 14.80 14.40 13.00 12.10 19.40 16.00 15.20 12.90 14.30 15.00 15.10 14.60 13.80 14.30 14.80 20.30 16.10 20.90 15.80 17.50 12.40 15.70 15.00 Leisure - Hospitality 9.04 10.80 8.53 10.90 12.00 8.40 10.20 13.10 9.78 9.67 9.16 9.83 8.84 8.64 9.46 10.40 10.00 9.14 9.63 9.54 8.78 10.80 10.20 13.00 Other services 4.28 3.72 3.74 3.46 4.14 3.72 4.67 4.22 3.95 3.46 4.38 3.86 3.82 3.91 3.92 3.31 3.38 4.54 3.78 4.25 4.31 3.12 4.24 3.87 Government 20.40 17.10 18.90 17.10 17.40 14.70 15.40 15.10 17.20 19.20 14.80 15.00 16.90 19.30 18.40 18.90 17.10 19.90 13.50 15.70 15.20 22.60 16.50 20.60 Total (000s) 1866.6 2405.4 1160.0 14060.6 2255.3 1623.5 417.3 7271.5 3879.9 606.8 5663.1 2830.4 1477.8 1335.9 1790.4 1905.7 593.4 2548.0 3210.6 3935.6 2675.7 1090.0 2650.4 427.3 NE NV NH NJ NM NY NC ND OH OK OR PA RI SC SD TN TX UT VT VA WA WV WI WY 2011 (%) Mining - Log - Constr 4.35 5.90 3.64 3.39 7.94 3.59 4.62 10.20 3.65 7.69 4.68 4.49 3.44 4.38 5.13 4.12 7.53 6.37 4.81 5.15 5.08 8.82 3.41 16.80 Manufacturing 9.90 3.38 10.60 6.60 3.66 5.30 11.10 6.00 12.60 8.30 10.30 9.90 8.80 11.80 9.60 11.50 7.90 9.40 10.30 6.20 9.50 6.60 16.20 3.15 Trade - Transportations 20.70 18.90 21.30 21.20 16.50 17.10 18.50 21.80 18.80 18.10 19.30 19.20 15.90 19.10 20.20 21.00 19.90 19.30 18.80 17.10 18.60 17.90 18.60 18.40 Information 1.79 1.12 1.78 1.92 1.77 2.94 1.75 1.81 1.50 1.56 1.99 1.59 2.21 1.41 1.57 1.66 1.85 2.44 1.67 2.01 3.68 1.39 1.69 1.35 Financial activities 7.36 4.62 5.54 6.51 4.10 7.87 5.17 5.35 5.47 5.13 5.68 5.44 6.68 5.24 6.88 5.10 6.05 5.71 4.04 4.96 4.87 3.62 5.77 3.73 Professional activities 11.00 12.40 10.50 15.60 12.40 13.10 13.00 7.66 12.70 11.20 11.50 12.50 11.80 12.40 7.02 12.10 12.70 13.20 8.31 18.00 12.00 8.27 10.30 6.13 Education - Health 14.50 9.22 18.20 15.80 15.30 19.90 13.90 14.30 16.70 13.30 14.50 20.30 22.50 11.80 16.10 14.40 13.50 13.20 20.10 12.70 13.50 16.30 15.00 9.36 Leisure - Hospitality 8.71 28.20 10.10 8.68 10.50 8.82 10.20 8.86 9.40 9.17 10.20 8.96 10.70 11.40 10.40 10.00 9.88 9.41 10.90 9.45 9.64 9.67 9.10 11.30 Other services 3.88 2.96 3.63 4.24 3.41 4.26 3.97 3.95 4.13 3.86 3.56 4.47 4.79 3.76 3.83 3.82 3.50 2.81 3.28 5.07 3.80 7.27 4.95 4.14 Government 17.80 13.30 14.80 16.10 24.40 17.10 17.80 20.10 15.10 21.70 18.30 13.10 13.20 18.60 19.20 16.30 17.20 18.20 17.70 19.30 19.20 20.10 15.10 25.70 Total (000s) 944.0 1125.0 626.4 3856.1 804.1 8683.3 3922.4 394.2 5083.0 1550.3 1618.1 5687.0 460.2 1832.1 406.2 2656.2 10557.1 1208.0 299.6 3680.3 2820.4 753.8 2740.8 285.7 AL AZ AR CA CO CT DE FL GA ID IL IN IA KS KY LA ME MD MA MI MN MS MO MT 1990:01-2011:12 (%) Mining - Log - Constr -1.49 -1.41 0.05 -1.31 0.91 -0.73 -1.54 -3.19 -1.52 -1.14 -0.97 -0.43 0.54 -0.18 -1.70 -0.35 -1.45 -1.64 -0.02 -0.58 -0.41 0.23 -0.29 0.97 Manufacturing -9.50 -5.70 -10.10 -6.80 -5.50 -8.30 -7.10 -5.18 -8.50 -4.50 -7.20 -7.60 -3.90 -4.30 -6.40 -3.80 -8.90 -4.70 -8.20 -8.40 -4.80 -12.20 -7.40 -2.63 Trade - Transportations 0.10 -1.30 -0.80 -0.50 -2.20 -1.50 -1.40 -2.10 -1.10 -2.10 -1.60 -1.70 -1.60 -2.90 -0.40 -2.80 -1.20 -3.10 -2.00 -1.30 -2.50 -0.30 -2.20 -3.50 Information -0.58 -0.68 -0.66 -0.07 -0.26 -0.66 -0.06 -0.53 -0.34 -0.42 -0.71 -0.57 -0.55 -0.71 -0.32 -0.44 -0.53 -0.52 -0.35 -0.41 -0.57 -0.30 -0.65 -0.46 Financial activities 0.06 0.25 -0.19 -1.13 -0.51 -1.16 -0.40 -0.53 0.06 -0.14 -0.69 -0.53 1.13 -0.01 0.36 -0.49 0.38 -0.72 -0.32 0.03 0.46 -0.29 0.06 0.23 Professional activities 4.20 4.67 3.98 3.10 3.30 1.60 -1.20 8.70 4.60 3.80 3.80 3.90 2.74 3.50 3.80 3.00 3.32 3.70 3.30 3.90 2.30 3.14 3.20 4.32 Education - Health 3.70 5.68 4.52 4.03 3.60 7.20 6.50 3.90 4.70 5.90 4.80 5.20 2.60 4.20 2.80 4.50 7.90 5.70 5.50 5.40 6.20 4.40 4.80 3.20 Leisure - Hospitality 2.53 0.10 1.49 2.06 1.00 1.74 2.20 -0.10 1.73 0.14 1.71 1.58 0.57 0.61 1.27 2.40 1.00 0.63 1.69 1.16 0.33 4.80 1.40 1.20 Other services 0.54 -0.17 0.13 0.12 0.46 0.18 1.28 -0.08 0.89 0.47 0.46 0.02 -0.71 0.15 -0.30 -0.34 0.15 0.11 0.52 0.60 0.04 -0.24 0.34 -0.02 Government 0.40 -1.30 1.70 0.50 -0.80 1.70 1.60 -0.70 -0.60 -1.90 0.30 0.10 -1.00 -0.30 0.90 -1.60 -0.80 0.50 -0.20 -0.30 -1.10 0.90 0.70 -3.40 Total (000s) 230.8 922.3 236.2 1560.4 734.4 3.9 69.8 1898.2 888.1 221.5 375.5 308.3 251.5 244.1 303.2 318.1 58.2 374.6 222.5 -10.9 539.8 153.5 305.2 130.0 NE NV NH NJ NM NY NC ND OH OK OR PA RI SC SD TN TX UT VT VA WA WV WI WY 1990:01-2011:12 (%) Mining - Log - Constr 0.37 -4.30 -0.95 -0.77 -0.10 -0.43 -0.97 5.00 -0.66 1.01 -0.66 -0.42 -0.69 -2.20 0.26 -0.49 0.37 1.35 -1.24 -1.72 -0.95 -1.05 -0.60 2.40 Manufacturing -3.40 -0.52 -8.80 -8.00 -2.89 -6.70 -15.30 0.15 -9.10 -4.90 -6.00 -8.50 -12.20 -10.80 -2.00 -11.30 -5.40 -4.90 -6.30 -7.20 -6.20 -6.40 -6.60 -1.45 Trade - Transportations -2.30 1.00 -0.20 -1.70 -3.80 -2.00 -1.80 -2.50 -0.90 -2.80 -1.80 -1.00 -1.20 -0.60 -2.10 -1.00 -2.30 -1.90 -1.00 -2.80 -1.60 -3.60 -1.40 -2.60 Information -1.22 -0.65 -0.29 -1.39 -0.07 -0.54 0.09 -0.44 -0.57 -0.37 -0.16 -0.52 -0.01 -0.09 -0.33 -0.38 -0.64 0.59 -0.46 -0.55 1.37 -0.51 -0.25 -0.50 Financial activities 0.69 -0.49 -0.95 0.07 -0.66 -1.62 0.66 0.56 0.25 -0.58 -0.11 -0.84 0.78 0.62 0.96 -0.07 -0.41 0.90 -1.20 -0.01 -0.41 -0.53 0.36 -0.26 Professional activities 2.60 2.90 3.30 3.50 2.60 2.60 5.30 3.16 3.40 3.00 3.30 3.70 2.40 3.80 3.16 4.50 3.70 2.90 2.81 6.00 2.80 3.08 3.60 1.66 Education - Health 4.30 3.59 6.10 5.80 6.50 6.80 6.50 0.80 5.60 3.50 4.20 6.00 7.30 5.40 3.30 4.80 4.00 3.90 6.60 4.40 3.90 5.20 4.60 2.86 Leisure - Hospitality 0.52 -2.80 1.20 1.26 0.10 1.83 2.50 -0.33 1.21 1.15 1.50 1.54 2.80 2.80 0.30 1.90 1.52 0.67 -0.30 1.37 0.74 2.08 0.41 -0.70 Other services 0.36 0.26 0.15 0.95 -0.35 0.87 0.72 -0.93 0.49 0.31 -0.12 0.41 1.25 0.68 -0.99 1.70 -0.20 -0.72 0.21 1.04 -0.38 1.85 -0.14 0.77 Government -1.80 1.10 0.50 0.20 -1.40 -0.80 2.30 -5.40 0.30 -0.40 -0.20 -0.50 -0.60 0.30 -2.70 0.30 -0.60 -2.60 0.80 -0.70 0.60 -0.10 0.10 -2.10 Total (000s) 213.1 504.1 118.2 220.6 223.7 468.8 796.9 128.4 200.7 366.4 362.5 514.1 6.1 290.2 117.7 460.2 3456.1 484.5 41.8 786.0 677.5 123.9 449.4 87.2 Notes: See Table 1 for U.S. States Abbreviations. 17 PART B: DATA PROPERTIES Table 5: Panel Unit Root Tests ULogistic i,tULogarithmic i,tσi,tσpur ged i,t Statistic Prob Statistic Prob Statistic Prob Statistic Prob IPS -3.654∗0.000 -3.423∗0.000 -52.395∗0.000 -53.100∗0.000 CIPS -2.767∗0.003 -2.709∗0.003 -32.150∗0.000 -32.012∗0.000 Notes: ∗ indicates rejection of the null hypothesis at 5% significance level. The 5% critical value for the IPS statistics is -1.645 and the 5% critical value for the CIPS statistics is -2.12. Table 6: Cross-Sectional Dependence Tests ULogistic i,tULogarithmic i,tσi,tσpur ged i,t CD-test 127.32∗126.07∗152.29∗140.85∗ P-value 0.000 0.000 0.000 0.000 LM-test 406.71∗398.91∗538.87∗465.68∗ P-value 0.000 0.000 0.000 0.000 Abs(Corr) 0.237 0.235 0.281 0.260 Notes: ∗ indicates rejection of the null hypothesis at 5% significance level. Table 7: Poolability Tests ULogistic i,tULogarithmic i,t Regions Time Regions Time Chow test 3.530∗5.591∗3.446∗5.397∗ P-value 0.000 0.000 0.000 0.000 Notes: ∗ indicates rejection of the null hypothesis at 5% significance level. 18 PART C: ESTIMATION TABLES Table 8: Assumptions about Panel Estimators Parameter Heterogeneity Homogeneity Heterogeneity Cross-Sectional Correlation Independence POLS, FE, AB GMM, BB GMM, PMG MG Dependence FE-DK, CCEP CCEMG, AMG Notes: POLS – Pooled OLS, FE – Fixed Effects, AB GMM – Arellano and Bond (1991), BB GMM – Blundell and Bond (1998), PMG – Pesaran et al.’s (1999) Pooled Mean Group, MG – Pesaran and Smith’s (1995) Mean Group, FE-DK – Driscoll and Kraay’s (1998) Fixed Effects, CCEP – Pesaran’s (2006) Pooled Common Correlated Effects, CCEMG – Pesaran’s (2006) Mean Group Common Correlated Effects, AMG – Bond and Eberhardt’s (2009) Augmented Mean Group. Table 9: Lilien’s Index and Sectoral Shifts: Aggregate Estimates ULogistic tULogarithmic t OLS GMM OLS GMM OLS GMM OLS GMM OLS GMM OLS GMM OLS GMM OLS GMM ULogistic t−10.987∗∗ 0.989∗∗ 0.986∗∗ 0.987∗∗ 0.987∗∗ 0.989∗∗ 0.986∗∗ 0.987∗∗ (162.04) (159.29) (148.42) (144.52) (162.04) (159.29) (148.42) (144.52) ULogarithmic t−10.987∗∗ 0.989∗∗ 0.986∗∗ 0.987∗∗ 0.987∗∗ 0.989∗∗ 0.986∗∗ 0.987∗∗ (161.66) (159.41) (148.61) (145.19) (161.66) (159.41) (148.61) (145.19) σ9 t6.296∗∗ 5.889∗∗ 5.817∗∗ 5.507∗∗ (2.68) (2.72) (2.65) (2.72) σ13 t3.915∗∗ 2.794∗3.648∗∗ 2.685∗ (2.14) (1.73) (2.15) (1.78) σ9pur ged t6.296∗∗ 5.889∗∗ 5.817∗∗ 5.507∗∗ (2.68) (2.72) (2.65) (2.72) σ13 pur ged t3.915∗∗ 2.794∗3.648∗∗ 2.685∗ (2.14) (1.73) (2.15) (1.78) ∆Log(Mt)0.654 0.509 0.766∗0.709∗0.843∗∗ 0.685∗0.870∗∗ 0.783∗0.616 0.466 0.718∗0.649 0.790∗∗ 0.631 0.814∗∗ 0.720∗ (1.56) (1.22) (1.82) (1.68) (2.10) (1.68) (2.12) (1.88) (1.56) (1.19) (1.81) (1.63) (2.09) (1.64) (2.11) (1.83) Ht108.632∗∗ 97.610∗∗ 124.876∗∗ 118.708∗∗ 151.694∗∗ 137.887∗∗ 153.013∗∗ 138.789∗∗ 101.985∗∗ 91.359∗∗ 116.757∗∗ 110.605∗∗ 141.766∗∗ 129.024∗∗ 142.974∗∗ 129.904∗∗ (2.67) (2.43) (2.94) (2.65) (4.45) (3.84) (4.07) (3.29) (2.68) (2.45) (2.96) (2.67) (4.46) (3.89) (4.10) (3.35) Notes: t-statistics in parentheses. All estimations were carried out using Newey-West HAC robust standard errors. ∗ and ∗∗ denotes significance at the 10% and 5% significance levels, respectively. 19 Table 10: Lilien’s Index and Sectoral Shifts: Pooled & Heterogeneous Parameter Estimates ULogistic i,tULogarithmic i,t POLS FE DK FE AB GMM BB GMM PMG MG CCEP CCEMG AMG POLS FE DK FE AB GMM BB GMM PMG MG CCEP CCEMG AMG ULogistic i,t−10.993∗∗ 0.988∗∗ 0.988∗∗ 0.990∗∗ 0.990∗∗ 0.987∗∗ 0.989∗∗ 0.986∗∗ 0.983∗∗ 0.968∗∗ (984.34) (817.13) (208.88) (540.08) (534.57) (655.80) (1159.56) (344.70) (369.07) (190.84) ULogarithmic i,t−10.993∗∗ 0.988∗∗ 0.988∗∗ 0.990∗∗ 0.990∗∗ 0.987∗∗ 0.989∗∗ 0.986∗∗ 0.983∗∗ 0.968∗∗ (994.27) (829.04) (208.34) (549.13) (540.20) (673.34) (1160.11) (354.52) (374.14) (191.14) σi,t0.630∗∗ 0.846∗∗ 0.846∗∗ 0.694∗∗ 1.022∗∗ 1.055∗∗ 0.859∗∗ 0.504 0.435∗∗ 0.902∗∗ 0.584∗∗ 0.787∗∗ 0.787∗∗ 0.649∗∗ 0.952∗∗ 0.982∗∗ 0.805∗∗ 0.467 0.404∗∗ 0.843∗∗ (2.40) (2.71) (2.21) (3.08) (5.00) (5.18) (5.59) (1.39) (2.28) (5.75) (2.41) (2.72) (2.21) (3.12) (5.02) (5.21) (5.61) (1.39) (2.28) (5.72) ∆Log(Mt)0.414∗∗ 0.459∗∗ 0.459∗∗ 0.475∗∗ 0.594∗∗ 0.459∗∗ 0.436∗∗ 0.344∗∗ 0.392∗∗ 0.434∗∗ 0.434∗∗ 0.450∗∗ 0.561∗∗ 0.433∗∗ 0.411∗∗ 0.323∗∗ (8.77) (9.53) (2.54) (9.27) (11.00) (9.67) (14.97) (11.76) (8.74) (9.49) (2.56) (9.26) (10.93) (9.62) (15.07) (11.85) Ht24.415∗∗ 25.102∗∗ 25.102∗∗ 26.996∗∗ 28.643∗∗ 24.742∗∗ 23.678∗∗ 21.320∗∗ 22.907∗∗ 23.561∗∗ 23.561∗∗ 25.321∗∗ 26.862∗∗ 23.222∗∗ 22.127∗∗ 19.850∗∗ (17.95) (17.66) (2.60) (18.27) (19.73) (17.16) (17.23) (12.64) (17.76) (17.47) (2.62) (18.10) (19.59) (16.98) (17.27) (12.59) Notes: t-statistics in parentheses. All estimations were carried out using White heteroskedasticity robust standard errors. ∗ and ∗∗ denotes significance at the 10% and 5% significance levels, respectively. Table 11: Purged Lilien’s Index and Sectoral Shifts: Pooled & Heterogeneous Parameter Estimates ULogistic i,tULogarithmic i,t POLS FE DK FE AB GMM BB GMM PMG MG CCEP CCEMG AMG POLS FE DK FE AB GMM BB GMM PMG MG CCEP CCEMG AMG ULogistic i,t−10.993∗∗ 0.988∗∗ 0.988∗∗ 0.990∗∗ 0.990∗∗ 0.987∗∗ 0.989∗∗ 0.986∗∗ 0.983∗∗ 0.968∗∗ (984.34) (817.13) (208.88) (540.08) (534.57) (655.80) (1159.56) (336.77) (369.07) (190.84) ULogarithmic i,t−10.993∗∗ 0.988∗∗ 0.988∗∗ 0.990∗∗ 0.990∗∗ 0.987∗∗ 0.989∗∗ 0.986∗∗ 0.983∗∗ 0.968∗∗ (994.27) (829.04) (208.34) (549.13) (540.20) (673.34) (1160.11) (346.28) (374.14) (191.14) σpur ged i,t0.630∗∗ 0.846∗∗ 0.846∗∗ 0.694∗∗ 1.022∗∗ 1.055∗∗ 0.859∗∗ 0.493 0.435∗∗ 0.902∗∗ 0.584∗∗ 0.787∗∗ 0.787∗∗ 0.649∗∗ 0.952∗∗ 0.982∗∗ 0.805∗∗ 0.456 0.404∗∗ 0.843∗∗ (2.40) (2.71) (2.21) (3.08) (5.00) (5.18) (5.59) (1.34) (2.28) (5.75) (2.41) (2.72) (2.21) (3.12) (5.02) (5.21) (5.61) (1.34) (2.28) (5.72) ∆Log(Mt)0.416∗∗ 0.461∗∗ 0.461∗∗ 0.477∗∗ 0.596∗∗ 0.461∗∗ 0.439∗∗ 0.346∗∗ 0.393∗∗ 0.436∗∗ 0.436∗∗ 0.451∗∗ 0.562∗∗ 0.435∗∗ 0.414∗∗ 0.325∗∗ (8.79) (9.55) (2.55) (9.31) (11.04) (9.71) (14.94) (11.82) (8.75) (9.51) (2.57) (9.30) (10.97) (9.65) (15.04) (11.90) Ht25.246∗∗ 26.220∗∗ 26.220∗∗ 27.912∗∗ 29.992∗∗ 26.135∗∗ 25.014∗∗ 22.433∗∗ 23.679∗∗ 24.599∗∗ 24.599∗∗ 26.178∗∗ 28.119∗∗ 24.518∗∗ 23.343∗∗ 20.920∗∗ (19.01) (18.94) (2.69) (19.07) (20.40) (18.85) (19.21) (13.23) (18.82) (18.73) (2.71) (18.89) (20.25) (18.64) (19.39) (13.32) Notes: t-statistics in parentheses. All estimations were carried out using White heteroskedasticity robust standard errors. ∗ and ∗∗ denotes significance at the 10% and 5% significance levels, respectively. Table 12: Purged Index and Sectoral Shifts: Alternative Sectoral Decomposition ULogistic i,tULogarithmic i,t σ13 pur ged i,tσ10 pur ged i,tσ9pur ged i,tσ7pur ged i,tσ4pur ged i,tσ13 pur ged i,tσ10 pur ged i,tσ9pur ged i,tσ7pur ged i,tσ4pur ged i,t DK FE CCEMG DK FE CCEMG DK FE CCEMG DK FE CCEMG DK FE CCEMG DK FE CCEMG DK FE CCEMG DK FE CCEMG DK FE CCEMG DK FE CCEMG ULogistic i,t−10.988∗∗ 0.983∗∗ 0.988∗∗ 0.983∗∗ 0.988∗∗ 0.983∗∗ 0.988∗∗ 0.983∗∗ 0.988∗∗ 0.982∗∗ (207.69) (348.69) (208.22) (354.00) (208.88) (369.07) (209.21) (370.71) (211.26) (348.82) ULogarithmic i,t−10.988∗∗ 0.983∗∗ 0.988∗∗ 0.983∗∗ 0.988∗∗ 0.983∗∗ 0.988∗∗ 0.983∗∗ 0.988∗∗ 0.982∗∗ (207.21) (353.83) (207.74) (359.07) (208.34) (374.14) (208.64) (375.53) (210.60) (354.42) σpur ged i,t0.484∗0.354∗∗ 0.703∗∗ 0.381∗∗ 0.846∗∗ 0.435∗∗ 0.884∗∗ 0.429∗∗ 1.277∗∗ 0.654∗∗ 0.450∗0.328∗∗ 0.655∗∗ 0.352∗∗ 0.787∗∗ 0.404∗∗ 0.821∗∗ 0.399∗∗ 1.189∗∗ 0.606∗∗ (1.75) (2.01) (1.97) (2.15) (2.21) (2.28) (2.79) (2.62) (3.41) (2.24) (1.75) (2.01) (1.98) (2.15) (2.21) (2.28) (2.80) (2.61) (3.42) (2.25) Notes: t-statistics in parentheses. All estimations were carried out using White heteroskedasticity robust standard errors. ∗ and ∗∗ denotes significance at the 10% and 5% significance levels, respectively. 20 PART D: FIGURES (a) σ13 tand σ9 t(b) σ9 tand σ9 t pur ged Figure 1: Unemployment Rate and Lilien’s σtfor the U.S. Agregate, 1990:M1–2011:M12. (a) σ13 i,tand σ9 i,t(b) σ9 i,tand σ9 i,t pur ged Figure 2: Unemployment rate and Lilien’s σi,tfor the 48 U.S. States Average, 1990:M1–2011:M12. 21 Figure 3: Unemployment rate and Lilien’s σ13 i,tand σ9 i,tfor the 48 U.S. States, 1990:M1–2011:M12. 22