Elsevier

Economics Letters

Volume 123, Issue 2, May 2014, Pages 109-112
Economics Letters

Sluggish US employment recovery after the Great Recession: Cyclical or structural factors?

https://doi.org/10.1016/j.econlet.2014.01.032Get rights and content

Abstract

This paper aims at investigating the causes of the observed departure of employment path from the GDP movements occurred in US in the late of 2008 onwards. Starting from a production function approach, and assuming that the TFP growth is explained by variables linked to the business cycle, we are able to formulate an extended version of Okun’s law based on cyclical factors. Out-of-sample forecasting for the period 2008 onward shows that predicted US employment is on average 1.7% above the observed one, meaning that this gap cannot be attributed to identified cyclical factors.

Introduction

One of the macroeconomics “core ideas” that had been empirically confirmed before the recent crisis is Okun’s law (Blinder, 1997). This simple rule-of-thumb suggested a relation between unemployment (or employment) and GDP (Okun, 1962). Economic predictors and policy makers frequently used this rule to assess the evolution of employment using GDP forecasts. From the late of 2008, however, the path of employment exhibits important departures from the GDP movements (Daly and Hobijn, 2010, Elsby et al., 2010, Bernanke, 2012). Fig. 1 shows the growth rates of US employment against GDP over the period 1970Q1–2013Q2. From 2008Q4 employment growth is below the regression line (i.e. employment is overpredicted) and continued on into the recovery period.

The empirical employment–GDP discrepancy raises an important question about the causes, as highlighted by Bernanke (2012): Is the low level of employment primarily the result of pronounced dip in cyclical factors (i.e. shift in aggregate demand and/or cyclical variables such as wages, prices, capacity utilization, etc.)? Or is instead the result of structural factors (i.e. mismatch between supply and demand for jobs caused by the presence of unskilled workforce, offshoring of manufacturing, low labor mobility, etc.)? Understanding the sources of this divergence raises important policy implications. If shifts in economic conditions predominate, then fiscal and monetary policies should be effective in supporting the employment recovery; if the causes are structural, then other policy measures are needed, such as programs to retrain workers or to promote their mobility towards areas where jobs exist.

This paper aims at investigating the causes of the employment–GDP empirical failure in the aftermath of the 2008’s crisis going deeper into the characteristics of Okun’s relation. We start from a classical log-levels version of the production function and evaluate the impact of GDP and TFP on employment. The idea is that through the well-known “capitalization effect” (i.e. technological progress favors employment by generating opportunities for profits) the TFP growth increases jobs creation, as demonstrated by Pissarides and Vallanti (2007) for US labor market.

We assume that the movement of TFP is captured by trade openness and the ratio of wages to machinery prices. In general, trade openness favors the circulation of ideas, knowledge and technologies, affecting in this way the TFP. Since openness may present non-monotonic (U-shaped) impact on growth (see  Rivera-Batiz and Romer, 1991, Baldwin and Forslid, 1999), we introduce a quadratic term for openness into the regression as in Edwards (1998). The wages expressed in terms of machinery prices (i.e. the factors price ratio) affect the pace of mechanization and technical progress; when labor is more expensive than capital, there is an incentive to introduce a technological innovation (Acemoglu, 2003).

Since all the variables (i.e. employment, GDP, trade openness and wages to machinery prices ratio) are non-stationary in log-levels, we estimate a long-run relationship using the Dynamic OLS (DOLS) approach. The ECM version of this model is then compared to other specifications of Okun’s law considered in the empirical literature. Our model performs better compared to alternative specifications of Okun’s law in explaining the sluggish pattern of employment after the 2008. Since our model is based exclusively on variables linked to the business cycle, the misprediction of the employment (1.7% on average) may be attributable to structural factors.

The paper is organized as follows. Section  2 presents our model specification, Section  3 shows the estimation results and the comparison of our model with other alternative versions in terms of the employment explanation after 2008. Section  4 concludes.

Section snippets

Model specification

We start by considering the following production function: Yt=AtKtαLtβ where A is the stock of knowledge, Y is the output, K is the capital stock and L is the number of employers.

We assume that At=A0egt, where A0 is the initial Total Factor Productivity (TFP) and gt is the growth rate of A. We can specify the evolution of A by making g a function of trade openness (open, the sum of imports and exports as a percentage of GDP) and wages to machinery prices ratio (wm, wages as a percentage of

Estimation results and model performance

To estimate the model in Eqs. (6)–(7), we use the data (covering the period 1970Q1–2013Q2) taken from Federal Reserve Economic Data (FRED). All data expressed in log-levels (emp and y) and as a ratio (open, open2 and wm) are non-stationary over the period under investigation1 and for this reason we study the presence of a long-run relation. We first estimate the long-run relationship of Eq. (6) using DOLS. This estimator deals with the

Conclusions

We estimate an extended version of Okun’s law starting from a classical production function. TFP is assumed to be explained by trade openness and wages to machinery prices ratio. Using the Dynamic OLS (DOLS) approach, the ECM version of our model is then compared to other specifications of Okun’s law considered in the empirical literature. Our model performs better in explaining the sluggish pattern of employment after the 2008. Since our model is based exclusively on variables linked to the

Acknowledgment

The authors acknowledge financial support under the project MISURA, funded by the Italian MIUR.

References (11)

There are more references available in the full text version of this article.

Cited by (6)

  • GDP-employment decoupling in Germany

    2020, Structural Change and Economic Dynamics
    Citation Excerpt :

    Third, a decoupling of employment from GDP growth implies a relevant consequence: productivity slows down (Fig. 1). While output per worker unexpectedly rose during the Great Recession in the U.S. (Lucchetta and Paradiso, 2014), it fell and hardly recovered in Europe, e.g. in Great Britain or Germany. The data do not fit the idea of a clear positive correlation between the growth rates of real output and productivity per employee (Verdoorn, 1949) which has also been analysed as an outcome of the Great Recession (Blundell et al., 2014, Ball, 2014, Barnett et al., 2014, Pessoa and van Reenen, 2014).

  • Detecting unemployment hysteresis: A simultaneous unobserved components model with Markov switching

    2016, Economics Letters
    Citation Excerpt :

    What changed with the Great Recession? We build our contribution on several discrete pieces of work, on the NAIRU and unit roots (Canarella et al., 2013; Leon-Ledesma and McAdam, 2004), correlated UC models (Morley et al., 2003), identification of the causality structure between trend and cycle shocks (Weber, 2011), asymmetry of unemployment with respect to Okun´s law (e.g. Lucchetta and Paradiso, 2014; Owyang and Sekhposyan, 2012), and regime switching UC models (Morley and Piger, 2012; Sinclair, 2010). We use monthly seasonally adjusted unemployment rates from 1960:1 to 2015:6 (in line with German data availability).

View full text