قیمت های مهم، دستمزد منصفانه، و همکاری جنبش های بیکاری و رشد بهره وری نیروی کار

This paper studies the co-movements of unemployment and labor productivity growth for the U.S. economy. Measures of co-movements in the frequency domain indicate that co-movements between variables differ strongly according to the frequency. First, long-term and business cycle co-movements are larger than short-term co-movements. Second, co-movements are negative in the short and long run, but positive over the business cycle. A New Keynesian model that combines nominal rigidity on the goods market (sticky prices) and real rigidity on the labor market (fair wages) is shown to be quantitatively consistent with the observed co-movements both in the long term and over the business cycle. However, the model fails to explain the short-term co-movements.

The long-standing debate still goes in macroeconomics over the relationship between unemployment and labor productivity growth. This paper studies the importance of periodicity for a better understanding of the relationship. I analyze the behavior of unemployment and labor productivity growth in the short and long run and argue that such an approach is necessary to reconcile the (apparently) contradictory views on this relationship. The first view is that productivity growth increases unemployment. It has been put forward for the U.S. economy by Blanchard (1989) and Blanchard and Quah (1989) who show that technological shocks first increase unemployment and by Evans (1989) who finds that shocks that instantaneously increase unemployment have a positive long-term effect on output.1Gali (1999) demonstrates that this result2 is still valid for total hours worked (which decrease after a positive technological shock). This evidence supports the New Keynesian's view of fluctuations and contradicts the real business cycle approach of fluctuations which was first proposed by Kydland and Prescott (1982) and Long and Plosser (1983), and later applied to unemployment dynamics by Hansen (1985) among others. The second view is that productivity growth decreases unemployment. It aimed to account for the ‘roaring nineties’ experienced by the U.S. economy. Ball and Moffitt (2002) and Staiger et al. (2002) explain the exceptionally low unemployment rate of the 1990s with the equally exceptional productivity gains over the same period. The study of steady state properties in the tradition of Pissarides (2000) and Aghion and Howitt (1994) also provided evidence to support this view. Whereas the relation between growth and unemployment at the steady state is theoretically indeterminate,3 the empirical studies of Hoon and Phelps (1997), Blanchard and Wolfers (2000), and Vallanti (2004) suggest that permanent growth increase pulls down unemployment.4 The first view induces positive co-movements, whereas the second induces negative co-movements. This paper argues that both views are relevant: positive and negative co-movements can coexist, because they are associated with cycles of different periodicities. To make the distinction between the different periodicities, from the short to the long run, the co-movements are studied in the frequency domain by means of spectral analysis. Spectral analysis has become very popular in macroeconomics for describing the dynamic properties of time series as well as the co-movements between series (see, for example, Watson, 1993 and Diebold et al., 1998, and Wen, 1998).5 The usefulness of spectral analysis is twofold for the purpose of this study. First, it gives an overall view of the co-movements of unemployment and labor productivity growth, which can be appraised according to different frequencies. Second, one can evaluate models on their ability to reproduce empirical co-movements at different frequencies. The distinction can then be drawn between the short run – which corresponds to the highest frequencies – , the long run – which corresponds to the lowest frequencies – , and the business cycle – which corresponds to the medium frequencies. I begin by describing the empirical co-movements of unemployment and labor productivity growth for the U.S. economy. Measures of co-movements in the frequency domain indicate that these two variables are closely related and that the sign of their co-movements differs strongly according to the frequency studied: they co-move negatively in the short and long run yet positively over the business cycle. In addition, co-movements are concentrated in the low and medium frequencies rather than in the high frequencies. Interestingly, the same pattern (with the opposite sign of course) is obtained for co-movements between labor productivity growth and hours worked instead of unemployment. The use of spectral analysis therefore provides an overall view of the co-movements of unemployment and labor productivity growth. I then turn to theory to attempt to explain these empirical facts. I study a New Keynesian explanation of these facts based on the interaction of nominal and real rigidities. Nominal goods prices move sluggishly because firms face a quadratic cost of price adjustment, as originally suggested by Rotemberg (1982).6 As emphasized by Blanchard (1989) and Blanchard and Quah (1989), and more recently by Basu et al. (2004) and Gali (1999), nominal rigidities constitute a relevant mechanism to explain the short-term negative effects of technological improvements on the use of labor. Real rigidity on the labor market comes from the fair-wage hypothesis as originally suggested by Akerlof (1982) and recently extended by Collard and de la Croix (2000) and de la Croix et al. (2000). The extended fair-wage model leads to a persistence in the workers’ wage aspiration that positively links a current increase in productivity with future employment increases. The general idea is close to the proposition of Blanchard and Katz (1999) that introduces past wages in the reservation wage to explain wage dynamics. The two types of rigidity have recently been combined by Ball and Moffitt (2002) (who provide an estimation of the Phillips Curve) and by Danthine and Kurmann (2004) (who give results for the business cycle). The model is estimated to reproduce the empirical spectra of labor productivity growth and unemployment. Next comes the assessment of the model's ability to replicate the empirical co-movements between the variables. When a positive productivity shock hits the economy, the model predicts that an unemployment increase is followed by an unemployment decrease. Hence, the co-movements are positive at high and medium frequencies and then become negative at low frequencies. This theoretical timing of events appears consistent with the empirical relationship between labor productivity and unemployment in the long run and over the business cycle, but not in the short run. This conclusion is confirmed by the study of two models. A model with fair wages and flexible prices can only account for the long-term co-movements whereas a model with sticky prices and indivisible labor supply can only account for the business cycle co-movements. The remainder of the paper is organized as follows. Empirical facts are described in Section 2. The model is exposed in Section 3. Results are presented in Section 4. Section 5 concludes.

The co-movements of unemployment and labor productivity growth for the U.S. economy were studied by means of spectral analysis. Co-movements are positive over the business cycle and negative in the short and long run. This led to a theoretical explanation based on New Keynesian mechanisms: fair wages for the labor market and sticky prices for the goods market. The combination of these two rigidities provides a satisfactory explanation for the empirical co-movements in the long run and over the business cycle, but not in the short run. Due to real rigidity on the labor market, the rates of unemployment and of labor productivity growth co-move negatively in the long run, yet because of nominal rigidity on the goods market, they co-move positively over the business cycle. In a certain sense, the articulation between real and nominal rigidities is in line with Blanchard and Quah's (1989) initial recommendations: Nominal rigidities can explain why in response to a positive supply shock, say an increase in productivity, aggregate demand does not initially increase enough to match the increase in output needed to maintain constant unemployment; real wage rigidities can explain why increases in productivity can lead to a decline in unemployment after a few quarters which persists until real wages have caught up with the new higher level of productivity. (p. 663). Finally, I would like to emphasize the absence of search and matching frictions in the model although they are very popular in macroeconomics, especially when studying the relations between labor market and technological progress. To explain negative co-movements between productivity and unemployment, the consequence of workers’ aspirations in the fair-wage model has been given prominence over the capitalization effect that operates in the matching model of unemployment with growth. As discussed by Pissarides (2000), this effect requires strong assumptions on preferences (namely linear utility with consumption), which seem to contradict the standard macroeconomic models. To explain the positive co-movements between productivity and unemployment, the consequence of nominal rigidities on demand has been favored instead of the reallocation process associated with technological diffusion studied by Aghion and Howitt (1994). In that sense, this paper falls in a long tradition, resuscitated by the New Keynesian models that Blanchard (1989) applied to unemployment and Gali (1999) to total hours worked.