Technology Shocks and Labor: An Analysis Using Medium-Run Identification

Diploma Thesis by Almuth Balleer
March 18, 2004

Abstract:

The question how technology shocks affect aggregated hours worked is widely discussed in the literature and delivers a touchstone for Real-Business-Cycle theory. Initiated by Galí (1999), the impact and dynamic propagation of single shocks on macroeconomic variables are of interest. In this context, the response of labor to technology shocks needs special treatment, since it has not uniquely been determined so far. Usually, “long-run identificiation” is applied to disentangle the disturbances in the macroeconomic system, meaning that technology shocks are restricted to be the only influence of productivity in the long run.

While there exists a large body of research already, this thesis considers two recent approaches that shed some new light on the debate. Uhlig (2003a) introduces medium-run identification as an alternative to long-run identification. He states that there may exist more sources in the variation of productivity than technology only and allows for labor hoarding in his model. Fisher (2002) assumes investmentspecific technology as a complement to the neutral technology analyzed by Galí. He identifies the investment-specific shock as the only influence on the investment price, while both kinds of technology are the only driving forces of labor productivity in the long run.

In this thesis, I first derive an identification method that decomposes the forecast revision variance using Cholesky as proposed by Uhlig (2003a). I then solve the model introduced by Fisher as well as an extension with labor hoarding. Next, I derive identifying restrictions for the shocks from the solution of the model using forecast revision variances and test them on artificial data simulated from the model specification. Finally, I identify technology shocks in real data. My results can be summarized as follows: The theoretical impulse responses from the model solution are qualitatively matched by Fisher’s empirical ones. The response of labor after an innovation in investment-specific technology is understated in the model. The identification strategy proposed by Fisher is valid for the original model and the labor hoarding model, if there are no other permanent shocks influencing productivity. I show that medium-run identification could be used instead and derive an alternative identification strategy that identifies neutral technology as the only influence on productivity and both technology shocks as the only driving force of labor in the medium run. I show that this strategy works equivalently well.

In real data, my Fisher identification delivers responses of productivity and labor that fall after an innovation in investment-specific technology, which strongly contradicts Fisher’s own findings. This is probably due to a different data measure and should be handled with care; however, it sheds light on the specification’s sensitivity to the investment price series employed and indicates the desirability of an alternative specification. I find that the alternative specification in fact identifies the same shocks as Fisher. However, the response of labor is negative on impact and in the long run after an innovation in neutral technology. While this contradicts RBC theory, one may argue that, due to the strong rise of labor after investment-specific technology, which is quite robust across specifications, technology as a combined measure of neutral and investment-specific components induces hours to rise. This defends technology-driven business cycle theory.