Comment

Abstract

The main idea of this (very nice) paper is that taxes can be a long-run driver of labor productivity. This raises an intriguing hypothesis: could it be that a recent structural vector autogression (SVAR) based empirical literature that finds that positive innovations in labor productivity have a negative effect on hours workedmay be actually picking up the effects of taxes rather than technology per se? A first issue concerns the question,which taxes? InMertens andRavn’s paper, it is general income taxes. But an immediate implication of a standard neoclassical growth model is that it is only capital income taxes that affect labor productivity in the long run. Hence in principle we should measure capital income taxes as distinct from labor income taxes. In the following I use data from Chen, İmrohoroğlu, and İmrohoroğlu (2009), who extend McGrattan and Prescott’s (2009) measures of capital and labor taxes. Figure 1 shows that the long-run behavior of these two types of taxes has been in fact radically different. Even within capital income taxes, however, and as emphasized by McGrattan and Prescott (2009), it is important to distinguish between three types: (i) taxes on corporate income, (ii) taxes on corporate distributions, and (iii) investment tax credits. McGrattan and Prescott show that, in the 1990s, it is mostly the second category of taxes that has undergone major variations: on average, from 41.1% to 17.4%. Measuring capital and labor tax shocks. To the extent that capital and/or labor income tax changes are believed to be permanent, their first differences should follow random walk processes of the form Δτk;t 1⁄4 εn;t; Δτn;t 1⁄4 εk;t;