Productivity Differences and Technology

Let us first start with a brief overview of productivity and technology differences within countries. This overview will help us place the cross-country differences in productivity and technology into perspective.

that productivity and technology differences are ubiquitous even across firms within narrow sectors in the same country.

18.1.1. Productivity and Technology Differences within Narrow Sectors. A large literature uses longitudinal micro-data (often for the manufacturing sector) to study labor and total factor productivity differences across plants within narrow sectors (for example three-digit or four-digit manufacturing sectors). For our focus, the most important pattern that emerges from these studies is that, even within a narrow sector of the US economy, there is a significant amount of productivity differences across plants, with an approximately two or three-fold difference between the top and the bottom of the distribution (see, for example the survey in Bartelsman and Doms, 2000, for a summary of various studies and estimates). In addition, these productivity differences appear to be highly persistent (e.g., Bailey, Hulten and Campbell, 1992).

There is little consensus on what the causes of these differences are. Many studies find a correlation between plant productivity and plant or firm size, various measures of tech­nology (in particular IT technology), capital intensity, the skill level of the workforce and management practices (e.g., Davis and Haltiwanger, 1991, Doms, Dunn and Troske, 1997, Black and Lynch, 2004). Nevertheless, since all of these features are choice variables for firms, these correlations cannot be taken to be causal. Thus to a large extent the determinants of productivity differences across plants are still unknown. In this light, it should not appear as a surprise that there is no consensus on the the determinants of cross-country differences in productivity.

Nevertheless, technology differences appear to be an important factor, at least as an approximate cause, for productivity differences. For example, Doms, Dunn and Troske (1997) and Haltiwanger, Lane and Spletzer (1999) document significant technology differences across plants within narrow sectors. Interestingly, as emphasized by Doms, Dunn and Troske (1997) and Caselli and Coleman (2001), a key determinant of the technology-level or the technology adoption decision seems to be the skill level of the workforce of the plant (often proxied by the share of non-production workers), though adoption of new technology does not typically lead to a significant change in employment structure. These results suggest that, consistent with some of the models discussed in Chapters 10 and 15, differences in the availability of skills and skilled workers might be an important determinant of technology adoption (and development) decisions. We will return to the role of skills in productivity and technology differences in the cross-country context below.

The distribution of productivity differences across firms appears to be related to the entry of new and more productive plants (and the exit of less productive plants). For example, consistent with the basic Schumpeterian models of economic growth discussed in Chapter 702

14, Bartelsman and Doms (2000) and Foster, Haltiwanger and Krizan (2000) document that entry of new plants plays has an important contribution to industry productivity growth. Nevertheless, entry and exit appear to account for only about 25% of average TFP growth, with the remaining productivity improvements are accounted for by continuing plants. This suggests that models in which firms continually invest in technology and productivity (for example such as the model of step-by-step innovation in Section 14.4 in Chapter 14) may be important for understanding the productivity differences across firms and plants and also for the study of cross-country productivity differences.

18.1.2.

The classic paper in this area is Griliches’s (1957) study of the adoption of hybrid corn in the US. Griliches showed that the more productive hybrid corn diffused only slowly in the US agriculture and that this diffusion was affected by the local economic conditions of different areas. Consistent with the theoretical models presented so far, his study found evidence that the likelihood of adoption was related to the contribution of the hybrid corn in a particular area, the market size and the skill level of the area. The importance of these factors has been found in other studies as well (see, for example, Mansfield, 1998). Another important result of Griliches’s study was to uncover the famous S-shape of diffusion, whereby a particular technology first spreads slowly and then once it reaches a critical level of adoption, it starts spreading much more rapidly. Finally, once a large fraction of the target population adopts the technology, the rate of adoption again declines. The overall pattern thus approximates an S curve or a logistic function. Jovanovic and Lach (1989), among others, show how this type of diffusion process can emerge as an equilibrium of an industry model with knowledge spillovers.

The important lesson for our focus here is that productivity and technology differences are not only present across countries, but also within countries. Moreover, even within countries better technologies do not immediately get adopted by all firms. In light of these patterns, the presence of significant productivity and technology differences across countries should not be 703

entirely surprising. Nevertheless, the causes of within-country and cross-country productiv­ity and technology differences might be different, and despite the presence of within-country differences, the significant cross-country differences do pose a puzzle that requires investi­gation. For example, within-country productivity differences might be due to differences in managerial (entrepreneurial) ability or related to the success of the match between the man­ager and the technology (or the product). These types of explanations would be unlikely to account for why almost all firms in many less developed countries are much less productive than the typical firms in the United States or other advanced economies. Motivated by the evidence briefly surveyed here, I will discuss both models in which technology diffuses slowly across countries and also models in which productivity differences may remain even when instantaneous technology diffusion and adoption are possible.

18.2.