Innovation and Productivity in Services: Evidence from Chile

Abstract

We analyze empirically the firm-level relationship between innovation and productivity in the Chilean service sector using the manufacturing sector as a benchmark. We find that manufacturing and service industries have similar determinants of the probability of introducing technological innovations. We also find a positive effect of technological and nontechnological innovation on labor productivity for both sectors. However, there are some differences in the quantitative importance of some determinants of innovation. Our findings help to characterize the different stages of the service industry’s innovative process and its effect on an emerging economy, providing useful information for policy design.

KEY WORDS:

• innovation
• productivity
• services

Acknowledgments

We thank attendees of the seminars at INTELIS at the University of Chile, MEIDE 2013, and LALICS 2013 for valuable comments and suggestions, and Matias Caamaño and Rolando Campusano for their invaluable research assistance

Notes

1. According to World Bank World Development Indicators (data.worldbank.org), services as a percentage of GDP represent today close to 60 percent of value added in Latin America, reaching close to 70 percent in large countries such as Brazil.

2. The fact that National Institute of Statistics provides a disaggregation for the service sector only at the one-digit level makes it impossible to not consider a subsector such as real estate, which is not considered a part of KIBS.

3. In the setup of our database, we also drop plants that do not have data for our relevant estimations. In particular, some data points that do not declare sales or employment are dropped from each survey.

4. A more detailed analysis within KIBS indicates that this is particularly driven by the large expenditure of subsector K (real estate and business activities), which has more than double the expenditure than manufacturing as a percentage of sales.

5. Within KIBS, the subsector K has the highest percentage of firms spending in R&D (16.8 percent).

6. We define innovation intensity as innovation investment per employee.

7. This model may be estimated using alternative econometric techniques such as asymptotic least squares, as actually done in the original article by Crépon et al. (1998). However, recent works on this issue tend to prefer the less computationally intensive techniques of estimation (Griffith at al. 2006; Hall et al. 2008).

8. The question of product or process innovation is actually asked in more detail, requiring firms to declare whether these innovations were just technological improvements of were new to the firm or the market as a whole. Our dummy takes a value of one if the firm answered positively to any type of innovation and novelty of it.

9. All regressions incorporate sector fixed effects at the one-digit level.