Indicators of Product Sophistication and Factor Intensities: Measurement Matters
What is it about?
The widely used PRODY indicator by Hausmann et al. [What you Export Matters; Journal of Economic Growth: 12, 1–25] allows the calculation of product sophistication and factor intensities at high lev- els of disaggregation. In this paper I deal with its measurement, its conceptional limitations, its sensitivity to subjective decision making, its alternatives, and how to reap the best possible benefit from its usage. I introduce a theory-based alternative and investigate the sen- sitivity of empirical results with respect to different measurements. In particular, I re-estimate the main results of Hausmann et al. with 75 variants of measurement. Small changes to the sample, the aggrega- tion level, or the construction of the indicator can make a difference. Moreover, I offer two STATA programs that ease the calculation of various PRODY variants.
Why is it important?
In this paper I emphasize the importance of measurement for indication of product sophistication and factor intensities in general, and for its usage in growth regressions in particular. My critical review of existing indicators should raise awareness about the sensitivity of results and should motivate and guide future researchers to calculate and choose the indicator properly. Beyond that, I introduce the CPRODY indicator. It has some advantages, because it controls for distorting trade effects that may bias the weighting scheme which is used in every product sophistication and factor intensity indicator. I invite researchers to use my Stata commands in order to conduct more sensitivity analysis with respect to their product sophistication and factor intensity measurement, respectively. My software applications reduces the technical barriers in calculation and speeds up the calculation in general. Finally, I hope that this paper motivates “economic research workers to take the trouble to ferret out the necessary information to understand in detail the particular measurement characteristics of the data they use”[Renfro, 2009, p.7]