Is Newer Better? Penn World Table Revisions and the Growth Literature

Versions 6.1 and 6.2 of Penn World Tables (PWT) have essentially the same methodology and the same underlying data, but report significantly different growth numbers. It is not the case that 6.2 is better or worse than 6.1; it is inherent in the PWT approach that some data will vary a great deal across versions and there is no way to determine which of the existing versions is “best”. We examine 13 leading studies of growth for “Table Invariant” results that are robust across different versions of PWT. Table Invariant results are common in studies examining cross-sectional or very long run data; in contrast, results based on higher frequency data are less likely to be robust in this sense, and annual data are particularly problematic. This lack of robustness can be attributed in part to poor data quality in lower income countries, which has been flagged in the Tables’ own health warnings. The lack of robustness is also due to the PWT’s methodology for constructing and compiling the data. This methodology renders data for countries with relatively small total GDP and data distant from the current benchmark year especially variable. We propose an alternative way to use PWT data that might be less prone to some of these issues.