How Large is Large? Preliminary and relative guidelines for interpreting partial correlations in economics

An essential part of empirical economics research is the identification of the size of an empirical effect. Partial correlations offer a convenient statistically based measure of the strength of an economic relationship. A key question arises in their interpretation: When is a partial correlation large? This paper draws upon the observed distribution of 22,000 partial correlations from a diverse group of economics fields. The median absolute partial correlation from these fields is 0.173, which under Cohen’s (1988) conventional guidelines for zero order correlations is a small to moderate effect. The paper develops new guidelines for key qualitative categories (small, medium and large). According to the new guidelines, partial correlations that are larger than ± 0.33 can be deemed to be large. This is considerably different to Cohen’s guideline of ±0.50 for zero order correlations. Researchers and meta-analysts should exercise caution when applying Cohen’s guidelines to describe the importance of partial correlations in economics.