Publication bias in FDI spillovers in developing countries: a meta-regression analysis

We used a recent meta-regression-analysis (MRA) method to investigate the publication bias of the intra-sectoral foreign direct investment (FDI) spillovers in a large sample of developing countries. Recent meta-analyses on this topic suggest that publication bias is not a problem for this field. Using a much larger sample of studies this article, however, finds substantial evidence of publication bias. Evidence suggests a preferential tendency to publish the article if studies reject the null hypothesis, that is, when the investigations produce positive and statistically significant findings. We collected 1450 spillover estimates conducted by 93 researchers from 69 primary empirical studies dealing with 31 developing countries for the period 1986–2013. We found that the FDI spillover effects are suffering from significant and meaningful publication bias. The bias ranges from 0.51 to 1.34, implying that the empirical effect size appears substantially much larger than the actual spillovers. Our results also indicate that for this field of research, publication bias is largely due to self-censorship rather than the censorship by reviewers and editors of journals.