A Replication of “Bank Competition and Financial Stability: Much Ado About Nothing?” (Journal of Economic Surveys, 2016)

This study replicates Zigraiova and Havranek’s (2016) meta-analysis of banking competition and financial stability. It performs three types of replications: a “Reproduction” replication where Z&H’s data and code are verified to reproduce the results of their study; a “Repetition” replication where the studies used by Z&H are independently recoded and then re-analyzed; and an “Extension” replication, where 35 additional studies on banking competition and stability are examined to see if they confirm Z&H’s findings. Our replication analyses confirm Z&H’s main finding that competition in the banking sector is unrelated to financial stability. However, we do not confirm their finding of publication selection in the “Extension” replication, and both “Repetition” and “Extension” replications fail to confirm most of Z&H’s findings about the effects of data, estimation, and study characteristics on estimated competition effects. While our “Repetition” replication produced a high match rate with Z&H’s data, we found that the meta-regression analysis was substantively impacted by discrepancies between the original and recoded data. We conclude that meta-regression analyses should be viewed with caution. We also demonstrate that “best practice” estimates can be precarious and highly dependent on variable specification and how “best practice” is defined. They likewise should be viewed with caution. Our study highlights the value-added that replication brings to meta-analyses. While some of Z&H’s results were found to not be robust, the fact that all three types of replications confirmed Z&H’s main finding about competition and stability provides strong evidence that this finding can be regarded with confidence.