ANOTHER LOOK AT ‘BANK COMPETITION AND FINANCIAL STABILITY: MUCH ADO ABOUT NOTHING’?

• Published in: Journal of economic surveys Vol. 34; no. 2; pp. 344 - 371
• Main Authors: Bandaranayake, Samangi, Das, Kuntal K., Reed, Robert W.
• Format: Journal Article
• Language: English
• Published: Oxford Blackwell Publishing Ltd 01.04.2020
• Subjects: Analysis; Bank competition; Bayesian model averaging; Best practice; Competition; Financial stability; Meta-analysis; Publication bias; Repetition; Replication; Robustness; Studies
• ISSN: 0950-0804; 1467-6419
• DOI: 10.1111/joes.12352

This study replicates Zigraiova and Havranek's (2016) meta‐analysis of banking competition and financial stability. It performs multiple 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‐analysed; an ‘Extension’ replication where additional studies on banking competition and stability are analysed; and a ‘Robustness Analysis’ where we check Z&H's results using an alternative empirical procedure. Our analysis strongly confirms Z&H's main finding that competition in the banking sector has an economically negligible effect on financial stability. This result is consistently confirmed across a variety of replication analyses. Most impressively, we confirm their finding even when we analyse a completely independent set of 35 studies not included in Z&H's meta‐analysis. Our results for Z&H's other findings are less supportive. As the first comprehensive replication of a meta‐analysis, this study also provides insights into the robustness of meta‐analysis. We find that meta‐regression analysis, where estimated effects are related to data, estimation, and study characteristics, is sensitive to how data are coded and to the choice of estimation procedure; and that this sensitivity extends to ‘best practice’ estimates.