Matthieu Chavaz, David Elliott
Review of Finance, Volume 30, Issue 3, May 2026, Pages 1071–1108, https://doi.org/10.1093/rof/rfaf071
Whether banks’ retail and investment banking activities should be separated has been debated since at least the Glass-Steagall Act of 1933. In the US and Japan, banks must split these activities into separate subsidiaries, whereas European and Canadian banking groups can operate as integrated “universal” banks.
A key obstacle to identifying the impact of structural separation is that plausibly exogenous shocks to bank structures are rare. In this paper we exploit the UK ring-fencing reform, which requires large universal banks to separate into retail and non-retail subsidiaries from 2019. For identification, we exploit differences in the extent to which banks must restructure given their pre-existing business models, and differences in the degree to which individual loans are affected depending on their maturity and origination dates.
Structural separation mainly aims to protect retail customers and taxpayers from investment banking risks in crisis times. However, this separation also changes the funding structure of a wide range of universal banking activities in normal times. Our key finding is that as a result, separation involves a range of side effects for credit supply, competition, and risk-taking in important credit markets—even where those markets are not directly targeted by the reform.
After separation, any activity in the non-retail entity loses access to retail deposit funding. These deposits instead become available for activities in the retail entity—such as retail lending. In response, affected banks rebalance towards retail mortgage lending. This increases household credit supply, but also increases the concentration of the mortgage market in the hands of large universal banks. This erodes the market share of smaller banks, and pushes them to rebalance towards higher-yielding but riskier mortgage lending. Meanwhile, affected universal banks respond to the loss of deposit funding in the non-retail entity by reducing syndicated lending to large corporates.
Taken together, our results suggest that assessing the merits of structural separation requires weighing its intended financial stability benefits in crisis times with its side effects for credit markets in normal times. The shift by affected banks towards mortgage lending reduces the cost of credit for households. But the increased mortgage market concentration could reduce competition in this market over the longer term. The increased concentration also casts doubt on the idea that separation reduces the systemic importance of the largest banks, or their political clout. Meanwhile, the increased risk-taking by small banks suggests that separation does not unambiguously improve financial stability across the banking system. Internationally, our results suggest that persistent differences between regulatory regimes close to ring-fencing (as in the US and Japan) and regimes that allow integrated universal banks (as in the EU and Canada) can matter for the pattern of credit supply and competition across retail and capital markets.