Banks are drastically adjusting their branch networks in response to rapid digitalisation, fierce competition from FinTech, and the continuing expansion of online and mobile banking. An underappreciated fact is how unequally these dynamics are playing out within countries. Early evidence suggests that banks’ remaining branches increasingly cluster together, as new branches open in economically strong centres while branches close down in sparsely populated areas. The emergence of ‘banking deserts’ – localities almost entirely devoid of branches (Morgan et al. 2016) – has raised concerns about adverse local effects on firms’ funding costs (Bonfim et al. 2021) and on small-business lending and employment opportunities (Nguyen 2019).
In recent research (Qi et al. 2021), we assess how the introduction of new technology to share borrower information affects the spatial clustering of bank branches. Our setting is emerging Europe, where several countries in rapid succession introduced public credit registries and private credit bureaus – systems that allow banks to share digital information about the repayment behaviour and current indebtedness of their clients. We use the staggered introduction of these systems as shocks that pushed banks towards new clustering equilibria.
A simple spatial model of bank branching
To structure our empirical work, we first built a spatial oligopoly model of branch clustering. The main intuition is that while clustering increases the likelihood that an entrepreneur will visit a locality and successfully obtain a loan, inter-bank proximity also implies that competition that is more vigorous. If the first (market-size) effect dominates, banks earn higher profits by locating themselves closer to each other so that they attract more clients. If the second (price-cutting) effect dominates, banks try to decrease competition by dispersing their branches geographically.
We then use this model to think through how clustering is affected by the introduction of a formal mechanism through which banks share hard (codified and transferable) borrower information (Pagano and Jappelli 1993). Information sharing among banks affects the equilibrium level of branch clustering as it eliminates the distance threshold beyond which entrepreneurs cannot successfully apply for loans. In other words, when borrower information is shared, entrepreneurs can, in principle, apply in every locality (as long as transportation costs are not prohibitive). This increased competition from distant localities incentivises banks to make nearby localities more attractive. Through branch clustering, they aim to attract or retain relatively distant entrepreneurs that are in search of deeper credit markets in which they can apply for a loan from a wider variety of banks. Using this model, we derive three main testable hypotheses:
1. Information sharing increases branch clustering because banks can attract some distant borrowers that were previously too opaque to lend to
2. Information sharing increases the likelihood that banks open branches in new localities where they were not previously active
3. Information sharing spurs relationship banks, which rely heavily on soft information in the absence of information sharing, to cluster more when compared with transactional banks
To test our model predictions, we use the introduction of information sharing regimes as country-level shocks that push banks towards a new clustering equilibrium. This requires time-varying data on branch locations for countries that introduce information sharing (either through a public credit registry or through a private credit bureau) at different points in time. We have access to information on the geographical coordinates of 56,555 branches owned by 614 banks in 8,536 localities (towns and cities) across 19 emerging European countries. The data paint a precise, complete, and gradually changing picture – reflecting branch openings and closures – of the banking landscape during the years 1995 to 2012. Figure 1 depicts the spatial branch distribution in these countries at the start and the end of our sample.
Figure 1 Distribution of localities with bank branches in 1995 (left) and 2012 (right)
In terms of methodology, we implement a ‘difference-in-difference-in-differences’ framework, with the treatment (presence of information sharing) varying across countries and years. We then compare how, within the same country, the introduction of information sharing differentially affects branch openings across localities with different numbers of pre-existing bank branches.
What do we find?
We find that electronic information sharing has a strong positive effect on branch clustering. Banks are more likely to open new branches in localities where they do not yet operate but where relatively many other banks are already present. This branch clustering is stronger for relationship banks and in countries where information sharing is more effective. Additional analysis of data on bank-firm relationships shows that, in line with a reduction in geographical credit rationing, information sharing allows firms to borrow from banks that are more distant.
One may wonder whether our findings could mostly reflect branch clustering in specific parts of countries. A secular urbanisation process can induce a disproportionate increase in the opening of new bank branches in urban areas. We may then pick up clustering forces in urban areas that are largely unrelated to (but coincide with) the introduction of information-sharing regimes. Yet, when we conduct an event study, we find in fact sharp changes in clustering behaviour right after the introduction of information sharing regimes. This already partly dispels concerns about gradual trends driving our estimates. To investigate this issue further, we split our sample into localities with different population sizes. We find that while our estimates point to a somewhat stronger effect of information sharing in larger localities, the impacts in more rural areas are both highly significant and economically sizeable.
Ongoing digitalisation and the advance of FinTech are putting pressure on banks to prune their bricks-and-mortar branch networks – and to do so in a strategic way. To help understand this process, we develop an intuitive framework in which banks rationally trade off the market-size and price-cutting effects of geographical clustering. We then test our model predictions in a rich international context, using the introduction of digital information-sharing as country-level shocks that shift the relative advantages and disadvantages of branch clustering. We observe how these shocks play out at a very disaggregated level (that of individual towns and cities), across a large number of countries. Succinctly put, we find that the digital availability of hard borrower information leads to a branch-clustering equilibrium in which it becomes much more important for banks to be close to each other than to be close to their borrowers.
Our findings also imply that local credit markets have become more homogeneous in terms of composition but less homogeneous in terms of size. While the public availability of hard information leads to further clustering of banks in well-served locations, other (smaller) locations lose out as access to credit deteriorates further. Assessing the real-economic impacts of such spatial variation in access to credit due to information sharing is a promising avenue for future research.
Bonfim, D, G Nogueira and S Ongena (2021), “‘Sorry, we’re closed’, Bank branch closures, loan pricing, and information asymmetries”, Review of Finance, forthcoming.
Morgan, D P, M L Pinkovsky and B Yang (2016), “Banking deserts, branch closings, and soft information”, Federal Reserve Bank of New York, working paper.
Nguyen, H L Q (2019), “Are credit markets still local? Evidence from bank branch closings”, American Economic Journal: Applied Economics 11(1): 1-32.
Pagano, M and T Jappelli (1993), “Information sharing in credit markets”, Journal of Finance 48(5): 1693-1718.
Qi, S, R De Haas, S Ongena and S Straetmans (2021), “Move a little closer? Information sharing and the spatial clustering of bank branches”, CEPR Discussion Paper 15829.