Patrick Coen

Home     Research     CV

Working papers

A Structural Model of Interbank Network Formation and Contagion
with Jamie Coen (LSE)

The interbank network, in which banks compete with each other to supply and demand financial products, creates surplus but may also result in risk propagation. We examine this trade-off by setting out a model in which banks form interbank network links endogenously, taking into account the effect of links on default risk. We estimate this model based on novel, granular data on aggregate exposures between banks. We find that the decentralised interbank market is not efficient, primarily because banks do not fully internalise a network externality in which their interbank links affect the default risk of other banks. A social planner would be able to increase surplus on the interbank market by 13% without increasing mean bank default risk or decrease mean bank default risk by 4% without decreasing interbank surplus. We propose two novel regulatory interventions (caps on aggregate exposures and pairwise capital requirements) that result in efficiency gains.

Information Loss over the Business Cycle

The business cycle induces turnover in mutual funds: they exit in recessions and enter in recoveries. The effect of this firm turnover on welfare depends on a key trade-off: on the one hand, the business cycle “cleanses” the market of low quality exiting funds and replaces them with entrants that may on average be higher quality. On the other hand, the entrants have no returns history and so investors have less precise beliefs about their ability, where this “information loss” leads to misallocation that harms welfare. I examine this trade-off by estimating a structural model in which rational investors form and update beliefs about competing mutual funds that endogenously choose to enter and exit the market. I estimate this model using data on US mutual funds. I find that the business cycle has material, persistent effects that are negative in the short-term but turn positive as the effect of information loss decays over time. Finally, I use my estimated model to consider how subsidies can optimally trade off information loss and cleansing in a temporarily predictable recession such as that resulting from the Covid-19 pandemic. I find that aggregate surplus is maximised by subsiding older funds, so as to preserve their socially valuable returns history.

A Structural Model of Liquidity in Over-the-Counter Markets
with Jamie Coen (LSE)

We study how firm heterogeneity determines liquidity in over-the-counter markets. Using a rich dataset on trading in the secondary market for sterling corporate bonds, we build and estimate a flexible model of search and trading in which firms have heterogeneous search costs. We show that the 8% most active traders supply as much liquidity as the remaining 92%. Liquidity is thus vulnerable to shocks that restrict active traders’ willingness to trade: if the 4% most active traders stop trading, liquidity falls by over 60%. Bank capital regulation reduces the willingness of these active traders to hold assets and thus reduces liquidity. However, trader search, holdings and intermediation respond endogenously to reduce the welfare costs of regulation by 30%.  These costs are greater in a stress, when these margins of adjustment are constrained. The introduction of trading platforms, which homogenise the ability of traders to trade frequently, improves aggregate welfare but harms the most active traders who currently profit from supplying liquidity.

Work in progress

Experimentation, Transparency, and the Speed of Learning in the Mutual Fund Industry
with Jamie Coen (LSE)