A paper written earlier this year entitled, "A Lending Network under Stress: A Structural Analysis of the Money Market Funds Industry," studies funding supply shocks in the money markets. Author `Paula Beltran from UCLA explains, "In this paper, I study the transmission of an aggregate funding supply shock in a lending network and quantitatively assess the implications for the allocative efficiency of funding provision of the US Money Markets Funds Industry. I build a tractable model that features banks and funds that bargain over the terms of trade subject to an incomplete network of existing counterparties and bilateral bargaining. I discipline the model using data on the funds' portfolio." (Note: Thanks to those who attended our European Money Fund Symposium, in Paris! We hope you enjoy Day 2, and safe travels home. Thanks to our speakers and sponsors! Watch for excerpts and quotes from the event in our daily news next week and in our October Money Fund Intelligence newsletter.)
She continues, "I show how to identify the key parameters of the model by exploiting granular shocks of connected agents. Taking as primitives the observed changes in assets under the management of prime funds at the onset of the COVID-19 crisis, the model accounts for 85% of the drop in total lending and 70% of the increase in price dispersion. I show that the allocation is inefficient. Faced with the same drop in asset under management and taking as given the network of bilateral counterparties, a central planner would reduce lending by 9% instead of 14% in equilibrium. Finally, I use the model to examine the effectiveness of the Overnight Repo Repurchase Facility."
The paper states, "Global funding markets experienced acute distress in March 2020 when the COVID-19 'dash for cash' drained the supply of funding. Severe dislocations in the cost of funding resulted in interest rates spikes in several funding markets.... Commercial paper spreads with respect to the T-bill increased by 150 basis points on average around March 15. These dislocations did not subside until unprecedented policy measures were implemented to restore liquidity in key funding markets."
It continues, "Many of the funding markets that were under significant stress in March 2020 are decentralized, and funding provision in these markets relies on a network of bilateral relationships. The March 2020 events were not isolated. Funding dry-ups occurred in other financial crises; for example, during the 2008 Global Financial Crisis. Many papers highlight the importance of the interconnectedness in financial markets in 2008 (Di Maggio, Kermani, and Song, 2017; Eisfeldt et al., 2019), but leave unanswered two fundamental questions regarding large aggregate funding shocks in a lending network. First, how much of the total funding provision and dispersion in the cost of funding after a large aggregate funding supply shock can be explained by network frictions? Second, how much allocation inefficiency results as a consequence of market power within a lending network under stress?"
Beltran writes, "This paper addresses these questions quantitatively in the unsecured funding market, where U.S. Money Market Funds provide a significant source of dollar funding to global banks (FSB, 2020). This $9 trillion industry was under significant stress in March 2020. Prime money market funds were subject to a liquidity withdrawal comparable to the run experienced by Money Market Funds in September 2008 (Anadu et al., 2021a). This episode provides a suitable setting to study the role of network frictions in the transmission of funding supply shock in the context of a highly concentrated industry."
She tells us, "I build and estimate a model of bilateral unsecured funding within the network of banks and funds, using data from before March 2020. To estimate the main parameters, I use data on the funds' portfolio from 2011 up to January 2020 and rely on granular variation in interest rates and funding provision at the bilateral level. I use the model to produce the counterfactual changes in interest rates and funding supply in this lending network after introducing a large aggregate shock to the funds' assets under management as observed in March 2020."
The paper states, "The model is designed to capture two main features of a lending network composed of funds and banks. First, the model captures limited connectivity between banks and funds. Second, the model features bilateral market power that affects the funds' portfolio choice. Together, limited connectivity and the distribution of market power determine the cost of funding."
Beltran also says, "Limited connectivity in my model comes from two sources: an exogenous network of possible counterparties and concentration risk. First, I assume an exogenous network as in Eisfeldt et al. (2019). Banks and funds interact through an exogenous network that constrains the agents' set of counterparties. This assumption captures relationship frictions between banks and Money Market Funds. During the COVID-19 crisis, very few new relationships were created: the fraction of trades corresponding to new bilateral relationships was less than 1%. Therefore, an exogenous network is plausible in the short term and implies that the preexisting network of counterparties will shape the outside option of agents and dispersion in terms of trade in the model."
She continues, "Second, funds face concentration risk, which captures that funds are subject to strict counterparty limits by regulation. Besides aversion to aggregate risk, funds have an additional cost of large bilateral exposures. The costs of bearing aggregate and concentration risk govern the network effects present in the model, since they determine the elasticity of substitution across different counterparties. Moreover, they play different roles in the model. The marginal cost of risk involves all of the marginal units of risky positions. Meanwhile, the marginal cost of concentration risk gives a fund incentives to smooth its exposure across banks within its network of counterparties. Concentration risk prevents equalizing the cost of aggregate risk across counterparties, and its prevalence is larger as the number of counterparties reduces."
The study explains, "Funds can invest in three types of assets. They can lend to banks, hold Treasuries, or hold securities in the Overnight Reverse Repo Repurchase Facility (ON-RRP). I assume that the latter has no risk; meanwhile, Treasuries and unsecured lending are subject to aggregate risk. In the model, funds face a two-stage problem: In the first stage, they determine how many Treasuries to hold. Then, in the second stage, funds and banks meet simultaneously. This setting has an important consequence: Funds internalize the effects of holding Treasuries on the negotiation results with their counterparties. Also, the ON-RRP increases the bargaining power of funds, because it raises the value of the outside option for funds."
Beltran writes, "I depart from competitive pricing and assume that connected agents negotiate the terms of trade in a bilateral bargaining process characterized by heterogeneous relative bargaining power. Funds and banks meet and decide the terms of the contract in a Nash-in-Nash bargaining process. However, how do they split the surplus depends on their relative bargaining power. Two consequences of this assumption are worth noting. First, this bargaining process distorts prices as a signal of the marginal cost of funding. In this context, prices will be the average of the bank's benefit and the fund's cost of funding with respect to their outside option. Second, market power will affect the funds' portfolio choice: A low market power incentivizes funds to reduce the funds available for negotiation by internalizing the price of aggregate risk-taking."
Finally, she adds, "My model predicts an increase in cost of funding, a rise in interest rate dispersion and a drop in funding provision comparable to those observed in the data. Price dispersion, measured by the interquartile range, increases from 22 to 66 basis points in the model. At the same time, the median rate in my model rises 61 basis points. The model predicts a 14% fall in aggregate lending, which is close to that observed in the data of about 16%. A reduction in loanable funds reduces the funds' supply of funding available for banks. I find that the allocative efficiency worsens after the shock. A planner subject to the same preferences and regulatory constraints would allocate 22% more funds to banks than the decentralized solution. The planner would reduce lending by only 9% from February to March 2020."
