IMF working paper no. 16/64
15 March 2016
Source: International Monetary Fund
Christoph Aymanns, University of Oxford
Carlos Caceres, International Monetary Fund
Christina Daniel, International Monetary Fund
Liliana Schumacher, International Monetary Fund
Understanding the interaction between bank solvency and funding cost is a crucial prerequisite for stress-testing. In this paper we study the sensitivity of bank funding cost to solvency measures while controlling for various other measures of bank fundamentals. The analysis includes two measures of bank funding cost: (a) average funding cost and (b) interbank funding cost as a proxy of wholesale funding cost. The main findings are: (1) Solvency is negatively and significantly related to measures of funding cost, but the effect is small in magnitude. (2) On average, the relationship is stronger for interbank funding cost than for average funding cost. (3) During periods of stress interbank funding cost is more sensitive to solvency than in normal times. Finally, (4) the relationship between funding cost and solvency appears to be non-linear, with higher sensitivity of funding cost at lower levels of solvency.
The main results from the empirical analysis can be summarized as follows.
(1) Linear panel estimation shows that our solvency measure (comprised of the leverage ratio, Tier 1 capital ratio, and total regulatory capital ratio) is negatively and significantly related to both average and wholesale measures of funding cost. While the magnitude of the relationship is small, shocks to bank solvency can have a non negligible impact on the bank’s profitability.
(2) On average, the coefficient linking funding cost and solvency is larger in magnitude for wholesale funding cost than for average funding cost. This is consistent with the notion that wholesale investors that provide unsecured loans to banks are more sensitive to counterparty risk than retail investors whose deposits are covered to a large extent by deposit insurance.
(3) Re-estimation of the relationship between funding cost and solvency on a yearly basis shows that wholesale funding cost is more sensitive to solvency during periods of economic stress than in normal times. This finding is relevant from a stress testing perspective. The fact that the strength of the interaction between solvency and funding cost is state contingent introduces an additional factor of uncertainty into the stress testing process: when a severe enough crisis hits, historically estimated average sensitivities may be rendered invalid and underestimate the true sensitivity.
(4) There is evidence that the relationship between funding cost and solvency is non-linear, with higher sensitivity of funding cost at lower levels of solvency. While further work will be required to better quantify this non-linear relationship, our results suggest that banks with already low levels of solvency will be hit harder by funding cost increases making them even more susceptible to exogenous shocks.
These results have a number of implications. Increasing capital ratios may help banks reduce their cost of funding. Depending on the cost of raising additional capital it may be beneficial for banks to increase their capital in order to drive down their average funding cost. While this is a theoretical possibility, it is unclear whether the magnitude of the reduction in funding cost identified here would be sufficient to convince banks to increase their capital level.
Estimates of the sensitivity of bank funding cost to solvency shocks are critical for stress-testing, but the fact that the sensitivity of wholesale lenders to solvency shocks varies considerably over time and appears particularly elevated during times of crisis makes it hard to predict how funding cost will respond in future crises.
Furthermore, the non-linear estimations performed in this paper suggest that solvency should be viewed as an indicator of banks vulnerability to wholesale funding shocks. This adds to the view that the level of solvency is a determinant of funding cost. Banks with low capitalization are at risk of stark wholesale funding cost increases. However, whether or not such a funding cost increase materializes is likely to be conditional on other bank specific events that are notoriously hard to predict. This alternative interpretation suggests the following approach to funding cost stress testing. Rather than focusing on the mean of a heavily skewed distribution of funding cost conditional on a given level of solvency, an alternative approach would consider the probability of a particular funding cost spike conditional on a level of solvency.
Given an exogenous shock to bank solvency derived from a satellite model, a stress tester may use the funding cost elasticity, derived from the linear panel estimation, to compute changes in the overall interest expense of a stressed bank. However, it should be noted that the estimated increase in interest expense obtained through this procedure is likely to be conservative, and may underestimate the actual effect of a solvency shock in times of crisis. Based on this conservative estimate, the net interest margin of the stressed bank could be recomputed in order to project bank’s earnings following the solvency shock. This would ultimately allow for the computation of a secondary solvency shock via the funding cost channel.
Nevertheless, a better grasp of the nonlinearities at play in the response of funding cost to solvency shocks is crucial for improving the calibration of stress test models. As mentioned above, estimates of funding cost elasticity based on linear models are likely to underestimate the actual impact of solvency shocks on ‘weak’ banks. In particular, those models are not able to estimate a threshold value at which banks may be shut out from wholesale funding markets altogether. Furthermore, the models used in this paper probably place too strong emphasis on the mean of a skewed distribution of funding cost when outliers represent the events of actual importance for risk management and regulation.
Better data may help resolve some of the estimation issues. While the data available for this study allowed for a rough proxy of wholesale funding cost and the identification of linear sensitivities, an improved dataset could help make the conclusions of this study more robust and actionable. In particular data on wholesale funding cost (e.g., actual bond yields) and events of extreme funding liquidity tightening (including cases where liquidity completely dried out) would be useful. While data on the latter are rare, it is crucial to understand the nonlinear response of funding cost and the identification of a potentially state contingent solvency threshold for access to wholesale funding markets.