Viral V. Acharya and Nada Mora
A Crisis of Banks as Liquidity Providers
Journal of Finance | Volume 70, Issue 1 (Feb 2015), 1-43

Our paper investigates whether the 2007-09 crisis was a crisis of banks as liquidity providers, reducing credit and increasing the fragility of the financial system. A widely accepted notion is that banks have a natural advantage in providing liquidity to businesses through credit lines and other commitments established during normal times. Banks are thought to function well as liquidity providers by combining deposit taking and commitment lending activities. Both activities require banks to hold liquid assets to provide liquidity-on-demand (Kashyap-Rajan-Stein (JF 2002)). Banks conserve on liquid asset buffers in meeting liquidity demands, provided deposit withdrawals and commitment drawdowns are not too highly correlated. Supportive evidence comes from previous crises such as in 1998 following the LTCM hedge fund failure. Even when market stress led to significant drawdowns, banks met the increased credit demand because of flight-to-safety flows from depositors as shown by Gatev-Strahan (JF 2006).

1: Deposit rates: Failed minus nonfailed banks
Weak institutions offered higher deposit rates in the run-up to failure.

In 2007 to 2009, however, the banking system was itself at the center of the financial crisis. Questions were raised about the solvency of the banking system as a whole, which was exposed to toxic credit instruments as documented by Acharya-Schnabl-Suarez (JFE 2013) and others. As the solvency risk of a bank increases, it might seek to attract deposits by offering higher rates such as Washington Mutual during the crisis: “The fact that Washington Mutual is now owned by Chase is very positive, because they were a huge outlier on rates” (Ken Lewis, then Chairman and CEO of Bank of America, cited in American Banker October 9, 2008).

Figure 1 shows this more generally by drawing on weekly survey evidence of deposit rates offered by failed banks. The figure plots the difference from the rates of banks that did not fail, over a one-year period prior to failure. The x-axis is the time to failure, where failure also covers cases of near-fails based on stock return performance. Weak institutions offered substantially higher CD rates in the run-up to failure. For example, the average difference on their 12-month CD rates reached close to 60 basis points. This finding aligns with the literature on market discipline where depositors identify insolvent banks and demand higher returns.

When the liquidity provision mechanism of the banking system as a whole breaks down

But the main result of this research is that the onset of the crisis was a crisis of banks as liquidity providers in the aggregate; not just of the weakest banks. It makes sense ex ante for banks to combine deposit taking with commitments, but banks may experience higher liquidity demand coming from both depositors and firms. It is important to understand how banks adjust to such a shock.

2: Market stress and net flows into deposits at commercial banks
As the VIX increased, deposits at commercial banks shot up.

Core deposits increased by only $90 billion up until end-2008Q2. This increase fell short of the increase in core deposits in a comparable period before the crisis and also fell short in relation to the surge in deposits in previous crises such as LTCM. More importantly, deposit inflows fell short in relation to increased loan demand and drawdowns during this period. Core deposits eventually increased in the banking system as a whole by close to $800 billion by early 2009, but only starting in 2008Q3 when they grew by $272 billion in just one quarter. Lending growth outpaced core deposit funding growth from the ABCP “freeze” in August 2007 to just prior to Lehman's failure. The aggregate loan-to-deposit shortfall is shown in Figure 2 below based on quarterly snapshots of weekly commercial banking data. For example, the cumulative difference between the increase in lending and in deposits reached $239 billion by end 2008Q2 and widened to more than $300 billion in the weeks just prior to Lehman's failure.

We argue that the weak deposit funding position of banks and its sharp reversal following Lehman's failure is explained by investor perception of greater risk in bank deposits relative to instruments offering similar liquidity and payments services. Investors piled into securities with more explicit government backing than bank debt, especially as most deposits were over the deposit insurance limit at the start of the crisis. Investors preferred government-backed securities because of concerns about the banking sector's health and the lack of information about exposures to the subprime shock. The rise in aggregate risk also reduced banks' ability to diversify shocks across business and deposits (Acharya-Almeida-Campello (JF 2013)).

For example, money market funds specializing in government securities funds saw greater funding inflows than prime funds. The flow into government funds accelerated in the aftermath of Lehman's failure and Reserve Primary Fund's “breaking the buck” when there was a run on a range of prime funds. The flight out of prime funds of $400 billion in the two weeks following Lehman's failure went to government funds as well as to deposits in the banking system of close to $200 billion as seen in Figure 2. Concurrently, the TARP backstop helped reverse the anemic deposit position of the banking system. TARP increased insurance limits from $100,000 to $250,000 and fully insured noninterest-bearing accounts, among other measures. The inflow of deposit funding enabled by explicit government backing finally allowed the banking system as a whole to close its loan-to-deposit shortfall.

Liquidity demand risk and individual bank behavior in the crisis

To understand liquidity provision by individual banks, we test whether a bank at greater risk of credit line drawdowns offers higher rates if it does not gain enough deposits to match its funding needs. The coefficient β2 on the interaction term of crisis and liquidity demand risk in the equation below is the one of interest. Because our thesis is predicated on the reversal in the aggregate liquidity shock in September 2008 and the primary data are quarterly Call Reports, crisis is empirically represented by two dummy variables, crisis1 (2007Q3-2008Q2) and crisis2 (2008Q3-2009Q2). We also confirm the findings by examining CD rates from a weekly proprietary survey. The key measure of a bank's vulnerability to liquidity demand is the ratio of unused commitments to the sum of unused commitments and on-balance sheet loans. Deposit Ratei,t = β1 · liquidity demand riski,t – 1  + β2 · liquidity demand riski,t – 1 × Crisist +   + Bank FEi + Time FEt + Other Controlsi,t + ei,t

We also test whether banks took additional actions to meet increased drawdowns. Banks will likely be forced to adjust by cutting back on new credit if deposit funding and interbank borrowing are not sufficient. We test for both claims. Other backup actions include running down liquid assets and seeking government-sponsored borrowing such as advances from Federal Home Loan Banks (FHLBs).

Our results show that the aggregate liquidity shock particularly hit banks at greater risk of credit line drawdowns. For example, a 0.1 increase in a bank's unused commitment ratio (roughly a 1 standard deviation or the difference between the 75 and 25 percentiles) raised a bank's large-time deposit rate by 5.9–7.1 basis points in the first year of the crisis. But crucially, despite scrambling for deposits by raising rates, commitments- exposed banks experienced weaker deposit growth. Even the growth of deposits that are commonly considered a stable source of funding such as core deposits was limited.

That banks honored their existing credit lines drawn by firms beginning in August 2007 was possible only because of explicit, large support from the government and government-sponsored agencies such as FHLB advances and Federal Reserve liquidity facilities. For example, advances from the FHLBs covered 65% of non-deposit borrowing growth at commitments-exposed banks during the first year, and the widening shortfall between their on-balance sheet loans and deposits was closed halfway with government-sponsored borrowing.

We conduct several tests to rule out the alternative hypothesis that commitments-exposed banks were simply those with greater solvency problems. The results indicate that solvency problems, such as real estate related exposure, were independent risk factors, whose effect persisted into the latter part of the crisis. In contrast, the funding pressure on commitments-exposed banks coincided with the shifts in aggregate deposit funding. But liquidity risk interacts with solvency risk as predicted by theory. Commitments-exposed banks with weaker fundamentals were more vulnerable to the onset of the crisis than equally exposed banks with stronger fundamentals.

To conclude, the role of banks as liquidity providers was itself in crisis during the crisis from 2007 to 2008. Unlike previous crises, banks did not expand total loans and credit lines. And in the absence of a liquidity backstop by the government and the Federal Reserve, financial disruptions and business liquidations would have likely been much greater.