Deposit Rate Advantages at the Largest Banks

While uninsured deposits are senior to subordinated debt, both are uninsured and unsecured sources of funding for the bank.

The vast majority of banks do not have public equity and even those that do are typically traded only infrequently. An even smaller portion of banks issue bonds, even fewer of which have liquid CDS markets.

Throughout the paper, bank refers to a depository institution (including thrifts) consolidated at the level of the regulatory high holder. In the baseline definition, we use $200 billion in assets as the large-bank threshold, though alternative thresholds are explored. See Section 4.1.

Observationally similar phenomena have been described variously as “too-systemically-important-to-fail” (Ueda and Weder di Mauro 2013), “too-complex-to-fail” (Herring 2002), “too-difficult-to-fail-and-unwind” (Kane 2009); and probably many others. For the purposes of this paper, we harbor no preference for one description over another. Here, we test for differences in deposit pricing based on size as measured by assets, though asset size is likely correlated with measures of systemic importance, connectedness, and complexity.

A natural extension would be to examine the $250K-$100K MMDA spreads, as well. However, the $250K deposit data is currently too sparse for meaningful tests.

To the best of our knowledge, the only other paper using this data set is Ben-David et al. (2011), who show that poorly capitalized banks counterintuitively paid lower deposit rates on CDs in 2009–2010.

Since banks enter the data only when a competitor requests their information, self-selection is not a concern.

http://​www.​rate-watch.​com/​faq.

Our data excludes various sweep products that would prove problematic. For example, banks routinely sweep funds from retail checking accounts into non-retail MMDAs at the end of business days and back again before the start of the next business day, often to reduce their reserve requirements (see Anderson and Rasche 2001; VanHoose and Humphrey 2001). Alternatively, some banks participate in MMDA sweep programs (e.g., Insured Cash Sweep (ICS) from Promontory) in which the bank reallocates any amount over the SMDIA across a bank network so that the entire principal and interest will be insured. However, the ICS contract stipulates that banks cannot “offer or promise any particular interest rate” on the principal, as specific rates are the obligation of the destination institution. Therefore, the explicit rate offers observed in RateWatch would preclude the possibility of either type of sweep account in the data.

For more details, see 12 CFR §204.2(d)(2); Reg. D, 45 FR 56018, Aug. 22, 1980.

Weighting by deposits produces similar estimates for nearly all banks. For those banks where it makes a difference, the result is driven by branches offering deposit rates but recording no deposits in the Summary of Deposits data. Banks offering deposit rates yet seemingly holding no deposits may be the consequence of consolidated deposit recording across multiple branches in the SOD data.

We also run the analysis using only deposit rate data from within a week of the quarter end, and the (unreported) results are similar.

This is consistent with Heitfield and Sabarwal (2004) and Park and Pennacchi (2009), among others.

A list of the variable abbreviations can be found in Table 1.

This is in contrast to the typical difference-in-difference approach that differences across time. However, this third level of differences is the subject of Section 4.3, below.

¹⁷Kennickell et al. (1996) estimate that a decrease in the deposit insurance limit from $100K to $25K would not be associated with a dramatic change in many non-wealth household characteristics.

Examining the differences surrounding the policy change adds a third level of flexibility (similar to a difference-in-difference-in-difference technique). However, in that case, it is important to note that there are confounding factors that limit the interpretability.

If present, double violations may work against a finding of a large bank advantage. For example, if large depositors are more rate-sensitive (i.e. resemble wholesale more than retail deposits) than are small depositors, then the uniformity assumption would be violated. However, this would bias estimates against a large bank advantage, as small depositors accept lower rates at the largest institutions, which would drive up the $100K–$25K MMDA spread.

There is empirical evidence that this is not the case (as discussed in in the previous paragraph), however this supposition acts a useful illustrative tool.

Source: FDIC. https://​www.​fdic.​gov/​bank/​individual/​failed/​index.​html. Recovery values come from FDIC dividend data and the set of banks with uninsured depositor losses come from examining those banks with a PI or PO resolution method.

In unreported analysis, we find that conditional on failure, there is no systematic relationship between size and the probability that uninsured depositors face losses. Consequently, it is unlikely that the results given here are driven by differential treatment of uninsured depositors across bank size in the event of failure.

There are a number of extreme values observed for asset growth. Thus, asset growth rates are Windsorized to a floor of −50 % and a ceiling of 100 %.

Under the unified supervisory regime instituted by FDICIA, banks are evaluated according to six components of safety and soundness. Each letter in “CAMELS” refers to one of these components: Capital adequacy, Asset quality, Management, Earnings, Liquidity, and Sensitivity to market risk.

For example, it may be the case that banks increase deposit rates in a systematic way across deposit account sizes independently of risk. This may be a result of using a rule-of-thumb pricing strategy (behavioral) or may be the result of a bank’s strategic marketing to larger accounts. Additionally, as MMDAs have limited withdrawals, the $25K to $10K spread could also reflect a liquidity premium. However, given the relative liquidity of MMDAs, this difference more likely reflects factors beyond liquidity.

Using the difference between the $100K and the $10K MMDA rates provides another measure of the risk premiums. However, whatever non-risk-related differences exist between products would be less pronounced between $100K and $25K MMDAs because of their closer deposit minimums. Thus, these products provide a better measure of risk premiums. Nevertheless, applying the same analysis as given below except exchanging the $10K for the $25K yields similar results.

MSA-level results for the panel analysis are reported below. However, cross-sectional results are left unreported.

Comparable results are obtained using market-based notions of the Z-score, including market capitalization and option implied volatility (albeit, with a smaller sample size).

Ignoring zeros the result is even larger and significant, but forces the exclusion of one large bank which always posts a zero premium in the New York City MSA during the sample.

For example, uninsured depositors took losses at IndyMac in July 2008.