«Research Division Federal Reserve Bank of St. Louis Working Paper Series Understanding the Accumulation of Bank and Thrift Reserves during the U.S. ...»
4.2.2 Banks diﬀerentiated by size We next diﬀerentiate banks by size and consider whether large banks had diﬀerent cash and reserves accumulation responses than small banks. We deﬁne large banks as the top 2 percent of banks measured by assets and small banks as those with assets below the 95th percentile (leaving an intermediate group between large and small). In 2008:Q3, there were 148 banks classiﬁed in the top 2 percent; by 2010:Q2, there were 128 banks in this category, a 16 percent decline. This attrition reﬂects mergers, acquisitions, and failures. Since we use a pooled sample rather than a panel, sample attrition is unlikely to have a signiﬁcant eﬀect. Fig. 9 shows the cross-sectional distribution of excess reserves accumulation as a ratio to required reserves for small versus large banks.
Some noteworthy diﬀerences are reported in Tables 4 and 5.26 We focus on the CLAD results, though the Tobit results are similar.27 First, we ﬁnd that large and small banks have a similar response to an increase in the opportunity cost of holding cash and ER: A 0.1 percent increase in the opportunity cost (measured as the diﬀerence between the yield on 1-year Treasury bills and the IOR) is consistent with a 1 percent decline in ER and cash holdings. The response is slightly smaller for small banks; tests of the equality of coeﬃcients across groups reject equality at the 1 percent conﬁdence level.
Second, we ﬁnd a huge response to an increase in the penalty rate (measured by an index of interest rates on Treasury bill repos) for large banks and a much smaller response by small banks (the responses are signiﬁcantly diﬀerent at the 1 percent conﬁdence level). For small banks, a 0.1 percent increase in the repo rate is consistent with a 3 percent increase in their ratio of ER and cash to deposits. We ﬁnd the response of large banks to a 0.1 percent increase in the penalty rate is consistent with an increase in ER and cash of 40 percent. In the previous regressions with “all banks,” the strong response to the penalty rate is driven primarily by the response of large banks.
Returning to our discussion of disruptions in the repo and federal funds markets in 2008:Q3–2009:Q2 and 2009:Q4–2010:Q2, we ﬁnd that small banks experienced a much smaller decline in trading volume in this market than large banks. While there was a decline in volumes of small bank activity, the decline, particularly between the end 26 Approximately 2 percent of the observations for the large bank sample and 0.15 percent of the small bank sample are censored.
27 The large coeﬃcients on the 1-year Treasury bill yield minus the IOR measure of the opportunity cost for large banks are similar to the coeﬃcients in the CLAD regression when exponentiated.
of 2009 and mid-2010, was much smaller than the decline in large bank activity. For large banks, trading volume dropped by more than half (in 2010:Q2, trading volume was 27 percent of what it was in 2008:Q3 for large bank trading). For small banks, trading volume was 51 percent of what it was in 2008:Q3.
To determine the eﬀect of a larger capital ratio on ER and cash accumulation, we consider the total eﬀect, which is the sum of the coeﬃcient on the adjusted capital ratio and the coeﬃcient on the interaction term for various levels of loan loss provision (exponentiated). In considering the eﬀect of the capital ratio on large banks, holding the loan loss provision at 0, we ﬁnd a 1 percent increase in the capital ratio results in about a 25 percent increase in the ratio of ER plus cash to deposits (focusing on the two speciﬁcations where there is signiﬁcance). Holding the loan loss provision at zero, for small banks we ﬁnd a 1 percent increase in capital adequacy results in a 4 percent decline in ER and cash holdings. These diﬀerences are statistically signiﬁcant at the 1 percent level. What causes this discrepancy? To answer this question, we consider the total eﬀect ﬁrst for large banks and then for small banks, focusing on the coeﬃcients for the regression that includes “bad loans 2,” since the coeﬃcients on the relevant covariates are signiﬁcant for both large and small banks for this speciﬁcation.
When loan loss provisions are held constant at the 50th percentile, for large banks the total eﬀect of a 1 percent rise in capital adequacy is a 27 percent increase in the ratio of ER and cash holdings to deposits. With loan loss provisions at the 75th percentile, the eﬀect is a 19 percent increase. With loan loss provisions at the 95th and 99th percentiles, the eﬀect is an 11.5 percent decrease and a 42.4 percent decrease, respectively, in cash and ER holdings. For small banks, the pattern is the opposite.
When loan loss provisions are held constant at the 50th percentile, the total eﬀect of a 1 percent rise in capital adequacy is a 3 percent decrease in the ratio of ER and cash holdings to deposits. With loan loss provision at the 75th percentile, the eﬀect is a 2 percent decease. However, at the 90th, 95th, and 99th percentiles of loan loss provision, there is a 2.7 percent increase, a 3 percent increase, and a 13 percent increase, respectively, in ER and cash holdings.
The behavior of small banks makes sense: As loan provisions increase, the probability of near-future write-downs rises, so we observe a small positive relationship between the capital ratios (which may be increasing due to the increase in loan loss reserves) and ER and cash accumulation—the high loan loss provisions reﬂect a risky position for small banks.
The behavior of the large banks is more diﬃcult to understand. Loan assets are reported on bank balance sheets net of the loan loss reserve. When the bank increases its loan loss provision (an expense item against proﬁts on the income statement), its loan loss reserves increase by the same amount. Therefore, higher loan loss provisions mean lower net assets (net of loan loss reserves), increasing the capital ratio. When a loan is written oﬀ, loan receivables are decreased by the size of the loan and the loan loss reserve is also reduced by the amount of the loan. These actions should not change net asset positions in the quarter in which the charge-oﬀ occurred. However, in the next quarter, the loan loss provision needs to be rebuilt; therefore, the loan loss provision increases. Banks may increase their loan loss reserves when the probability of imminent losses is higher. These accounting facts make disentangling the relationships among capital ratios, loan loss provisions, and ER and cash accumulation more diﬃcult.
Then, why do large banks with high loan loss provisions decrease their ER holdings?
When we examine large bank behavior for the broader category of problem loans (bad loans 3), we ﬁnd a 4 percent and 22 percent decline in ER at the 95th and 99th percentiles, respectively, of loan loss provisions for a 1 percent rise in the capital ratio— a smaller eﬀect compared with the previous speciﬁcation, but we still ﬁnd the unusual decrease in cash and reserve holdings. Another consideration is that large banks may use a diﬀerent strategy to increase their loan loss provisions than small banks. Large banks may have high loan loss provisions as an attempt to smooth income (i.e., for the tax savings they generate in times of reduced income). Thus, their loan portfolio is not as risky at the 99th percentile as the loan portfolio of small banks and therefore they are not concerned about holding more ER and cash. Alternatively, these results may reﬂect TARP-CPP funds. Most large banks received TARP funding during this period and the diﬀerential behavior of ER holdings may be due to higher equity holdings (in addition to lower asset holdings) from TARP investments.
We can examine the eﬀect of loan loss provisions at a given capital ratio to attempt to answer these questions. When the capital ratio is held constant at the 50th percentile, the total eﬀect eﬀect of the loan loss provision on excess reserve and cash accumulation is a 41 percent decrease for large banks and a 30 percent decrease for small banks. For a capital ratio at the 90th percentile, the reduction in ER and cash is 60 percent for large banks and 46 percent for small banks. For a capital ratio at the 95th, the reduction is 70 percent for large banks and 57 percent for small banks. Thus, the diﬀerential eﬀect of increasing capital ratios for a given loan loss provision must be related to some diﬀerence in how large and small banks account for loan loss provisions.
Finally, considering the eﬀect of distressed loans on reserve and cash accumulation, we ﬁnd larger eﬀects for large banks, in the range of 0.6 to 1 percent, while the eﬀect for small banks is between 0.1 and 0.2 percent. Large banks have signiﬁcantly higher ratios of bad loans to deposits than small banks, and the variation across large banks is signiﬁcantly higher than across small banks (the standard deviation of the ratio of bad loans 1 to deposits for large banks is 132, whereas it is 0.03 for small banks). These eﬀects are signiﬁcantly diﬀerent at the 1 percent conﬁdence level.
In summary, we ﬁnd that large banks have a much stronger response to increases in the penalty rate than small banks, a stronger precautionary accumulation motive, and the relationship among loan loss provisions, capital ratios, and ER is signiﬁcantly different between large and small banks — a fact that is possibly explained by accounting issues or diﬀerent sources of liquidity.
Our results can be compared to those of Ashcraft et al. (2011), who examine highfrequency (intradaily) movements in ER balances rather than the lower-frequency quarterly data we use. They ﬁnd that small banks appear to have credit constraints that prevent them from actively borrowing in the interbank market as large banks do; they also have limited borrowing or lending at the end of the day and hold larger intradaily and overnight reserve balances. Controlling for balance sheet characteristics, small banks are reluctant to lend at the end of the day, potentially due to the unpredictability of their payment shocks and the large ﬁxed cost to enter the interbank market.
Ashcraft et al. (2011) ﬁnd that in response to higher uncertainty about payments, banks—especially small banks—were reluctant to lend ER when balances were high and borrowing banks were more aggressive in bidding for borrowed funds when balances were low. Even though large banks may be relatively unconstrained and able to access funds easily on the market, aggregate reserve balances can become stuck in the accounts of small banks at the end of the day, leading unconstrained large banks to also keep precautionary balances. Our results are consistent in the sense that the stronger response of large banks to increases in the penalty rate and distressed loans could be caused by lack of available liquidity in the interbank market.
Thrifts are savings and loans institutions with separate charters from domestic commercial banks. Thrifts were originally created with a special function: to channel loans to the housing market. Their loan portfolios and the types of securities they can hold are also more closely regulated than commercial banks (Kwan, 1998). For example, thrifts have restrictions on the percentage of consumer and commercial and industrial loans in their asset portfolio and the percentage of nonconforming loans secured by residential or farm property that banks do not have (Oﬃce of the Comptroller of the Currency, 2013). There are also a number of other restrictions on lending that would reduce the risk of thrifts’ loan portfolios. In addition, in order to maintain its status as a qualiﬁed thrift lender, a thrift is required, among other things, to maintain qualiﬁed thrift investments equal to at least 65 percent of its asset portfolio. These investments include loans to purchase, reﬁnance, and so on, domestic residential or manufactured housing, home equity loans, educational loans, small business loans, and loans made through credit cards, as well as securities based on mortgages on domestic residential or manufactured housing. Our results likely reﬂect these restrictions.
Thrifts respond similarly to all banks to the opportunity cost of holding ER: A 0.1 percent increase in the opportunity cost is consistent with a 1 percent decrease in ER and cash. Thrifts have a large positive response to increases in the penalty rate (the repo rate evaluated at the mean)—larger even than the response of large banks.28 We do not have a comparable measure of equity holdings for thrifts, so we use the Tier 1 capital ratio as a measure of capital adequacy. We ﬁnd that thrifts behave similarly to small banks in terms of the relationship among Tier 1 capital requirements, loan loss provisions, and ER and cash accumulation: A 1 percent increase in the capital ratio when loan loss provisions are zero is consistent with a 17 percent decrease in ER and cash accumulation. Holding loan loss provisions at the 50th percentile, a 1 percent increase in the capital ratio generates a 16.5 percent decrease in ER and cash holdings. As the loan loss provision rises, a 1 percent increase in capital adequacy has a sequentially smaller negative eﬀect on ER and cash accumulation. When loan loss provisions are at the 99th percentile, a 1 percent increase in capital adequacy generates only a 1 percent decrease in ER and cash holdings.
Thrifts’ reserve and cash accumulation response to an increase in distressed loans is similar to the response of large banks: A 1 percent increase in bad loans generates a 6 percent increase in ER and cash as a ratio to deposits.
In previous research, Contessi and Francis (2011) found the lending behavior of thrifts during the ﬁnancial crisis behaved quite similarly to that of small banks. We ﬁnd that thrift cash and ER accumulation patterns are similar to those of small banks except in two dimensions: The response to the penalty rate is much more similar to that of large banks, and the responsiveness to bad loans is much greater than that of 28 We could not separately identify the eﬀect of the opportunity cost and penalty rate using the last observation in the quarter for the index of Treasury repo rates, which was our penalty rate for banks.