Empirical Evidence Supporting Capital Strength

Do Tougher Bank Capital Requirements Matter?
New Evidence from the Eighties

Adam B. Ashcraft

Download the complete article in PDF format

The premise of bank capital regulation is a fear that the presence of imperfectly-priced deposit insurance permits banks to view portfolio and leverage risk as compliments, creating incentives for excessive risk-taking at the taxpayer's expense. While banks certainly would fail in the absence of deposit insurance, the concern is that banks are less likely to carefully reverse bad shocks on their own and are more likely to increase asset risk when they don't have to pay a default premium on their liabilities to depositors. Bank capital regulation has traditionally tried to curb these incentives either by limiting banklev erage or by tying allowable asset riskto the actual level of capital.

Formal leverage requirements in the United States were introduced by the regulators for all but the largest banks in 1981, and then for the multi-nationals in response to the Latin American debt crisis in 1983. Differences in these requirements across regulators and bank size were eliminated in 1985. Since 1988 the centerpiece of commercial bank regulation in developed countries has been the Basle Accord, an international agreement by bankr egulators in G-7 countries as to what constitutes bank capital and bank capital adequacy. Basle was not only important in eliminating most differences in standards across countries that put U.S. banks at a disadvantage in their own market, but also was fundamental in its attempt to limit the ability of banks to transfer risk off the balance sheet. Congress supplemented risk-based capital requirements in 1989 by requiring the banking agencies to take prompt corrective action when banks fail these standards, presumably increasing the regulatory costs of inadequacy, to convince institutions of the need to carefully reverse bad shocks. There were certainly other important reforms, in particular the introduction of risk-based deposit insurance premiums by the FDIC, but it is fair to say that the safety and soundness of commercial banks over the last two decades has been largely guarded through the regulation of bank capital.

The time series of aggregate capital ratios for commercial banks is illustrated in Figure 1, which describes a 200 basis point increase in the ratio of equity to assets and almost a 300 basis point increase in the primary capital ratio since formal capital standards were introduced and subsequently increased. How much of this increase in capital ratios was caused by tougher adequacy standards is the crucial question of this paper. While the analysis below focuses on changes in leverage requirements and not directly on an admittedly more interesting policy change like the implementation of risk-based capital, it is certainly relevant to any evaluation of the Basle Accord. Baer and McElravey (1993) estimate that the capital shortage created by the 1985 increase in standards was similar in magnitude to that created by the introduction of risk-based capital 6 In addition, several strategies that have been employed to evaluate the Accord were developed when economists evaluated these earlier standards. In a recent survey of the literature, Jackson et. al. (1999) claim that at best researchers have reached a broad consensus that in the 1980s and 1990s, relatively low capital banks tended to increase their capital ratios more than better capitalized banks.

Early studies by Keeley (1988) and Shrieves and Dahl (1992) reach exactly this conclusion for large commercial banks and Bank Holding Companies in the first half of the 1980s. Wall and Peterson (1987) claim that changes in capital ratios are better explained by a regression model with regulatory variables than one containing information possessed by the market. Using similar methods, Jacques and Nigro (1994) and Wall and Peterson (1994) reach similar conclusions about risk-based capital. The main limitation of these studies is their failure to appropriately use data before the policy change, implicitly focusing on identifying consequences of the level of standards and not the effect of tougher capital regulation. This latter question really is the only one identified in the data, as the former presumes that there is some way of mapping the observed outcomes of high capital banks subject to leverage standards to the counterfactual outcomes of (federally-insured) low capital banks in the absence of capital regulation. I find this a difficult place to start given a concern about weakly capitalized banks gambling with taxpayers money is the motivation for capital regulation.

It is worth stating explicitly that this paper evaluates how much of the change in capital ratios after 1985 can be attributed to tougher capital regulation, and does not attempt to evaluate pre-existing capital standards. The fundamental identification problem is that tougher adequacy standards generally affect all commercial banks so that there is no true control group. A finding that low capital banks generally increased their capital ratios relative to high capital banks after an increase in capital standards is only relevant in answering this question if such mean-reverting dynamics did not exist before the policy change. This possibility is strongly rejected in Figure 2, which graphs for the population of insured commercial banks the regression-adjusted one-year change in the primary capital ratio given last year's capital ratio. Mean-reverting dynamics are present throughout the period. A direct consequence of this result is that any evaluation of tougher capital adequacy standards that simply compares the outcomes of banks by level of capital as measured before the change in overestimates the effects of a policy change. As the capital ratios of low capital banks rise over time relative to high capital banks, presuming any differences between these groups is fixed implicitly assumes away the presence of mean-reversion and loads the natural adjustment of low capital banks on top of any effects from the change in policy. Moreover, an estimator which differences out the pre-existing dynamics between low and high capital banks eliminates all measured effects of tougher standards in 1985.

There are at least two potential problems with this strategy. First, relatively high capital banks are being used as a control group to identify what would have happened to low capital banks in absence of a policy change. If the control group also reacts to tougher standards in the same direction by increasing capital ratios, we will be underestimating the consequences of the policy change on the low capital group. The time-series in Figure 2 generally support this concern, illustrating that the capital ratios across the entire distribution of banks seem to increase after each policy change. Second, identification is based on the absence of time-varying factors that affect the adjustment of low versus high capital banks. If low capital banks adjust primarily by increasing capital growth and relatively higher capital banks adjust primarily by reducing asset growth, as appears to be the case in the data below, time-series variation in the cost of external finance or loan opportunities could affect these groups differently. Moreover, I argue in the analysis below that when bad shocks temporarily reduce bank capital, there is a change in the composition of low capital banks toward banks with higher target capital ratios. When changes in capital regulation are either prompted by or accidentally timed near these shocks, exploiting differences in initial leverage to identify the likely effects of regulation is especially dubious.

The only way to properly identify the causal effect of interest is to locate a large group of banks with similar asset powers that are lending in similar markets, and to randomly increase capital requirements for one subset of this group. While such deliberate experimentation by the regulators is generally not possible, it turns out that something pretty close to this occurred by accident in the mid-1980s. In particular, I exploit the plausibly exogenous elimination of differences by Federal Reserve System membership status in leverage requirements for community banks (those with less than $1 billion is assets). In particular, until early 1985 when leverage requirements were made uniform across the regulators, member banks effectively had requirements of 7 percent while nonmember banks had requirements of 6 percent. I analyze the 100 basis point increase in minimum leverage requirements by the FDIC in 1985, using member banks as a control group. Controlling for pre-existing dynamics across membership status and permitting precautionary behavior, I corroborate the finding above tougher leverage standards in 1985 had little impact on bankb ehavior. Moreover, I find that despite lower minimum capital standards in 1985, non-member community usually banks raised their capital ratios as much as member banks, suggesting that the level of capital requirements was also unimportant. In any case, analysis of this natural experiment does not change the basic message from above: tougher leverage requirements did not matter.

This paper provides important lessons when evaluating the impact of risk-based capital requirements on bankb ehavior. Several economists pointed to the ahistorical portfolio shift from loans into securities that occurred in the early 1990s as evidence that Basle had important real effects. Figure (3) documents this shift 1989-1994, which is clearly too large to be explained by the recession when differencing across 1980-1982. On the other hand, there is a clear problem with this interpretation in more recent data. While risk-based capital requirements have if anything gotten tougher over the last decade, commercial banks have completely reversed this portfolio shift. While advocates of risk-based capital standards could reasonably argue that standards have been eroded over time as banks discovered or created loopholes in them there is another interpretation of these facts: tougher standards were simply not important and that the observed portfolio shift was in response to an ahistorical level of bankf ailures.

When banks view FDIC guarantees as a call option on the value of assets, there are incentives for excessive risk-taking. This implies that any minimum capital requirement is binding as banks seekto maximize their leverage. I interpret the evidence above as a strong rejection of this simple call option model of bankb ehavior, suggesting that the moral hazard incentives for banks have been largely overstated and that there are possibly market-based incentives for banks to hold capital. Identifying and quantifying these incentives is thus the next important step in properly evaluating the likely effects of past bank regulation and of course when designing regulation for the future.

Institutional details are provided in Section 2 and the data employed is described in Section 3. Analysis of the 1985 standards is performed in Section 4, of differential changes across membership status in Section 5. Directions for future research are outlined in Section 6.

Download the complete article in PDF format