Regulatory Capital Arbitrage and the Potential Competitive Impact of Basel II in the Market for Residential Mortgages

See Basel Committee on Banking Supervision 2004.

Two alternative sets of rules—the Foundation approach and the Standardized approach—incorporate more risk sensitivity than Basel II but stop short of the variations in risk sensitivity of capital requirements associated with the AIRB approach. They are not being implemented in the U.S.

See, for example, Vice Chairman Ferguson’s address, “Basel II: Some Issues for Implementation,” at the Institute of International Finance, New York City, June 17, 2003, available online at http://​www.​federalreserve.​gov/​boarddocs/​speeches/​2003/​20030617/​default.​htm.

Our attention is focused primarily upon competition among banking and savings organizations subject to Basel II. The current role of the two large government-sponsored enterprises (GSEs)—Fannie Mae and Freddie Mac—and the potential impacts of heightened competition for residential mortgages between them and the adopters are discussed in more detail by others (Frame and White 2004, 2007). By our understanding, these impacts are not deemed central to the decision facing the regulators about Basel II.

The process can be viewed as an example of the concept of the “regulatory dialectic” (Kane 1981) and is regularly cited in the banking literature as a concept to describe the “cat and mouse” game between banking organizations and their regulators (e.g., Kovakimian and Kane 2000; Cabral dos Santos 1996).

There is a substantial theoretical literature on the relationship between capital regulation and bank risk taking. The literature generally suggests that banks will increase portfolio risk in response to a binding regulatory capital requirement (Allen 2004). Under special conditions, this relationship need not hold; for instance, if relative risk weights under the regulatory standard align with relative economic capital (Rochet 1992).

Although this is a simplifying assumption, it does reflect a belief embedded in our paper that the cost of debt to banking institutions is likely to be a below market rate due to the nature of the financial safety net available to them—payment systems, deposit insurance, and, for some, the implicit guarantee associated with the “too big to fail” option. More specifically, we are simply assuming that deposits are the dominant form of debt and that the deposit rate paid by banks is less responsive to changes in its leverage ratio than it would be in a world without the safety net. So incorporating an adjustment in the cost of debt into this model would alter the specific outcomes we report, but it would not alter the bank incentive to maximize leverage. It is this reasoning that underlies our emphasis upon the maximization of leverage.

For instance, concavity could be introduced by specifying economies of scope on the operating cost side, or diversification benefits that affect economic capital. We do not believe that introducing concavity in other ways would affect the basic points we wish to emphasize. In particular, introducing the benefits of diversification raises another thorny problem regarding distribution of the capital surplus (Myers and Read 2001). The Myers and Read study summarize and add to the literature on the issue of how insurance companies and, more generally, financial institutions can allocate a capital surplus among divisions or asset classes. Although they do propose a rule for such intra-firm allocations, they acknowledge the complexity of the issue, especially for infra-marginal changes in investment decisions.

For a more intricate model with similar implications, see the previously cited study by Repullo and Suarez.

The two other major components are the cost of debt financing (typically approximated by the cost of deposits in the case of banks) and the cost of originating and servicing the mortgage (largely operating costs and the cost of requisite infrastructure.)

The GSE typically purchases loans from one or more originators and then packages them into an MBS that is sold to an investor. The GSE retains the credit risk on the pool of mortgages (that is, it provides a credit guarantee, exclusive of the portion that is assigned to mortgage insurers, if applicable) and it receives a “guarantee fee” in return. The interest rate risk is transferred to the investor who purchases the MBS, who in turn receives coupon payments. In essence, this particular securitization process involves the sale of credit risk protection or a credit guarantee to the investors in the MBS in exchange for a guarantee fee.

A GSE must also hold 250 basis points for a prime MBS held in a GSE portfolio. Since 45 basis points is associated with the credit risk guarantee associated with all of its MBS whether held in portfolio or not, the implicit minimum regulatory capital charge for bearing the interest rate risk of the MBS held in portfolio is 205 basis points (250–245).

As of year end 2003, the GSEs held about 950 billion dollars of the 3 trillion dollars in outstanding GSE MBS as of year-end 2003, representing a 31 percent share, while banks and thrifts held about 960 billion dollars of these MBS plus collateralized mortgage obligations backed by GNMA as of year-end 2003 (from the website of the Federal Deposit Insurance Corporation: http://​www2.​fdic.​gov/​sdi/​main.​asp)

Separately, U.S. banks are subject to a set of “leverage” requirements (not part of the Basel Accord) that define required capital in terms of non-risk-adjusted assets. These vary by category of capitalization. For example, a “well-capitalized” banking organization has at least total capital in excess of 10 percent of risk-weighted assets and Tier 1 capital in excess of 5 percent of total assets; “adequately capitalized” ratios are 8 and 4 percent, respectively.

The study is available at http://​www.​federalreserve.​gov/​generalinfo/​basel2/​docs2003/​asset-correlation.​pdf.

Since not all of these meet the credit quality requirements of the GSEs the actual percentage of conforming loans sold likely exceeds 80%. We calculated the percent sold for loans originated for purchase or refinance of one-to-four family, owner-occupied properties in 2003.

The differential also is found to be robust to controlling for factors such as size of the banking organization, borrower income, and geographic region.

One possibility is that the costs associated with capital arbitrage transactions with the GSEs are smaller than those associated with other non-banks, due, for instance, to economies of scale, established channels or relationships between individual banks and the GSEs. Another is that both banks and the GSEs may have cost of debt or informational advantages relative to other non-banks. A fundamental premise of our analysis is that regulatory capital arbitrage between adopters and non-adopters under Basel II will be less costly than is currently the case between banks and nonbanks other than the GSEs. Reasons why we expect this to be the case include the existence of established origination networks of adopters and of correspondence networks between non-adopters and adopters, and a relatively level playing field with respect to the cost of debt and information. We note disagreement on this premise between ourselves and the Federal Reserve study cited in footnote 4.

Indirect affirmation of this view is found in a recent study that evaluates the probability that the originator of a mortgage will either hold or sell it (Ambrose, B. W., LaCour-Little, M., and Sanders, A. B., “Does regulatory capital arbitrage, reputation, or asymmetric information drive securitization?” University of Kentucky, August 2004). The estimated coefficients of their empirical model of the decision to hold or sell a loan can be used to infer the sensitivity of this choice to the lender’s cost advantage. We performed such calculations and concluded that the model suggests an elasticity of loan sale three or higher. Although this elasticity measure is not identical to the one we have in mind, the values calculated for it are consistent with a highly competitive market structure.

Other ways of reducing the risk of this investment such as options could be included as capital substitutes; we simply assume that the bank chooses the least costly way of hitting its risk tolerance targets with capital or capital substitutes.

We use total capital in our example but the argument could also be made in terms of tier 1 capital. The amount of tier 1 capital required by non-adopters would be 200 bps in this case.

See the website of the Office of Thrift Supervision for these two tables: http://​www.​ots.​treas.​gov/​pagehtml.​cfm?​catNumber=​10

We do not include an explicit cost of transferring the mortgage. They are likely to be quite small at this point, although we do in the case our discussion of newly originated loans, case 2. We could include such costs at this point as well even though they are not essential to case 1. They are also likely to be quite small given the extensive network of mortgage brokers who may simply end up selling more loans to the adopters and bypassing non-adopters more frequently, and possibly even in the case of seasoned loans, given the existence of established correspondent networks between large and small banks.

In contrast, economies of scale and direct channels from originators may allow the GSEs to accomplish unbundling at relatively low cost.

Another possibility is that the optimal mechanism will be affected by the Basel II rules that pertain to capital requirements for securitization. The particular type of process for the credit transfer would not affect the essence of our story.

Tables with our calculations are available upon request of the authors.

The ex-ante ROE remains 15% for the entire amount of the investment by non-adopters because we used this assumption in the calculation of the credit guarantee fee.

A detailed explanation of the assumptions embedded in our estimates is available in a separate appendix available from the authors.

Using the definition offered by the FDIC—lenders with at least 50% of their assets in the form of residential mortgages and mortgage-backed securities—243 commercial banks (among 7,600) fit this description.

See, for example, the comments of William Longbrake on behalf of Washington Mutual at: http://​www.​federalreserve.​gov/​SECRS/​2003/​November/​20031106/​R-1154/​R-1154_​67_​1.​pdf