BI: Incentive of Capital Requirements
The Practical Incentive Effects of Different Approaches to Capital Requirements
Several approaches to setting bank capital requirements are competing for policymakers’ favor: the current risk-weighted approach embedded in Basel III and variants; a leverage ratio; and the use of stress tests. In practice, as is currently true in the US, some combination of these will be used.
One of the key questions in considering capital requirements is what the incentive effects on banks will be and therefore how their behavior will change with different modifications to the requirements. This paper examines the issue by answering the following questions:
- Why would banks change their behavior based on capital requirements?
- What are the incentive effects which are desired by policymakers?
- How will banks decide how to respond to capital requirements?
- What are the likely effects of different capital regimes?
- How will multiple approaches interact when used together?
In offering answers to these questions, I will be drawing on both my theoretical analyses and discussions with other analysts and also on the roughly two decades of experience I had earlier in my career as an investment banker advising financial institutions.
A key point of this paper is that the sophisticated banks that dominate the financial system will directly incorporate the effects of capital regimes into their internal pricing models. This will result in the incentive effects flowing directly through to almost all decisions about business mix and pricing. Thus, incentive effects will be much more than academic constructs, they will play through into actual business decisions via banks’ internal markets much as changes in monetary policy trigger substantial changes in market prices in the economy as a whole.
Why would banks change their behavior based on capital requirements?
Banks view capital as more expensive than other forms of funding and therefore attempt to minimize its use while achieving the desired, or required, level of safety. At first blush, the rationale is obvious. Common shareholders of banks historically indicated that they required expected returns of roughly 15% annually, although they seem to have been willing to accept something closer to 10% after the fact. In the current market, double-digit expectations still seem to be built into pricing, despite historically quite low inflation levels and interest rates, perhaps in part reflecting the great uncertainty currently surrounding the banking business and its regulation. In contrast, deposit funding is far cheaper, with explicit rates of close to zero currently, plus allocated costs for branch networks, advertising, etc. that add perhaps another percentage point or two. Bank debt issued in the capital markets is somewhat more expensive, but still much cheaper than the required return on equity.
However, Modigliani and Miller, two Nobel Prize laureates, proved years ago that, under very specific conditions, a firm’s total cost of funding should not change with changes in its mix of funding sources. There are some academics who argue that this is true for banks as well, when viewed from the point of society as a whole, and therefore policymakers should require much higher capital levels than is currently the case. I have written extensively in opposition to this view. Regardless of where one comes out on that argument, it is important to realize that virtually no one argues that banks as private actors are indifferent to capital requirements, in part due to the tax deductions that makes interest payments cheaper than stock dividends. Therefore, they should, even in theory, attempt to minimize capital usage within other constraints.
More importantly, banks clearly show through their actual behavior that they consider equity to be substantially more costly than debt or deposits. It is indisputable that, in practice, they make strong efforts to minimize their capital usage within the constraints of their other objectives and regulatory requirements.
What are the incentive effects which are desired by policymakers?
The Basel Committee on Banking Supervision recently released an excellent discussion paper on bank capital requirements that summarizes three key objectives for capital requirements: risk-sensitivity; simplicity; and comparability. Risk-sensitivity is crucial, because the core reason for banks to have capital is to protect against the risk of bad outcomes, especially insolvency. Therefore, the level of capital needs to be commensurate with the risk. However, simplicity and comparability are also important, because they help make it possible in practice to be comfortable that capital is indeed adequate and that required capital levels are reasonably similar across banking jurisdictions and within them. (If requirements differ significantly, there is a danger of regulatory arbitrage, with business moving to the laxer jurisdictions.)
The core approach of Basel III, the third version of the global Basel Accord on Bank Capital, is to calculate the capital needs of a bank by weighting each asset category according to its risk and then adding the totals. This achieves a great deal of risk-sensitivity and would probably continue to be the single approach used by most jurisdictions, if policymakers and others thought the risk weightings were perfect. Instead there are a whole series of problems that have considerably harmed the credibility of the risk weights.
Policymakers are now generally looking for a combination of approaches, including fixing problems with the risk weightings, that will retain the risk-sensitivity of the risk weightings while countering the worst potential problems. In particular, Basel III adds, as a supplemental measure, a leverage ratio calculation that is fairly insensitive to risk. The US has used a leverage ratio as a second constraint for some years.
In brief, policymakers want a capital adequacy regime that captures the bulk of the differences in risk across asset categories while ensuring that there is a certain minimum level of total capital per dollar of assets, (and asset equivalents for off-balance sheet items), even if the risk weighting on its own would allow lesser levels of capital. The leverage ratio constraint is intended to counter the dangers of “gaming” of risk weightings by the banks as well as the intellectual errors that can lead banks and their regulators to believe that certain items, such as sovereign debt, have low levels of risk when in reality the risks are higher than that. (Although I am not sure this can be proven, my intuition is that the biggest dangers from faulty risk weightings come from assigning zero or near-zero weights to assets that may truly be relatively low risk, but should have risk weights well above zero.)
Therefore, the desired incentives are for banks to set aside capital in proportion to the risk of different assets while also ensuring that total capital is sufficient to handle a shock whereby banks lose a few percentage points of value across the board, to reduce the dangers from consistent misperceptions of risk levels.
How will banks decide how to respond to capital requirements?
Because they view capital as costly, banks try to use the lowest amount that also allows them to meet all of their safety objectives. Conceptually, this means that they use the highest minimum capital requirement from a set of different measures, including:
Regulatory requirements, plus a buffer. Banks will be forced by regulators to hold at least certain minimums, plus banks will add a margin for error to avoid serious risk of slipping below these minimums if they take losses.
Rating agency requirements, plus a buffer. Similarly, banks usually target a certain credit rating that carries with it some minimum capital requirements. Higher ratings make it easier to attract customers and counterparties, but bring with them greater capital needs and costs.
Economic capital requirements. Banks run their own models to determine how much capital they need in order to operate with their desired level of safety, as measured by the probability of insolvency in any given year. They may be willing to accept a risk of failure every 1,000 years or perhaps 200 years, depending on their conservatism. The models for economic capital requirements tend to be similar to the ones used for regulatory risk weightings, but there can be substantial differences in detail.
Other constraints. In theory, equity investors or other constituents might have stiffer requirements for the bank than these other measures, but it is generally not the case in practice.
At any given time, one of these constraints will be the binding one, meaning that it requires more capital than the others. The incentive structure for the bank depends significantly on which constraint is binding. It appears to be the case that the binding constraint for many banks going forward will be the regulatory capital requirements, which is why the choice of regulatory capital regime may have major incentive effects. This is a quite different situation from the case a few years ago, when relatively low regulatory capital requirements generally meant that large banks determined their capital either primarily through ratings targets or from their own economic modeling.
In practice, capital requirements affect decisions by sophisticated banks through the formula used to price capital throughout the bank. Banks generally try to ensure that each of their units and each of the transactions of those units earn the bank enough profit to more than cover the target cost of capital multiplied by the amount of capital allocated to that unit or transaction. Thus, they effectively set up an internal market for capital in order to maximize the efficiency of their firm-wide capital allocation, in much the same manner that our economy as a whole relies on market mechanisms to maximize efficiency.
In theory, a bank would calculate all possible combinations of activities and use a complicated program to determine which set of activities would produce the maximum profit for a given level of capital and to determine whether additional capital should be raised or some capital returned to shareholders.
In practice, this is infeasible. Instead, the bank will generally construct a much simpler formula to be used in pricing transactions that attempts to roughly calculate the marginal impact on total capital requirements of a particular activity. If risk-weighted capital is the binding constraint, then the appropriate risk weighted asset (RWA) figure for that activity is used. If the leverage ratio were the binding constraint, then the amount of assets and asset-equivalents related to the activity would be fed in.
However, the amount of capital for a given activity may be radically different depending on which capital requirement is binding at a given time. This creates a risk that the simple approach just described would cause a sharp enough change in the bank’s behavior, and indeed business model, that the originally binding constraint may cease to be the binding one going forward. One could imagine a bank caroming back and forth in its business strategy as the binding constraint shifts over time, which will occur not only due to its own decisions, but market factors that are out of its control, including movements in the financial markets that may raise or lower its total capital level.
It seems probable that banks will therefore include multiple factors in their capital pricing and allocation formula in order to reflect different potential constraints, with greater weighting on the constraint that is most binding. When I was working at J.P. Morgan prior to the crisis, the pricing formula for a transaction included a capital charge that related not just to risk-weighted assets, but also to the total assets involved in the transaction. I understood that this was intended to reflect the fact that one or more of the rating agencies were concerned about the leverage ratio, in addition to risk weightings, and therefore J.P. Morgan wanted to ensure there were incentives to pay attention to total assets as well as RWA. It seems likely that such an approach will be common going forward. The formulas will presumably also be tailored to take account of the likely impact of a given activity on the capital required to meet the criteria of stress tests, such as the Comprehensive Capital Analysis and Review (CCAR) exercise. The latter will generally be of similar nature to the RWA approach, but the effective weights may differ from the ones employed under Basel III.
In essence, the capital pricing formula will be intended to move the bank in the direction of the theoretically optimal mix of activities that would take place under the binding capital constraint. This will often be a “corner solution” where any further movement of activity in the same direction would cause the originally binding constraint to produce a lower capital requirement than a new binding constraint, such as moving from leverage as the constraint to RWA as the constraint. Thus, the formula will create incentives for decisions on the margin that are similar to what the bank would do if the binding constraint were the only constraint, but with automatic stabilizers to avoid excessively rapid changes that could lead to instability and excessive movement in strategic and tactical direction.
It should also be noted that there will be other regulatory elements that affect the pricing formula used in evaluating bank activities and individual transactions. In particular, the new Basel III liquidity constraints incorporated in the Liquidity Coverage Ratio and Net Stable Funding Ratio will limit the degree to which banks could respond to the leverage ratio as a constraint by reducing high-quality, and low return, liquid assets such as short-term Treasury securities. Banks will almost certainly include a charge in their pricing formula for activities that create a drag on the LCR or NSFR and a credit for those that ease the liquidity constraints.
What are the likely effects of different capital regimes?
The most obvious differences in incentive effects relate to those that are created when RWA is the binding constraint as compared to when the leverage ratio is. The RWA calculations incentivize banks to undertake those activities where they are sufficiently well paid for the level of risk undertaken, as measured under the Basel III rules. All else equal, this encourages lower-risk activities, unless the pricing advantage of taking on more risk outweighs the additional capital need. In consequence, this also encourages those activities for which the regulatory risk weights are too low, such as may be the case with some sovereign debt and some mortgage transactions, and away from ones where the RWA is too high, as may be the case with some trade finance transactions.
On the other hand, the leverage ratio is relatively insensitive to risk and therefore encourages higher risk activities at the expense of those with lower risk, since the same capital can earn a higher expected return associated with the higher risk. Thus, the leverage ratio pushes banks away from sovereign debt and towards higher risk loans. One of the reasons many support the use of leverage ratios as a complement to RWA calculations is in the hopes that the leverage ratio’s penalization of activities with excessively low risk weights will appropriately offset the built-in incentives in the RWA calculations.
How will multiple approaches interact when used together?
The principle effect of using multiple regulatory capital approaches, such as the leverage ratio alongside the RWA calculations, is likely to be to push banks towards “corner solutions”. That is, they start with their existing business mix and move in the direction encouraged by the binding constraint until they are on the edge of finding that the other constraint becomes the binding one. The reason that there is likely to be movement all the way to the corner is because the incentive effects of the two radically different capital approaches push towards very different business mixes. As long as the initially binding constraint remains binding there are likely to be powerful incentives to keep moving towards optimization under that constraint by shifting business mixes. It is only when the corner is hit, that the incentives for further movement in that direction vanish.
This push towards the corners could be reduced if banks view the potential volatility of business mixes as too harmful and incorporate in their capital pricing formulas a substantial weighting for the initially non-binding constraint in order to reduce the tendency to change strategies. This would have a cost, however, by producing sub-optimal business mixes at each point in time.
The actual changes in business mix are hard to determine from the outside and, indeed, may be quite hard to predict from the inside. The direction of change will be fairly clear once one knows the initially binding constraint, but it will be harder to know how far the bank will go in that direction and in what ways the individual business units will achieve the new targets. In this sense, it may be like raising fuel efficiency standards for cars. Some things will be clear. All else equal, there will be an attempt to reduce the weight of cars and to improve the efficiency of engines. However, there will remain many different ways in which the overall goal can be achieved.
The banking industry is correct to point out that very low risk activities, such as holding Treasury securities, will be severely penalized by a leverage ratio constraint as compared to RWA constraints, and that this is very likely to reduce bank’s holdings of those securities and related activities such as securities financing transactions. However, it may be difficult to discern what the total mix of changes will be within the bank, since not all changes will occur in regard to the extreme cases and there are other constraints such as the liquidity rules or specific business needs that may limit the response through the most obvious actions. Areas of activity where banks dominate may also see price responses that fully or partially offset the increased cost of the capital allocated to that activity.
There is much that could be learned by trying to construct internal pricing rules that mimic those we would expect banks to use. Asset mixes, with associated pricing, could then be introduced to see what banks are likely to do in those situations. These mixes and prices would necessarily be fairly simplified, but they could still be broadly realistic. This type of exercise could help considerably in finding an appropriate calibration of the various regulatory standards.
(Emphasis added, citations omitted, link added)