Publications

Therefore, if banks are not to be allowed to fail, it is essential that corrective action be taken while the bank still has a manageable cushion of capital. This is particularly crucial since low or negative capital often tempt bank managers to try desperate remedies such as offering high rates of interest on deposits to fund credit to high-risk borrowers or ';gambling for resurrection';, to use a phrase coined by Dewartipont and Tirole (1994). Even the Basle Committee has strongly endorsed the need for supervisors to take timely corrective action when banks fail to meet capital adequacy ratios or other prudential requirements. Yet, one of the commonest complaints about bank supervisors is that they intervene too late in problem banks.ⁱ This has led many observers to suggest that interventions should be guided by rules rather than be left to the discretion of the supervisors. A recent study by Jones and King (1995) applies the definition of Basle Accord to a confidential data set of US banks in the ‘eighties and finds that a number of institutions which would have exceeded the 8 per cent hurdle would, in fact, with a high probability, have been insolvent within two years.

The case for automatic rules rests on the premise that they lead to prompter action, which is important as the costs of restructuring a bank are likely to rise, the longer the action is delayed. Several arguments can be advanced to support this case. Forbearance, or expecting that the problem will solve itself, is always a tempting option, especially given the usual lack of precise information about the extent of a bank’s problem. If a large number of banks are simultaneously in trouble, political economy considerations might prevent contemplating the short-run costs of radical action. Alternately, supervisors may fear that intervention in one bank could spark a run on others, as occurred in Indonesia in November 1997 (Sheng, 1996). As a consequence, rule-based methods of intervention, especially if enshrined in legislation, may be particularly helpful for supervisors to take decisions based on established procedures and principles.

The best-known examples of rules are the compulsory quantitative triggers (in relation to bank capital levels) for action by the supervisors set in the 1991 US Federal Deposit Insurance Corporation Improvement Act (FDICIA) (Table 3.1). Studies by Shrieves and Dahl (1992), and more recently, Peek and Rosengren (1997) find that excessive risk-taking among undercapitalised banks is, at least, partially constrained by regulation, while Gilbert (1992) finds that the length of time a bank is undercapitalised prior to its failure does not affect its ultimate resolution cost.

Similar rules have been adopted in some industrial economies and in a number of emerging economies (Table 3.2). Once capital falls below 8-9 per cent, such rules typically require banks to draw up plans for recapitalisation, limit or prohibit dividends and impose limits on risk-taking. Restrictions often involve limiting new acquisitions or restricting interest rates on deposits. When capital falls to very low levels, the authorities can force mergers or acquisitions, or proceed to closure. Such rules, however, would be rarely applied to a large bank-in such an eventuality, some observers believe that greater discretion would inevitably condition supervisors’ responses (the ';too-big-to-fail'; argument).

1. Based on current minimum CAR of 8 per cent.

2. CAR: Capital Adequacy Ratio.

Source: Hawkins and Turner (1999)

After examining country experiences, one can surmise that supervisory authorities in different countries incorporate a blend of discretion and rules. For example, the less rigid criteria under some discretionary regimes are accompanied by some quantitative ‘triggers’, with the actions taken often similar to those in a rule-based system. Also, some of the rule-based systems are yet to be tested in a banking crisis: some more discretion may have to be used in practice. Finally, there is also the broader issue of what can be done to strengthen the political incentives to implement banking reforms. A banking crisis itself should lead to the adoption of an improved incentive or supervisory framework. Caprio and Klingebiel (1996) while detailing a number of success cases (Chile, Argentina and Hong Kong), observe that out of the 64 cases of bank restructuring, such success stories were few and far between.ⁱⁱ

In view of the above considerations, various actions are being contemplated by the regulators to detect early warning signals of crisis. One such measure which has been widely discussed in the literature has been Prompt Corrective Action (PCA). The major focus of PCA is to detect incipient signals of distress and promptly undertake remedial measures. In developed countries, the major focus of PCA is on capital ratios and this is not without justification. As Dahl and Spivey (1995) have observed, PCA (a) enables to reduce the losses for deposit insurers by discouraging healthy banks from becoming undercapitalised, and, (b) enables to reduce the number of failures among undercapitalised banks.

While the adoption and implementation of PCA focused attention on bank capital ratios, two issues merit further attention. First, did PCA cause banks to increase their capital ratios, or is the increase attributable to some other factors such as bank income levels in the early 1990s? Second, a number of theoretical and empirical studies suggest that increasingly stringent capital standards in general, and PCA in particular, may have the unintended effect of causing banks to increasing their level of portfolio risk.

In order to achieve our objective, we use a simultaneous equation model along the lines developed by Shrieves and Dahl (1992), and later modified by Jacques and Nigro (1997) to examine the possible consequences of PCA on bank capital and portfolio risk levels. However, unlike prior studies on this topic, by using a simultaneous equation model, the endogeneity of both bank capital and risk is explicitly recognised in our approach, and as such, the impact of possible changes in bank capital ratios on risk in a bank’s portfolio can be examined. Table 3.3 summarises the major papers that examine this issue.

affiliates and affiliated banks and others. Finally, once a bank is critically undercapitalised, it faces not only more stringent restrictions on activities, but also the appointment of a conserver (receiver) within ninety days of becoming critically undercapitalised^v.

It needs to be recognised that the ratio of 8.0 percent for the Capital to Risk Asset Ratio (CRAR) is recommended as a minimum ratio and there is a view that higher capital adequacy ratio may be necessary in select cases and in select markets. Therefore, in our judgement, while banks with CRAR of 8.0 percent may be considered as sufficiently capitalised, banks with a CRAR of 10.0 percent or above and tier–I capital not below 6 per cent may be considered as adequately capitalised. The FDIC Improvement Act of 1991 in the US considered banks with risk-weighted capital of 10.0 percent and a tier-I capital of 6.0 percent and above as well capitalised, while adequately capitalised institutions have minimum thresholds of 8 per cent and 4 per cent, respectively. Banks that do not meet the BIS capital adequacy criteria and have risk weighted capital less than 8 per cent and tier-I capital below 4 per cent can be considered as undercapitalised. Banks with even lower CRAR may be classified as ';significantly undercapitalised';. Although there is not much to distinguish between significantly undercapitalised and critically undercapitalised banks, we prefer to make a distinction between the two.^vi Table 3.4 summarizes the information.

Table 3.5 shows the number of public sector banks by PCA zone over the period 1997 to 1999. The choice of the period is dictated by two considerations. The first was the availability of data for the same. More importantly however, since the introduction of prudential norms pertaining to capital adequacy standards, in 1992-93, banks were initially in a state of flux, adjusting their balance sheets in response to the new guidelines. Consequent upon the introduction of off-site returns data in 1997, it is now possible to check the feasibility of implementation of a framework for PCA in the Indian context.

As Table 3.5 reveals, at year-end 1996, 19 public sector banks were classified as absolutely/well capitalised. In 1997, the number increased to 25. By 1998, only one bank was in the undercapitalised category; the capitalised banks, numbering 26, by contrast, accounted for 76 per cent of total assets of SCBs. The figure has remained at 26 for 1999, with these banks accounting for 78 per cent of total assets.

It seems reasonable to assume that the introduction of capital adequacy standards have, in general, been effective in increasing the capital ratios of banks. This position has however, not been without its critics. For one, it has been argued that regulatory capital standards may have led banks to increasing levels of portfolio risk. Research by Koehn and Santomero (1980) and Kim and Santomero (1988) have shown, using the mean variance framework, that regulatory capital standards cause leverage and risk to become substitutes and that as regulators require banks to meet more stringent capital standards, banks respond by choosing assets with greater risk.^v Thus, increases in minimum capital standards by bank regulators cause banks to increase not only their capital ratios, but also have the unintended effect of causing them to increase their level of risk.

While one primary purpose of early liquidation/closure is to prevent banks from taking increasing levels of risk as they approach insolvency, research by Davies and McManus (1991) demonstrates that early closure may fail to protect the deposit insurance fund from losses because it creates incentives for banks to increase portfolio risk by increasing their holdings of high-risk assets. An even more recent study by Dahl and Spivey (1995) on undercapitalised banks regarding identification of factors that influenced bank recovery to a position of adequate capitalisation notes that (a) there appears to be only a limited capacity for banks to ‘correct’ positions of undercapitalisation by growth limitations or dividend restrictions and (b) the impact of profitability on recovery is greater, the longer a bank remains undercapitalised. As such, the design of the PCA standards has important implications not only for optimal capital levels, but also for the level of risk, and ultimately, the safety and soundness of the banking system as a whole.

To examine the possible impact of the PCA standard on bank capital ratios and portfolio risk levels, a simultaneous equation framework is developed, on the lines of Shrieves and Dahl (1992) and is modified to incorporate PCA zones. In their model, observed changes in bank capital ratios and portfolio risk levels are decomposed into two components, a discretionary adjustment and a change caused by an exogenously determined random shock, such that

(3.1)

(3.2)

where DCAP and DRISK are observed changes in capital ratios and risk levels for bank j in period t, D^dCAP and D^dRISK represent discretionary adjustments in capital ratios and risk levels and E and U are exogenous shocks. Recognising that banks may not be able to adjust their desired capital ratios and risk levels instantaneously, the discretionary changes in capital and risk are modeled using a partial adjustment framework. As a result:

(3.3)

(3.4)

Thus, observed changes in bank capital ratios and portfolio risk in period t are functions of the target capital ratio CAP^*_j,t and target risk level RISK^*_j,t, the lagged capital ratio CAP_t-1 and risk levels RISK_t-1 and any random shocks. The target capital ratio and risk levels are not observable, but are assumed to depend upon some set of observable variables, including the size of the bank (SIZE), bank’s net income to total assets in period (t-1) (TOTINC), changes in portfolio risk (DRISK_j,t) and capital ratios (DCAP_j,t), while the exogenous shock that could affect bank capital ratios or risk levels is the regulatory pressure brought about by PCA.

Specifically, SIZE is measured as the natural log of total assets. As noted by Shrieves and Dahl (1992), size may have an impact on a bank’s capital ratios and the level of portfolio risk because larger banks have greater access to capital markets. Following Dahl and Shrieves (1990), the ratio of net interest income to total assets in period (t-1), TOTINC, is included to recognise the ability of profitable banks to increase their capital ratios by using retained earnings. In addition, as noted by the partial adjustment model, lagged capital ratios and risk levels are included to measure the fact that banks adjust their capital ratios and risk levels to their target levels over time.

To recognize the possible simultaneous relationship between capital and risk, DCAP_j,t and _DRISK_j,t are included in the risk and capital equations, respectively. Shrieves and Dahl (1992) note that a positive relationship between changes in capital and risk may signify, among other possibilities, the unintended impact of minimum regulatory capital requirements or even managerial risk aversion, while Jacques and Nigro (1997) note that a negative relationship may result because of methodological flaws in the capital standards underlying PCA.^vi Empirical estimation of the simultaneous equation system requires measures of both bank capital and portfolio risk. Following previous research, portfolio risk is measured in two ways: using the ratio of total risk weighted assets to total assets (RISK) and gross non-performing loans as percentage of total assets (GNPA)^vii. Avery and Berger (1991) have shown that RISK correlates with risky behaviour, while other studies (Berger, 1995, Shrieves and Dahl, 1992) using non-performing loans (GNPA).

Since regulatory influence is a cornerstone of the hypotheses involving regulatory costs and minimum capital standards, a binary variable (PCAA) reflecting the degree of regulatory pressure is also included as a determinant of target capital and risk levels. This variable is defined as follows:

PCAA=1, if a bank is adequately/well capitalised; otherwise=0.

These variables allow banks across different PCA zones to respond differently, both in capital ratios and portfolio risk. A priori, banks in undercapitalised group would be expected to have the strongest response because PCA imposes penalties on their activities. Furthermore, adequately/well capitalised banks, PCAA, may increase their capital ratios or reduce their portfolio risk if they perceive a significant penalty for not being considered well capitalised, or if they desire to hold a buffer stock of capital as a cushion against shocks to equity (Wall and Peterson, 1995). Besides being included as a separate variable, PCA is included in an interaction with the lagged capital ratios. The use of this term allows banks in different PCA zones to have different speed of adjustment to their target capital ratios. As such, banks in the undercapitalised PCA zones would be expected to adjust their capital ratios at faster rates than well/adequately capitalised banks. An alternative method of accounting for the ‘regulatory effect’ is to stratify the sample relative to the regulatory benchmark total capital, tier-I capital and the leverage ratio levels, and estimate the model separately for each subset of banks. This approach has the advantage of allowing all model coefficients to differ across the capital strata, and provides additional insights as to the role of regulatory influence as a determinant of the observed relationship between changes in risk and capital.

Given these variables, equations 3.1 and 3.2 can be written as^viii:

(3.5)

(3.6)

where x_j,t and z_j,t are error terms, and PCAA*CAP_j,t-1 is the interaction term, which allows a bank’s speed of adjustment to be influenced by the PCA zone in which the bank is in^ix.

Since the right hand side of both equations include endogenous variables, simultaneous estimation of equations (3.5) and (3.6) is carried out by two-stage least squares. Under the null hypothesis that changes in risk and capital do not influence one another, the coefficients m₃, m₅, g₂ and l₂ will not be significantly different from zero.

It has been stated earlier that data set for the present study spans from 1997:Q1 to 1999:Q4. In this present context, the analysis for a particular year needs to take into account the data for the year as well as those for the previous as well as the subsequent year. In view of the above, the present exercise focuses on 1998 (which uses data for 1997, 1998 and 1999). With greater availability of data, it is possible to carry out the analysis for other years as well. As noted earlier, a significant decline occurred in the number of all types of undercapitalised institutions post introduction of capital adequacy norms. Alternately, in studying the impact of risk based capital standards, Haubrich and Watchel (1993) note that because the composition of bank portfolios can be changed quickly, and because banks appear to have experienced a period of learning, the impact appears more clearly after the implementation date. The same argument may be true for PCA, although learning by banks may be less significant with regard to PCA because all of the capital ratios defined in the PCA standards had been in effect since at least December 1990.^x

The study examines 27 public sector banks using year-end data for 1998. In carrying out the above exercise, the paper uses confidential supervisory data for Indian public sector banks to address some of the issues addressed above. The non-performing assets and capital adequacy are obtained from the Report on Trend and Progress of Banking in India (various years). The procedure used for estimation is the two-stage least squares method, which recognises the endogeneity of both bank capital ratios and risk levels, and unlike ordinary least squares, provide consistent parameter estimates.