On the Real Dangers of Basel II for Emerging Economies

Transcription

1 On the Real Dangers of Basel II for Emerging Economies Andrew Powell 1 Universidsad Torcuato Di Tella, Buenos Aires Revised, June Abstract In contrast to much of the literature, I argue that emerging country sovereign lending is unlikely to be affected by Basel II and that concerns of pro-cyclicality are overdone. On the other hand, if lead regulators insist on the consistent implementation of the IRB approach, then lending to the emerging country private sector may be affected, foreign banks will be placed at a competitive disadvantage and global banking may retreat back towards international banking. More generally, there are a set of cross border issues related to, but that go beyond Basel II of increasing importance to emerging country financial sectors and their regulators, that are far from resolved and that are in urgent need of attention from the international financial community. Key Words: Basel Accord, Banking Regulation, International Lending. JEL Codes: F34, G28, F33. 1 The author is Professor in the Escuela de Negocios, Universidad Torcuato Di Tella in Buenos Aires, apowell@utdt.edu. The author wishes to thank Jerry Caprio, Mark Carey, Giovanni Majnoni, Michael Gordy, Patricia Jackson, Lorna Martin, Pascual O Dougherty, Rafael Repullo and Eric Rosengren for very useful discussions. Naturally, all mistakes remain my own.

2 I. Introduction The BIS consolidated banking statistics indicate that there are US$ 1.92 trillion dollars of foreign claims from BIS reporting banks to developing countries 2. This is a significant number, around 26% of total domestic credit in the countries concerned but this figure rises to 69% for Latin America and 78% for developing Europe 3. Figure 1 shows the regional and sectoral breakdown. The BIS only gives a sectoral breakdown (public sector, banks and non-financial private sector) for what is referred to as international claims, there is no breakdown for the local claims of international banks in local currency 4. Of the US$ 1.13 trillion of international claims on developing countries, US$203bn are on the public sector, US$344bn on developing country banks and US$570bn on the nonfinancial private sector. Figure 1: Foreign Claims of BIS Reporting Banks by Region and by Sector US$ bn Africa and Middle East Asia and Pacific Europe Latin America and Caribbean Bank Public Non Bank Unallocated Local 2 Figures as of December There are an additional US$1.4 trillion of claims on Offshore Centers including US$537 on the Cayman Isles, US$300bn on Hong Kong, US$158 on Singapore, US$58bn on Bermuda and US$ 37bn on Panama. 3 Domestic credit is line 52 from the IMF IFS and here I use the BIS definition of developing countries and the BIS s regional breakdown for consistency with the BIS foreign claims data. This figure refers to IIIQ International claims include those that are truly cross border in any currency plus local claims in foreign currency. The sum of international claims plus local claims in local currency is referred to as, total foreign claims by the BIS. 2

3 While total foreign claims on developing countries have grown over the last 20 years, local claims in local currency have risen even more strongly. Before 1990 local claims were less than US$ 50bn, reaching US$ 100bn at the end of 1994 and now stand at a record of almost US$790bn. In terms of total foreign claims on developing countries, local claims have risen from about 5% in the 1980 s to a plateau of just over 40% in recent years 5. This marked increase reflects the rise in brick and mortar operations of BIS reporting banks in developing countries and is normally refereed to as globabilization, and is illustrated in Figure 2 below 6. Figure 2: The Growth of Total and Local Currency Claims On Developing Countries US$ bn Q Q Q Q Q Q Q Q Q % 40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% Total Foreign Claims Local Claims (% Total) These statistics suggest that changes in international bank regulation might have significant effects on developing countries, especially for Latin America and developing Europe. The rise in local claims in local currencies also suggests that changes in regulation, not only in the G10 countries, but also in how the subsidiaries and branches of G10 banks in 5 The focus on local lending in local currency underestimates the importance of loans booked on the balance sheets of subsidiaries and branches of BIS reporting banks in developing countries as this excludes loans in foreign currency booked in these local financial institutions. As mentioned, the BIS statistics do not distinguish between foreign currency loans that are cross border and foreign currency loans that are booked locally. 6 The growth in cross border lending more evident in the early 1990 s is referred to as internationalization rather than globalization. 3

4 developing countries are regulated are also of importance 7. In particular, Basel II, the new Accord regarding bank capital finalized during 2004 for implementation in Basel Committee Countries by 2007, will affect all countries in two distinct ways 8. First, to the extent that the new standards are binding on international banks or relax previously binding standards, they will affect the cost of capital and hence economic growth and welfare across the globe. Second, while no country is legally obliged to implement the Accord locally, more than 100 countries claim to have implemented Basel I and official reports suggest a similar number will implement Basel II 9. Basel II s quantitative requirements (so-called Pillar 1) include several alternatives and how emerging countries should and will implement these hard components of Basel II remain open questions. In this paper, I first estimate the potential effect of Basel II on the cost of capital for countries in Latin America section II. In contrast to several previous papers the headline result is that Basel II may have no effect on the cost of capital for most emerging country sovereigns, except for those with the lowest credit ratings. This result is however dependent on the mapping between credit ratings and default probabilities employed. Subject to the same qualification, no additional pro-cyclicality may be introduced. Section 3 is devoted to whether and how emerging countries may wish to implement Basel II locally. It is argued that many emerging countries fall between two stools as the Standardized Approach yields little in terms of linking capital requirements to risk and yet many countries may lack the expertise to implement the more advanced Internal Rating Based (IRB) approaches. Given this situation, an intermediate Centralized Rating Based approach (CRB) is proposed. However, many emerging countries are likely to introduce the IRB approach for the more sophisticated local banks and for the branches and subsidiaries of international banks. In section 4, I then consider more specifically the implications of IRB implementation. If IRB is applied across the globe by an international bank using current calibrations this may lead to significant increases capital requirements. It is suggested that this would lead to a retreat from the recent trend towards global banking back towards international banking. However it is suggested that IRB calibration should be revised. Section 5 discusses a number of cross border issues that are yet to be resolved. Section 6 concludes suggesting among other issues that, under certain 7 In this paper I do not enter into the differences between subsidiaries and branches. See Cerutti et al (2005) for a discussion of how foreign banks have expanded abroad. 8 See Basel Committee for Banking Supervision (1988) and (2004) for the old and the new Accord. These, together with literally hundreds of comments are available on 9 See Financial Stability Institute (2004) 4

5 circumstances, the lead regulator approach should be reconsidered. Section 6 details the conclusions. 2. Basel II: On the Cost of Capital for Emerging Countries A set of earlier papers considered the effect of Basel II on emerging countries cost of capital 10. These papers all employed the following "cost and return" of funds equation. This assumes that banks are risk neutral, that the market for lending to emerging country sovereigns is perfectly competitive and that banks make a specified required return on capital: r( 1 k1 ) + ( r + c) k1 = ( r + s1)(1 p) lp (1) where r is the risk free (say Libor) interest rate, c is the cost of capital expressed as a spread, k 1 is the current percentage of the loan financed by capital, s 1 is the current lending spread, p is the probability that the loan will be defaulted on and l is the loss given default expressed as a percentage of the loan amount 11. There is an implicit assumption that deposits are insured at no cost to the bank and hence depositors do not require a premium over the risk free rate. Consider what would happen to spreads if the portion of the loan financed by bank capital were to change. Suppose that k 2 is the new portion of the loan financed by capital and s 2 is the new spread, rewriting the same equation but replacing s 1 by s 2 and k 1 by k 2, and then subtracting equation (1), we can solve for the implied required change in spread as: ( s 2 c ( k2 k1) s1) = (2) (1 p) There are various ways to use this approach to estimate the effect of Basel II on spreads depending on which variables in equation (2) are taken as endogenous, and calculated from market prices, and which are considered as exogenous. Previous work has assumed that Basel I and Basel II both bind, and hence that k 1 and k 2 are exogenous. This surely overestimates the effect of Basel II on spreads. In this paper I take the initial capital, k 1, as endogenous and calculate it from equation 1 using an estimate of the current market spread as s 1. The relevant equation for k 1 is then: 10 See Deutsche Bank (2001), Reisen (2001), Griffith Jones (2001), Powell (2002) and Weder and Wedow (2002). 11 See Repullo and Suarez (2004) for the derivation of such an equation from a more fully elaborated equilibrium loan-pricing model. 5

6 ( r + s1)(1 p) lp r k1 = (3) c The rationale for endogenizing capital in this way is that, assuming a competitive market and risk neutrality, this economic capital would give the minimum leverage ratio that would make lending to that particular sovereign viable at the current market spread. Suppose now Basel II implied an increase in required capital due to the new regulations. Then we can calculate, using equation 2, the required increase in the spread such that banks would be willing to lend to that sovereign. This does not necessarily mean that the increase in spreads would be observed as banks might decide not to lend. The interpretation is then that at the current spread banks are lending and the new spread is then the required spread for an IRB bank to continue to lend given risk neutrality and a competitive international banking. With endogenous economic capital, the following logical possibilities must be taken into account considering the effect of Basel II on required spreads: A) Basel I and Basel II do not bind and hence Basel II has no effect. B) Basel I does not bind but Basel II does bind and hence Basel II will increase required spreads. C) Basel I and Basel II both bind and hence Basel II has an effect on spreads - either to increase or decrease. D) Basel I binds but Basel II does not and hence Basel II will decrease spreads. The potential effect of Basel II is then calculated in the following steps: 1. A selection of countries in Latin America for which Emerging Market Bond Index (EMBI) spreads and (Standard & Poor s) ratings are available is chosen The average sovereign EMBI bond spread over 2004 is taken for each country if that country had the same rating over the period. Otherwise, the EMBI spread is used for the period in 2004 for which the rating employed applied. 12 We exclude Argentina that was classified as SD selected default at the time of writing. 6

7 3. Economic capital is then calculated using a default probability consistent with the country rating. The mapping from credit rating to default probability employed is a 12-month Standard and Poor s mapping. An LGD of 45% is employed (as that assumed in Basel II s Foundation IRB approach), a risk free rate of 4%, and a required rate of return of 18% The calculated economic capital is then compared to Basel I and Basel II IRB capital requirements to consider which alternative A to D) listed above is relevant. Using the same mapping as in step 3, the Basel II IRB required capital ratio is calculated using the curve as stated in the new version of the Accord, page If relevant, the effect on the spread is then calculated using equation (2) above and using the relevant endogenous capital as k 1 and the relevant Basel II capital requirement as k Table 1: effect of Basel II IRB on Required Sovereign Spreads Country Country Probability of Average Basel I Basel II Basel II Change Rating Default Spread Binds? IRB Binds? in Spread Chile A 0.050% 0.83% % % Mexico BBB % 2.20% % % Colombia BB 0.940% 4.63% % % Panama BB 0.940% 3.47% % % Peru BB 0.940% 3.60% % % Brazil BB % 4.40% % % Venezuela B 8.380% 4.59% % % Ecuador CCC % 8.30% % % The results are presented in Table 1 for a selection of countries in Latin America. The bottom line is that in most cases neither Basel I nor Basel II bind and hence Basel II will have no effects 15! Only for the two countries in the sample with the lowest credit rating (Venezuela and Ecuador) will the Basel II, IRB approach result in higher spreads, where both Basel I and Basel II bind. It may seem counter-intuitive that Basel I binds only for these higher risk countries. The rationale is that in the cases of countries 13 The sensitivity of the results to these assumptions is discussed below. 14 In the case of Chile we find Basel I binds but not Basel II so k 1 is Basel I s 8% and k 2 is the endogenous capital. 15 As Weder and Wedow (2002) remark, this appears to be consistent with the views of some insiders of the Basel II process and discussion. 7

8 with higher ratings (Mexico and Colombia etc), the spreads are high relative to the assumed default probabilities. Hence it is economic in these cases to lend with lower leverage with more capital. The effect of Basel II would be an increase in spreads of 134 and 301 basis points above today s levels, if Venezuela and Ecuador respectively wished to borrow from Basel II regulated banks. The case of Chile is also of interest. Here, for the parameters employed, Basel I binds but Basel II will not bind and so spreads will be reduced by some 58 basis points. Comparing these results to previous results in the literature, Reisen (2001) and Griffith-Jones (2001) both suggest that that the effects of Basel II will be significantly larger. These authors assume Basel I and Basel II both bind (k 1 and k 2 exogenous) and use equation (1) to solve for an endogenous cost of capital, c. This results in very high estimates of c and (hence) very significant effects for Basel II IRB through equation (2). Powell (2002) also assumes Basel I and Basel II both bind but also that the cost of capital is exogenous. This reduces the effect of Basel II although the effects are still quite significant especially for the IRB approach. However, there is a potential inconsistency between the assumed cost of capital and the market spread. Weder and Wedow (2002) updates the Powell (2002) estimates for the new flatter IRB curves and consequently finds reduced effects of Basel II. Moreover, these authors also conduct an empirical analysis that leads them to conclude that Basel I may not bind when it comes to lending to emerging country sovereigns backing up the approach taken here. The results also question the importance of the debate regarding emerging country asset correlations 16. As reviewed below the IRB curve assumes a particular correlation between assets as a function of default risk apparently calibrated on a portfolio of G10 country corporates. While the argument that asset correlations between emerging and developed countries are lower than those within a typical G10 country, is surely valid, if Basel II, as written, does not bind then it makes little difference whether the requirement is reduced by adjusting the curve for lower correlations. Adjusting the correlations would however reduce the effect of Basel II in those cases where it does bind Venezuela and Ecuador in the sample above for the S&P 12 month default probabilities. However, the results presented are naturally dependent on the various assumptions employed. Having said that, the results are not particularly sensitive to the assumption on the required rate of return on capital or on the riskless rate. Reducing the required return to 15% implies smaller increases in spreads given IRB for Venezuela to 105bpts and Ecuador to 16 See for example Griffith Jones, Segoviano and Spratt (2003) 8

9 237bpts. For those countries where Basel I and Basel II do not bind, there are no effects. Increasing the required rate of return on capital to 21% increases the effect of IRB for Venezuela and Ecuador to 162bpts and 366bpts respectively. Again, there is no effect for the other countries where the requirements do not bind. The results are more sensitive to the assumed default probability mappings. Weder and Wedow (2002) present an interesting comparison of 12 month S&P, 12 month Moody s and 3 year S&P default probabilities 17. These are all based on corporate default histories and hence there is uncertainty as to whether they are truly appropriate for considering sovereign risk. The problem is that the information on sovereign default is (fortuitously) very limited. Hu et al (2001) discuss the estimation of a rating transition probability matrix for sovereigns notwithstanding the limited data using a mixture of naïve rating transition experience, a probit model and a Bayesian system to update priors to ensure a smooth result. The implied 12-month default probabilities from Hu et al (2001) are quite different to the 12 month S&P corporate default probabilities employed above and give higher default probabilities especially for lower ratings. They are closer to the 3-year S&P estimates quoted by Weder and Wedow (2002). Table 2 shows the results if the Hu et al (2001) default probabilities are employed. Larger changes in spreads are then required for Venezuela and Ecuador respectively and there is also now a significant effect for Brazil with an increase in 102 basis points in required spreads See also Moody s Investor Services (1999) and (2000) and Standard & Poor s (2000) and (2002). Powell (2002) uses the mapping quoted in Jackson (2001), which gives results similar to the 12 month default probabilities. 18 Brazil was upgraded from B+ on Sept 17 th 2004 by Standard and Poor s. If the B+ rating is used which Hu et al (2003) indicate has a default probability of 14.7% more significant changes in required spreads are obtained. A Venezuelan bond was placed under selective default in January 2005 but here we use the B long term currency rating through the relevant period in

10 Table 2: Effect of Basel II using Hu et al Default Probabilities Country Country Probability of Average Basel I Basel II Basel II Change Rating Default Spread Binds? IRB Binds? in Spread (Hu et al) Chile A 0.000% 0.83% % % Mexico BBB % 2.20% % % Colombia BB 2.500% 4.63% % % Panama BB 2.500% 3.47% % % Peru BB 2.500% 3.60% % % Brazil BB % 4.40% % % Venezuela B % 4.59% % % Ecuador CCC % 8.30% % % To give some feeling of the overall sensitivity of the results to the assumptions regarding default probabilities and spreads, Table 3 presents some further results for different sovereign spreads and different assumed default probabilities. This table is created with Brazil in mind which had an average sovereign spread of 4.4% from 9/2004 until 12/2004 and the 12 month S&P and Hu et al default probabilities are 1.3% and 6.4% respectively. The 12-month Moody (corporate estimated) default probability is 2.47%. Table 3: Changes in Required Spread, Sensitivity Analysis Default Probability 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 8.0% 9.0% 10.0% 1.0% -0.04% 0.25% 0.45% 0.63% 0.79% 0.96% 1.12% 1.28% 1.43% 1.58% 2.0% 0.00% 0.25% 0.45% 0.63% 0.79% 0.96% 1.12% 1.28% 1.43% 1.58% Spread 3.0% 0.00% 0.00% 0.12% 0.63% 0.79% 0.96% 1.12% 1.28% 1.43% 1.58% 4.0% 0.00% 0.00% 0.00% 0.00% 0.55% 0.96% 1.12% 1.28% 1.43% 1.58% 5.0% 0.00% 0.00% 0.00% 0.00% 0.00% 0.28% 1.01% 1.28% 1.43% 1.58% 6.0% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.01% 0.76% 1.43% 1.58% 7.0% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.51% 1.27% 8.0% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.27% 9.0% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 10.0% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% At the bottom left of the Table, with higher spreads and lower default probabilities, Basel I and Basel II do not bind. At the top right, both Basel I and Basel II bind and then only the assumed default probability is important in determining the effect of Basel II - as that is used in the IRB 10

11 capital requirement calculation 19. Along a diagonal, Basel I does not bind but Basel II does bind, and then both the spread and the default probability are important in determining the effect of Basel II; as the spread determines the initial endogenous economic capital. At 1% spread and 1% default probability both Basel I and Basel II bind but the result is a reduction in capital requirements and spreads. As can be seen from the Table, while the base case for Brazil (a spread of 4.4% and an S&P 12 month default probability of 1.33%) gives no effect of Basel II, a spread of 4% and a perceived default probability of 6% (close to the Hu et al 6.4% figure) yields a 96 basis point effect. The 12-month corporate S&P corporate inspired default probabilities are probably too low whereas the Hu et al default probability might be considered an upper bound. The range of spreads from 2% to 5% and default probabilities from 1-5% gives a potential range in the increase in required spreads from zero to 80 basis points. This appears a realistic range. On Pro-Cyclicality and Circularity There has been much discussion regarding Basel II and pro-cyclicality. The above methodology can also shed some light on this issue. In the previous section we considered the rating and the spread of a country as a constant. In practice ratings and spreads change over time. Economic capital will then also change over time. The appropriate question is not then whether Basel II introduces pro-cyclicality but whether it increases it. To shed light on this, using the same methodology we can consider what would have happened to the Argentine spread leading up to the Argentine crisis in In other words we calculate for each month on the basis of Argentina s spread and rating (and 12 month S&P default probabilities) at that time, the economic capital, and test whether Basel II IRB would have been binding and implied an increase in capital. To cut to the conclusion, we find that neither Basel I nor Basel II IRB bind for the whole of 2001, so that Basel II implementation would have had no effect on required spreads for IRB banks to lend leading up to the Argentine crisis. 19 As discussed previously, in this part of the table as Basel I binds banks may not find it profitable to lend at the current spread. The change in the spread should then be interpreted as the change required such that, under Basel II, banks would be willing to lend. 11

12 70.000% Figure 3: Argentina: Simulated Spreads under Basel II (with Hu et al (2001) probabilities) % % % % % % 0.000% Jan-01 Feb-01 Mar-01 Apr-01 May-01 Jun-01 Jul-01 Aug-01 Sep-01 Oct-01 Nov-01 Dec-01 Actual Spread Simulated Spread However, once again if we take the Hu et al (2001) default probabilities we get very different results. Figure 3 plots the results in that case. In January, February and March 2001, spreads would have been unchanged. However, as Argentina was downgraded in April 2001 from BB- to B+, according to the Hu et al (2001) the default probability would have risen from 6.4% to 24.7% and while actual spreads rose to almost 10% they would have had to rise to 15% in order for Basel II IRB regulated banks to have continued lending. The difference between actual and simulated spreads from about 5% to 9% as Argentina is downgraded again in June And this difference rises to almost 20% in October of the same year. Again, however, the Hu et al (2001) estimated default probability mappings should be viewed with caution. In particular the jump from 6.4% to 24.7% default probability on the downgrade from BB- to B+ appears considerable. Perhaps of more concern than potential pro-cyclicality of Basel II for sovereign lending is potential circularity. Argentina s ratings reflected perceptions of Argentina s access to international markets to roll over its debts. But access to international capital markets is also, to some extent determined, by a country s rating. The point is not that Argentina s problems were purely of a liquidity nature. However, the use of ratings may well introduce some measure of circularity. If the circularity is serious enough this might just spell the difference between being able to roll-over and buying time to adjust and a costly default. There is an interesting issue here regarding whether the Standardized Approach or the IRB approach should be pushed for sovereign lending. While the quantitative effects are smaller, a potential danger of the SA is that there are very few international credit rating agencies, and if one 12

13 agency simply gets the rating wrong, and this leads to a change in the relevant capital requirement, then this might have quite significant effects 20. An alternative would be to introduce a simplified IRB approach for sovereign credits whereby banks would simply be asked to use their own ratings but according to a standardized scale. While the use of external ratings for corporates might be justified given that there are many corporates and hence if all banks conducted their own rating this might lead to costly assessment duplication, there are fewer sovereigns and hence the potential costs are lower. Moreover it might be argued that, an assessment of a corporate is more of a science whereas that of a sovereign is inevitably as subjective as assessing political risk. If a bank is lending (perhaps greater than a stipulated minimum) to a sovereign, arguably it should have an internal rating. An international body (as the OECD is doing for the export credit agency ratings, an obvious candidate being the BIS) could collect those internal ratings across banks from Basel Committee countries and publicize the average ratings for countries and other statistics. This would increase the number of opinions and possibly lessen the concern, whether correct or not, that, official, export credit agency ratings may suffer from political pressures 21. In this paper, and following much of the literature, I only consider changes in the cost of capital for international lending to sovereigns 22. As evidenced in the figures in the introduction, international banks lend more to banks and non-financial corporates than to sovereigns. Clearly the change in the overall cost of capital to emerging economies will depend on the change in required capital for international lending to the private sector as well as to sovereigns. The above estimates should not be thought of as an estimate of the overall cost implications of Basel II - as such it may be a gross underestimate. However, there remains considerable uncertainty regarding the change in the cost of capital for the private sector in large part related to the cross border issues of Basel II. We discuss this further in section 4 below. To conclude this section, if standard (corporate) default probability mappings are used, Basel II will have little effect on most emerging sovereigns but will only affect the cost of capital in the case of the countries with poorer ratings such as Venezuela and Ecuador. If estimated 20 The introduction of the ratings by export credit guarantee agencies and their publication is to be welcomed in this regard. These ratings are to be published on the OECD website, 21 Still it is recognized that increasing the number of opinions may have quite subtle effects as illustrated in recent papers employing global games. 22 Powell (2002) does attempt to estimate costs to banks and non-bank corporates but the methodology adopted is almost surely also an underestimate. 13

14 sovereign default probabilities are employed, we get more significant effects including an increase in required spreads of up to 80 basis points for Brazil a BB- risk. 3 Implementing Basel II in Emerging Countries The Basel Committee for Banking Supervision (henceforth the BCBS) has acknowledged that developing countries will likely need more time to implement Basel II than the 2007 deadline. Moreover, recent statements by the IMF and the World Bank have suggested that, in terms of the on-going Financial Sector Assessment Program (FSAP), Basel II implementation will not be considered as a requisite 23. Basel II itself includes various alternatives. In Pillar 1 (Quantitative Requirements), these include (i) the Standardized Simplified Approach, (ii) the Simplified Approach, (iii) the Foundation Internal Rating Based Approach and (iv) the Advanced Internal Rating Based Approach. There are then also different choices regarding (a) credit risk mitigation techniques, (b) securitization risk and (c) operations risk. Table 4 provides a brief details. Countries will then have to decide whether they stay on Basel I or, if they do move to Basel II, which of the many alternatives on offer, should be adopted. 23 One view however is that many developing country supervisors will not wish to be considered as lagging behind this new standard and some country authorities may be concerned that the market may punish them for non-implementation even if the International Financial Institutions do not. 14

15 Table 4: Pillar 1 Alternatives Under Basel II The Approaches Basic Credit Risk Credit Risk Mitigation Securitization Risks Operational Risk Measurement Technique Simplified Standardized Export Credit Agencies Simple: risk weight of SSA banks can only invest Basic Indicator: ( Trade collateral subsitutes that of (cannot offer Capital=15% Gross Income Directorate, ECA page) claim. enhancements or liquidity facilities). Riskweight=100% Standardized Approach Export Credit Agencies Simple: (as above). Standardized: uses export Basic Indicator. Or or Credit Rating Agencies Comprehensive: exposure credit agency ratings Standardized Approach where (eg: S&P, Moody's, Fitch) amount reduced subject to (only investing banks can Bank Capital = weighted sum claim and collateral haircuts. use below BB+) of gross income across activities IRB Foundation Banks' internal ratings Comprehensive, then IRB Approach: Investing More sophisitcated banks will be for default probability LGD adjusted given banks may use bank expected to graduate to the and Basel II formula reduction in exposure and Ratings according to a Advanced Measurement Approach sets capital requirement capital requirement given standard scale. Originators where capital requirement given by (Loss Given Default 45% by Basel formula may use Supervisory own risk measurement system. for Senior and 75% Subord). Formula IRB Advanced Banks set internal rating Own model determines LGD As IRB Foundation As IRB Foundation (default probability), LGD and EAD and capital Exposure At Default and requirement given by forumula Maturity. Capital requirement still given by Basel formula. Powell (2004) and Majnoni and Powell (2004) argue that many developing countries may fall between two stools. On the one hand, the Standardized Approach (SA), that allows the use of credit ratings from private agencies will yield little in terms of linking capital requirements to risk. A serious problem for many emerging countries is that markets for credit ratings are very shallow and hence the SA will simply reduce to a flat credit charge for most borrowers. Of course, adopting the SA may create incentives for ratings markets to develop but this brings its own dangers in terms of companies buying a good rating, provoking a race to the bottom in rating quality. A similar argument might be made for some G10 regional or smaller banks - and this might lie behind the recent decision of the US to keep many of its banks on Basel I. However, such banks are not likely to be systemic in a G10 country. The situation in some developing countries is different large parts of systemic banks portfolios are largely unrated. On the other hand, many countries may lack the expertise to implement the more advanced IRB approaches. The IRB approach calls for banks to develop a rating methodology for all borrowers including the development of a comprehensive scale for ratings. Under the Advanced IRB approach, the bank must also estimate Loss Given Default and Exposure at Default, whereas, under the Foundation IRB these other parameters are determined by the supervisor. These parameters are then fed into a formula provided 15

16 by the BCBS to calculate the appropriate capital requirement. This formula is per instrument and is then understood to be an approximation to the output of a credit portfolio risk model under specific assumptions regarding distributions, loan risk correlations and a statistical tolerance value. Specifically Basel II IRB is now calibrated subject to a 99.9% statistical tolerance value, with the assumption that the bank already has provisions to cover the mean of the distribution (expected loss) and capital should cover the distance between the mean and the 99.9 percentile of the distribution (unexpected loss). The entire distance between the origin and the 99.9 percentile (expected plus unexpected loss) is sometimes referred to as the Value at Risk (in this case up to a 99.9% confidence limit). In Figure 4, I illustrate the Basel II IRB approach graphically. In an Appendix I discuss the derivation of the IRB curve and various manipulations used in the next section. Figure 4: The Basel II IRB Curve, Expected and Unexpected Losses Expected + Unexpected loss = Value at Risk Basel II IRB Capital = Unexpected Loss Probability Expected Loss Unexpected Loss 99.9% of Distribution Mean of Distributio n Potential Loss The IRB approaches call for significantly more sophisticated bank and supervisory resources. In particular the bank must develop its rating system and mapping to default probabilities and develop techniques to check the robustness of those and other parameter estimates. The supervisor must in turn have the human resources and systems to monitor these processes. However, the data on the IMF and World Bank completed FSAP s illustrate that many countries (including G10 but especially 16

17 developing) are far from being fully compliant with the Basel Core Principles for Effective Banking Supervision and, on average, developing countries lag their G10 counterparts 24. Of particular concern is the lack of (i) effective consolidated supervision 25, (ii) supervisory independence, resources and authority and (iii) effective prompt corrective action. If supervisors lack resources and the basics of effective bank supervision, correcting this should be the first priority and more complex rules on capital requirements may well be counter productive. However, full BCP compliance is too strict a precondition for moving to Basel II IRB after all many G10 countries are not compliant with all the BCPs. In general a country should be BCP compliant to the degree required to implement the appropriate alternatives chosen within the Basel II framework. Given this situation, Powell (2004) and Majnoni and Powell (2004) suggest a Centralized Rating Based (CRB) Approach, at least as a transition measure. Under this approach the supervisor dictates a rating scale and asks banks to rate borrowers according to that centralized scale. Each rating would then correspond to a probability of default and, combined with other loan information, that rating would imply a capital charge. This system would have the drawback that each bank would be forced to use the same scale that may not then be the particular scale most appropriate to the borrowers of that bank. For example a bank specialized in a particular type of lending or a particular sector would not necessarily wish to use the same scale as a more general bank or one specialized in another business. The rating scale could be devised to be appropriate for the larger institutions so for countries with a more concentrated banking sector the costs would be minimized 26. However, the great benefit of the approach is that the supervisor would be able to monitor and control banks ratings and hence monitor and control their capital sufficiency in relation to risk much more effectively. In particular, the supervisor could very easily monitor banks average ratings, banks ratings for the same borrower, banks ratings for the same type of borrower, banks ratings for the same type of loan, banks ratings in the same economic region etc. These kinds of comparisons combined with simple procedures for spotting outliers and keeping a track of the different banks ratings of the main borrowers from the financial system are extremely valuable tools for a bank regulator. Naturally, for countries that 24 See World Bank (2002) and Powell (2004) 25 Basel II also introduces a significant change to the level of consolidation required for banking supervision from the bank itself to its holding company. As many countries do not comply with more modest versions of consolidated supervision; these countries remain far from the spirit of the Basel II proposals. 26 A slightly more complex version could have a different centralized rating scale for different portfolios. 17

18 had already developed a bank rating for the purposes of provisioning, this proposal would build very neatly indeed on those systems. This methodology could not truly be called the IRB approach as internal in IRB is normally thought of as referring to the scale and not just the rating. However, the same type of minimum criteria as discussed in Basel II s IRB could be thought of as the minimum criteria for this system for example in terms of the number of rating buckets and the history of information. Moreover, Basel II s IRB curve could be used to calculate the capital charge based on the centralized ratings and a mapping of those ratings to default probabilities. The centralization of the rating scale provides another advantage here as the mapping and the calibration of the curve could be then easily be checked on a bank by bank and on a system wide basis by the supervisor using actual loan data. Furthermore, there is a simple way for a country to adopt a CRB approach and be fully compliant with Basel II at the same time. In particular, a country could adopt the Standardized Approach (SA) but still employ the CRB approach to calculate the total Value at Risk. Then the difference between the total CRB calculated Value at Risk and the capital charge given by the Standardized Approach could be used as an estimate of the forward-looking provisioning requirement appropriate on that loan. Moreover, the CRB approach could be used as a precursor to IRB. Once the CRB approach was working the supervisor could then work with banks to approve their own rating scales and rating methodology using the basic CRB approach as a reference tool. Many countries are then facing quite a difficult choice when it comes to Basel II implementation. Official indications are that most countries will adopt Basel II. Many will surely implement the Standardized Approach at least initially for most banks. However, several larger emerging economies are likely to implement the IRB approach at least for a subset of their banks. As reviewed in the introduction, foreign banks have become extremely important in many developing countries and these are precisely the more sophisticated internationally active banks that will be planning to implement Basel II and most likely IRB on a global consolidated basis. If this is the case, the implication is that a developing country host regulator must either a) accept an IRB methodology agreed between the bank and its lead regulator or b) insist on maintaining Basel I or a different Basel II alternative or a local IRB approach within its jurisdiction and hence insist on a double regulatory regime 27. In view of the stress on the 27 We discuss below adaptations to the IRB approach that might be required by local regulators. 18

19 lead regulator model, the potentially high regulatory costs for banks operating in many different jurisdictions and the potential for regulatory duplication, the BCBS, in its short document on the high level principles for cross border supervision BCBS (2003), comes down on the side of the first alternative While there is wording regarding the legal requirements to supervise banks in their jurisdiction, BCBS 2004 clearly sides with the lead regulator model. 19

20 4 Implementing IRB Across the Globe In the new Accord there is considerable emphasis that banks should adopt a consistent approach globally. This implies that international banks will be under pressure to implement a consistent IRB methodology across all the countries where it operates. There is a danger that this may lead to high capital requirements, especially in those countries with lower sovereign ratings. BBVA (2003) for example argue that a small or medium-sized enterprise (SME) with no external rating might attract an internal bank rating 2 or 3 notches below that of the sovereign. This implies that an SME in Brazil, where the sovereign local currency rating is BB might be expected to have a rating of B+ or B indicating a default probability of 2.91% to 8.31% and a capital requirement of 8.3% to 13.4% 29. If the same logic is applied to SME s in Uruguay (local currency rating B), Bolivia (local currency rating B-) and Ecuador (CCC) then the current Basel IRB formula would give capital requirements well in excess of 20%. Lying behind this discussion is a fundamental question as to what is a standard for bank solvency? Basel I suggested a standard in terms of an absolute level of minimum capital and hence for different risks across countries this offered different degrees of protection in terms of the risks of bank failure. Implicitly the implication was that Basel I offered a lower degree of protection in a riskier emerging country than in a typical G10 economy. This is also the case for Basel II s Standardized Approach. Basel II IRB, as written, changes this to a standard in terms of a statistical tolerance value (99.9%) that implies different levels of capital dependent on estimated risk. This is a significant change, and for international banks operating in many countries of different risk, may have profound effects. In particular if Basel II IRB as written is implemented on a globally consistent basis this is likely to lead to significantly higher capital requirements than Basel I or the Basel II standardized approach 30. Moreover, if Basel II IRB is implemented on a globally consistent basis and local regulators maintain Basel I or implement Basel II s standardized approach, then international banks may be at a competitive disadvantage in that capital requirements under IRB may be significantly higher than under 29 This assumes S&P 12 month default probabilities, sales of 25 million euros and applying the correlation correction for SME s. 30 Still it is possible that IRB capital requirements will not bind in the sense that international banks already have capital in excess of Basel I and possibly similar to Basel II IRB for local lending in developing countries. However, there are reasons to believe that this is less likely to be the case than for sovereign lending. We discuss these reasons in the next section. 20

21 the standardized approach for the default probabilities found in developing countries. The implication is that global banking may then retreat towards international banking where foreign banks operating under a lead regulator essentially only lend to higher rated local banks and selected corporates in emerging economies. Indeed the situation may resemble an extreme version of that suggested in Repullo and Suarez (2004) where the more sophisticated foreign IRB banks lend to higher rated clients and the less sophisticated local SA banks lend to SME s and retail clients. This would then reverse the process of deeper penetration of foreign banks in emerging economy markets that we have seen over the last years. BBVA (2003) suggest an adjustment to the IRB formula as a solution to the above, replacing the 99.9% tolerance value with the default probability appropriate to the rating of the sovereign. Brazil with a BB rating might then adopt IRB with a 99.06% tolerance limit. This yields capital requirements for SME s of 2 or 3 notches below the sovereign of 5.18% or 9.58% respectively. While appealing as a simple rule of thumb, this procedure gives rather low requirements given that Brazil has already adopted Basel I with a basic capital requirement of 10%! More generally, Majnoni and Powell (2004) using actual corporate loan performance data from public credit registries in Argentina, Brazil and Mexico find that (1) Basel IRB will result in significantly higher capital requirements than Basel I but that (2) the formula will not yield the 99.9% level of protection as advertised. One explanation is that correlations between default risks are typically higher within an emerging economy than within that of a developed country. Thus an estimated probability of default and expected loss in an emerging economy may lead to a higher unexpected loss for a particular tolerance value than that for a G10 nation. For example, Majnoni and Powell (2004) find that in pre crisis Argentina the average default probability for all corporate loans in the public credit registry was 9.6% and unexpected losses in the Basel IRB formula to be some 21.5% whereas the Basel IRB formula yields 17.76%. Following this argument, an alternative route is to adjust both the correlation assumptions in the IRB formula and the tolerance values 31. This has the advantage that the curve will fit better the mean default probabilities and unexpected losses found in a typical emerging economy. To illustrate, in Figure 5, we then plot the Basel IRB curve and an adjusted curve that roughly fits the experience of Argentina - with higher assumed covariances. It can be seen that this second curve (labeled 99.9% Hi 31 In the Appendix the technical details of adjusting the statistical tolerance and the covariance assumptions in the Basel IRB curve is discussed. 21

22 Correlation) then suggests that capital requirements to obtain 99.9% protection are significantly higher than the original Basel II curve and much higher than Basel I s 8% or Argentina s basic 11.5% capital charge at that time. Now suppose that Argentina applied a 97.09% standard but with this correlation structure (this is a B+ standard and in line with Argentina s current top bank ratings). This gives the curve labeled 97.09% Hi Correlation and is a little flatter than the current Basel IRB curve. In the next section I develop a specific proposal employing these types of adjustments to the IRB curve. Figure 5: The Effect of Altering the Correlation and Statistical Tolerance in the IRB Curve 30.00% 25.00% 20.00% Capital Requirement 15.00% 10.00% 5.00% 0.00% % % % % % % Default Probability Tol 99.9% Tol 97.09%) 99.9% Hi Corr Hi Corr 22

23 5 Unresolved Cross Border Issues The tradeoff of adjusting the IRB curve for different locations is between consistency across an individual (albeit large international) financial institution, and hence reduced incentives for regulatory arbitrage, versus consistency within a country (and hence appropriate calibration and a level playing field within a host country). Currently, the focus on the lead regulator approach would lead to greater consistency across an institution but would result in significantly higher capital requirements for emerging country SME s given their higher default probabilities, a curve that is probably not well suited to the relevant distribution of credit losses in most countries of the world and potentially significant effects on bank competition. While consistency within an institution is surely at a premium, there may be other ways to obtain it 32 and also it is more likely appropriate for lending to sovereigns and high rated corporate clients, where assets can be booked on or offshore with greater ease. Regulatory arbitrage is less of an issue when considering lending to SME s and retail clients. Moreover, these sectors are typically the ones where regulation matters. In the first section we showed that Basel II is unlikely to bind for higher rated sovereigns and by extension for higher rated corporates. However, if markets are reasonably competitive for SME lending, requirements are more likely to bind. These arguments suggest that some further adjustments particularly for SME lending should be considered 33. The details of IRB implementation and testing has been at the center of a debate in the European Union that has recently decided on a college of supervisors approach such that supervisors from different jurisdictions will be involved in supervising an international bank s credit risk assessment systems and IRB implementation 34. A controversial clause in the arrangements gives the lead supervisor the ultimate power if a dispute arises that is not resolved within 6 months 35. As some jurisdictions have legal responsibilities for bank regulation (for subsidiaries or branches of foreign banks and other domestic institutions), local regulators may resist 32 For example, a 99.9% standard could be adopted for assets originating in BCBS member countries and a different standard could be adopted for assets originating in a developing country where the trade off between stability and the availability of capital may be different. 33 These arguments might also be extended to retail and to other business lines all assuming that they are relatively competitive. If the sectors are not competitive then it would be economic to lend with more capital and minimum capital regulations might then not bind. 34 See for example the thoughtful speech by Callum Macarthy (Chairman of the Financial Services Authority, 29/9/04), McCarthy (2004). 35 Clearly this goes beyond IRB testing and implementation. 23