First, Do No Harm. A Hippocratic Approach to Procyclicality in Basel II. Michael B. Gordy. Federal Reserve Board

Transcription

1 First, Do No Harm A Hippocratic Approach to Procyclicality in Basel II Michael B. Gordy Federal Reserve Board michael.gordy@frb.gov May 2009 Based on Gordy & Howells, J. of Financial Intermediation The opinions expressed here are my own, and do not reflect the views of the Board of Governors or its staff. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

2 Motivation Under Basel II, the capital requirement on a loan is increasing in the default probability (PD) of the borrower. Capital takes a double-hit in an economic downturn: charge-offs on defaulted loans reduce available capital; required capital on surviving portfolio rises as downgrades outnumber upgrades. To restore capital ratio, bank must either sell assets, raise capital or curtail lending. market for loans is illiquid, so prices likely to fall significantly when many banks selling at once; raising new equity is expensive in a downturn; reduced lending would amplify the recession. Traditional role of banks as shock absorbers for the market economy may be hampered. Distinct in principle from capital volatility, though for large institutions credit cycle is main source of capital volatility. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

3 Is Basel II procyclical? Literature reports wide range of predictions for volatility of Basel II capital requirements over business cycle. Basel II not yet fully in force, data on impact of crisis remain scant. Mildly skeptical that Basel II will materially increase the procyclicality of bank lending behavior. Basel II not intended to be a binding constraint for large, market-oriented institutions. Economic capital is also procyclical and may be difficult to dampen by regulatory measures. Nonetheless, better to evaluate corrective measures proactively. Otherwise, national regulators may pursue informal measures on an ad hoc basis. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

4 Objectives Many potential ways to fix procyclicality. To judge among them, recall objectives of Basel II: Close alignment between regulatory capital and economic capital. Founding objective of Basel II. Deviations restore incentives for regulatory capital arbitrage. Bank s information systems should be harnessed but not hijacked. Maintain integrity of use test. Avoid distortion of bank internal risk management processes. Market disclosures of IRB capital ratio should be informative. IRB capital ratio as first standardized modern metric of portfolio risk. Formulae derived from asymptotic approximations to 99.9% VaR in widely-used risk-factor models. An imposed common standard allows more reliable comparison of creditworthiness between banks and across time. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

5 Overview on potential measures Proposed solutions fall in three broad classes. Smooth input PDs (through-the-cycle rating). Flatten risk-weight function mapping PD to capital. Time-varying standards for minimum capital adequacy. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

6 Through-the-cycle ratings Most obvious way to dampen volatility of capital is to dampen the volatility of obligor PDs. Most obvious way to do this is to dampen volatility of ratings. Rating stability is characteristic of through-the-cycle (TTC) rating systems. Basel II appears to mandate TTC ratings: Banks expected to use horizon greater than one year in assigning ratings ( 414). Borrower rating must represent bank s assessment of borrower s ability and willingness to contractually perform despite adverse economic conditions ( 415). M.B. Gordy (FRB) Procyclicality in Basel II May / 22

7 What does through-the-cycle mean? Several competing plausible definitions for TTC, e.g., Treacy & Carey (Fed. Res. Bull.,1998): TTC rating as measure of ability to survive trough of business cycle. Löffler (JBF, 2004) proposes that TTC ratings respond only to permanent shocks to the firm, ignore transitory shocks. Altman & Rijken (JBF, 2004) emphasize microstructure goal of rating stability, as well as long horizon. Cantor & Mann (Moody s, 2003) explain that the prime objective of [rating] agencies is to provide an accurate relative ranking of credit risk at each point in time. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

8 Problems with TTC ratings Statistic validation requires precise quantitative characterization of rating system objectives. In practice, TTC is too murky. TTC ratings not very useful for active portfolio management. Portfolio models assume ratings are PIT. If not, then VaR forecast is biased. Same holds for model that underpins Basel II. Pricing/trading applications require that ratings reflect all available information. Large, sophisticated banks generally aim for PIT ratings. Contradicts spirit of other Basel II requirements Banks must use all relevant and material information in assigning ratings ( 411). Banks must update the borrowers rating in a timely fashion ( 426). Degrades information in changes in IRB capital ratio over time. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

9 Flattening the risk-weight function Relative to CP2, final IRB rules flattened risk-weight function by specifying asset correlation as declining function of PD. Reduced volatility of capital requirement by about 16%. Given available data, cannot estimate asset correlations with much precision. Formal test between CP2 and final specifications would lack power. Economic story is lacking. Asset correlation as declining function of firm size is more defensible, partly recognized in Basel II. Further significant flattening of risk-weight function likely to invite regulatory capital arbitrage. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

10 Smoothing the output Kashyap & Stein (2004) observe that the socially optimal quantile in capital rule should reflect trade-off between private cost of capital and social cost of bank failure. Basel II fixes quantile at 99.9%. Optimal quantile may vary over business cycle. In this spirit, suggest countercyclical rule for buffer above existing minimum requirement. Avoid renegotiation of Pillar 1 formulae. Application under Pillar 2, so more flexible. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

11 Countercyclical index for capital buffer Let C ijt be the Basel II minimum capital requirement for the holdings of bank i in jurisdiction j at date t, expressed in the units of currency of the bank s home country. Propose that the capital buffer, B ijt, associated with this portfolio be a time-varying multiple of C ijt. The multiplier, α jt, is the countercyclical index for jurisdiction j at date t. The index is given by a pre-committed function of observed macroeconomic or market variable. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

12 Choosing a scaling factor If pre-committed function, then require an index that is publicly observable; available at least quarterly and without large revisions; strongly correlated with the credit cycle; and difficult for banks to manipulate. Market-based credit spreads most natural choice. CDX in US, itraxx in Europe, or rating-based average spreads. Proxy based on vendor EDF indices (say, KMV or Kamakura). Backward-looking measures such as aggregate charge-off rates might be adequate for smaller bank-dominated markets. Market volatility indices as measures of fear, which correlates with funding liquidity. Equity index or aggregate credit growth measures are possible, but may be difficult to align reliably with current position in credit cycle. Macro-variables not ideal, subject to regime shifts due to changes in productivity, labor laws, etc. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

13 Example: Index based on market spreads In North America and Europe, CDS indices offer natural measures of state of credit cycle. Propose: α jt = Q N (δ j (µ j log(s jt ))) Spread S jt is index spread, e.g., quarterly average IG CDX. Q is the maximum buffer as a percentage of the baseline requirement, set by Basel Committee. Location parameter µ j determines spread at which α jt crosses target mean of Q/2. δ j controls the sensitivity of α jt to S jt, determines variance of α jt. Link function N is normal cdf. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

14 Countercyclical index as function of spread Q µ = 70,δ = 2 µ = 70,δ = 1 µ = 100,δ = 2 Q/ spread (bps) M.B. Gordy (FRB) Procyclicality in Basel II May / 22

15 European investment grade CDS index itraxx Main quarterly average MA of quarterly averages spread (bps) Sep08 20Mar08 20Sep07 20Mar07 20Sep06 20Mar06 20Sep05 21Mar05 20Sep04 19Mar04 19Sep03 20Mar03 20Sep02 M.B. Gordy (FRB) Procyclicality in Basel II May / 22

16 Historical performance of the proposed specification Q Q/2 µ = 70,δ = 2 µ = 70,δ = 1 22Sep08 20Mar08 20Sep07 20Mar07 20Sep06 20Mar06 20Sep05 21Mar05 20Sep04 19Mar04 19Sep03 20Mar03 20Sep02 Q Q/2 quarterly average MA of quarterly averages 22Sep08 20Mar08 20Sep07 20Mar07 20Sep06 20Mar06 20Sep05 21Mar05 20Sep04 19Mar04 19Sep03 20Mar03 20Sep02 M.B. Gordy (FRB) Procyclicality in Basel II May / 22

17 Implied volatility of S&P Lehman Brothers VIX Asian Currency Crisis LTCM Enron M.B. Gordy (FRB) Procyclicality in Basel II May / 22

18 Advantages Transparent, as α jt disclosed to public. No distortion of Pillar 3 disclosures. Applied uniformly to all bank assets within jurisdiction. Will tend to work best for large institutions whose performance is highly correlated with state of macroeconomy. Very difficult for individual institution to abuse. For banks, zero cost of implementation. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

19 Tilting the playing field? When national economies not in synch, some countries will have higher α jt than others. So? Smoothing by any other means would do the same, only with less transparency. Buffer indexed by host country (not home) to minimize competitive distortion within-market. Average α jt over cycle could vary across countries. For example, due to national differences in long-term behavior of indices. Certainly an issue, but at least transparent. Can remediate over time by recalibration of µ j,δ j. Problem is inherent in Basel II, not just this proposal. Difficulty of PD validation leaves wiggle room for regulator inclined to laxity. If competitive concerns insurmountable, then Basel Committee can specify uniform international α t based on global factor. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

20 Autoregressive smoothing of a capital buffer Autoregressive process that adjusts the buffer towards the mean level of Q/2. B i,j,t = γ ((1 + g i,j,t )B i,j,t 1 (C i,j,t (1 + g i,j,t )C i,j,t 1 )) + (1 γ) Q 2 C i,j,t Impose bounds 0 B i,j,t Q. g ijt is the rate of asset growth for bank i in jurisdiction j in period from t 1 to t. Smoothing parameter γ set by Basel Committee, controls speed of adjustment towards Q/2. Advantages: Simple to implement, automatic, transparent. Disadvantage: Presumes stability of the bank s business model. Poorly suited to large, complex banks that can quickly alter risk profile through trading activities or through acquisitions and dispositions of business units. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

21 Are the models excessively procyclical? An underexplored question is whether bank models of economic capital (and the Basel II model) are excessively procyclical. Ratings-based models (e.g., CreditMetrics) assume time-homogeneous, serially independent transition processes. Most structural models (including KMV) assume deterministic default boundary. Implies that changes in PDs are serially independent. Empirical literature suggests firm-specific target for leverage ratio (Hovakimian, Opler & Titman, JFQA, 2001), implies mean-reversion in ratings and PDs. Next generation portfolio models could be less procyclical. Collin-Dufresne & Goldstein (JF, 2001) extend structural model to case of mean-reverting stochastic default boundary. Stochastic default intensity models invariably allow for mean-reversion. M.B. Gordy (FRB) Procyclicality in Basel II May / 22

22 Conclusion Key advantage of countercyclical indexing is transparency. Likely to be most effective and most easily implemented for large, well-diversified banks. So far assumed that the countercyclical index would be specified as a mechanical function of macroeconomic or market indices. Mechanical function limits discretion, which is desirable under ordinary circumstances. Makes changes in the capital buffer more readily understood and anticipated by market participants. Reduces scope for competitive lenience across jurisdictions. However, should beware Goodhart s law: Any statistical relationship will break down when used for policy purposes. Should be leeway for jurisdictions to depart from the mechanical rule under exceptional circumstances. Can be conceived along lines of the systemic risk exception governing prompt corrective action in the US. M.B. Gordy (FRB) Procyclicality in Basel II May / 22