b1201_Chapter-09.qxd 9/7/2011 9:13 AM Page 121 b1201 Macroprudential Regulatory Policies 9 Macroprudential Regulatory Policies Downloaded from www.worldscientific.com by NATIONAL UNIVERSITY OF SINGAPORE on 10/27/17. For personal use only. COMMENTARY ON MACROPRUDENTIAL AND OTHER POLICIES E. Philip Davis* National Institute of Economic and Social Research In providing commentary on this topic, we begin with a National Institute perspective, highlighting work done partly for the UK regulator, the Financial Services Authority, and now emerging in journals. Our research shows that OECD banking crises differ from emerging market ones. OECD crises were absent up to 1973, but since then they have been very common, albeit with a gap for most datasets from 1995 to 2007. In our work looking at predictors of such crises using logit models (Barrell et al., 2010a, b), we found a role for bank capital ratios and bank liquidity ratios in determining those crises, and apart from that, house price bubbles and current account deficits. Our results contrast with those from the World Bank and IMF, with global samples of crises dominated by emerging market crises that highlight variables such as GDP growth, M2 as a proportion of reserves, and credit growth as crisis predictors. One reason for this is that data for bank capital and liquidity are unavailable for emerging markets, nor for the most * E. Philip Davis is a Senior Research Fellow at the UK’s National Institute of Economic and Social Research and an Associate Professor at Brunel University, West London. 121 b1201_Chapter-09.qxd Macroprudential Regulatory Policies Downloaded from www.worldscientific.com by NATIONAL UNIVERSITY OF SINGAPORE on 10/27/17. For personal use only. 122 9/7/2011 9:13 AM Page 122 b1201 Macroprudential Regulatory Policies E. P. Davis part are house prices. But we would argue that we are also detecting fundamental differences in crisis determination for OECD countries, which are less vulnerable to banking crises driven by currency crises than are emerging market countries. In other work for the FSA, we used a crisis model to generate estimates of benefits from tighter regulation to offset against the wider spreads that you get from increasing capital ratios (Barrell et al., 2009). This then gave a suggested optimal degree of regulatory tightening. The paper I want to talk about mainly in my commentary is called “Calibrating Macroprudential Surveillance” (Barrell et al., 2010c). Basically, what we did in that paper was to look at variables that have been used in the past in crisis models (e.g., Demirgüç-Kunt and Detragiache, 2005), testing down to a simplest model possible and using a short data period from 1980 to 1997, a long way before the subprime crisis. What we found is, as was mentioned earlier, that variables such as credit growth, output growth, and M2 to reserves are not significant; instead, our specification search ends with the four variables that I mentioned earlier, namely bank leverage (unweighted capital adequacy), net liquidity ratios, house price growth, and the current balance. We did a number of robustness checks to see whether it was justified to exclude credit growth and GDP growth, such as additional lags. But in each case, they dropped out. Our economic argument for this result is that credit growth may accompany a favorable as well as an unfavorable period for financial stability. Credit growth may be rapid during a period when the output gap is not growing and when supply is growing with demand, but equally in a period when there are major imbalances and excess demand. What distinguishes such a more difficult period for financial stability are trends like rapid house price growth and a large deficit in the current balance. They show when rapid credit growth is likely to lead to a financial crisis. Our model works well; we get 75% of crises correct and 72% of non-crisis periods. As an illustration of the model’s use in macroprudential surveillance, we are able to use the model to forecast b1201_Chapter-09.qxd 9/7/2011 9:13 AM Page 123 b1201 Macroprudential Regulatory Policies Macroprudential Regulatory Policies Downloaded from www.worldscientific.com by NATIONAL UNIVERSITY OF SINGAPORE on 10/27/17. For personal use only. Commentary on Macroprudential and Other Policies 123 up to 2008 (using actual values for right-hand side variables), and we find that it was flagging the subprime crisis quite strongly for the UK, US, France, and Belgium. It also highlights risk in Spain. Possibly the reason Spain didn’t have a full-blown banking crisis was that the authorities had introduced dynamic provisioning that protected the banks from the losses they incurred. There are also relevant results for macroprudential regulation. Because the model includes policy variables, namely liquidity and leverage on the right-hand side, we can invert it and then calculate what levels of those variables would be needed to get the crisis probability down to 1% for each country, i.e., a crisis every 100 years (as compared to every 16 years in the sample). We show, on average, over 1998 to 2008, i.e., out-of-sample, what level of leverage and liquidity would be needed to get us down to this 1%. The key result is that a rise in leverage and liquidity of 3.7% points would be sufficient to get 1% on average across the OECD. However, there are differences between countries in the degree of tightening that is warranted. A number of countries need much more extra capital than 3.7% in order to get their maximum crisis probability down to 1%. That relates in turn to different levels of macroeconomic volatility, different levels of crisis risk, as well as the initial positions in terms of leverage and liquidity. This aspect has not been discussed much yet in policy circles, but it implies potential externalities from more risky to less risky countries if all have the same Basel III level. Some countries, including the UK and US, need more capital and liquidity than the OECD average to get their crisis probabilities to 1%, and in turn to avoid externalities to other countries when effects of their crises spill over. As was mentioned earlier, we find credit insignificant as a crisis predictor, and we argue that countercyclical provisioning should be based on house prices and the current account. In Barrell et al. (2010c), we show calculations of extra capital needs based on 5% higher house prices and show that the need for extra capital in order to keep the crisis probability down does build b1201_Chapter-09.qxd Macroprudential Regulatory Policies Downloaded from www.worldscientific.com by NATIONAL UNIVERSITY OF SINGAPORE on 10/27/17. For personal use only. 124 9/7/2011 9:13 AM Page 124 b1201 Macroprudential Regulatory Policies E. P. Davis up over 1998–2008, but in a very non-linear way. It, of course, depends on all the variables in the model and there is no scope for a simple “rule.” This is an area of interest for further research. With the above as background, we can address now the topic of the session directly — looking first at the interaction of macroprudential with monetary policy. We’ve shown that the overall country adjustment in prudential policy to reduce crisis probabilities depends partly on macroeconomic volatility. So one argument one can make is that if monetary policy can reduce house price bubbles and imbalances in the current account, it impacts on macroprudential adjustment. On the other hand, it is more doubtful that macroprudential regulation will significantly affect the macro economy, and hence the demands on monetary policy. I’m referring here to our own calculations, but also the Basel calculations that were made in FSB (2010). According to simulations with macroeconomic models in various countries, 1% more capital and liquidity seems to take around 0.1% off GDP, which is not huge, and only 0.03% if all countries moved together. So in other words, the effect of regulatory tightening on the macro economy is small so long as the tightening is gradual, which I think is a point relevant to monetary policy. As the BIS point out in their recent Annual Report, there are some benefits to monetary policy of a more active macroprudential policy. Less financial crises imply less economic fluctuation. If crisis risk can be reduced, interest rates are less likely to become ineffective due to financial distress, and also there will be less need to cut interest rates for financial stability reasons in the downturn with possible inflation risks. Conflict between monetary and macroprudential seems most likely if it’s an ex-post issue i.e., there is already a crisis. In advance of the crisis, the optimal policies move in the same direction — to avoid bubbles that threaten financial instability and that are also going to be inflationary at some point. So both macroprudential and monetary policies should be tightened. There is also notable interaction of macroprudential policy with microprudential regulation in that crisis probabilities depend b1201_Chapter-09.qxd 9/7/2011 9:13 AM Page 125 b1201 Macroprudential Regulatory Policies Macroprudential Regulatory Policies Downloaded from www.worldscientific.com by NATIONAL UNIVERSITY OF SINGAPORE on 10/27/17. For personal use only. Commentary on Macroprudential and Other Policies 125 on the stringency of microprudential regulation. The weaker the regulation, the more adjustment needed for macroprudential aims. As mentioned, some countries will need tighter average regulation than the OECD/Basel III average. Whether or not that is the case, the macroprudential stance of a country is a benchmark for all banks, but there would also be a need for discretionary adjustment at a micro level for riskier banks and the larger banks. Finally, we argue that the use of estimated credit-to-GDP gaps as recommended by the BIS is not necessary for calculation of the need for countercyclical buffers as credit is not significant as a crisis predictor — and does not even Granger-cause property prices in most countries, according to our research. There are also some points to raise on competition policy. Systemic institutions that are of concern to macroprudential policy as being “too big to fail” (Haldane, 2010) are also likely to have market dominance. This suggests a synergy with competition policy, notably in recommending the unwinding of some of the mergers that took place under duress in the crisis (the UK example is Lloyds/HBOS). What I think is more questionable is whether proposals for dividing universal banks have a competition policy justification, unless they dominate both commercial and investment banking. Equally, it is not clear that the outcome of the subprime crisis shows the universal bank model’s weakness — Northern Rock and Lehman’s were “specialized” institutions. We now comment briefly on the papers. The Reichlin and Pill paper is an excellent analysis of the ECB’s extraordinary monetary policy and its impact on interest rates in lending. One technical point is that it would have been helpful to have the standard error bands in the first part of the paper as well as the second part because it was not clear whether the effects were significant. Turning to some more substantive points, a first is on wholesale funding. The paper takes the dependence of banks on wholesale funding as given. That’s been a big issue for central banks to provide liquidity to compensate for the dependence of banks on the wholesale funding markets that have been b1201_Chapter-09.qxd Macroprudential Regulatory Policies Downloaded from www.worldscientific.com by NATIONAL UNIVERSITY OF SINGAPORE on 10/27/17. For personal use only. 126 9/7/2011 9:13 AM Page 126 b1201 Macroprudential Regulatory Policies E. P. Davis chronically unavailable. The question one has to raise is whether the banks should have been in that position in the first place, and therefore, is the microprudential regulation of liquidity actually sufficient? There’s an argument that there should have been tighter regulation of the dependence on wholesale funding. Banks could have been prevented from taking on such a high proportion of non-customer funding and that should also have been viewed at the macroprudential level (for example, the aggregate customer funding gap of the banking system). The wholesale funding of the whole system should’ve been looked at and it could be looked at countercyclically. So that’s a substantive suggestion drawing from the paper. Equally, the paper assumes that the interbank spread of zero prior to the crisis was in some way normal. All the central banks assumed that as well, but economic logic states that a zero LIBOR versus Central Bank rate spread implies gross underpricing of risk. Why wasn’t there always a positive spread? Was it perhaps because the expectations of the market that the banks would always be bailed out and therefore, a zero LIBOR spread was in some sense rational? Or does it also show disaster myopia both in banks and in policy circles — that a risky position with large proportions of wholesale funding and a zero spread is somehow seen as normal. We need to view such patterns from first principles in future. There’s a further issue of unwinding of extraordinary ECB policies. Banks have become dependent on central bank funding and need incentives to adjust balance sheets away from wholesale funding. Can exit from extraordinary policies be accomplished without a major credit contraction? Then there’s the broader issue of integrating central bank liquidity policy and bank regulation. Do money market desks talk to regulators enough? How can one gauge the overall stance of macroprudential policy? Is the tradeoff limited? How was the liquidity support integrated with inflation concerns of central bank? The nature of ECB’s extraordinary measures deserves further attention. Why did they not buy up government bonds (as in b1201_Chapter-09.qxd 9/7/2011 9:13 AM Page 127 b1201 Macroprudential Regulatory Policies Macroprudential Regulatory Policies Downloaded from www.worldscientific.com by NATIONAL UNIVERSITY OF SINGAPORE on 10/27/17. For personal use only. Commentary on Macroprudential and Other Policies 127 the UK) or also private sector assets (US) as more effective means of reducing long rates and supplying credit directly than buying long-term money market instruments? Our work at NIESR suggests that as a consequence, the ECB got a lower impact on GDP for its special measures. Has any counterfactual work been done to see if this was more or less effective to stick to such assets, given the differing financial structure in the Eurozone? And finally, how different would the outcome have been if Basel III/countercyclical macroprudential policies were already in place? Would the conclusion be that it was a liquidity and not a capital crisis for most Eurozone countries, so it would have made little difference? In the Király paper, it was good to highlight the needs for cross-country coordination of macroprudential policies, and one only has to consider the crisis in Iceland to show how important those things can be. The local example of Austria and Hungary was very helpful. But the question arises — why wasn’t product regulation, banning foreign currency mortgages, used to complement macroprudential? Arguably, they should have never been allowed to have been sold in Hungary in the first place. Their suggestion that we need global credit growth based macroprudential policy is harder to sustain. Countries with slow economic growth and very little cross-border lending would consider themselves penalized. They would say, why should we suddenly have this buffer based on global credit growth? The focus needs to be on the countries where the cross-border lending is coming from. This paper also raises the issue of cross-border wholesale funding. In the UK, this rose from around zero in 2000 to around £400 billion in 2008. This points again to the need for macroprudential regulation on the liability side of liquidity, not just on the asset side, which seems to me to be the major focus at present. On the other hand, it’s worth adding that in our research at NIESR we tried very hard to get wholesale funding into our crisis models, and we found that it couldn’t readily be found to be a stable predictor. b1201_Chapter-09.qxd Macroprudential Regulatory Policies Downloaded from www.worldscientific.com by NATIONAL UNIVERSITY OF SINGAPORE on 10/27/17. For personal use only. 128 9/7/2011 9:13 AM Page 128 b1201 Macroprudential Regulatory Policies E. P. Davis In the paper, it is argued that loan to value limits need to be fixed. We suggest they should be flexible over the cycle, lowered, for example, during periods of house price boom as is the case in emerging market countries that use such ratios actively. The authors talk about the need for measuring a macroprudential stance. I would argue that our work at NIESR cited above gives at least an initial approach to that. Our models generate probabilities of crises in each period that in effect measure how tight the stance of macroprudential policy is. For example, using a rolling probabilities model, we can forecast one year ahead; then the indicator of the stance is the excess of the crisis probability in the next year over an acceptable level of risk. The model also shows how it could be reduced (albeit not immediately). The authors discuss the fiscal rules of the stability and growth pact, but they didn’t elaborate on that. We would contend that there’s an important role of fiscal coordination in macroprudential, notably in terms of a need for fiscal flexibility to respond to future banking problems. This is an important argument for the fiscal consolidation that’s underway at present; we need to be ready for any future financial crises and that consolidation can be helped by coordination. The Greek example shows the risk of contagion from uncontrolled fiscal policies with cross-border lending, and is another argument for coordination. To illustrate their argument further, I want to highlight Barrell et al. (2010d) on cross-border effects of crises. What we did was to look at the weighted incidence of ongoing crises elsewhere in our crisis model and see what the effect is on our probabilities elsewhere. We came up with a table which unsurprisingly shows that a crisis in the United States has a big impact globally. Its impact depends not only on the size of the country, of course, but on all the other coefficients and variables within the other countries. So that’s an indicator of induced changes to crises probabilities that are affected by crises elsewhere, and that seems to me very relevant to this discussion about cross-border coordination. b1201_Chapter-09.qxd 9/7/2011 9:13 AM Page 129 b1201 Macroprudential Regulatory Policies Commentary on Macroprudential and Other Policies 129 Macroprudential Regulatory Policies Downloaded from www.worldscientific.com by NATIONAL UNIVERSITY OF SINGAPORE on 10/27/17. For personal use only. References Barrell, R., E. P. Davis, T. Fic, D. Holland, S. Kirby and I. Liadze (2009). Optimal regulation of bank capital and liquidity: How to calibrate new international standards. FSA Occasional Paper, No. 38, October. Barrell, R., E. P. Davis, D. Karim and I. Liadze (2010a). Bank regulation, property prices and early warning systems for banking crises in OECD countries. Journal of Banking and Finance, 34, 2255–2264. Barrell, R., E. P. Davis, D. Karim and I. Liadze (2010b). How unique was the subprime crisis? Discussion paper, No. 351, National Institute of Economic and Social Research. Barrell, R., E. P. Davis, D. Karim and I. Liadze (2010c). Calibrating macroprudential policy. Discussion paper, No. 354, National Institute of Economic and Social Research. Barrell, R., E. P. Davis, D. Karim and I. Liadze (2010d). Contagious effects of banking crises in OECD countries. Mimeo, National Institute of Economic and Social Research. Demirgüç-Kunt, A. and E. Detragiache (2005). Cross-country empirical studies of systemic bank distress: A survey. IMF working papers 05/96, International Monetary Fund. Financial Stability Board (2010). An assessment of the long-term economic impact of stronger capital and liquidity requirements. Basel, August. Haldane, A. (2010). The $100 billion question. Speech, Institute of Regulation and Risk, Hong Kong, March 30.