The Paradox of Liquid Loans discussion by Leonardo Gambacorta Economic Outlook and Monetary Policy Department Bank of Italy The Transmission of Credit.
Published byModified over 7 years ago
Presentation on theme: "The Paradox of Liquid Loans discussion by Leonardo Gambacorta Economic Outlook and Monetary Policy Department Bank of Italy The Transmission of Credit."— Presentation transcript:
The Paradox of Liquid Loans discussion by Leonardo Gambacorta Economic Outlook and Monetary Policy Department Bank of Italy The Transmission of Credit Risk and Bank Stability Conference London – 22 nd May, 2008 These slides represent the views of the discussant only and not necessarily those of the Bank of Italy
L. GambacortaLondon - May 22nd, 2008 Structure of the discussion 1.Main novelty of the paper 2.The analytical part: too many models! There is little that justifies the econometric specification 3.Data and the econometric model. Check of the robustness of the results. What are their policy implications?
L. GambacortaLondon - May 22nd, 2008 1. Main novelty of the paper So far the literature has found that the lead bank in a syndicated loan retains a larger share when there may be significant moral hazard (Sufi, 2007, JoF). Does this more concentrated syndicate structure translate into better outcomes? Yes, but so far the literature has analyzed short-run measures (spreads at origination, Focarelli et al., 2008, JME) Investigate if greater lead bank exposure and the resulting higher monitoring effort raises the probability of a project success (something similar in Dahiya et al. 2007, JoB, but no syndicated loans)
L. GambacortaLondon - May 22nd, 2008 2. The models: are they necessary? The authors present 4 models to justify the results of the econometric part: –pure free riding –no free riding but private information –liquidity shocks and sunspots –credit risk transfer with and without reputation These models are trivial and disconnected. The reader has the impression that they are used to explain only one driver of the story at a time
L. GambacortaLondon - May 22nd, 2008 Example 1: Are we sure that there is always under provision of monitoring effort? Game between two banks in the free-riding model There are two pure strategies asymmetric equilibria with both monitoring and no duplication of effort (not nice if you want to model a world with free-riding…)
L. GambacortaLondon - May 22nd, 2008 Example 2: Are monitoring incentives in syndicated loans modelled in a proper way? In the pure free riding multiple-lending model the bank is indifferent between monitoring and not monitoring if: Therefore the authors conclude that if n↑ than s↓ and the “free riding non-cooperative outcome is not efficient”, which is trivial. A syndicated loan is quite different from multiple-lending. In particular, loan distribution matters, there is a strong incentive in delegated monitoring, there are economies of scale in monitoring. However, all these aspects should be explicitly modelled.
L. GambacortaLondon - May 22nd, 2008 Example 3: The nature of the equilibria In the model with secondary market, the authors deviate from Parlour and Plantin (2007) considering two possible market beliefs. For example, if the market is not aware that the intermediaries will stop monitoring a bank will do not monitor if: This is not an equilibrium in the long run because the market cannot be cheated forever. The authors mention that “due to the difficulties in determining whether a default was “unlucky” or due to “shirking”, it may be sustainable in the short run.” However, this is not an equilibrium but an ad hoc hypothesis.
L. GambacortaLondon - May 22nd, 2008 2. Data and the econometric model Data on syndicated loans are collected from Loan Pricing Corporation’s Dealscan (79.054 deals) Data on firms are taken from Compustat Bond information default are from Moody’s (26,000 issuers). Massive data work! –From the initial number of issuers that recorded a default, 1,200 over 26,000 issuers, after filtering the authors obtain a final dataset of only 115 defaulting firms with information on their lenders’ exposure on syndicated deals. Since the authors assume that the remaining firms - for which they do not have information - do not default, how could this bias the results? –Final dataset of less than 6,000 loans over the period 1997-2007. In most of the specifications 1,500 loans are used. Probit model; corroborated with IV estimations
L. GambacortaLondon - May 22nd, 2008 Robustness of the results (1) 1.The lead bank share exerts a negative influence on the likelihood of default. However, this effect is unstable and it is marginally significant when IV are used (see last column of Table 2) 2.Another result that seems unstable is that “Loan syndicated during upturns in the business cycle are more likely to default”. This effect indeed vanishes when the IV estimator is considered (see column 7 of Table 3)
L. GambacortaLondon - May 22nd, 2008 Robustness of the results (2) 3.The authors should consider also that syndicated loans may be securitized. Obviously, a bank’s monitoring incentive changes a lot if, after the origination, the tranche is sold to an institutional investor. This calls for some work on tracking the borrower-lead syndicate relationship dynamically 4.Some results seems implausible: i.e. the authors claim that “a marginal increase in the lead bank’s share raises the probability that a borrower will have an investment-grade rating in the future by 16% (on a mean of 57.9%).” A marginal increase in the lead bank’s share cannot determine such a big impact!
L. GambacortaLondon - May 22nd, 2008 Conclusions This is a nice paper that analyzes a very important topic, also in the light of the recent turmoil The authors have worked very hard but a last effort is needed to: –Link the analytical to the econometric part –Refocus their discussion on what is the main message of the paper –Provide useful policy implications for their results