English Appendix
Payment defaults andthe change in the trade cycle
Models for evaluating credit-insurance risk in view of the changing trade cycle in the Spanish economy
The world finance system is now
entering what seems to be a
deep slump. It started off as a
liquidity squeeze but is now
hardening into a credit squeeze
as confidence falls in the finance
institutions themselves.
FRANCISCO ARENAS ROS
MAPFRE CAUCIÓN Y CRÉDITO
THIS SLUMP HAS SEVERAL STRIKING FACTORS:
- The involvement of practically all types of risk: liquidity, credit, market, operational, reputational,…
- The difficulty of quantifying the effects, given the fragmentation of the risk.
- The value of the losses already acknowledged by the companies themselves, in a continual trickle. —And, last but not least, the fact that a large part of the architecture developed in recent years, precisely to improve risk management, has failed to do its job: the sophisticated structured risk assessment models, the major credit-rating agencies, the supervisory authority of some countries, the corporate governance rules of some major finance institutions…
It is too soon to draw any conclusions about this
slump, among other reasons because it is not over yet
and because the stakeholders are still trying to bring it
to a quicker end and soften its effects. Nonetheless I
regard it as necessary to kick off with this initial
reflection, given that one of the triggers of these
problems was precisely the use of certain structuredproduct
assessment models that supposedly meant
converting a bad underlying risk into securities with an
AAA rating.
In this article I am going to address the credit rating models used to evaluate ordinary financial operations carried out by finance institutions and by credit insurers.These models, which could be regarded as less sophisticated, are unlikely to be as severely besieged by the abovementioned questions and problems.
I am going to concentrate on credit insurance but I consider that some conclusions of this article could be of interest to professionals who, over the next few months, will have to tackle a rise in delinquency as the Spanish economy swings into the slump phase. In credit insurance the damage covered is the final loss experienced by the insured as a result of its debtors’ insolvency, on deferred-payment commercial operations.
NEW CREDIT RISK MANAGEMENT (BASEL II / SOLVENCY II)
Credit risk has undergone a veritable revolution in the last decade, initiated in the banking field but based on the use of actuarial and statistical methodologies (expected loss/unexpected loss).
The possibilities offered by new technologies in terms of the mass information processing and generation of scenarios, the breakthroughs made in market risk management (mainly the VaR – Value at Risk) and the need for modelling complex finance products (derivatives, options,…) have all led to a greater sophistication in the credit risk management tools wielded by the banks, since the appearance of Creditmetrics, developed by J.P. Morgan in the mid nineties.
This new credit risk management was given a great boost by the Basel II accord («International Convergence of Capital Measurement and Capital Standards: a Revised Framework») published in June 2004.The long gestation period of this accord, involving a groundbreaking collaboration system between supervisory authorities and private finance institutions, has helped to bring the new risk management culture to a much wider audience. Credit risk, or default risk, has been dealt with very thoroughly in the so-called First Pillar (Minimum Capital Requirements), as befits its central character among banking risks.
In the insurance world credit risk plays a more secondary role, apart from one particular arrangement that affects the default of reinsurance companies; due to its specific nature it is known as the «reinsurance risk». For this reason less interest has been shown in this aspect in drawing up the future solvency norm for European insurance companies, «Solvency II». Nonetheless there is an insurance class in which credit risk is the central concept and its management can be considered to be almost like underwriting risk. Obviously I am referring here to credit insurance.
As in Basel II the provisions of Solvency II are likely to provide for at least 2 credit risk calculation methods: one standard applicable to all companies and another only for these companies that have their own validated models; this would be promoted in the interests of lower capital consumption.
STATISTICAL CREDIT-INSURANCE PREDICTIVE MODELS
The development in credit insurance of internal credit rating models may have the aim of guaranteeing the company’s solvency, also allowing the application of a more favourable legislation.
But they may also be conceived as a pure management tool to improve profitability and the service to insureds. Companies with suitable systems for calculating the expected loss corresponding to each debtor will be able to discern the more profitable segments and cut down the loss ratio by rejecting those debtors in which the risk premium is not covered.
The concept of expected loss has 3 basic components:
- Probability of Default (PD).
This is the likelihood of the insured’s debtor defaulting on its credit obligations. It is an anticipated default rate, recording the quality effect (solvency).
This is the best indication of expected loss, since it may vary widely in any given portfolio of debtors. I.e., there may be a glaring difference between a debtor with a good rating and another with a bad rating; for example, it can be a factor of one thousand between a debtor with a PD of 0.01% and another of 10.00%. - Exposure at Default (EAD).
This is the sum pending payment at the time of default. It records the quantity effect (volume). It represents the use of the rating granted by the insurer at the moment of default.The coverage percentage provided for in the policy should also be applied to determine EAD. - Loss Given Default (LGD).
The loss given default is the loss that would finally be incurred in the event of default. It represents the probability of not recovering the debt once the default has been forthcoming. It would be the inverse of the recovery rate. There is a fourth component of the expected loss, namely Maturity (M).Nonetheless it is of less importance in credit insurance since the coverage of commercial operations usually involves shorter terms.
In Spain credit insurance is especially favourable terrain for the application of predictive statistical models based on the historic defaults of each company:
- The term to maturity of commercial transactions is short, normally a maximum of 120–150 days; nearly always less than one year. This means that rating transition matrices do not have to be used.
- Guarantees (collateral) exist only in exceptional cases, meaning that conditional probabilities do not have to be used.
- There is a reasonable culture of financial transparency, so financial information databases with a reasonable quality and depth can be obtained. In credit insurance the credit risk does not reside in the insured but rather in the insured’s debtor, so the relation is indirect and information on the risk is sometimes hard to come by.
Spain does present a problem in terms of the data time-span, however. In theory a period of ten years should be enough to ensure that the variations caused by trade cycles are properly captured by the models. In the Spanish case this does not quite hold true because of the buoyant economy in recent years. If we go back ten years we record only the modest default rate peaks of the 1999 and 2002 «blips».
This begs the question of whether models built up on historical data could be useful in view of the Spanish economy’s current downturn.
CHANGE IN TRADE CYCLE / SLOWDOWN
In Spain the downturn was clearly recorded in Q1
of 2007.The slowdown has been constant since then,
though its intensity and duration is as yet uncertain.
The long predicted «gentle touchdown» of the «real
estate balloon» in fact occurred, its effects then
exacerbated by an early summer credit squeeze.This
put an even sharper brake on real estate activity and
residential building, with knock-on effects on
confidence and demand.
In these circumstances, and especially in the current pre-election context, we need to weigh our words very carefully. Use of the terms «slump», «recession», «slowdown» or «turbulence» are sometimes telltale indications of a pundit’s ideology.Regardless of the expressions we use, there is no doubt that the economic situation is now having slump-like effects on the default rate, and these effects are bound to increase in the future.Whenever I explain these terms I am reminded of a quote by Woody Allen who categorically declared that money did not make him happy but it produced a feeling that was so very similar that he could hardly tell the difference.
We who work on default-related issues are already suffering an upsurge but the starting point was practically an all-time low. For diverse reasons the default effect is likely to be pretty similar to that experienced in 1992-1993, even if default in percentage terms does not reach the heights of that time.
Under these circumstances we are once more prompted to wonder whether it is worthwhile using models that have not been designed to reflect cases as adverse as those they are now to be applied to.
CHANGE OF CYCLE MODELS
I personally am in favour of their use, although this should obviously be more prudent than in ordinary circumstances.
The current slowdown will produce a greater default probability than that predicted by the models and a greater severity (as the recovery rate falls); conversely, risk exposure may also fall as insurers draw in their horns.
The increase in the default probability would occur on two fronts:
- On an individual level: it would increase in companies belonging to sectors most affected by the new circumstances; in brief it would be those with the sharpest drop in activity, with raw material price increases that could not be passed on and those hardest hit by the credit squeeze.
- At portfolio level: the joint expected loss will increase as a result of the correlation.As well as the correlation deriving from the trade cycle there is another correlation driven by the construction activity.This activity is very significant in the Spanish case and has been the driving force in good years behind the so-called «economic miracle»; at this juncture, however, it is about to show its negative effects.
The question of default correlations is perhaps
one of the points that still needs to be better dealt with
in the new credit risk management and there is still
some way to go in modelling terms. Basel II, for
example, assumes standard correlations for SMEs
ranging from 12 to 24 per cent.Whether these
estimations are realistic, the coming months will tell.
In any case it is more important than ever to underline certain key ideas on the use of evaluation models:
- a) The models do not substitute the analyst’s
expert judgment:
The models do not substitute but round out the work of the analysts, allowing them to take their decisions on firmer ground in the ever-changing economic environment.This medley of criteria makes implementation and monitoring more difficult, since control needs to be kept over the direct decisions of the system and over those in which qualitative elements have been phased in by the analyst; these elements have to be accounted for and recorded.
Analysts are those responsible for risk decisions and are bound to question model evaluations they disagree with. - b) Analysts have to be familiar with model weaknesses:
It is important for analysts to be aware of which circumstances are not properly recorded by the models; for example, the financial expenses in a moment of rising interest rates. An exhaustive monitoring is also necessary to pinpoint new weaknesses as a result of the new circumstances; for example, increase of raw material prices with a significant effect on a given sector. - c) The models enable tasks to be prioritised:
A task prioritising system becomes even more important in difficult times.When the number of companies to be analysed is very high and the number of warning signs increases, this facility is always crucial.
To wind up, we could say that the shortfalls of the car have to be made up by the driver. In this case we have to bear in mind that we are driving firstgeneration vehicles without ABS or four-wheel drive. This calls for cool heads in poor driving conditions.