The Great Recession introduced the public at large to a previously seldom mentioned type of bond called Collateralized Debt Obligation or CDO. CDO’s are asset-backed securities whose purpose was to attract more cash to the market by breaking up loan portfolios into different classes of debt or levels of risk and return.
Classifying the bonds according to risk AA to BB and maturity dates of 1, 3, 5 and 10 years the number of options for investors looking to target risk and return coupled with a choice of time lines attracted more money to CDO’s. Investors saw less risk with an investment that had a predictable cash flow and defaults.
CDO’s served their purpose in the economy by attracting more capital to the market that was in turn loaned back into the consumer debt market at lower rates. Credit Derivatives or an insurance policy on the debt helped calm investors nerves if the loans within the asset class began to default at higher rates or an unexpected number of the loans paid out early, the insurance would kick in paying investors their expected return.
The debt market collapsed as the CDO’s became increasingly complex and the rates of default far exceeded expectations for the class. Bond ratings on this investment product were increasingly divorced from the actual risk. When the insurers like AIG were overwhelmed by the number of calls on the Credit Derivates that they were the guarantor of, the market froze up.
A question the market has to ask itself is how did investment risk deviate so far from its historical ratings and what factors contributed to it.
The complexity of the debt offerings and the mixing of assets classes in addition to lack of market oversight were factors but at the root of the cause is the financial institutions move away from basing risk in large part on credit bureau scores to internal scores.
Credit Risk profiling/modeling proliferated within the last decade as statistical software such as SAS and SQL developed the capability to allow lenders to extract or mine data from their existing portfolio and determine the likely hood of a particular applicant defaulting on their loan. The software drills down into the data systematically, detects important relationships, co-factors, dependencies and associations between multiple variables, and assigns values to segments of the loan portfolio. The analysis can identify profiles of high and low risk loans through a systematic analysis of all available data.
An auto lender for example keeps loan records on motor vehicle purchase in its database including default information: occupation, income; vehicle type, manufacturer, model, year make, price, loan amount, default amount, etc. The lender wishes to know which types of loans for motor vehicle purchases are at the highest risk, i.e., highest default ratio by probability. The credit bureau score in contrast is looking at the consumer current financial situation and by nature is forward-looking.
Credit Scoring is backward looking by its nature. A scoring model analyzes historical data and assumes that current offerings with similar characteristics will behave in the same manner. The downside rears its ugly head when the market as a whole does not perform according to the criteria assumed in the scoring model. Scoring models such as a Neural Network can predict either relative default levels or expected default levels with surprising accuracy in presumed non-risky segments of business. When the old rules no longer apply in a deteriorating economy, scoring models may accelerate default rates when the criteria of the model is assuming a stable or growing economy.
The investment bankers who packaged debt for sale on the open market with a grade of AA in 2006 and once again bundled loans from the same lender using the same credit risk factors in 2008 in reality offered an AA rating investment that contained a much higher default risk. Until scoring models are able to foresee the economic future and adjust the modeling criteria accordingly, the ability of rating agencies such as Moody’s to accurately asses the risk of a CDO will be in question.
Credit Risk modeling is not going away in fact it is being used to decision and price loans throughout the consumer finance industry. Credit Risk Analysts and Collectors are the primary positions that banks and other financial institutions are currently hiring.
Banks and financial institutions have tightened lending activities as delinquency and charge off have increased and the availability of capital to make new loans has shrunk precipitously, especially in light of the market for CDO’s drying up. What is to prevent the cycle from repeating itself again when the economy stabilizes? The pressures to grow loan receivables, revenue and earnings per share will not disappear, once again creating a situation where additional risk will be taken for greater returns with the unspoken specter of runaway defaults lurking in the background
The answer for this quandary will be debated by minds greater than mine but until scoring models are able to adapt to financial conditions, the human element needs to remain an essential part of the loan underwriting decision process. When the institutional investor does not question whether an AA bond is truly an AA bond, the credit markets will have overcome a significant hurdle and will be on the road to recovery.
Credit This! 