The New Lending System - Putting common sense into compliance and risk assessment

The Federal Reserve recently proposed the most significant changes to the mortgage provisions of the Truth in Lending Act (TILA) since it was enacted. Mainly, this proposal would impose new limitations on payments to mortgage lenders to curb predatory lending practices including originators’ steering borrowers to mortgage products that aren’t in borrowers' best interest. The new provisions are expected to improve the borrower’s loan affordability. In other words, the Federal Reserve wants to ensure that lenders make the right loans to the right borrowers.

In our new lending system book, failure to qualify the borrower’s loan affordability is identified as one of the root causes that contributed to the financial crisis. Many underwriting systems have incrementally worked off the partial and static information throughout the decision process. As a result, consumers were overcharged or approved for the loans that were not affordable. Many subprime loans were, in fact, made by steering “high risk borrowers”, who were vulnerable, due to their future financial conditions (such as future high debt to income ratio), to unaffordable or “high risk loans” (such as interest only ARM).

Improving borrower’s loan affordability should be one of the key lessons learned from the crisis. To qualify and improve the borrower's affordability, we first need to be able to define and measure it (as Clark put it in his recent blog “You cannot manage what you cannot measure”). The key aspect here is the dynamic nature of the borrower loan affordability and the circumstances which can change rapidly relatively to the markets. Therefore, a complete loan affordability definition and measurement should go beyond current borrower’s qualifications, monthly payment, property appraisal, etc. This requires a lending system to be forward-looking and be able to effectively predict borrower’s future loan affordability and financial vulnerability, and how they will be impacted by future states of the market and economy.

It has been estimated that Federal Housing Administration endorsements reached almost $100 billion in the second quarter, another record level that was up over 20 percent from the first quarter of 2009. This is encouraging news and with U.S economy emerging from recession, we expect to see the mortgage lending business return to normal levels. At the same time we need to ask a basic question “What do we do differently?” More specifically, are loans being made using new or fundamentally improved lending systems or simply the same broken systems with bandages? Obviously, given the lessons learned from the subprime crisis, many lenders have tightened their underwriting policy and become more careful with their loan origination practices. However, merely adjusting the current underwriting policy is not enough. A sustainable lending business demands fundamental changes in the loan underwriting framework. Future lending success will hinge upon the lender’s ability to measure and effectively predict the borrower’s future loan affordability. A recurrence of the financial crisis cannot be prevented unless the root problems are fixed.

Clark and I recently spoke on a SAS annual business forecasting conference. We had a chance to discuss issues with attendees from banks regarding business forecasting. The focus was what we can learn from what happened in financial markets? How to restore confidence in business and economic forecasts? One specific question was how to improve forecasting on consumer default in wake of financial crisis and economic recession?

Obviously, one of the biggest challenges facing business forecasting is appropriately adjusting models in light of the structural changes caused by the current economic recession. The recession and crisis have changed the world. Weakened market demands have shrunk both services and manufacturing industries. Consumers’ spending behavior is significantly different from that of prior to 2007. This is not just because consumers have to cut down their spending because of degraded financial condition. People have also been leaning from the crisis and improving their financial literacy to better manage their finance. For example, as financial markets and loan products become more transparent, consumers become more responsible and prudent in borrowing. “Regulatory intrusions” are also expected to have significant impact on lenders and borrower’s behaviors. For example, recent Federal banking regulators’ rules to curb deceptive credit card practices. Investors are more realistic about their investment strategies and expectations on risk taking.

So, what do these changes mean to business forecasting?

Two traditional approaches, net flow rates and vintage curves, have been used widely as simple tools in loss forecasting. Net flow rates can depict the change of loan or accounts from one delinquency status to another in a given month for a specified future date, while vintage curves are often used to track credit losses over loan life cycle. Usually, separate net flow rates or vintage curves often are required for different loan segments such as line of business, region, channel, etc.

A drawback associated with both approaches is they do not tell you specifically which factors cause delinquencies or changes in default severity. Some may argue that net flow analysis is a simplified version of so called Markov chains/transition matrix approach. Therefore one may incorporate some cross-sectional variables (such as score bands, months on books) or macroeconomic variables (such as interest rate, unemployment rate, consumer price index, etc.) into the model to make it more interpretable (i.e. a hybrid model).

But still those variables only explain the outcomes at aggregated or summary level. It is difficult to associate the outcomes directly to loan or account level variables such as borrower’s characteristics (loan-to-value and debt-to-income ratios, employment status, etc.). Therefore it does not provide much insight in terms of identifying early warning indicators and mitigating losses at account/borrower level.

With CCAF, each loan or account is classified into a handle cell, which can be directly associated with values of a set of primary/secondary factors (e.g. 5 C’s of Credit plus a secondary factor). Therefore, when an account changes its delinquency status, it is easy to identify and monitor which factors are directly associated with or contributing to the movement.

In addition, since handle-based structure can incorporate both cross-sectional variables and borrower profile variables into homogenous handle cells (groups) for a more holistic and comprehensive view, the process of generating net flow rates can be significantly simplified. This also reduces the number of vintage curves that otherwise need to be created by using various cross sectional variables.