by Dana Wiklund

Time for organizations to validate their predictive models

News
Jul 29, 20092 mins

As the global economy struggles in recession, lenders across industry verticals should closely track and validate their scoring models. During economic times of rapid expansion and contraction, scoring models can degrade in their ability to make optimal loan recommendations. Sound population and scorecard tracking and validation practices are part of a robust enterprise risk management program.

The vast majority of today’s lending for both consumer and small business is done through the use of credit scorecards and predictive credit criteria. Statistical analysis of scoring systems relies on a predictive model development and then ongoing, the population being evaluated mostly resembling the original development sample. When populations change, predictive models of all types become unstable and loose their predictive abilities.

Over the past two years there has been a sea change in the consumer and small business economies alike. Delinquencies have risen across all loan types, working capital has become a precious commodity for businesses of all sizes and there have been demand shifts throughout the economy. These events all point towards an urgent need for all financial institutions to validate their predictive models. Validations require seasoned loan traunches but there are actions lenders can take to understand how their models are reacting to changes in populations. Review of early delinquency rates across score intervals, population and attribute stability analytics as well as analysis of approval and override rates can yield valuable early warning signals that predictive models might be faltering

Risk and finance executives within the financial services industry should be looking at the scope of predictive models across their enterprises — risk models, response models, profitability and propensity models as well as behavioral risk models and recovery models. Predictive models used across a credit life cycle are sensitive to changes in the macro economy. Technology solutions providers to financial services are advised to see where solutions for data organization and business analytics can be leveraged to help clients evaluate changes to their populations. Ignorance of changes to the payment behaviors of new and existing populations due to a changing economy will result in damaging financial performance later on.