How Credit Scores Predict Your Behavior
May 25, 2013 — -- Credit scores, as with most attempts to make sense of our world, operate on the assumption that the near future will look a lot like the recent past, and that going forward, people can be expected to behave pretty much as they always have. That is, credit scores base their predictions of the future on their knowledge of the past.
To leverage this knowledge of the past with the idea of building a crystal ball view into the future, the people who build credit scores thoroughly analyze credit bureau information to understand how consumers have gotten to where they are now, and, most importantly, which of the credit bureau data can be used to predict specific future credit behavior for all consumers.
Since most people will exhibit the same behaviors, good or bad, for years at a time, it's not hard to build a scoring model that simply predicts that people with bad credit today will have bad credit tomorrow, and that people with good credit today will have good credit tomorrow. The trick in credit score building is to predict which consumers with problem histories today are most likely to turn the corner and become low risk -- and thus profitable -- for lenders in the future, as well as identifying credit applicants who may look good now, but who appear to be headed for financial trouble.
The information used to observe such consumer credit behavior is found in the vast quantities of credit bureau data held by the three major credit bureaus: Equifax, Experian and TransUnion.
A key feature of the credit score building process is a "two-snapshot" method of looking at the data. By observing credit reports of a large sample of consumers as of two points in time -- the beginning and ending of a two-year period -- scoring analysts are able to simulate a laboratory-like study of consumer behavior:
- Snapshot 1 -- a set of credit reports for a large set of consumers at one point in time, e.g. January 1, 2010
- Snapshot 2 -- another set of credit reports for the same consumers as in Snapshot 1, only two years later, e.g. January 1, 2012
The score developers observe the 2012 credit behaviors for these consumers, such as how they pay and how much they owe, and then take a look back at the 2010 data in search of past behavior patterns leading up to what was seen in 2012. For example, if a high proportion of people who filed for bankruptcy during that two-year period of 2010-2012 showed late payments on their 2010 reports, while at the same time a high proportion of people who paid all obligations on time during that two-year period had very few late payments as of 2010, it could be concluded that people with late payments today are more likely to file for bankruptcy over the next two years than people with no late payments on their current credit reports.
This kind of information is said to be predictive of future behavior, and is likely to make up the credit scoring factors that become the characteristics of a credit scoring formula. The most predictive of these factors are placed in a "score card" with points assigned for weighting according to their predictive value -- with the most predictive characteristics assigned the most possible points. The sum of all points achieved for these factors becomes the credit score representing the risk for a particular consumer.
