How Does Stuff Get In Your Credit Score Anyway?

By now we should all be familiar with the breakdown of what affects your credit score. 35% based on payment history, 30% based on amounts owed, 15% based on length of credit history and 10% each, new credit and types of credit used.  But what does “payment history” really mean and how does a credit score company decide what exactly goes into a score?

Scores, or more accurately the models that produce the scores are made up of a series of individual characteristics, specific pieces of information that are weighted according to how well they predict delinquency.  So under “payment history” there might be characteristics like “how many payments missed in the last 12 months” or “how long ago was the most recent delinquency”.  (Disclaimer: these are fictional characteristics since the actual ones are generally highly guarded secrets.)  These characteristics are then assembled in such a way as to provide a comprehensive model that predicts delinquency in a specific way.  So which characteristics get in and which don’t?

How well does it predict. The first cut is always how well the specific characteristic predicts the desired outcome – typically 90 day delinquency for common credit risk scores.  The most predictive characteristics always go to the top of the list with some consideration for making sure the models aren’t too heavily weighted towards too few characteristics.

Data restrictions. Credit scoring models can only use data that’s available.  In the case of credit scoring models that’s what’s provided by the credit reporting agencies (CRAs).  Why don’t credit scores use income, or employment or residence type?  Because it’s not reported to the CRAs.

Legal restrictions. Some characteristics are just plain illegal! For example the Equal Credit Opportunity Act makes it illegal to discriminate according to age or sex when making lending decision. Even if this information was collected by the CRAs (some of it is or can be derived) it’s off limits for modeling.

Stability. From a practical standpoint you don’t want characteristics that even though predictive, swing wildly from one period to the next.  Scores should remain relatively stable over time absent significant changes in the underlying  behavior. A good example of this would be bank balances.  By definition they tend spike on payday and slowly dwindle until the next payday.  You wouldn’t want to be scored the day before payday!

They have to make sense. If you’ve ever been turned down for credit you know that the lender is required to provide for you reasons why you were declined.  In credit score parlance these are called Adverse Action Reasons and they’re typically driven by the characteristics used in the model if a score was used.  In fact, when you are provided scores  either because you were declined or you purchased them, these reasons also drive the explanations you get for what’s good and bad about your information.  Modelers love to perform statistical acrobatics to squeeze out that last drop of predictiveness but how do you explain to a consumer that they were declined because the natural log of their balance divided by their credit line cubed wasn’t optimal? (Purely hypothetical I have no idea if that’s predictive!)

While it’s mostly a mystery exactly what’s in a credit score there are some boundaries built in to ensure they’re suitable for their intended purpose.

John Ulzheimer is the President of Consumer Education at SmartCredit.com, the credit blogger for Mint.com, and a Contributor for the National Foundation for Credit Counseling.  He is an expert on credit reporting, credit scoring and identity theft. Formerly of FICO, Equifax and Credit.com, John is the only recognized credit expert who actually comes from the credit industry.  Follow him on Twitter here.

Categorised in: Credit Monitoring, Credit Report, Credit Score, Improving Credit, Money & Identity

This post was written by John Ulzheimer