The Beginning of the End for FICO Scores?

by John Birge, Chief Credit Officer

FICO scores have become a hot topic in marketplace lending. Over the past few months, a number of such lenders have criticized FICO and announced that they will no longer use the score for underwriting, favoring instead new data inputs to determine a borrower's creditworthiness.  This trend was summarized in a recent Wall Street Journal article titled “Silicon Valley: We Don't Trust FICO Scores."

 Eschewing a score that has been ubiquitous since the 1980s and is used by banks fits squarely within the broader theme of the disruption of traditional underwriting.  So is this the beginning of the end of the FICO score or is this largely hype, what I refer to as the “marketing of underwriting?”

To answer this question, it’s important to understand what FICO is and perhaps more critically, what it is not.

 

What FICO Is

FICO is a statistical model that uses consumer credit bureau data to predict how likely one is to pay their obligations in the near future. The data variables come from one of the “big three” consumer credit bureaus - Equifax, Experian or TransUnion. So in practice, individuals actually have three unique FICO scores, though they are normally fairly close to one another.

 Per FICO, there are five key types of variables that drive the score:

  • Payment history
  • Amounts owed
  • Length of credit history
  • Credit mix
  • New credit

One of the main criticisms of FICO is that it only considers historical credit events.  This, of course, is true!  And the inputs do tend to focus on negative items such as delinquencies, judgments and liens. This combination of looking backward and focusing on adverse actions, while far from perfect, has proven to be remarkably predictive, performing well across a variety of lending products and various economic cycles.

 

What FICO Is Not

Since FICO sources its input variables from the consumer credit bureaus, it has no “direct knowledge” of any of the following data points which the bureaus do not track:

  • Income
  • Debt-to-income ratio ("DTI") or cash flow
  • Employment history
  • Education history
  • Assets including retirement accounts
  • Checking or debit card behavior 

It is very common for lenders to use income and some form of DTI in conjunction with FICO to make underwriting decisions.

 What about employment and education history?  While most underwriters believe this information contains incremental predictive value, utilizing this data for credit decisioning could present serious consumer regulatory concerns, specifically with respect to the Equal Credit Opportunity Act (ECOA) and the Fair Credit Reporting Act (FCRA).

 It should also be noted that FICO does not include other variables that have become fashionable lately:

  • Social behavior (e.g., Facebook data, LinkedIn connections)
  • Session statistics (e.g., dragging a slider to the max loan amount)
  • Text mining variables (e.g., whether a person has a gmail or yahoo email domain)

These types of variables are certainly intriguing from a “big data” and behavioral science perspective and have become part of what I refer to as the “marketing of underwriting.” But utilizing this data could present potential regulatory issues. And of course there is the question of whether the data actually adds value, which is discussed in this Wall Street Journal article - “Facebook Isn’t So Good At Judging Your Credit After All.”

 One argument for delving into nontraditional data is that 40 million Americans don’t have a FICO score due to no or “thin” credit history.  And perhaps not surprisingly Millennials represent a disproportionate percentage of this group. For this population new and promising data sources include positive rental payments and positive payment of utility bills. These variables are generally regarded as compliant from a regulatory point of view. I would not be surprised if future versions of FICO incorporate this data.

 

Why Some Marketplace Lenders Believe They Don’t Need FICO to Underwrite

 Depending on the product, marketplace lenders may target segments that are known to have high FICO scores. For example, some marketplace lenders loan only to people who have graduated from elite educational institutions with degrees that afford high earning potential. In a sense, these lenders rely on the admissions departments at the nation’s best colleges and universities to do their underwriting for them. An underwriter would simply need to check a few additional data points related to derogatory payments and DTI ratios to finalize an underwriting decision. In statistical parlance, this population is so super-prime and homogenous that FICO doesn’t provide any “orthogonal” or incremental predictive value.

 It should be noted that companies that no longer use FICO for underwriting typically do still rely on the score for loan sales and securitization, so FICO is in fact still an important variable for these lenders.

 

Conclusion

FICO is not perfect but it is has demonstrated consistent performance across a variety of products and economic cycles i.e., it is statistically robust – a very desirable attribute. Used in conjunction with business judgment about current and future economic conditions, the score is very effective.  In addition, FICO is constantly being improved. For example, a next version of the score could include positive remittances for rental payments and utility payments. While marketplace lenders – and traditional lenders too – will continue to identify and leverage new, compliant data sources to enhance their underwriting models, I expect most lenders will continue to rely on FICO as a significant component of their underwriting policy.