How many times have you heard a politician or regulator say they use data to make decisions? That sounds great until you look a little deeper. Three trendy approaches to data analysis — disparate impact theory, proxy data and behavioral economics — completely destroy the credibility of any government official who claims to use data to drive objective decision making.
Disparate impact, a theory affirmed by the Supreme Court in June, says judgments can be made according to the results of particular practices, regardless of the intent of those instituting those practices. So if lenders turn down a disproportionate number of mortgage applications from potential borrowers of a particular ethnic background, the lenders are guilty of discrimination even if they treat all applications in the same color blind manner.
Proxy data is used by some analysts to make educated guesses about the ethnic backgrounds of customers based on their addresses and surnames. This approach is used when no formal data is collected about the ethnicity of particular loan applicants. The Consumer Financial Protection Bureau has used proxy data to target auto lenders. Proxy data is an extremely weak analytical approach; I think of my own home where our adopted Hispanic children would surely be miss-categorized based on their Swedish last name.
While traditional economics are based on rational decisions leading to predictable outcomes, behavioral economics dispenses with the traditional connections between concepts such as pricing and demand. Behavioral economists will argue that people often act irrationally, giving economists license to predict economic outcomes on a random basis. The result is usually a syllogism supporting a pre-conceived theory advocated by the economist long before beginning the research. President Obama signed an executive order in September directing federal agencies to use insights from behavioral science.
Daniel Moynihan said we are entitled to our own opinions but not our own facts. Since his Senate days in the 1970s and 1980s, it seems we have embraced methodologies that blur the line between facts and opinions.
It is hard enough to do anything when we all agree on the facts but when those facts can be manipulated to mean whatever the most powerful player in the group wants, we have a problem. This is manifesting itself in banking in a number of ways, particularly with respect to rules around loan loss reserve provisioning and UDAAP compliance, not to mention the housing and auto lending examples cited above.