Data Collection: It's an Idea Past Its Prime
The Federal Reserve has spoken. Monitoring data may only be collected under controlled circumstances and for purposes of self-analysis. This position is responsive to concerns about compliance burden, additional cost, imposition on many consumers, and concerns about the actual value of monitoring information collected on a voluntary basis.
Don't expect the consumer advocates to take this loss quietly. Consumer groups want data - which means they also want more data collection. In fact, some groups asked for mandatory collection on all loans. Data is their most promising and profitable tool for making discrimination cases.
Collection of monitoring data - information about the race, ethnicity, and gender of credit applicants - is an old idea. Collection of monitoring data for lending is more than 30 years old. Data collection was a remedy demanded by the coalition of civil rights and consumer advocacy groups that sued the bank regulatory agencies (back in the 1970s) for alleged failure to enforce the Fair Housing Act.The same groups asked for data collection when Regulation B was being rewritten in 1976-77. They got the data for housing loans. They have been trying ever since to get information on other types of loans. Now they want data collection on loan applications including those taken by telephone, through mail, and over the Internet.
Never mind that most of the horses are already out of the barn. When a lender doesn't see or talk to a customer, the lender has limited opportunity to find out or deduce who or what the applicant is. And if the applicant cannot be forced to answer the questions - as in a policy of no monitoring data means no loan - then we have an increasing number of applications for which there can be no monitoring information.
It is all very well and good to talk about the interesting analyses that could be performed with data. It is quite another thing to actually have data that is of sufficient quality to be worth analyzing. Lenders have to deal with reality. And the reality is that most consumers don't like being asked to provide their race and ethnicity. Some even object to being asked what their gender is.Given the opportunity, somewhere between 20% and 30% of consumers will choose not to answer these questions. This is based on the response rate for a trial data collection program asking for voluntary monitoring information in face to face interviews. This shortfall was "remedied" by requiring lenders to complete the monitoring information in a face-to-face interview. If the customer is not in a face-to-face interview, the lender cannot make the response for them.
However the information does or does not get filled out, we are left with a seriously deficient database. We need to figure that a quarter of the data we actually have is the loan officer's guess. The information here is at least reasonably related to what the loan officer perceives.
For the non-face-to-face applications, the information will be missing. We also need to take into account that we don't know why the applicants refused to provide the information. They could be minorities who fear that their application will get second-class treatment if the lender knows. Or they could be non-minorities who are offended by the question.
When we don't have information about this proportion of applications, there are two problems for analysis. First, we don't know whether to make assumptions about those who didn't respond in order to fill in the gaps. Second - and this is more difficult - we don't know whether those who did respond are an accurate representation of the applicant population. In short, we have no way of being sure of what we actually do have.
Data collection is an idea past its prime. It is time to look for other solutions to the discrimination problem. Training, procedures, and file analysis remain our best tools.
Copyright © 2003 Compliance Action. Originally appeared in Compliance Action, Vol. 8, No. 2, 3/03
First published on 03/01/2003