Sunday, November 23, 2008


Katy tagged me. I am supposed to write six things about me. I am slow in responding. I already revealed 5 things. By my arithmetic, I owe one more.

When I was 16, I worked as a credit analyst for the consumer credit division of Citicorp, now Citigroup. Yes, they let high school students do that. No, I don't see any connection between their imminent doom and my past employment or their reliance on high school and college students who can pass a simple timed math test.

I was mad at several things that happened while I worked for Citicorp, but in hindsight, I can see that they treated us decently overall and I learned many, many valuable lessons there.

I was originally sent to Citicorp by a temp agency. I didn't even know what I would be doing until after I showed up at their office. The job paid above minimum wage and needed workers during evening and weekend hours. That made it an unusual temporary office position and ideal for students. They sent many of the temp agency workers away after one week trial. After my trial period 0f XXX hours, Citicorp hired me directly.

The job involved taking "instant credit" applications from stores off a fax machine, running a credit check through one of the 4-5 major credit information services (via telex), and then scoring the applicants with a score sheet which I recognized instantly as an algorithm. We were supposed to do all this within 60 minutes of receipt of the fax, except under exceptionally busy shopping times when a manager made the call to suspend the 60 minute turnaround guarantee.

People shop after work or on weekends. Citicorp exploited time differences by hiring workers in the Pacific time zone. It was a perfect fit for students who were content to work those hours.

After I learned a few arcane things*, it was an easy job. I was surprised by how many people were let go because they couldn't score the credit information and add up the points accurately, even with the aid of an adding machine and a scoring sheet. I was also surprised by the number of others fired for snooping--running credit checks on people who didn't apply for credit. That was not spelled out in my initial training, and I was also curious, but I just assumed that would be verboten. The managers probably also assumed that was a given, but we were mistaken in our assumption. Privacy and ethics were then added to the initial training. ;-)

That they had so much difficulty finding enough people who could score credit, following an algorithm that was pretty easy to interpret, encouraged management to change the process. They sent some people to run the telexes only. If the faxes were illegible or were missing information, others called the stores for the required data.

I did that for a few weeks, but I kept losing my voice after we moved from an old office building with operable windows to a newly built office building with inoperable windows. They allowed smoking in the old building, but the windows were kept open. In the hermetically sealed new building, which had plenty of off-gassing from the paint, building materials, new carpet and cubicles, the air quality was awful. I called in sick repeatedly. The office-wide smoking ban went into force after the managers got sick.

They moved me to another section that only scored credit. There were no smokers in that section.

My early exposure to credit scoring algorithms taught me the things that one can do to boost and mess up credit ratings. I must have internalized that because, when we applied for our home mortgage, my FICO score was higher than Mark's, even though he had higher earnings and a longer earning history.

We also ignored negative information from medical services for employed people. Why? My manager told me that people shouldn't suffer because their health insurance didn't pay in a timely manner. He said that health insurance companies were often slow to pay so they could earn the extra float. Or that insurers and hospitals disputed balances. Anyway, it was a frequent enough occurrence that a policy was formulated and followed to cover it. Notice the assumption 25 years ago that every employed person would be covered by health insurance.

During the summer, when I was there during normal business hours, I also spent time with people who fell in the gray areas. That is, they had good cash flow, but moved or changed jobs frequently. Would they be called sub-prime now? I don't know. Credit scoring algorithms are proprietary and I only have experience with an early version of them. Even then, I saw that there were problems. Some people that appeared to be good bets were not**. Others could be good credit risks, but did not score highly enough.

There are some highly skilled and highly paid contingent workers that move frequently with their contracts. They used to have a difficult time establishing credit, but it can be done with extra legwork on the part of the lender. I called banks to confirm bank balances; contingent workers often hold large bank balances to tide them over in between contracts.

I also called former landlords to confirm that they paid their rent on time and moved for reasons other than financial duress. I called former employers to confirm that credit applicants left because they had highly specialized skills that were no longer needed once everything was up and running.

None of these people had any idea they were talking to a 16 year old. It didn't really matter because I could do things that others couldn't, or wouldn't think to do.

I also saw that the bulk of the people there could be replaced by software, which they eventually were. They had to lay people off, and I went in an early round.

I was mad at the time, because my manager told me, right at the end of my shift that it was my last day. I had no warning. But, it really would have been dangerous to let someone who knew they were leaving to remain there. There was so much potential for mischief-making in a place like that; it was prudent to let people go with no warning.

What wasn't prudent was the order they laid people off. The manager said that corporate told him to lay off the people who worked less than 20 hours per week first. Those were the students. Assuming that the manager told the truth, that was a lame decision by corporate. The students were quicker with the math than the "lifers" and we also worked only the busiest hours.

Had I been more savvy, I would have asked for severance pay. I just packed up a box of my stuff and left.

Today, I work in a computer science research group as one of several "domain experts". Sometimes, we are asked to evaluate surprising computer output. Is the problem with the way the software algorithm is implemented? Or perhaps the algorithm does not accurately reflect what is actually happening? Again, I gravitate toward the gray areas. Why work on a problem that has already been solved?

* Back then, credit reporting bureaus were more regional than national. Each region had one or two bureaus that had more information than the rest. Citicorp was paid a flat fee for evaluating each application, but paid a fee to each bureau for each report we pulled. Thus, profitability depended upon us accurately guessing which credit bureau would most likely contain useful information about an applicant. This is especially important for people who have moved. Moreover, young adults may be "missing" data in one database, but not the others, or they genuinely may not have any credit history.

** The algorithm gave married people more points. Yet, divorce is a common reason for unpaid bills. Better credit rating algorithms should take into account the stability of a marriage. I learned early that young marriages had an exceptionally high failure rate and were disastrous to credit.



  1. That was really interesting. Thanks!

  2. This was very interesting! I work a little bit in credit reporting law and the history of the industry is a good background for my work.

  3. I love that your manager was able to decide to ignore bad information from medical services. No such flexibility once the entire process was replaced with a computer...