Here’s a direct link to the article. It doesn’t give a very clear explanation of what the credit card companies are actually doing, but I think that there are a couple reasonable applications of technology like this:
[ol]
[li]If you’re a new customer at a given card company, they don’t have any historical data to use in rating your risk. But if they know what sorts of purchases indicate risky and/or safe customers, they can compute an accurate risk assessment more quickly by comparing your behaviors to those better known customers than they can by waiting for you to make or miss a few payments. This sort of technique is the industry standard in insurance.[/li][li]They can keep an eye out for sudden changes in a given customer’s behavior. If MeanOldLady has a good history of making payments, it doesn’t really matter if she buys a lot of liquor. But if she’s very recently started buying a lot more liquor than she used to, that’s interesting and might indicate that she’s fallen on hard times and is at risk for missing payments.[/li][/ol]
Maeglin, how well do these notions match up with what you know that the companies are actually doing?
Let’s read into it a little more carefully.
I don’t know who these “lenders” are, but I do know who they are not. They are not the vast majority of global credit card issuers who have no insight whatsoever into what consumers buy. This is why I found this article so disingenuous.
[quote]
[li]If you’re a new customer at a given card company, they don’t have any historical data to use in rating your risk. But if they know what sorts of purchases indicate risky and/or safe customers, they can compute an accurate risk assessment more quickly by comparing your behaviors to those better known customers than they can by waiting for you to make or miss a few payments. This sort of technique is the industry standard in insurance.[/li][/quote]
This is only somewhat true. When you are a new customer, the credit card issuer will access your credit bureau data to evaluate your risk. Issuers are potentially quite good at making credit decisions based on bureau data alone. The risk rules can be relaxed when future expectations about the performance of the portfolio are very positive and when issuers have ready access to cheap cash. When that dries up, obviously there is a bit of a problem.
Keep in mind that credit card issuers can only update your bureau information when you apply for a new card or for some other credit-related product. If that does not occur, risk can only be assessed based on transaction patterns. Keep in mind that no issuer has any more information about the transaction than is printed on your bill, and in many cases, the issuer has even less.
[quote]
[li]They can keep an eye out for sudden changes in a given customer’s behavior. If MeanOldLady has a good history of making payments, it doesn’t really matter if she buys a lot of liquor. But if she’s very recently started buying a lot more liquor than she used to, that’s interesting and might indicate that she’s fallen on hard times and is at risk for missing payments.[/li][/quote]
We are getting more astray here. First, transaction behavior is obviously very noisy. I spend all day trying to model it to make tiny refinements in my business, and I am very frequently just wrong.
Second, we probably have no idea whether or not MeanOldLady is buying liquor. In some states IIRC, liquor can only be purchased at state-licensed grocery stores. So we know that she transacts in a grocery store, but we don’t know whether she is buying butter or beer.
Third, there are very few individual behaviors that are truly indicative of upcoming likely default. Buying liquor is not one of them. For example, suppose there were a program that let superprime cardholders pay their mortgages via credit card. Usually they would do this to collect buckets of card rewards and they pay their bill in full at the end of every month. A big warning would be if that cardholder suddenly started revolving his mortgage payments.
In reality, ongoing credit risk is assessed against hundreds of portfolio metrics. An uptick in a consumer’s spending at Walmart is not itself enough to justify an intervention. But if the consumer exhibits many behaviors consistent not with post hoc reasoning or backfilled narratives but with the actual behavior of defaulters, a credit decision might occur.
Many lenders use “behavioral scorecards” for this. These are essentially statistical models packaged for business decision-makers to understand. Ongoing consumer behavior is evaluated statistically and credit “scores” are produced. In general, the more a consumer’s behavior conforms to the behaviors of cardholders who have defaulted, the higher the score will be. The business sets the intervention thresholds depending on several factors, including principally the size of and the spread on the receivable and the performance of the rest of the portfolio. All of this works very nicely until it doesn’t.
Let’s say that shopping at Huddy’s correlates strongly with defaulting on payments and debt restructuring and other indications of being financially out to sea. Would a lender be justified in taking special measures regarding someone who gets a new card and makes a number of purchases at Huddy’s? That seems to be what the article is saying, anyway, but I want to be sure I understand it correctly.
Good for you that your employer doesn’t do it, but other lenders do.
American Express, for example used behavior analysis to lower customers’ credit lines until just a few months ago.
And a subprime lender called CompuCredit (marketer of cards with names like Freedom Card, Aspire, Imagine and Fingerhut) got sued by the Federal Trade Commission (PDF) for using a behavioral scoring model.
[
(From [URL=“http://www.creditcards.com/credit-card-news/how-shopping-can-affect-credit-1282.php”]here](http://www.creditcards.com/credit-card-news/how-shopping-can-affect-credit-1282.php))
So much for “patently false.”
Wouldn’t it depend on the type of POS system used by the merchant where you shop?
Why yes, it would. So I’m calling bullhonkey on this one.
When I go into a big, major grocery chain, they scan each item by code and those items itemize on the screen as the checker goes through your stuff. So in that case, the Big Brother-esque scenarios are, on the surface, plausible, because some electronic record is made, however momentarily, of the individual items being purchased.
But I work part-time at a store… a liquor store, actually, though a very nice one, I swear.
It may be relevant only b/c some people are saying booze is a risk factor, otherwise I’d leave that out.
It’s a small, independently-owned little shop, run by a very nice lady with zero experience owning a retail outlet. So it’s a little ad hoc. There is no master inventory, for example, just a paper booklet with items and the prices, but it doesn’t have everything we sell and half the prices have been scratched out and changed. (Okay, so more than a little ad hoc.) There is also no scanning machine - we add up the prices on an honest-to-God old-fashioned calculator. If they pay cash, we take the money, hand back change - no receipt. If they pay with a card, we punch the total into a little swiper and they swipe and then sign.
I give those details only to make this point clear: At NO point is there any record of the exact item bought. Your $25.99 could be going for one nice bottle of wine or as much of the cheapest rotgut you can carry out the door … but the government, insurance industry, or credit card companies can’t know the difference.
This is not the only retailer like this in the area. So how could this psychological profile-building possibly be a pervasive part of the credit card industrial complex?
The simple fact that a custom bought something at a liquor store is also NOT proof of “risky behavior” since the store sells healthy, innocuous things like juice and bottled water.
So, again … the Big-Brother type stuff? Bull-honkey.
No, it has nothing to do with the POS system. The credit card companies are using what’s called a merchant category code, or MCC. These codes are set up by the IRS and identify the kind of business. This is what the card companies are using in their modeling and tracking.
There have been a series of articles about this phenomenon on Consumerist lately. Remember, Consumerist is owned by Consumer’s Digest, so I think it has some weight of authority.
http://consumerist.com/5310575/dont-put-these-items-on-your-credit-card-if-you-can-help-it
http://consumerist.com/5256326/your-credit-card-company-is-building-a-psychological-profile-of-you (which links to the NYTimes article someone linked to earlier)
http://consumerist.com/5115522/amex-lowers-your-credit-limit-if-you-shop-where-deadbeats-shop (includes link to AJC article: http://www.ajc.com/news/content/business/stories//2008/12/21/creditcards_1221.html?cxntlid=homepage_tab_newstab&imw=Y)
American Express is a great example. I know a great deal about the Amex business for professional reasons. One point in this article is worth mentioning:
This is exactly what I said in my reply to ultrafilter. There is nothing proprietary about this. I would also like to reiterate that I do not work in this part of the business and I do not know anything about our specific practices aside from what has been disclosed. I do know about the standard behavioral statistical techniques used across the industry and are often peddled by software and analytics firms. Google it yourself.
What is patently false is that anyone decides arbitrarily what sorts of behaviors might lead one into distress. These are derived entirely from data. In a standard scorecard model with a reasonable cut-off, it would typically require many behaviors to accumulate before any individual is flagged for credit intervention. The problem is, the statistics have been very poorly articulated by the press, by Congress, and by people at large. It is understandable since they are somewhat complex, but it is absolutely maddening.
Here’s the article on the same topic from Marketplace:
http://marketplace.publicradio.org/display/web/2009/07/08/pm_new_redlining/
You are half right. Many POS and property management systems used by large merchants do in fact transmit additional data via their POS. Look at your cc bill sometime from a large merchant, like a department store. Itemized on your bill might be the department or even the item you bought. There are a lot of technological dependencies, but sometimes we do capture this stuff. Most of the time, though, it is barely intelligible and the information from no two merchants is really comparable.
Another massive pet peeve of mine is below. This is more misleading reportage in action to generate RO.
Thanks to robust regulation and consumer protection, credit card companies cannot access bureau data unless consumers apply for more credit or new products. The fact that Kevin Whoever has a a high credit score is invisible to the credit card company. If we did not have those consumer protections, well, credit card companies would not have to spend so much time trawling through transaction data. Understandably, most people would not like the alternative very much.
And defaulters very often have a perfect payment history. Until they don’t.
To be fair, credit card companies themselves have been in a real bind trying to articulate these decisions to their own customers. It is very hard to tell them they their credit was actioned because their risk score reached a critical threshold or that their patterns were highly correlated with existing defaulters. Unfortunately the people who wrote the letters did not do justice to the statistics and left a lot of people quite upset.
But the fact that your store is a liquor store is the important factor. If a person has made a purchase a week there forever, and has a good credit history, there is unlikely to be any impact. However if a person has just started making purchases there, and switched from shopping at Nordstroms to shopping at WalMart, there might be something going on.
Perhaps some of the confusion is mixing up general cards with store cards. My Macys bill has the type of thing I bought there on it, and I’d certainly expect them to have full details of my purchases in my profile. If I used a general card there, the data wouldn’t go out.
The last time I looked at the fine print on a CC application, it said they reserved the right to periodically run a general inquiry (not a full report) on my credit worthiness in order to see if my credit limit ought to be raised or lowered. That was a few years ago - are you saying that recent changes in regulations has removed this language from CC company terms of service?
Good thing all those porn shop purchases show up as convenience store charges. whew
Those are “soft” inquiries. They are not actually all that informative. When you get some crap in the mail saying that you are"pre-approved", it is always contingent on a successful “hard” inquiry which yields much more information.
I think you’re picking nits, parsing the articles to an unreasonable degree. The point is that someone with a high score should not have anything in his record, including those “soft enquiries” that you mentioned, to merit having his credit revoked or reduced. The very existence of “credit scores” was something most consumers didn’t know about not all that long ago. The fact that you could even FIND OUT what your credit score is is a recent phenomenon.
What this points to is a decades-long trend in the cc industry to make up rules about who they will and will not give credit to, how much credit they can have, and what rate they will be charged, and those rules are not disclosed to the customer. In addition, the cc company can change the rules at the drop of a hat, based on an “indicator” that is meaningless in many cases. It’s basically stereotyping, and stereotyping which can effectively change someone’s life through no true negative actions of their own.
No one is saying that a cc company should not make a profit or should not manage risk. It is the WAY in which this is done that drives decent consumers up a wall.
Harvard Law Professor Elizabeth Warren was on Fresh Air a couple of years ago to talk about all the billion ways that cc companies figure out how to get money out of you. IIRC she doesn’t discuss the type of profiling we’re talking about here, but it’s still a VERY eye-opening interview – a lot of the issues she discusses are the type that lead up to the recent legislation. Warren has been on the program since then, and I haven’t heard the more recent interviews, but plan to listen and catch up.
Maeglin, I understand that from a professional’s point of view it makes a huge difference whether a decision is based on
A) A lot of empirical evidence showing that people who carry balances solely from liquor and gambling (or any other pattern of fact) tend to default.
or
B) Someone at the company reasoned out that people who had balances from liquor and gambling don’t seem like good risks.
But the thing is, to the consumer, it’s not that big of a difference. There is still that one person out there who is a flawless user of credit who gets his limit lowered because he matches the fact pattern. The empirical support may make a huge difference in whether or not a practice is legal. It doesn’t necessarily make any difference in terms of whether consumers like it. Consumers will never see the empirical data on other consumers and, even if they could, would not have the statistical background to evaluate it.
Combining the level of use of my credit cards for travel (air, hotel, car rental) and internet pr0n, were it not that most of that travel is to the Northern USA they’d conclude I’m into sex tourism 
I absolutely disagree. In this respect, details matter. Policy is being shaped that will have a resounding, multi-billion dollar effect on consumers. Exactly what the practices are and what they mean is extremely important.
I am struggling and failing to see why this is a problem. Lending money has to make money, else there will be nothing to lend.
There are big problems in the industry with disclosure around rate. I agree that this is something that badly needs to be cleaned up.
You are completely incorrect about the idea of stereotyping. We have no a priori beliefs here. And our indicators are far from meaningless. For every one person whose complains to the newspaper about a reduction in his credit line, a credit card issuer will successfully avoid losses on a thousand people headed for default. If our information gathering on our customers were any more invasive and tailored to the individual, people would hate it. So issuers take a data-driven approach to manage portfolios of customers.
But that is kind of what you are saying. To put it another way, “no one is saying you can’t drive the nail into the piece of wood, but using the hammer drives decent consumers up a wall.”