We all know FICO scores are something of a black box. But, there's one problem with them that has bothered me for some time and seems like a fatal flaw. Namely, scores are highly dependent on credit ratios / balances / limits. How can they accurately assess risk on these factors when CRA profiles do not collect information on income or assets. Obviously, the default "risk" associated with the debt of a person earning a high income with significant assets is much lower than the same debt owed by someone without these resources. Don't get me wrong. I don't want CRAs collecting any more information on us. I just don't see how they can create a valid model that neglects such an obviously important part of the equation. Any thoughts? Greg
They claim that income isn't really a good indicator in whether or not a person will pay. Generally speaking as a person earns more, they spend more. They also had the problem that when they would ask for say gross income people would report net, so it ended up being a very flexible number anyways. The thing that gets me though is that two of the major factor for default are divorce and medical problems--I'd like to see Fair Isaac try to predict those. Some of the new questions you would be asked on credit card apps would be "How do you feel, do you smoke?" and "How are things with your wife, and do you sleep in the same bed?" I shudder at the thought!
I agree with IndyGreg it bothers me that I have two credit cards with a limit of 1500.00 each and about a 1000 on each, and I get that the" limit is too high for my group" on my FICO scores. However, how would they know what group I am in and compare me to other people by age??? So, I owe 2000 on cards but I make 90K a year. The funny thing is I can pay off one card in one week. So why should MY scores be lowered because of this? Typically my 90K group should have larger CC limits since I have elected to keep my limits at a conservative amount why should I be penalized?
Real quick...my 2 cents. Word on the street is that doctors and lawyers have real bad credit because they are slow payers often. The may pay bills every 6 months.
It seems like a fatal flaw because you're assuming that the FICO score is a measure of financial health. FICO isn't really measuring anything about money; FICO measures *behavior* - more specifically, conformity of behavior. They go through their credit files and, using statistical analysis, decide which behavior patterns have historically caused the fewest losses. Although this behavior pattern is assesed by looking at credit mix, balance ratios, etc., they don't care about this data from a finiancial standpoint. The more your behavior as expressed in your credit file matches those patterns selected as causing the fewest losses, the higher your score. The more your behavior deviates -- even for sound financial reasons that would be applauded by any financial analyst on the planet -- the lower your score. The way it plays out in real life can be expressed in a bell curve: (this graph is supposed to look like a bell [[Bell curve doesn't work, imagine the stars are spread out to outline a bell shape]] | * | * * | * * | * * | * * | * * | * * | * * |* * ------------------------------------------- On the left hand side of the curve you have people with a low score because of derogetory information on their credit report. In the middle are the vast majority of creidit users who have a good score. For these people, FICO works: the majority match the pattern, thus have a good score, those who deviate because of negative credit items get a lower score. But what about the right hand of the graph: these are people who deviate from the majority *with no traditionally negative information on their credit reports*. This is where balance ratios, number of inquiries and other items that, depending on your personal balance sheet and income statement, may have no direct bearing on your financial health. Your choice is to conform your behavior (don't carrry balances, don't generate inquiries, etc.) or accept a lower score. The data forms a curve like this becuase of people who file surprise bankruptcies (never late, lots of credit, then bingo, BK with "no warning"). Credit issuers and CRA's have no concrete data with which to measure the likelyhood of surprise BK, so they guess based on behavior related to the very thinnest data set imagninable (like number of inquiries). If your credit report matches the surprise BK profile, you're out of luck scorewise, even with a million dollars in the bank. But you can conform your behavior so that it appears that you match the best profiles, thus improving your score. Of course, the more that people conform their behavior to meet the score the less validity the scoring model will have. It gets creepier and creepier, hmmm? ;-)
Racer, I think that's an excellent analysis especially about the "positive credit" people who deviate from the norm. You're so right. Nonsense like excessive inqs. have to then come up. Also, I think when excessive credit lines are mentioned but the person have a good report this is a case of "looking" for deviant factors. Stuff like too many bank cards, etc. seem to fall into this category too.
I think that FICO's secret problem is that they have a hard time keeping their models current with the financial industry. Imagine a few years ago when consumers became smarter and with the help of the internet, they began shopping for rates instead of just walking into their local bank. Instead of one inquiry for a car loan, suddenly a consumer might have 4 or 5 which could seriously impact their score. 10 years ago, opening a store charge card and getting 10% off or getting 0% financing for a year was rare. Today, that's quite common so that the wise consumer would take advantage of these things. Doing this, though, may cause the consumer to have too many open accounts or inquiries although they are making decisions that are in their best interest. As financing becomes ever more complex, I'm sure that FICO has a difficult time separating "good" behavior from "bad" behavior. The first problem has been recognized and largely corrected although I'm sure it took them a good year to come up with a solution that was effective. As to the second problem, I'm not sure that there is an easy fix for it although I'm sure there are certain behaviors that are common amongst people who do this.
I think one of Fair Isaacs biggest problems is a basic statement that they always teach you in statistics-correlation is not causation. I'm sure their models correlate very well, but their ability to predict.. thats another story...
Your choice is to conform your behavior (don't carry balances, don't generate inquiries, etc.) or accept a lower score. Doing this, though, may cause the consumer to have too many open accounts or inquiries although they are making decisions that are in their best interest. As financing becomes ever more complex, I'm sure that FICO has a difficult time separating "good" behavior from "bad" behavior. I think one of Fair Isaacs biggest problems is a basic statement that they always teach you in statistics-correlation is not causation I'm sure their models correlate very well, but their ability to predict. That's another story... ================== In other words scoring just won't work ! ! ! !