The Tyranny of the Credit Score
nytimes.com
nytimes.com
There was a time, before credit scoring became ubiquitous, when you would walk into a bank and get a loan from a loan officer directly, who would have very little to go on besides what you immediately present. That meant that women and minorities, even ones who would be excellent credit risks, were less likely to be able to get credit.
Sure, the current system has lots and lots of flaws (mostly from reducing everything to a single immutable number), but there's something fundamentally good behind the intention of judging people based on their concrete financial activity instead of just on the color of their skin and/or the cut of their suit.
The relationship, as far as I can see, is the same as that of credit: they bill me, and I send them electronic funds. What am I missing?
Each business is different, and a competent risk analyst will build an algorithm based off the portfolio's actual history.
The problem is the credit score might have meaning in some cases, like unsecured debt, to show how likely you are to payoff what you have little incentive to return. It's a measure of all financial behavior, but all loans are not the same.
A credit score isn't going to tell you how likely somebody is to make their mortgage. Unless there are really desperate circumstances, most people will pay their mortgage. A middle-of-the-road credit score really doesn't tell you much about how likely somebody is to have desperate circumstances. That makes it a poor predictor.
The above implies that a company that does not use credit scores when making mortgage loans will outperform one that does.
What's the the name of your company that makes mortgage loans and does not use credit scores? More to the point, are you taking investment or borrowing money?
>The above implies that a company that does not use credit scores when making mortgage loans will outperform one that does.
The point is, while we don't really know how these algorithms work, we know they don't control for the type of loan. You just get one number. That one number is not always meaningful.
I paid for 30 days access to check my credit report a few months ago and I had to fix a whole slew of errors dating back almost 10 years. When I followed up to see if they had fixed the errors, I was told that since 30 days had passed, I had to pay again to get an updated look at my credit report. Talk about messed up incentives.
However, bear in mind that covers the report of the raw data. The credit score itself is not covered by that report, and they can make you pay again for that. I don't think knowing their exact score is all that useful, compared to making sure their data is accurate.
Imagine if we tried to apply this approach to other areas. For corporations, instead of reporting total assets, total liabilities, earnings, payments, and such, they instead give all this information to an "Accounting Disclosure Bureau", who produces an "Expected Profitability Index" based on those numbers. While the report contains other information as well, people usually just look at the EPI. Would that be a functional investment climate?
In the recent past, many people's credit scores were artificially inflated due to the low perception of systemic risk in the economy.
Now that that risk is better understood, the same person who had a 750 score might have a 650 score, etc., even if all payments have been made on time.
I'm curious about the incentive this creates. Suppose someone had a score of 680 and was paying everything on time and had $5000 of available credit. Now, he/she has a score of maybe 640 and no available credit.
What is the benefit to this person of continuing to care about any no-recourse debts in this case? He/she might imagine paying the debt off over 5 years as originally planned, or may consider abandoning/defaulting on the debt and saving the money instead, which would improve the person's financial picture over the next 5 years, and by 7 years out the default would be expunged anyway.
Were you not able to use this or was this in addition to that?
(Yeah it's messed up).
You don't get the score for free, but you do get a copy of your report (all the things listed on there that go into the calculation of the score).
It's one thing to say somebody is a smelly, poop-eating mongrel vs child rapist. One is an opinion and the other is a provable statement of fact. If, in fact, the child rapist epithet is false, it's libel.
Now, these companies are selling rights to loans, along with your information and a formula they use to equate it all. If the information is just not true and affects you negatively, "sue the bastards".
It's not published in a public place so libel wouldn't apply (my guess anyway). Now if they put it up in the local bar (like bad checks used to be before they cracked down) then you might have a case.
If the premise of the article is true, that people who should be getting credit are not, why can't we start a credit-extending company that does a better job of assessing risk?
One or more of three things must be true here. Either:
1) there is an untapped business opportunity here,
2) the premise of this article is false and the machine is working fine, or
3) there are regulatory barriers getting in the way of making the machine work well.
In other words, sure, if you've got $billions lying around and are willing to hold 30 year loans for their full term, feel free to start a business doing this. If you're like any other bank and you want to off-load it to Fannie/Freddy immediately or securitize them and sell them to hedge funds, you need to put a FICO score on every loan and that is the only thing anyone cares about.
It's obvious this is why our financial system is so screwed up. There are people with 800 credit scores that are upside down by hundreds of thousands of dollars, yet they can still get as much revolving credit as they want because of their credit scores. Then, there are people like me who have actual cash reserves in the tens of thousands and 0 debt, and refuse to hold revolving credit, who have sub-700 because we don't buy into the whole debt ponzi scheme. The system is broken and it will take a lot more than a new business model to fix it. It will literally take several acts of congress.
The issues are pretty much the same as mortgage and there is alot of room for innovation.
While we're talking about holding the debt, you could drop into microloans in the kiva style, and start with only four figures.
Maybe hedge funds need to get into this business. Or maybe I should find a way to show hedge funds that my new risk assessment is better.
Though one of the challenges is getting new types of hard-to-fake data. If my risk assessment looks at a person's email account or does machine vision on their facebook photos, those things are easily forgable.
Profits are too low. What does a US mortgage do nowadays, 6-10%? That's before costs and risk premium. With money tied up for 20 to 30 years.
Didn't the hedge funds just game Morningstar against AIG against FICO scores?
Goldman wasn't sued because of what they offered as investments, they were sued because they misrepresented the net positions of at least one major player, allegedly deliberately.
What would the other factors be that you'd measure them on to do better? Race? Religion? Education? I bet there are some algorithms that work better but they might be more messy. I bet if you looked at the credit history of a client's circle of 6 closest friends, the credit history of their parents, whether or not they were married, whether or not their circle of friends and families thought they were happily married and their attendance records at work and school and the last education level they had you could make some good predictions. I wouldn't be surprised if there were some religions that had better loan payers than others and some that had worse. I also wouldn't be completely surprised if there were races that tended to have better loan repayment habits and some that had worse, I'm fairly certain that it's true if you limit your samples to certain regions. I wouldn't be surprised if you could correlate a borrower's child being a good or poor student with the borrowers credit risk. I woulnd't be surprised if you could correlate the number of children had out of wedlock with credit risk. I don't know that'd we'd want large institutions using these kinds of factors to make lending decisions, we want it to be clinical, sterile, and somewhat blind which leads us back to something like the FICO score.
Now if they showed mortgage brokers gaming it, that might be a different matter, they're simply complaining about missing the cut-off. Trying buying groceries with just a couple dollars less than what the bill comes to..
If another company can just verify that "Yes, this person has paid all of their promised debts on these specific accounts over the last X years and currently has Y amount of debt that we know of" I think that would solve a lot of the confusion with these scores.
Companies use it because it works pretty well or well enough in most cases. Yeah there is some really bad data. However, half of that blame probably goes to the institutions reporting the data. (That's just a hunch)
However, because I've worked in the finance industry for many years, I've made it a point to check my credit and try to keep it clean, because I know the credit bureaus lack the incentive to do it.
I'm cursed because I have a common name and someone with the same name is very bad about paying his bills. I usually end up disputing something every other year or so.
If these credit companies suck at their algorithm, which they apparently do, I'm screwed. It doesn't help that they're most likely totally captured by the financial companies.
Yes, it's called an algorithm. I'm guessing she doesn't understand that.
"Their bottom line number is 620 — the company will buy mortgages only if the borrower has a credit score of 620 or above. Which means, given the current state of the mortgage market, that anyone with a score below 620 can’t get a mortgage. Even if that score is 619."
The thing is regulators are looking at that number. There is a cut off there. Especially now that the regulators are going through the bank books with a fine toothed comb.
Guess what, the worse you portfolio, the higher your capital requirements become. (Your margin of safety)
A company I worked for bought a huge portfolio of debt. Turns out the average FICO was way off or out of date. Very bad thing. The original company had booked some really bad stuff and was one of many that went down in the waves of financial institution failures. Bad mortgages finally got them.
He always used to recite the Five C's, Collateral (what percentage of the value of the property where you borrowing against), Capital (How much money do you have in the bank), Capacity (what is your income) Character (what type of person are you) and Credit (The infamous credit score).
When you are lending with your own money, only of those Five C's is worth a squat - and that's collateral. What percentage of the asset where you lending against. Anything north of 50% was considered risky.
They did stagger that law in by region, but it's completely phased in and has been for a while.
Bear in mind you are entitled to a dump of the raw data they have, but you are not entitled to the final number they call your score. They make you pay for that. In general, it isn't all that useful to know that score IMHO. (At specific times, perhaps, but not in general.)
I now seriously question anything further in that article. The idea that credit scores can predict car accidents is ludicrous. The author probably meant that credit scores can be used to predict claims on automobile insurance, but that, too, is just as incorrect. Consider that many people with low credit scores are going to buy the least expensive vehicles and insurance they can find. This means carrying liability-only coverage, which does not pay to the insured, just to anyone the insured might hit. That means that the insurer is now using the credit score of the insured to "predict" the decision of a completely unknown third party who may or may not even exist.
Credit scoring, when used outside the context of granting credit (and, often, even in that context), is simply a way to continue to gouge people who have, for whatever reason, a damaged set of credit reports long after that damage has occurred.
Do you find the idea that credit scores are correlated with car accident rates to be ludicrous?
I find it very plausible. Some people are careful, others are careless.
Of course, one should be very careful in using vague correlations to determine policy in specific cases. I bet I could draw up a correlation one way or the other between accident rates and skin colour, but you'd be in serious trouble if your insurance company tried to use that to set premiums. (Oddly though, they're allowed to take your sex and age into account.)