4 ms·

My question is, who is to say that X "FICO Score" correlates to anything too meaningful -- or that the score itself has not become harder or easier to obtain?

My question is, who is to say that X "FICO Score" correlates to anything too meaningful -- or that the score itself has not become harder or easier to obtain? Perhaps the data exists on this, but to my unresearched mind there's a lot of "fudging" potential within it all by itself. For instance, maybe the current 'borrower state' that gets you a 690 previously would have given you a 725; or maybe it's the opposite. I understand how averages work, but is there any oversight or auditing done on the rating agencies and the ratings themselves? It just seems ripe for 'gaming' at that level IMO; if banks need to sell financial products and they buy the ratings from the agencies, surely they can lean on an agency to 'fudge' numbers one direction or the other (I'm speaking in aggregates here)? Anecdotally, every time I look at my credit score I cringe -- I've never carried much debt, have always paid it off on time or early, and yet my credit score stays somewhat low, seemingly because I largely don't participate in "the system". Furthermore, every time I've ever applied for credit I end up with scores (and offers of credit) far in excess of the numbers I see when I either get my 'free credit report' from the 3 agencies, and they're similarly different than the numbers I see if I pay to see my reports from the 3 agencies. I don't know, it just seems like a "chicken-egg" problem to me, and one where the rating agencies are perhaps incentivized in a way that does not align with the "economy" as a whole -- and is instead much more closely with their high-value customers (lenders). I also feel like there's so much "missing data", which could probably tell quite the tale, for instance, if someone such as Google was to begin a 'ratings agency'. Half of my finds that to be a scary thought, the other half of me thinks that we'd probably end up with objectively better outcomes, both economically and probably, in many cases, even individually too.

adanto6840 7y ago
Hrm -- I'd love to be enlightened! I didn't word it very well, but my questions are primarily: What oversight is there on the ratings agencies themselves? How do we know the distribution of scores is the same today as 10 years ago (and proportionate to credit-worthiness)? Is there anything that prevents credit-ratings companies from under/over valuing the credit-worthiness for large swaths of the population? Is there historic data available to correlate "consumer credit score" to distressed notes -- and can that be done in a forward-looking manner, to exclude credit score 'fudging' as a factor, and to better be able to know that it's apples-to-apples comparison historically speaking. Otherwise, could it not simply be that borrowers previously known as higher-quality borrowers are simply being rated more poorly/cautiously, for instance? I'm not saying that is the case, but I've always felt that the incentives seemed strange. Are credit rating agencies paid (by lenders) based upon their prediction accuracy?
- kasey_junk 7y ago
  The lenders themselves spend a lot of effort on making sure that the credit ratings are good signals. When your business is extending credit (typically very low margin) you walk a very fine line between being overly permissive in lending and not lending enough. Minor movements of that line can be the difference between profitability and not. If a ratings agencies algorithm was shown to be changing to be too aggressive (either historically or against its peers) the lenders will dial back its importance in their own decisions. In the extreme to the point of not paying for it anymore. This internal competition does a pretty good job of making sure the algos represent the current state of the world. There is a question of about this system causing future systemic risk to be underpriced because no one is incentivized to investigate it fully and add it to their algorithm.
patio11 7y ago
You're being a little imprecise with your terminology here, which may have contributed to folks thinking that this was not a comment which moved the discussion forward. The ratings agencies (Moodys et al) rate debt, including particular securitization pools and/or corporate borrowers, but not consumer borrowers. FICO produces an algorithm which lenders use to rate consumer borrowers based on records retained by credit reporting agencies (CRAs). You, as a lender, can't call Equifax (a CRA) and say "Hey you said 690 but I think really 725." Their information says what it says regarding tradelines (the fact that they were delinquent on a CC 2 years ago is a fact on record that you can't alter), and you have no easy interface to modify that (though you can, and you do, encourage customers in marginal situations to do their own work on their credit). You ask why don't banks defraud the GSEs on a retail level for marginal loans. The short and perhaps unsatisfactory reason is that if the GSE is dissatisfied with a particular loan they'll put it back on you (you have to repurchase rather than securitizing) and that if you make a practice of this you're going to go to jail. Anecdotally, every time I look at my credit score I cringe The video game player in me understands that being told one has a score one wants to increase it, but, do you care about this number? Why? If you don't anticipate buying a house in the next 6 months it is all but irrelevant to you. If that is in your near term plans, then you have a fairly straightforward path to increasing it. (Stop paying off earlier than the reporting date, which can result in tradelines looking like they're dormant and therefore not counted towards your score.) Furthermore, every time I've ever applied for credit I end up with scores (and offers of credit) far in excess of the numbers... Yep, because the offers of credit are an actual fact about the world and the credit score is a point-in-time snapshot of a model which predicts a model that partially influences a decision that might be relevant to your interest. In the case where reality and your model differ, reality takes precedence, so why do you care what the model outputs?

adanto6840 7y ago

Hrm -- I'd love to be enlightened! I didn't word it very well, but my questions are primarily: What oversight is there on the ratings agencies themselves? How do we know the distribution of scores is the same today as 10 years ago (and proportionate to credit-worthiness)? Is there anything that prevents credit-ratings companies from under/over valuing the credit-worthiness for large swaths of the population? Is there historic data available to correlate "consumer credit score" to distressed notes -- and can that be done in a forward-looking manner, to exclude credit score 'fudging' as a factor, and to better be able to know that it's apples-to-apples comparison historically speaking. Otherwise, could it not simply be that borrowers previously known as higher-quality borrowers are simply being rated more poorly/cautiously, for instance? I'm not saying that is the case, but I've always felt that the incentives seemed strange. Are credit rating agencies paid (by lenders) based upon their prediction accuracy?

kasey_junk 7y ago
The lenders themselves spend a lot of effort on making sure that the credit ratings are good signals. When your business is extending credit (typically very low margin) you walk a very fine line between being overly permissive in lending and not lending enough. Minor movements of that line can be the difference between profitability and not. If a ratings agencies algorithm was shown to be changing to be too aggressive (either historically or against its peers) the lenders will dial back its importance in their own decisions. In the extreme to the point of not paying for it anymore. This internal competition does a pretty good job of making sure the algos represent the current state of the world. There is a question of about this system causing future systemic risk to be underpriced because no one is incentivized to investigate it fully and add it to their algorithm.

patio11 7y ago

You're being a little imprecise with your terminology here, which may have contributed to folks thinking that this was not a comment which moved the discussion forward. The ratings agencies (Moodys et al) rate debt, including particular securitization pools and/or corporate borrowers, but not consumer borrowers. FICO produces an algorithm which lenders use to rate consumer borrowers based on records retained by credit reporting agencies (CRAs). You, as a lender, can't call Equifax (a CRA) and say "Hey you said 690 but I think really 725." Their information says what it says regarding tradelines (the fact that they were delinquent on a CC 2 years ago is a fact on record that you can't alter), and you have no easy interface to modify that (though you can, and you do, encourage customers in marginal situations to do their own work on their credit). You ask why don't banks defraud the GSEs on a retail level for marginal loans. The short and perhaps unsatisfactory reason is that if the GSE is dissatisfied with a particular loan they'll put it back on you (you have to repurchase rather than securitizing) and that if you make a practice of this you're going to go to jail. Anecdotally, every time I look at my credit score I cringe The video game player in me understands that being told one has a score one wants to increase it, but, do you care about this number? Why? If you don't anticipate buying a house in the next 6 months it is all but irrelevant to you. If that is in your near term plans, then you have a fairly straightforward path to increasing it. (Stop paying off earlier than the reporting date, which can result in tradelines looking like they're dormant and therefore not counted towards your score.) Furthermore, every time I've ever applied for credit I end up with scores (and offers of credit) far in excess of the numbers... Yep, because the offers of credit are an actual fact about the world and the credit score is a point-in-time snapshot of a model which predicts a model that partially influences a decision that might be relevant to your interest. In the case where reality and your model differ, reality takes precedence, so why do you care what the model outputs?