33 Comments

There appears to have been some type of market risk reform for the housing market in the United States, called the Home Valuation Code of Conduct. There is no blindness at all to preventing risk fraud in small doses, but not enough political will to turn something like it into law on a larger scale. Also, The congressional bill HR 3044 seeks to impose a moratorium on this agreement.

Expand full comment

OK, that starts to shade into evil.

Expand full comment

Has the stock market actually "predicted nine out of the last five recessions"? Scott Sumner usually says 1987 is the one time the stock market crashed but we didn't get a recession.

Expand full comment

So far, I haven't seen anyone in the comment thread mention one of the most important details of ratings: credit rating agencies are not covered by Regulation Fair Disclosure (FD), so they can get information about debt securities that are not public knowledge (and that firms wouldn't disclose publicly due to concerns about revealing proprietary information). There may be reasons to question whether the ratings agencies used their additional information effectively, but a prediction market would not be a solution, at least given the existence of Reg FD.

Expand full comment

It is not that the Reserve and other regulators opposed the proposal quietly behind the scenes to get Congress not to pass it. It is that Congress passed a rule and then they refused to enforce it.

Expand full comment

Grade-inflated securities allow managers to take in more if what those managers are allowed to hold by government regulation must clear some grade threshold.

Imagine a world where everything is rated B, but regulation requires that you only hold A-rated stuff. Your volume of assets managed is 0. Now demand and get grade-inflation by one letter, so that everything is rated A. Now you're free to take in whatever you want.

Expand full comment

Regarding the relative accuracy of bond prices and credit ratings, bond prices are certainly swifter in their responsiveness to news (e.g., news of higher-than-expected levels of subprime mortgage defaults). On the other hand, the bond market is like the stock market, which has correctly predicted nine of the last five recessions. So if you look at, for example, LIBOR (not strictly a bond price, but it's the same idea) or AAA CMBS yields, the volatility over the past 18 months doesn't necessarily seem that meaningful.

It's a little like asking whether Intrade is more reliable than an average of polls. Normally, they track so closely that there's not much to choose. Intrade obviously moves more when there's breaking news.

Expand full comment

Er, buy high.

Expand full comment

If you buy low and then refuse to sell, speculators can't create a liquid market. So - though I suppose this is also obvious - you'd have to allow naked shorting of the market.

Expand full comment

Eliezer, Phil Gramm was an economist before he went into politics, so that could help explain how the idea got as far as it did.

David Smith/y81, perhaps the young workers were math whizzes from Bombay. But what about management? Might the bosses have had chummy connections with those of the firms they were rating?

Expand full comment

I know that derivatives are in the dog house but you can solve the illiquidity problem (which is real) with a futures market for the sub debt market (of course, futures markets are basically prediction markets). Which, BTW, everyone who understands the problem knows is the real solution if you can meaningfully precommit to "no bailout." It does not have to be an ironclad commitment, just substantial risk, to convey information about the downside.

Of course, you need a capped upside too - the reason that stock prices don't work is that they do not necessarily fall when firms take terrible risks (think call options), at least until the risks blow up. We actually have an example of this in Fannie and Freddie - bondholders were bailed out but preferred stockholders were not. Any future issues of GSE preferred stock will provide the risk-taking signal Robin seeks. Which is why I don't expect any future issues...sigh.

Expand full comment

I guess I was making the more modest argument 1) having a duopoly generate conspicuous validation has real value and 2) adding functionality to the bond price introduces new issues around fraud and gaming, especially in illiquid bonds. Lots of people seem to think agency ratings are useless or intrinsically corrupt.

But ultimately, I totally agree that bond prices should be used by anyone interested in evaluating a company's credit rating--it's clearly more important than anything else (bond prices Granger cause ratings changes). Anyone doing credit analysis finds a market price of credit as their most prominent datapoint.

Expand full comment

Eric, the issue is the relative accuracy of market prices and agency ratings, both under the assumptions that such estimates are relied on. Corruption goes with agency ratings, and with market prices that were relied on there would be attempts at manipulation via trades. That price manipulation would induce more trading and more liquidity by speculators attempting to profit from pricing errors. Shouldn't we at least *try* this market price method of estimating risk to see how well it compares to agency rating?

Expand full comment

There's no evidence, just anecdotal stories about poorly marked bonds. Unlike equities, bonds are often held a very long time, because the transaction costs are so much higher. I know that illiquid bond prices are a real problem, in that people in that space have a lot of protocols to protect themselves from relying on prices from brokers on bonds they quote but do not trade. Illiquid asset prices make volatility seem low (Andy Lo has done work here), merely because it seems like zero until a big move makes this untenable, which is why monthly return data should not be used to estimate Sharpe ratios on illiquid strategies.

I guess its part of a general problem. Say you have a prediction market, that has an open interest of $100k. You find it predicts some target pretty well, looking at 10 years of data. If you then use that market to make a decision worth $10MM, I would be wary that the past performance of that predictor would be the same.

Expand full comment

y81, the issue is the relative accuracy of the two signals. Yes of course they both make mistakes, but which makes bigger mistakes overall? One key clue: which signal first showed the news of the upcoming crisis?

Expand full comment

Of course, this totally undercuts Mr. Hanson's cronyism explanation, since the ratings agency staffers clearly hadn't lived in Davenport or Eliot House, being from Bombay. Mr. Smith's sociology is certainly closer to the truth than Mr. Hanson's.

Expand full comment