Would you have predicted this?
This paper investigates the impact of credit rating changes on the sovereign spreads in the European Union and investigates the macro and financial factors that account for the time varying effects of a given credit rating change. We find that changes of ratings are informative, economically important and highly statistically significant in panel models even after controlling for a host of domestic and global fundamental factors and investigating various functional forms, time and country groupings and dynamic structures. Dynamic panel model estimates indicate that a credit rating upgrade decreases CDS spreads by about 45 basis points, on average, for EU countries.
I would not have! Perhaps I am biased from living in a country where credit ratings are a contraryindicator of sovereign interest rates, and where municipal defaults inevitably lead to helpful comments from ratings agencies like "If the payment doesn't get made, we would downgrade the rating." Apparently, though, sovereign ratings matter, at least in Europe and at least at some points on the ratings scale.1
Not AA through AAA, though, obviously; and the paper includes a little model of how ratings changes affect CDS spreads at various ratings points suggesting that they matter most in the sort of BB/BBB crossover area, which makes sense.2 There are also some differences between country groups and over time: CDS spreads were mostly unresponsive (1 notch = 1.8bps of CDS) to credit ratings in 2005-2007, becoming far more responsive (1 notch = 42bps) during the 2008-2012 crisis period, and countries using the euro are far more responsive (1 notch = 45bps) than non-euro EU countries (11bps). The GIIPS countries are also more responsive (55bps) than the non-GIIPS euro countries (23bps).
The short story here is basically that the market's designated arbiters of sovereign creditworthiness seem to have some influence on the market's perception of sovereign creditworthiness, which when you put it like that would seem to be non-news except, y'know, it's S&P and Moody's and Fitch, come on. Beyond that, what do you make of the "controlling for a host of domestic and global fundamental factors" bit? It's not clear exactly what that means; the tables control for the local stock market index, a global commodities index, the price of crude oil, and the VIX as a proxy for general global crisis-ness, and the paper notes that the authors "pre-tested with a number of possible country-specific and global control variables including foreign exchange reserves, inflation, industrial production and unemployment" and found they were not consistently significant.
I guess there are two ways to read that. One is that, effectively, the ratings agencies are better arbiters of creditworthiness than a bunch of academics looking at historical macroeconomic data: that the ratings agencies bring knowledge or judgment or insight or whatever to their job that makes them, like, actually good at it. You'd sort of hope so? ("If you're so smart, Aizenman, Binici & Hutchison, why don't you start a ratings agency?" is probably not something that is said a lot, but still.)
The other is that ratings have their own self-fulfilling effect on market perceptions that is not justified by economic fundamentals: that investors sell sovereign debt on a downgrade even when the downgrade is, y'know, wrong.3 If the ratings agencies' downgrade decisions cause market reactions beyond what is justified by economic fundamentals, that's not entirely an advertisement for those agencies' work.
Joshua Aizenman, Mahir Binici, & Michael M. Hutchison: Credit Ratings and the Pricing of Sovereign Debt during the Euro Crisis [NBER, earlier free version]
1.If you prefer your variables standardized this might be of interest to you:
[A] one standard deviation rise in credit ratings lowers CDS spreads by -0.15 to -0.16 of a standard deviation, not dissimilar to the effect of a one standard deviation rise in equity prices (-0.11 to -0.15). Standardized changes in commodity prices and oil prices have smaller effects on CDS spreads, ranging from -0.07 to -0.12, while the VIX coefficients range from 0.03 to 0.10. Clearly, credit rating changes have economically important effects on CDS spreads, as well as statistically significant, even when controlling for domestic and global economic variables.
2.Here's a graph of how much a one-notch ratings upgrade improves (negative) or worsens (positive) a country's CDS spread, depending on the starting credit rating. The y axis is CDS in bps; the x axis is starting rating in small notches with AAA = 25, AA = 23, A = 20, BBB = 17, BB = 14, B = 11, CCC = 8, etc.
Does this look silly? It should. The basic idea is that CDS is very sensitive to ratings changes in near-default territory, pretty sensitive in the BBish area, and near-totally insensitive at AA and above and in single-B territory. The thing where ratings seem to be negatively correlated with CDS at around a triple-C is - as the authors put it, where the effect of a downgrade is "implausibly positive" - is not like the thing where U.S. interest rates fall with every downgrade. It's just, like, statistical nonsense. Splines.
3.Is the fact that CDS spreads are most sensitive to ratings changes at the BB/BBB margin evidence of that? If you're limited to investment-grade bonds then you can ignore an A- to BBB+ downgrade (or, of course, a BB- to B+ downgrade) but not a BBB- to BB+ downgrade. That difference has nothing to do with creditworthiness and everything to do with technical factors and unearned faith in ratings.