Sunday, September 11, 2011

Stock Returns by Debt Rating

Above are annual returns for portfolios formed every July 1, based on the Senior rating of the company. I only used non-financial companies because financial companies tend to be only with investment grade, and they are very different from a debt rating perspective (when I modeled default at Moody's, there was a clear non-financial focus because financial companies are very different).

The annual data are as follows for this period (Jul1975-Jun2011):
StockReturns(%) betaVolatility(%)
AAA 12.4 0.7817.1
AA 13.9 0.8116.1
A 14.3 0.8116.5
BBB 14.2 0.8217.9
BB 15.0 1.0423.4
B 8.6 1.4332.0
C -12.7 1.1844.9


It appears there's a reasonable story one could tell about returns from AAA to BB: higher returns, and higher intuitive measures of risk: beta, volatility. But for B and C rated stocks, the returns make no sense to standard asset pricing theory, because these are obviously risky stocks. I remember presenting this chart to an NBER conference around 2000, and the esteemed audience told me I was wrong; my data had to be incorrect. I was working at Moody's, so my ratings data was as good as it got. Anyway, I wrote it up and sent it to Journal of Portfolio Management, and the editor, Peter Bernstein, wrote back they weren't accepting submissions at that time. I thought that was an odd response. This avenue wasn't part of my day job, so I let it go, but I keep updating my data for fun.

The result is really corroborated by Campbell, Hilscher and Szilagyi (2005), who found distress risk to be negatively correlated with stock returns, which makes sense because volatility and leverage is inversely correlated with future returns, and cash-flow is positively correlated with future returns, so those are the main drivers of default risk.

Reality is that which, when you don't believe it, doesn't go away, so I don't really mind when people tell me I'm wrong on facts like this.

5 comments:

Anonymous said...

when you control for beta, does the rating still explain returns? i.e. is this just low beta anomaly in disguise?

Nick Gogerty said...

Just curious. was there control for survivorship bias in the data, and if so any chance of sharing the attrition rates?

Eric Falkenstein said...

I don't have it at the ready, but if a firm moved rating, it retained its portfolio memembership for the 12 month period. For firms that delist for any reason, I take their delisting return, and then merge their remaining capital pro-rata into the portfolio it's in.

Anonymous said...

Squinting at the graphic, looks like the C-rating effect lost its mojo in ~2000. Is that right?

Anonymous said...

This is a great idea that certainly deserves more attention. I echo Nick's sentiments in that I would love to see the attrition rate for each rating band. I also think a return range from mean for each rating band would be a real value add for this reseach.