Eugene Fama is interviewed by the Minneapolis regional Fed.
Region: How do you explain the equity premium puzzle [the idea that stocks should in theory provide only a 1 percent higher annual return than bonds, but have historically returned nearly 7 percent more]?
Fama: In terms of these consumption-based asset pricing model stories? What I say to the consumption people is: You’re telling me the premium should be about 1 percent a year. Well, you wouldn’t be able to tell the difference between that and zero over a 1,000-year period. And for a 1 percent a year premium, who do you know that would hold stocks? It’s this representative investor, but who is that guy anyway? I wouldn’t hold them. I don’t know anybody else who would. So there’s got to be something missing in those models.
Thanks to Tyler , who in turn thanks Mark Thoma, for the pointer.
READER COMMENTS
Brandon Berg
Dec 15 2007 at 11:23pm
Isn’t focusing on the 1% figure a bit of a cop-out? Maybe the model that says there should only be a 1% gap assumes an unrealistically low level of risk aversion, but let’s say the real gap at which the average investor is indifferent between the two is 3%. If it’s actually 7%, that still requires explanation.
Also, couldn’t one solve the equity premium puzzle simply by asking a representative sample of people who buy bonds why they do so?
Maniakes
Dec 15 2007 at 11:49pm
You’re telling me the premium should be about 1 percent a year. Well, you wouldn’t be able to tell the difference between that and zero over a 1,000-year period.
1.01 ^ 1000 = 20,959
By what criteria is a factor of 20,959 undiscernable?
dearieme
Dec 16 2007 at 8:15am
For some reason I become uncomfortable when people start banging on about a thousand years.
eric
Dec 16 2007 at 7:41pm
Is the point about 1% related to marty Weitzman’s piece where he notes that if you incorporate parameter uncertainty, you get a much higher risk premium.
DeLong on Weitzman here
Bob
Dec 17 2007 at 1:55pm
The point about 1% isn’t about the return you’d earn over 1,000 years, but rather about our ability to observe the premium in the data – if the “true” difference were 1%, the “paradox” would be why stocks don’t earn more than bonds (as you wouldn’t be able to support the claim with a high degree of statistical certainty from available data).
rmark
Dec 18 2007 at 4:02pm
Speaking of questionable model assumptions –
I remember a journal article a few years back in which the author noted that both long term corporate equities (stocks) and long term AAA corporate bonds had the same 7% return over some longer period of time, say 1926 – 1995. Entirely reasonable if business borrows up to no extra return. However equity investors received less than market returns due to broker, transaction, and dollar weighting costs, while bond investors had broker costs plus shorter, safer bonds, maintaining a rough equivalence of net return at the investor level of about 5%. The changing value of money comes into play based on the characteristics of the securities. Stocks reinvest part of their earnings, capturing inflation and making the net 5% a real return (8% nominal), while bonds zero reinvestment but higher cash flow loses to inflation, which at a past roughly 3% results in a 2% real bond return, reduced to 1% if corporate bonds are removed from the mix. My average investor model pretty much replicates the past US averages using only bond yield and reinvestment rate.
Comments are closed.