Monday, June 06, 2011

Equity Premium Survey Doesn't Register My Vote

Pablo Fernandez has made it his job to survey finance professionals to see what they all think the 'equity risk premium' should be. This is the expected annualized return one expects on equities over the 'risk free' rate, and is used for various cost-of-funding assumptions. In his recent survey most put it currently about 5.5%, 3% below the number from textbooks around 1990.

But here's something funny. Someone told me they responded to his annual survey with 0%, as I argued in my book, noting that with geometric averaging, survivorship bias to the US experience, taxes, transaction costs, and adverse timing, your average investor is at the discount rate Warren Buffet uses: Long Term Treasuries. This observation was not included in his survey because it was too extreme!

Now, I can see removing observations that appear mistakes, but 0% isn't crazy, just low. Ivo Welch is a pretty conventional finance professor, has new finance textbook (available free online here), and he argues that it is between 2 and 4%. Eugene Fama and Ken French argue it's around 3.5%. Dimson, Marsh and Staunton, 3.5%. These professors, meanwhile, are ahead of the professionals who naturally remember what they learned in business school, which for your average respondent was around 1990. So, I'm below the current estimate of orthodox and highly esteemed economists by about 3.5%, which is less than these guys are below the conventional wisdom circa 1990. Just looking at how this datum has moved, suggests my estimate is plausible. Forecast errors are not like lightening strikes in that they are more likely if they happened previously, not less, meaning, if the conventional wisdom has changed from 8.5% to 3.5%, it can go to zero.

This is why I like reading about science and epistemology in general because it's all well and good to say science is all about the evidence and people being 'reality based', and only idiots or bigots are prejudiced. Everyone only sees what they believe, and beliefs are not easily moved by contrary data: observations are theory-laden. The experts do not sees a risk premium below 2% because it just doesn't fit in their theoretical understanding of how things should be. To those who say that equities must generate a risk premium, otherwise no one would invest there, I would say that we don't pay premiums for 'smelliness' or other kinds of unpleasantness. That is, Lord Keynes wrote on page 162 of his General Theory:

For example, smelly processes command a higher reward, because people will not undertake them otherwise. So do risky processes.

As usual he didn't actually look at the data and instead just threw out a plausible assertion, because toilet cleaners do not earn more than other custodians, and the untouchables of India who monopolize sewage work are not compensated well at all. Further, we don't earn extra for the following: IPOs, penny stocks, options, long duration bonds, junk bonds, long-short sports bets, lottery tickets, volatility within the stock market, currencies, and commodities, just to name a few, so why should stocks be different?

Don't expect to make more money for taking risk, just know you have to take risk to make more money. If you don't understand the difference, you shouldn't be taking risk.

7 comments:

Anonymous said...

I think we need to clarify exactly what we mean by the risk premium.

I interpret it as the difference in the long run geometric return of equities over 30-year Treasury yields. Since I think large cap stocks will earn 7.6% and assuming today's 30-year Treasury yield of 4.26% is fair-valued, I'd say the risk premium is about 3.3%. But in the longer run, I expect 30-year treasuries to be around 4.8%, so I could see the premium being as low as 2.8%.

However, I notice in the comments that some people are using 3-month and even 1-year government bonds. Additionally, it isn't clear to me that some respondents aren't using arithmetic numbers, which bring about another issue--what is the standard deviation. A high arithmetic return with a very high standard deviation produces a very low geometric return.

Another thing: is it the premium over the yield or the return of long treasuries? (And some countries don't issue 30-year bonds, so maybe I should rescale the premium to the 10-year Treasury.)

Finally, it looks like some of the commentators are looking at cost-of-capital for specific firms. I believe this should be arithmetic return, but I could be wrong. The geometric return of a single stock is quite different from the geometric return of a market index fund like the S&P 500, but the arithmetic return could be the same.

Fernandez needs to clarify the precise definition of the MRP and the application of it. Because of the confusion, I find his surveys to be uninformative.

Anonymous said...

Excuse me but I'm almost sure that there's research out there showing that IPOs do have an excess return and it's pretty high, although after the first day pop they go back to average. In other words, there's alpha to be gained from buying into IPOs and selling on the first day of trading.

mOOm said...

It's the premium over the risk free rate. So 0% isn't means that it is lower than long-term treasuries.

mOOm said...

So the overnight or 90 day rate.

Eric Falkenstein said...

anon: first day pops on IPOs aren't something you can just buy, you have to have accumulated favors with your broker or be in a position to help someone (eg, Congressman). So, the one-day 12% IPO pop is not an 'equilibrium' result, just a way for insiders to make barter payments (ie, tax free). The return subsequent to the one-day pop, is an equilibrium result, and it's well below benchmark returns, even though it's pretty risky.

Kent said...

A lot can change in 30 years.
My understanding is that the 3 month T-Bill can be stated as risk free as it carries almost zero default risk.

jon e whiteford said...

experts do not sees a risk premium below 2%

a carpe type sees?