Tuesday, July 06, 2010
Aswath Damodaran on Applying the Equity Risk Premium
Aswath Damodaran has a set of lectures from his class at NYU available online. He's an expert on 'valuation', which is kind of like being an expert on 'efficiency'. Obviously, if you could make something specific more efficient, or know the true value of a certain stock, you can make a lot of money, so it's understandable he glosses over some things and instead just gives the kiddies a framework to become successful green-eyeshade equity analysts.
He's a very clear lecturer, not getting bogged down on second order adjustments that are technically correct but practically irrelevant. He's very unpretentious, which is refreshing. Alas, he seems oblivious to the fact that his method has zero empirical support, and not for lack of trying.
In contrast to those deriving risk premiums, he's all application, and he clearly assumes that a risk premium exists and must be accounted for. His number for the US, is currently around 4.5% (he has a website with this info here). While he notes that historically, the return on equities above the risk-free rate is pretty volatile, and that the geometric average is a couple points below the arithmetic average, he does not make a big deal about it, as if 2% doesn't matter much, and given the ultimate set of ad hoc decisions, this gives a flavor of where he's going. He mentions the survivorship bias of the US as well. Taxes, adverse market timing, and transaction costs, are presumably irrelevant, though estimates for each of these is on the same order of magnitude as the geometric vs. arithmetic difference, or the survivorship bias (see here). All those 2% adjustments add up.
He favors adjusting for risk even if no betas are used, however one likes to dice up risk, so that higher risk firms have higher expected returns (and thus, higher discount rates or costs of capital). But his preferred approach is bottom up betas and they involve many different assumptions. He outlines several methods, including one that takes the percent of revenues from various countries and sectors (consulting vs. software), and applies separate risk premiums based on these exposures. You can map these exposures into regions, which can have their risk premium derived from Moody's rating for government debt, and their implied currency-specific risk free rates. You can also use the relative aggregate volatility in that country and the equity premium implied by the currency's libor premium to the dollar.
He mentions risk adjustments to individual companies so that less diversified companies have a higher risk than diversified companies, manufacturing companies higher risk than service companies, companies that use futures to hedge their risk have a lower risk than companies that do not, or growth companies having higher risk than value companies. Of course, higher risk implies higher expected returns, presumably.
These all seem reasonable, but there's no empirical evidence that such distinctions show up in actual returns, in fact, most of his examples go the other way. Sure they all seem plausible, indeed, given an economist’s assumptions on the utility function—that utility is increasing in wealth at a decreasing rate—this is both a necessary and sufficient reason for there to be a risk premium, so presumably just construct an intuitive risk metric and expect higher returns. Alas, the risk premium does not appear. It’s simply not true that growth outperforms value, or certain stock sectors have higher returns, even countries do not have obviously higher returns, or that more volatile stocks have higher returns than low volatility stocks. A beautiful theory killed by data, it happens all the time.
I explain this as due to utility being primarily envious, about status as opposed to wealth. While correlated ideas, one produces risk premiums, the other not. If you want to get rich, find your niche, don't expect to make money merely by tolerating great pain (ever notice those 'dirty jobs' don't actually generate a premium--there's a surplus of masochists out there). The returns in equities is like returns derivatives, the risk free rate. There are exceptions, but as a general rule this is better than assuming that risk begets return. The benefits of this approach is not merely a more accurate though less ambitious risk premium, but the effort saved can then be applied to something more fruitful, such as actually looking at the company's business model and forecasting revenues.
At one point, Damodaran notes Warren Buffett's use of simple Treasuries for discount rates, seemingly making his bottom-up betas a huge waste of time, but you had to figure he's got an answer for that. He says that when the Oracle of Omaha projects cash flows, he calculates 'certainty equivalent' cash flows. Now, I know Buffett tries to be conservative, and has a 'margin of safety' in his hurdle rates, but I really doubt he generates 'certainty equivalent cashflows' in the academic sense. Such cash flows are not adjusted for imprecision, or volatility, but rather correlation with something like the S&P500 or GDP. Then, the adjustment is a linear function of this covariance. To say that's what Buffett is doing by making conservative cash flow assumptions, implicitly, is one of those assertions that seems plausible, but under scrutiny absurd.
Damodaran talks about applying this to specific companies overseas, where he's apparently an international expert, flying all over the globe. This highlights that credentialed experts are very useful when dealing with imprecise assumptions, just tell him the answer you need and you can get the objective datum you want.
With his bottom-up approach betas can be very complicated, yet never so much that a noob can’t understand each little step in his algorithm. I love the idea of taking a company like SAP, with years of history, and instead of using the obviously irrelevant top-down beta, use a less irrelevant-seeming 'bottom up' beta by separating the revenues by continent and then into consulting and software. As a consultant whose main input is the discount rate, if he merely applied the beta taken from regressions of the stock price it wouldn't seem worth $50k. Going over all the bottom up logic requires lots of face time with senior execs, something both consultants and insecure executives love.
Yet ultimately there's absolutely no evidence this approach produces either more accurate forward covariances with the aggregate market or better predicts expected returns than simply using a number like Libor (the Buffett approach). It's typical consulting, in that the process allows the consultant to give the answer a particular firm-insider want, all with the pretense of rigor and objective analysis. It's hardly more scientific than astrology in its actual implementation, even though its filled with many reasonable assumptions and interpolations. I can see how someone unaware of the failure of the CAPM, and any risk proxy, could find him totally convincing and convenient.
How can so many smart people be fooled like this? Well, like many popular erroneous beliefs, it is not obviously wrong. The false syllogism "you must take risk to get large returns implies risk begets return" seems true. It's popular among scientists, and their consensus has a lot of weight. It's consistent with other economic assumptions that have been around for a while. It creates a pedagogical structure that is perfect for teaching, as there are fundamental assumptions, mathematical implications, and empirical applications.
It's just wrong, a waste of time, like calculating different expected returns on options.