Many valuation mechanisms utilise a market-wide Equity Risk Premium. if you look at the algebra underlying the idea, it implies that the ERP is a catch-all aggregate measure of the differential between holding stocks versus holding a risk-free asset (ignore for the moment this ludicrous fiction that there is such a thing as a risk-free asset).
In my (not so humble) opinion, it's whackery of the highest order to refuse to explicitly recognise that some industries are inherently riskier than others. If you need convincing, just look at some industry level data. Some industries have more volatile output and profit series than others - that is inescapable. To the extent that volatility differentials are persistent, they ought to be incorporated explicitly.
The persistence of volatility differentials is the second thing that must be established - whether there is any economic configuration that could cause profits in bakeries to be as volatile as profits in, say, mining - without also causing the volatility in mining to go up.
It's easy enough to test whether or not the volatility of a given industry is persistent by performing the estimation on non-contiguous sub-samples of the historical data. I do that all the time - don't you?
The third variable that ought to be considered is the historical tendency for annual profit growth in an industry to be negative. Related to this is whether or not an industry has tended to produce aggregate losses.
The two things are related, but they are not the same - after all, it is possible to go out of business entirely, without ever incurring a loss... all that is required is for the return on capital to fall below the required rate of return, and remain there.
Think of the Textiles industry: you might be surprised to know that in aggregate, it did not have a single quarter of aggregate losses between 1977 and 1999, during a period of massively increased import penetration. Many firms went out of business, and profits in the industry were lower in 1999 than they were in 1985 - but they were never negative. The industry shrunk from a 4.8% of the share of total manufacturing profits to 2.5% (and from 1.9% of all profits in the entire Australian business sector to 0.8%).
What was happening to TCF at the time was that the return on capital was falling - but never to a level below zero. People who owned capital in the industry were smart enough to shut the doors before they actually started going backwards in nominal terms (although they went backwards quite hard in real terms).
So What, Idiot?
Well, doing an analysis like that mentioned above, can really point out some interesting things. things that go against what you might think a priori.
For example, a lot of people, if asked to rank Mining versus Banking/Insurance, would declare that Mining was riskier and the banks and insurance companies were full of smart people in nice suits. (They might have changed their tune in the last three weeks or so, but generally speaking, the bankers get let off the hook whenever they cock up... and plus, nobody gets rewarded for stating the obvious after the fact).
So let's go through the analysis.
Starting with the Mining sector.
It exhibited relatively low profits growth between 1977 and 2008 (an average of just 3.45% in nominal terms - less than inflation).
Its returns have been highly volatile (the volatility of profit growth is almost exactly 5x the average rate of profits growth).
Year-on-year profits growth was negative 40% of the time, but there were no periods of aggregate losses.
All up, you would not be surprised to find that I assign a relatively high industry risk premium to Mining stocks - and the smaller a stock, the higher the risk premium.
What would be your a priori guess for Banks/Insurers? If you were honest you would say that you think of them as 'conservative', and that you would guess that they would probably have had higher, less volatile profit growth than Mining companies.
To paraphrase Sherlock Holmes, to hypothesise in advance of the data is folly.
In fact, Finance and Insurance companies have had profits growth almost 4 times as great as mining companies (on average). Almost 16% a year (although it depends critically on the data subset selected - there was a spike at the end of the data sample. 11% is far more representative estimate).
However they have always exhibited profit behaviour that warranted a much higher industry risk premium than for miners.
The volatility ratio for Finance and Insurance is over 35 - that is, the volatility of profits is more than 35 times the average rate of profits growth. f you use my "sensible" estimate of the average rate of profit growth, the volatility ratio for BankAsurance is over 60.
Year on year profits fell in 50% of the estimation sample (quarterly data from 1985-1999); aggregate profits were negative 14% of the time. So one year in seven the banks actually lost money - and that is adding up all the profits in all the banks and insurance companies.
The high volatility is partly due to extraordinarily bad results from December 1997, when Financials and Insurers experienced losses equal to the entire previous year's aggregate profits. But that was the culmination of six straight quarters of declining profits, either side of the Russian default and the Asian currency crisis.
There was also a period of six consecutive quarters of profit declines in the later 1980s (1986-87) and recent history shows us that banks and insurers have not developed any special ability in their decision-making or risk management.
By contrast, mining stocks have had one stretch of four quarters of declining year-on-year profits (in 1991/92), but did not have a single quarter of aggregate losses - despite significant changes in economic conditions, changes in the national currency framework, and a global collapse in minerals prices.
One last Detail...
The last piece of detail is largely subjective. When I do the calculations for the sector-specific risk premia for the subscriber site, this bit of the analysis does not have much weight in the final calculation (it contributes less than 5 basis points to the differential between industries).
In short, we attempt to take account of situations in which the overall share of an industry becomes misaligned with some sensible equilibrium.
For declining industries, this requires a judgment as to whether the industry is in terminal decline (i.e., its eventual weight in the economy will go to zero and its output will be replaced by imports).
For advancing industries there is less requirement for subjectivity - an industry can not have a long run growth which is greater than the rate of growth of the economy, which in real terms is equal to the rate of population growth plus the rate of productivity growth.
Following this logic, we don't decrease a sector-specific risk premium if the sector's share of aggregate profits has been falling, but we penalise any sector whose share of aggregate profits gets too far out of equilibrium.
So anyway - this job is one of the things I have to do when I get back home, and I thought it made a nice little exposition of things about which most investors (and most analysts in fact) never think - if they do think about it at all, they tend to rely on their internal prejudices ('banks are conservative') rather than actually having a look at some data. Which is why the market ended up overinvested in banks. Which is why banks managements started getting paid ludicrous amounts of money, mostly in the form of stock and option grants. Which led to them having to do things to make stock prices go up. Which led to yet another of the bancassurance industry's once-in-a-decade massive stuffups.
So banks (and insurers) ought to have a greater risk premium than run of the mill industrials... because they are prone to blowing themselves up once every ten years, almost like clockwork.