Let's review the most popular approach, which is to build a supplyside model. There are three steps:
 Estimate the expected total return on stocks.
 Estimate the expected riskfree return (bond).
 Find the difference: expected return on stocks minus riskfree return equals the equity risk premium.
We'll keep it simple and sidestep a few technical issues. Specifically, we are looking at expected returns that are longterm, real, compound and pretax. By longterm, we mean something like 10 years, as short horizons raise questions of market timing. (That is, it is understood that markets will be over or undervalued in the short run.)
By "real," we mean net of inflation. Even if we estimated the stock and bond returns in nominal terms, inflation would fall out of the subtraction anyhow. And by "compound," we mean to ignore the ancient question of whether forecasted returns ought to be calculated as arithmetic or geometric (timeweighed) averages.
Finally, although it is convenient to refer to pretax returns (as virtually all academic studies do), individual investors should care about aftertax returns. Taxes make a difference. Let's say the riskfree rate is 3% and the expected equity premium is 4%; we therefore expect equity returns of 7%. Say we earn the riskfree rate entirely in bond coupons taxed at ordinary income tax rates of 35%, whereas equities may be deferred entirely into a capital gains rate of 15% (i.e., no dividends). The aftertax picture in this case makes equities look even better.
Step One: Estimate the Expected Total Return on Stocks
DividendBased Approach
The two leading supplyside approaches start with either dividends or earnings. The dividendbased approach says that returns are a function of dividends and their future growth. Consider an example with a single stock that today is priced at $100, pays a constant 3% dividend yield (dividend per share divided by stock price), but for which we also expect the dividend  in dollar terms  to grow at 5% per year.
In this example, you can see that if we grow the dividend at 5% per year and insist on a constant dividend yield, the stock price must go up 5% per year too. The key assumption is that the stock price is fixed as multiple of the dividend. If you like to think in terms of P/E ratios, it is the equivalent to assuming that 5% earnings growth and a fixed P/E multiple must push the stock price up 5% per year. At the end of five years, our 3% dividend yield naturally gives us a 3% return ($19.14 if the dividends are reinvested). And the growth in dividends has pushed the stock price to $127.63, which gives us an additional 5% return. Together, we get a total return of 8%.
That's the idea behind the dividendbased approach: the dividend yield (%) plus the expected growth in dividends (%) equals the expected total return (%). In formulaic terms, it is just a reworking of the Gordon Growth Model, which says that the fair price of a stock (P) is a function of the dividend per share (D), growth in the dividend (g) and the required or expected rate of return (k):
EarningsBased Approach
Another approach looks at the pricetoearnings (P/E) ratio and its reciprocal: the earnings yield (earnings per share Ã· stock price). The idea is that the market's expected longrun real return is equal to the current earnings yield. For example, at the end of 2003, the P/E for the S&P 500 was almost 25. This theory says that the expected return is equal to the earnings yield of 4% (1 Ã· 25 = 4%). If that seems low, remember it's a real return. Add a rate of inflation to get a nominal return.
Here is the math that gets you the earningsbased approach:
Whereas a growth factor is explicitly added to the dividendbased approach, growth is implicit to the earnings model. It assumes the P/E multiple already impounds future growth. For example, if a company has a 4% earnings yield but doesn't pay dividends, then the model assumes the earnings are profitably reinvested at 4%.
Even experts disagree here. Some "rev up" the earnings model on the idea that, at higher P/E multiples, companies can use highpriced equity to make progressively more profitable investments. Arnott and Bernstein  authors of perhaps the definitive study  prefer the dividend approach precisely for the opposite reason. They show that, as companies grow, the retained earnings they often opt to reinvest result in only subpar returns  in other words, the retained earnings should have instead been distributed as dividends.
Handle with Care
Let's remember that the equity premium refers to a longterm estimate for the entire market of publiclytraded stocks. Lately several studies have cautioned that we should expect a fairly conservative premium in the future.
There are two reasons why academic studies, regardless of when they are conducted, are certain to produce low equity risk premiums.
The first is that the studies make an assumption that the market is correctly valued. In both the dividendbased approach and earningsbased approach, the dividend yield and earnings yield have reciprocal valuation multiples:
Both models assume that the valuation multiples  the pricetodividend and P/E ratio  are correct in the present and will not change going forward. This is understandable, for what else can these models do? It is notoriously difficult to predict an expansion or contraction of the market's valuation multiple. The earnings model might forecast 4% based on a P/E ratio of 25. And earnings may grow at 4%, but if the P/E multiple expands to 30 in the next year, then the total market return will be 25%, where multiple expansion alone contributes 20%! (30/25 1 = +20%)
The second reason low equity premiums tend to characterize academic estimates is that the total market growth is limited over the longterm. You'll recall that we have a factor for dividend growth in the dividendbased approach. Academic studies assume that dividend growth for the overall market  and, for that matter, earnings or EPS growth  cannot exceed the total economy's growth over the long term. If the economy grows at 4% as measured by gross domestic product (GDP) or national income, then studies assume that markets cannot collectively outpace this growth rate. Therefore, if you start with an assumption that the market's current valuation is approximately correct and you set the economy's growth as a limit on longterm dividend growth (or earnings or earnings per share growth), a real equity premium of 4 or 5% is pretty much impossible to exceed.
Now that we have explored the risk premium models and their challenges, it is time to look at them with actual data. The first step is to find a reasonable range of expected equity returns; step two is to deduct a riskfree rate of return and step three is to try to arrive at a reasonable equity risk premium.

Investing
The EquityRisk Premium: More Risk For Higher Returns
Learn how the expected extra return on stocks is measured and why academic studies usually estimate a low premium. 
Investing
Calculating the Equity Risk Premium
Equity risk premium is the excess expected return of a stock, or the stock market as a whole, over the riskfree rate. 
Investing
How to Calculate Risk Premium
Think of a risk premium as a form of hazard pay for risky investments. 
Investing
Digging Into The Dividend Discount Model
The DDM is one of the most foundational of financial theories, but it's only as good as its assumptions. 
Investing
The 3 Biggest Misconceptions of Dividend Stocks
To find the best dividend stocks, focus on total return, not yield. 
Investing
Why Dividends Matter
Seven words that are music to investors' ears? "The dividend check is in the mail." 
Investing
The Risks of Chasing High Dividend Stocks
Dividend stocks offer enticing yields, but a lot can go wrong on the way to collecting that dividend payout.