Predicting is difficult-especially about the future. So runs the old joke.

Difficult or not, though, predictions are necessary in the money game, and there's the rub.

It's tempting to think that a prudent approach to diversified investing relieves us of forecasting. Saying "no" to predictions is popular in financial planning circles, where the idea of peering into the future and then acting on the analysis is dismissed as market-timing. It's a reasonable view, but it's also a bit misleading.

Expecting to invest without forecasting is akin to planning a day of swimming and expecting you won't get wet. Every investment decision-or at least every reasonable investment decision-requires a forecast, an assumption, an expectation. The embedded bet on the future isn't always conspicuous, but it's there.

Asset allocation certainly isn't immune to this rule. Ultimately, there's a forecast behind every decision on how much to hold of this or that asset class and how to design the mix generally. The only question is whether the forecast is based on reasonable expectations, and that's where things get tricky.

Take the popular plan of crafting a strategic asset allocation and sticking with it through thick and thin, supplementing it only with periodic rebalancing back to the initial asset weights. It's a sensible strategy on the surface, one that sidesteps market timing in favor of a dispassionate plan for picking up risk premiums, much as if they were lying on the ground like acorns, awaiting collection. Yet so-called static asset allocation strategies are rooted in forecasting. These forecasts are arguably flawed, but they are forecasts nonetheless.

The rationale for static asset allocations are long-run historical returns-for example, the U.S. stock market's annualized 10% total return for the past 80 years, which investors use as if it were a window looking into future performance. Even if that held true in the long run-a debatable proposition-there's reason to wonder if it will hold over, say, the next ten or even 20 years.

Returns, after all, are volatile in the short run, a fact that's too easily overlooked when we rely on long-run history as a guide. Even the rolling-ten-year return for U.S. stocks has been something of a roller-coaster ride, ranging from an annualized 3% to nearly 18%. Waiting for salvation in the long run is a nice idea on paper, but as Keynes said, in the long run we're all dead. In the short term, meanwhile, stuff happens, as the current climate so painfully reminds.

The lesson is that the future looks different, sometimes radically different, depending on from where you are looking. If you ignore this fact while crafting your asset allocation, you're asking for trouble-for instance, if you built your plan in early 2008 while looking only at long-run historical results. Obviously, that design would be something less than optimal now.

Risk premiums appear to be stable over very long periods of time, but that's mostly an illusion if we adjust for the real-world time horizons of clients. In fact, these premiums vary through time, as a large and growing body of academic research and real-world evidence tell us. The only question is how best to manage money in a world where the premiums fluctuate. The basic answer is that investors shouldn't rely totally on static asset allocation strategies.

Consider the outcomes delivered by a diversified group of asset classes in the period from 2000 to 2002 and compare those with the dramatically different outcomes in 2008. Last year, stocks, corporate bonds, REITs and commodities suffered steep losses. Only government bonds bucked the trend with positive returns. But the previous bear market was a different animal entirely. Although stocks took a beating from 2000 to 2002, bonds, commodities and REITs posted handsome gains. Why the difference? Valuation was one reason, and arguably the dominant reason.

Consider that the U.S. stock market peaked in March 2000. At that point, yields on REITs traded at an enticing 8.3%, or well above the 6.3% for the benchmark 10-year Treasury note at the time. By contrast, U.S. stocks at the time posted a paltry trailing yield of less than 1.2%, based on monthly data for the S&P Composite-an all-time low. Over the course of the next three years, REITs rose nearly 44%, while U.S. stocks retreated by 40%.

There was no repeat performance last year, of course, when REITs tumbled 39%, exceeding even the steep 37% loss for U.S. stocks. Why the difference? Valuation once again appears to offer a clue. At the last peak for stocks in October 2007, the REIT yield was 4.2%-slightly below the 10-year Treasury's 4.5%. Was there a compelling case to buy REITs when a risk-free Treasury note offered a higher yield? Apparently not. REITs looked compelling in early 2000, but they seemed overpriced in late 2007. Then again, saying as much at the time was tantamount to being labeled a market timer.

Labels aside, ignoring the price of risk isn't a strategy; it's denial. Even worse, it courts trouble when you're managing asset allocation. It's easy to say that market timing is a loser's game, but the truth is that it's naïve to compare market timing with adjusting asset allocation based on the expected price of risk.

"What is crucial to the success of diversification is the price you pay for initiating exposure to different asset classes," says Adrian Cronje, director of asset allocation at Wilmington Trust. "At certain times, you're paid to assume risk; at other times you're not paid to assume risk."

The price of risk changes, so asset allocation should respond to some degree to these changes, says Gary Brinson, a veteran strategist who co-authored a famous 1986 research paper that brought asset allocation to the fore as an investment concept. "You need to change asset allocation over time," advises Brinson, who is also president of GP Brinson, an investment firm in Chicago.

That's especially true in periods of extremes, when expected risk premiums are abnormally high or low. In those times, the future may be less uncertain than usual.

It's difficult for anybody to estimate a return, of course, and no one can be confident that the future is clear. A dose of humility is always required when you're making investment assumptions, especially these days when volatility and uncertainty are much amplified. But you would also be wrong to say that you somehow won't ever be bogged down by the messy work of prediction.

To be sure, there's value in considering the long-run return as one of many elements for crafting asset allocation. But by relying on it exclusively you needlessly increase risk, since this approach to diversifying among asset classes makes a tacit assumption that returns are stable.

The fact that expected returns vary through time warrants a dynamic asset allocation, or at least a partly dynamic portfolio strategy. There's a danger here, of course, in that predicting returns is prone to error. But so, too, is expecting stable returns to hold in any given year, or even over five or ten years. Perhaps doing some of both is a reasonable compromise. But that raises the question: How should financial advisors build return estimates?

In the proverbial 1,000-mile journey, it's best to begin with the first step, and that usually means equities, the primary asset class for generating risk premiums for most investors. There are many ways to calculate estimates for equity risk, but a reasonable place to start is with the so-called Gordon equation, one of the more unassuming (yet reliable) members of the discounted-cash-flow family of forecasting models. It's not a crystal ball, particularly if your investment horizon is less than ten years. But in the perennial quest to find context for thinking about future stock market performance, the Gordon equation deserves a spot in every strategist's analytical arsenal.

In its basic form, the equation tells us that the expected return for stocks in the long run is the sum of the dividend growth rate and the current dividend yield. A reasonable assumption, given that long-run investors will receive the current yield and any increases tied to higher dividend payouts. To the extent that actual returns differ from this basic formula, capital appreciation (or loss) is the reason. Forecasting that aspect of equity returns is virtually impossible, which leaves us to focus on what's clearer, as the Gordon equation does.

Financial planner William Bernstein, writing in his popular book The Four Pillars of Investing, says the Gordon equation is "an accurate way to predict long-term stock market returns." The history of the 20th century suggests as much. The average dividend yield from 1900 to 1999 was roughly 4.5%, which matches the compound growth rate of dividends. Based on the Gordon equation, the U.S. stock market should have delivered a 9% performance. The actual record is a bit higher, at 9.89%, Bernstein reports. "Not too shabby," he concludes.

What is the Gordon equation telling us these days? The outlook for equities has improved considerably in the last several years, as Figure 1 shows. From an outlook early in this decade that said stocks would return less than 4%-the lowest in the post-World War II era-the future looks more encouraging for equity investing. At the end of 2008, the Gordon equation was forecasting that the stock market performance would be around 9%, or slightly under its long-run historical record of 10%.

The prediction is heartening, given the losses of late, although it's important to remember that even the Gordon equation can only guess at what's coming. It might be an intelligent guess, but a guess nonetheless. Even so, the model's general trend through time is arguably more revealing than any particular numerical prediction at a given moment.

Take another look at Figure 1 and you'll see that the equation's forecast previously peaked in the early 1980s, which was another rough time for stocks and the economy. For the next two decades, a great bull market prevailed. As it unfolded, the outlook for equity returns fell, slowly but consistently, right up until the early 21st century. It's no accident that the Gordon equation's peak in the early 1980s and its trough in the early 2000s look like bookends for the intervening 20-year run-up. In short, higher prices generally reduce expected return, and lower prices boost the performance outlook.

We can debate exactly what the stock market will deliver in the future, but the Gordon equation's unambiguous message is that prospective returns for equities look quite a bit better now than they did earlier in the decade.

Then again, we have to be careful about forecasting the future without understanding the details of the model's inputs. Although the Gordon equation is a robust forecasting model, there's still ample room for interpretation-as well as mischief.

That's a nice way of saying that the model's predictions are only as good as the underlying assumptions.

With that in mind, let's consider the assumptions behind Figure 1. First, we're using the current dividend yield at last year's close, which was roughly 3.2%. According to the Gordon equation's forecast, we're then adding the current yield to the long-run growth rate in dividends. Here's where subjectivity plays a role: For our chart, we're using the rolling 10-year trailing dividend growth rate. But you can find any historical growth rate you want if you look. Consider these three figures for three different time periods:

6.2% (1945-2008)
5.7% (1979-2008)
3.5% (1877-2008)

Depending on which rate you choose, adding these numbers to the 2008-year-end yield of 3.2% produces a stock market return forecast ranging from 6.7% to 9.4%.

Perhaps the only thing we can say for sure about the Gordon equation's outlook these days is that it's predicting higher equity returns for the years ahead than it was in the late 1990s and early 2000s. Even so, there's no guarantee that the current prediction will prove accurate. Then again, there was no guarantee either in the early 1980s, when the Gordon equation was predicting strong returns; or in the late 1990s and early 2000s, when the performance outlook for stocks looked unusually meager. As it turned out, those general forecasts look pretty good in hindsight.

James Picerno is editor of The Beta Investment Report (BetaInvestment.com)