Multi-Asset / Wealth Management

Making Forecasts Reliable—and Useful

By Seth J. Masters February 06, 2012

If there’s one thing you know about capital-market forecasts, it’s that they’re usually wrong. So why bother forecasting—or paying attention to forecasts?

Investors have become inured to market projections that don’t pan out. And no one can consistently make accurate forecasts of asset-class returns. But wouldn’t it be nice to have a capital-markets model that provided viable projections of the likely range of returns?  And which dimensioned how bad an adverse environment would be?  A planning model incorporating that level of rigor could help investors get on—and stay on—the path toward reaching their long-term financial goals.  

To be successful, such a forecasting tool would have to factor in the lessons of history, without assuming that history will be repeated. Looking only in the rearview mirror is not the best way to see ahead.

To avoid extrapolating history, some forecasting models adopt a “Monte Carlo” approach, which randomly generates data for thousands of paths. The problem with most such “Monte Carlo” models, however, is that randomizers can create paths with unlikely combinations of events, such as high inflation and low interest rates.

We think a more robust approach is to base a Monte Carlo model upon the various drivers of asset-class returns. Inflation, for instance, is a driver of interest rates, which in turn affect bond yields and corporate profits, and hence fixed-income and equity returns. Thus, the fixed-income and equity return forecasts on any given path would always be consistent with the inflation and interest-rate assumptions on that particular path.

A plausible model should also pay careful attention to plausible sequences of returns, in our view. Bond yields, for example, are unlikely to jump from 1% to 10% in any given year along any one path.

Similarly, economic and market conditions at the time you’re making the forecast matter. Clearly, the long-term outlook for US Treasury bonds must be different when 10-year yields are under 2% (as they are now), from the outlook when they hit almost 16% (as they did in 1981).

“Normal” expectations should also be built into a forecasting model, since the drivers of asset returns tend to regress to the mean over time. But the model should embrace uncertainty, factoring in a degree of randomness. Markets can deviate from “normal” for long periods of time.

At AllianceBernstein, we’ve built a “Capital Markets Engine” based on these precepts.  Rather than produce single-point forecasts, it’s designed to project returns over various time periods for individual asset classes—and combinations of asset classes (or asset allocations). That approach allows an investor to choose the asset allocation most likely to meet his/her particular goals, circumstances and risk tolerance.

For example, consider three allocations over a 10-year period going forward: 100% global equities, a 60%/40% mix of global equities and global bonds, and 100% bonds. The Display below arrays our forecasted compound returns, ranging from market environments so good we’d expect to do better only 5% of the time (the top of the line), to scenarios so bad we’d expect to do better 95% of the time. The labeled returns correspond to our 10th, 50th, and 90th percentile outcomes. Not surprisingly, the range of uncertainty between very bad and very good outcomes shrinks as you reduce risk—but so does the median result.

We Project Our Forecasts as a Range of Outcomes

We believe that for most investors, focusing on median outcomes is often a mistake: By definition, the median carries a 50% chance it’s too high and a 50% chance it’s too low. If you’re planning your future financial security, you want to be very sure, not 50% sure, that you won’t run out of money.

In fact, a cautious investor might want a plan that worked even in a 90th -percentile outcome, which he/she would beat 90% of the time. Spelling out the inevitable risk/return trade-offs of any allocation is one of the benefits of our robust wealth-forecasting tools.   

The views expressed herein do not constitute research, investment advice or trade recommendations and do not necessarily represent the views of all AllianceBernstein portfolio-management teams.

Making Forecasts Reliable—and Useful
Back to a top