A Do-It-Yourself Forecasting Kit Updated

Forecasters who make predictions without regard to past experience have no benchmarks to distinguish between what is radically different about their expectations and what experience has already taught us. Although the future will never replicate the past, analysis of long historical periods is the basic tool for forecasters. In this article, comparison of a set of major U.S. macroeconomic variables and capital market returns over a sequence of 20-year periods (beginning with 1873–1893) highlights that inflation is critical. Real, inflation-adjusted data often bear only a vague resemblance to the nominal data from which they were derived. Forecasting the long-run future is an inherently perilous activity. Forecasts based on an extrapolation of the past are at the outside limits of peril. We all know that. But forecasters who make predictions without regard to past experience are simply playing games. Without any reference points, they have no benchmarks to distinguish between what is radically different about their expectations and what experience has already taught us. Those insights are the essential building blocks of a rational forecast.In this article, I compare a set of major U.S. macroeconomic variables and capital market returns over a sequence of 20-year periods beginning with 1873–1893 and ending with 1983–2003—which provides a total of 23 eras, of which 22 contain a 15-year overlap with their predecessors. This sequence of periods spans big differences in both fundamental economic data and capital market data.The starting point of analysis is to remember that the data that matter the most are the real data. When you think about the future, you have to think about inflation before anything else. The real data often bear only a vague resemblance to the nominal data from which they were derived.In the data, variability in real bond yields, with the standard deviation about double the mean, is startling. But the biggest surprise is in the long-run relationship between inflation and equity returns. Stocks have done better when inflation was high than when it was low. Stocks really have been a hedge against inflation—over the long run. The three periods with the highest rates of inflation were those beginning with 1963, with 1968, and with 1973. In those periods, inflation ran higher than 6 percent but the stock market averaged 9.2 percent nominal return a year. Furthermore, the coefficient of variation for equity returns over 20-year spans, real as well as nominal, is no greater than it is for real GDP. The primary reason is that the average return on equities over the long run has been such a big absolute number.The intimate link between money growth and inflation is clearly as expected: The directional movements in money and inflation have been almost precisely in step over periods as long as 20 years. It follows that money growth and stock returns have also tended to move in the same direction, although the relationship has been less consistent than with bonds.Two interrelated dangers await investors who are excessively fond of evoking the long run as a basis for projecting the future. First, the long run smoothes the data by averaging out the wild volatility we experience in the short run. One can be mesmerized by the smoothing (or “soothing”) process of the long run. Second, a powerful feature of past experience has recently disappeared from the scene: An average dividend yield of 5.1 percent with a standard deviation of only 1.3 percent throughout the period from 1872 right through to 1996 overshadows the 9.2 percent mean return of the 20-year periods.How should one use these statistics in framing a long-term forecast? Very gently. Few observations actually fall close to the averages. Nominal return on equity for only two periods are within 100 bps of the average. Average or below-average returns over the next 20 years are not in any way beyond normal expectations—but so are spectacular results.Perhaps the most important advice for considering the future is to remember that 20 years is a long time. Those who last that long into the future will get there one year at a time, one month at a time, one moment at a time. Keeping the 20-year outlook in view will be difficult, except perhaps during those brief quiet moments that come our way on widely separated occasions or, perhaps more often, in the middle of the night.