I recently outlined two models for estimating the US stock market’s return for the decade ahead. Let’s add a third model to the mix with the plan to take the average as a relatively robust forecast.
The previous two models (see here and here) used valuation to estimate ex ante performance for the S&P 500 Index. One used Professor Robert Shiller’s Cyclically Adjusted Price Earnings Ratio (CAPE), a widely monitored valuation measure of the US equity market. The second used Excess CAPE Yield (ECY), also defined by Professor Robert Shiller. The idea is to adjust the equity market valuation metric — CAPE — based on interest rates.
Today’s update introduces a third model, this time using a parsimonious set of economic and financial data to forecast equity market performance for the next ten years:
* US economic activity (gross domestic product)
* Inflation (headline consumer price index)
* M1 money supply growth
* Yield spread: Baa corporate bonds less US 10-year Treasury rate
This modeling approach, along with similar profiles, are widely used as a first approximation for estimating future equity performance. The reasoning is that fluctuations in the economic trend, inflation, monetary liquidity, the credit spreads influence stock market performance. As one spin on the modeling approach, Alpha Growth Capital recently outlined the case for using these four inputs to model the S&P 500. For convenience, we’ll call this model the 4-Factor S&P 500 (SP500-4).
For this example, the data starts in 1986. Running the numbers on the basis of using 10-year annualized market returns, GDP changes, inflation and M1 money supply growth, along with the current yield spread, generates the following results:
One quick observation: the low p-values in the far-right column suggest a statistically significant fit for the four variables. The relatively high multiple R-squared print is also encouraging for thinking that the modeling is moderately useful.
In comparison with the two previous models, SP500-4 estimates a substantially higher 10-year return. The point forecast is a stellar 15.2% annualized performance. That compares with around 1% and 7% for each of the other two models. The average of the three is in the high-7% range.
Clearly, SP500-4 is capturing economic and market dynamics that aren’t factored into the other two models. Which model is correct, or at least closer to reality? Unclear.
What we do know is that every model is wrong, but some are useful. Meanwhile, a long history of analysis shows that more models tend to be better for divining the future by way of combination forecasting. That’s partly because using a diversified set of models, each making different assumptions based on different data sets, is useful for minimizing error compared with any one model.
This is hardly the last word on projecting equity market returns, but it’s a reasonable start. The good news: there are other models to consider that can help to further refine the forecast.
Learn To Use R For Portfolio Analysis
Quantitative Investment Portfolio Analytics In R:
An Introduction To R For Modeling Portfolio Risk and Return
By James Picerno
