Is the Fed Model an Accurate Predictive Tool for This Market?

Over the past several weeks, there has been a blitz of media coverage surrounding the rise in Treasury yields and the implications for market returns. Although most of the coverage is qualitative, there are ways to quantify the relationship. One of the simplest methods is The Fed Model. As the name implies, it is rumored that this is one of the measures the Fed looks at, although it was officially named by Ed Yardeni. As with any simplistic model, it is contentious and should not be relied up exclusively to make market timing decisions.

In its simplest form, The Fed Model compares the forward earnings yield of the S&P500 to the yield on the 10 year Treasury Note. If the forward earnings yield of the S&P500 is greater than the yield on the 10 year Treasury Note then the market is implied to be undervalued by the percentage of the respective differential; the converse (overvalued) is of course calculated analogously. This gets us one step closer to quantifying the relationship, but not all the way. To more accurately quantify the relationship, let’s take a closer look at the differential between the forward earnings yield of the S&P500 and the 10 year Treasury Note yield versus the actual returns of the S&P500. This can be done by plotting the actual returns of the S&P500 versus the Expected Return as predicted by the Fed Model for multiple rolling periods and then creating a regression to obtain an equation that approximates a normalized Fed Model Return.

This gets us some pretty interesting results. At the beginning of 2007, the expected return of the S&P500 for the year, according to the normalized Fed Model, was 17.3%; it is currently 5.1%. Although this is still positive, the directional change, or velocity, and the acceleration of that change are things to worry about if the Fed Model offers any predictive powers. It should be noted that the correlation between actual returns and predicted returns of the normalized Fed Model is statistically significant. The Fed Model’s predictive ability tends to break down during bull markets, when valuation measures tend to be less useful and are replaced by growth and momentum indicators.

Interestingly enough, growth and momentum indicators are gaining strength as dominant forces over that last quarter. Most of the large quantitative research shops put out monthly letters detailing what is working among their factors. Over that last month, both Ford Equity Research and Sabrient Systems have made similar comments that their growth and momentum factors are outperforming the markets and their value factors are underperforming the markets. This same pattern can be seen in the Russell 2000, where the Russell 2000 Growth has returned 9.55% YTD and the Russell 2000 Value has returned 4.97% YTD. That’s right, growth has nearly doubled the performance of value this year. Based on these observations, one would assume that managers who have growth oriented models are likely to show stronger results thus far this year than value oriented managers.

One of two scenarios is likely to develop going into the second half of 2007. Either we are entering into a growth cycle where valuations are disregarded, such as in the late 90’s, and the Fed Model is not an accurate predictive tool (this would favor growth oriented managers), or we are headed for a period of slower market appreciation due to tension between earnings yields and Treasury yields, which may limit returns and cause investors to become more picky (thus favoring managers who have value oriented approaches). Current market volatility is likely a tug of war between these two forces.

Comments

[Will Rahal]

There is no doubt that the Earnings Yield and the (long-term)Treasury yield, are related.
The assumption however, that the ratio of EY and TY should equal one, is erroneous.

I have shown, that in an environment (as now) when the PPI is accelerating more rapidly than the CPI
the EY/TY should be greater than one and should have an upward directional bias.

[JohnnyBitchdog]

Alright, folks.....let's face some facts:

1. it is possible to regress the fed model results against market returns

2. presumably many people have done it

3. why are we chit chatting instead of simply providing regression results ?

John