The Fed Model Impels, but Does Not Compel

By now, most investors have at least a passing familiarity with the so-called Fed model, a valuation metric used to determine stocks' relative attractiveness to Treasuries. But does it work? What is its history as a market prognosticator? And is comparing stocks to Treasuries really a fair measure?

To recap, the Fed model compares the earnings yield (the inverse of the price-to-earnings ratio) of the S&P 500 to the yield of the benchmark 10-year Treasury note. The key to the model, as described in a 1997 monetary policy report to Congress prepared by the Fed, is its reliance on "consensus estimates of earnings over the coming 12 months." Forward earnings estimates, in other words.

Based on forecast 2003 operating earnings of $53.04 and Monday's closing price of 932.88, the S&P's earnings yield is 5.68% vs. the 10-year's yield of 4.18%. When the S&P's earnings yield exceeds the 10-year Treasury yield, shares are said to be undervalued, by about 27% at present, according to the Fed model. Yes, that's even after the market's big rally since Oct. 9.

Of course, one of the big criticisms of the Fed model is its reliance on analysts' projected earnings, which have proven to be notoriously overly optimistic, in some cases laughably so.

Furthermore, the value of operating earnings has been severely diminished by widespread abuses in the recent past, where companies were excluding as "extraordinary" any number of real expenses from their profit calculations, as detailed here.

On a GAAP, or reported, earnings basis, S&P 500 profits are forecast to be $39.50 next year, resulting in an earnings yield of 4.23%, meaning stocks are essentially fairly valued at current levels. Reported earnings sharply limit what "one-time" items companies are allowed to exclude from their bottom line. (Currently, there are no forecasts for so-called core earnings, an even stricter measure of profitability; still, the more stringent model means it likely would show shares to be sharply overvalued at current levels.)

Setting aside, for now, the debate over the worth of earnings projections and the value of operating earnings, fans of the Fed model say, quite simply: It works.

Reeling in the Years A history of the Fed model shows relative under- and overvaluation of stocks vs. the 10-year Treasury note

*Ratio of S&P 500 index to its fair value (12-month forward consensus expected operating earnings per share divided by the 10-year U.S. Treasury bond yield) minus 100. Monthly through March 1994, weekly after. Source: Prudential Securities, Thomson Financial

'Extremely' Valuable

"The Fed stock valuation model worked quite well in the past," wrote Edward Yardeni, chief investment strategist at Prudential Securities, in a report on the subject in August. "It identified when stock prices were excessively overvalued or undervalued, and likely to fall or rise," including:

Showing, with back-testing, the market was "extremely undervalued" from 1979 through 1982, a precursor to a major rally that lasted until the 1987 crash;

Reaching a then-record overvaluation peak in late summer of 1987, before the crash;