In this weekend’s Barron’s, Mike Santoli has a very simple criticism of the (so called) Fed Model for determining if equities are cheap or expensive (The Flaws in the Fed Model).

Stated simply, the “fabled stock-market predictor doesn’t work.

The idea here, once formalized as the “Fed Model,” is that stocks’ “earnings yield” (reported or forecast operating earnings for the S&P 500, divided by the index level) should tend to track the Treasury yield in some fashion. With this earnings yield now above 7%, based on a trailing price-to-earnings ratio near 13, this model and its various offshoots render equities a no-brainer buy. Or, if one prefers, that Treasuries are in a reason-defying bubble.

This simply doesn’t hold up in theory or practice. The Fed Model only “worked” as a predictor of market action in the 1980s and ’90s, when bond yields were steadily descending and stock values consistently rising as inflation and interest rates were slowly strangled. Both before that period and since, the Fed Model relationship has been mostly a non sequitur in terms of foretelling market performance.

For sure, the Fed Model is useless. But why it is useless is not well understood by most investors. Understanding this about the Fed Model or any other Wall Street hokum may help you look askance at other seemingly plausible, even persuasive arguments that actually fail.

Regarding the the Fed Model, it does not do what it claims to do. It does not tell an investor if stocks are cheap. Back in 2008, I made an attempt at explaining why: The key taker away is that it controls for two variables — not just one:

Note that the formula contains two variables: While it is commonly described as a way to evaluate when stocks are over- or under-valued, the other variable in the formula above is the forward S&P500 earnings consensus. SPX prices and the 10 year yield are the knowns, while BOTH valuation and forward earnings estimates are the unknowns.

Thus, the Fed model today might be telling you either of two things: When equities are undervalued — or when consensus earning estimates are simply too high.

-The Flawed Fed Valuation Model, February 5, 2008

Said differently, the Fed model assumes analysts consensus is accurate. That assumption has been the undoing of many an investor . . .



The Flaws in the Fed Model
Barro’s, June 9, 2012

Category: Investing, Valuation

Please use the comments to demonstrate your own ignorance, unfamiliarity with empirical data and lack of respect for scientific knowledge. Be sure to create straw men and argue against things I have neither said nor implied. If you could repeat previously discredited memes or steer the conversation into irrelevant, off topic discussions, it would be appreciated. Lastly, kindly forgo all civility in your discourse . . . you are, after all, anonymous.

18 Responses to “Why the “Fed Model” Doesn’t Work”

  1. GeorgeBurnsWasRight says:

    The Fed Model also ignores the emotional reaction of people in times of perceived economic crisis. As Mark Twain famously commented during one crisis, people become more concerned about the return OF their money than the return ON their money.

  2. dth20k says:

    Agreed — stocks trade more on psychology.

    Stocks offered a higher yield than bonds from the Great Depression through about 1958. This was considered normal, given the extra volatility associated with equities, the ease with which dividends on common stock could be cut, and common stocks low position in the capital structure in case of liquidation — not to mention the “death of equities” for an entire generation of investors (so widely touted recently).

    So the historical comparison surely matters. If the Great Depression (not the 1980s and 1990s) is the proper frame of reference, then we might not revert to bonds providing more income for another 25 years or so.

  3. Anyone who uses the Fed Model as a reason why “stocks appear cheap” immediately losses massive credibility. How it lives on is beyond me…..

  4. ToNYC says:

    If the Fed Model worked from 1913-2009 it was is an intellectual model-construct that received a ZIRP lobotomy in a desperate attempt to appear useful. In fact the AUSTERITY imposed on Seniors and indeed all saver’s lifetime accumulation of legal tender income destroyed the real, individual determinations of capital direction and thus any real solution. A “Dual mandate” controlled by damned fools is a race to the bottom.

  5. ilsm says:


    Federal reserve: arsonists running a fire department.

    While working for a government run insane asylum operated by the patients.

    Following a Philips curve to a pot of gold.

  6. constantnormal says:

    … the two “unknowns” (or estimates) are not exactly “independent variables” — they are so closely linked, they may as well be the same one … but the real sin here is in trying to predict the future from the past … as mediated by guesses about the future, from basically the same people working from the same assumptions and the same models of the markets/economy.

    Reality uses different models.

  7. carleric says:

    When about 80% of companies “beat earnings estimates” you have to realize that the game of anayst estimates is rigged. They have to low ball their best estimates to help generate sales (which is their real job). Any model based on just manipulated data is bound to be wrong all the tune.

  8. machinehead says:

    One can dispose of the ‘two variables’ problem by using a more-stable version of trailing earnings. Past peak earnings and a 10-year average of trailing earnings are two possibilities.

    The Fed used a forward earnings estimate in the model only because trailing 1-year earnings are quite volatile, and occasionally turn negative. Forward earnings estimates have never gone negative.

    Nevertheless, modified to have only one variable — stock valuation — the Fed model STILL doesn’t work. The root problem is that since bonds and stocks are only weakly correlated, neither constitutes a stable reference of value for the other.

  9. philipat says:

    But Santoli’s example uses the “As reported” (Inverse P/E) earnings as the example. Therefore, the consensus estimate is no longer a variable in his example. So can we, therefore, conclude that based on the present situation that either equities are cheap or Treasuries are way over-valued, having eliminated the consesnsus forecasr variables? 7% compared to 1.5% does seem like an inordinately large spread?

    The Fed Model actually seems to be less relevant when Interest rates are “Artificially lwo”, clearly the case now after miltiple QE and Twist operations. Further, I can’t see rates rising any time in the next 3-4 years, especially if, as seems likely, the Debt clock keeps on tickinhg higher because, if they did, debt repayments would become unsustainable. So I don’t see the Fed Model as being of much predictive value any time soon.

    Is there any Historical data on the Fed Model? It would be interesting to look at the relative disparities in relation to interest rates, particularly during the 1920′s and 1930″s.


    BR: Santoli discusses “earnings yield” as reported or forecast. Earnings yield as reported would tell if stocks WERE fairly valued for that prior quarter — not looking forward.

    The historical data is easily devised using historical S&P500 prices, Earnings, and 10 year yield — all of which are easily available. A chart of it is included at the link I provided. (Business insider discusses it further here)

  10. kaleberg says:

    Incomes and economic growth have flat-lined since the 80s, so people who want to invest have tended to buy stocks. Right now there is so much money sloshing around, and with incomes still flat or falling, no real investment opportunities, so I figure that stocks are the only game in town. You could buy private bonds and pray that the Fed doesn’t declare inflation at the first sign of growth. (“Inflation” in Fed speak is just economic growth in the form of rising wages.) Alternatively, you can lend to the government and pay them to hold your money. The Fed should really be issuing a lot more debt to meet the demand for it.

    Stock prices are what people are willing to pay. Saying stocks are under or over priced is rather meaningless. It’s like looking at the thermometer and saying the temperature is too high or too low for the weather. The only real decision is to buy or sell and at what price.

  11. gkm says:

    I’m not sure if folks understand how to value stocks but it’s a fairly straightforward thing. It is based on the expected future stream of cash flows discounted at an appropriate rate reflective of risk and opportunity cost.

    If stocks are undervalued as based on the variables of this equation as imputed by the analyst, then this is a divergence. You can either say 1) that stocks are a buy or 2) question the inputs of your model and conclude that based on the efficient market hypothesis something is indeed amiss because otherwise it should not exist naturally.

    My personal belief is that a divergence should exist because the imputed discount rate is not reflective of market forces. This is fact. The Fed is manipulating interest rates to make other assets more attractive. This doesn’t make them more attractive when you apply a correction factor ie if you believe things will normalize over time. It simply results in more short run volatility.

  12. boveri says:

    When is it that only the commenter Machinehead figured out that “One can dispose of the ‘two variables’ problem by using a more-stable version of trailing earnings?”

    Going one step further, there’s no need to throw out the entire baby because the relationship between the 10yr. rates and stock valuation based on earnings is not purely one to one. At the extreme we are seeing now, the Fed Model is telling us on one side that great economic uncertainty prevails (and for good reasons) and that therefore what appear to be great stock earnings are not to be trusted to last.

  13. SCTTD says:

    Back in the late 90s, I did some work around this model and concluded there might possibly be some level of relationship here but in the form of a curve, most likely a hyperbola. Played around with variables but could never quite find a formula for the curve that worked consistently. Santoli’s (and your conclusion) is faster and easier!

  14. kek says:

    The fed model has worked for me as one (of many) tools of relative and at extremes, absolute valuation. In 2000, the mania and bubble of S&P 5oo index investing gave a trailing earnings yield of 2.5%, with the 10 year at 6.6%. Looks like the last 12 years have brought us to a 180 from where we were in 2000. As equities were loved in 2000, they are hated today. 6.66% 10 year in in 2000, was the result of a buyers strike, 1.50% today, and they are lining up. The question today is can you believe in the earnings estimates to make the case today for the Fed Model. As there is labor slack in the economy, decreasing raw material costs, and ground zero RE markets showing signs of a bottom (AZ, Vegas & FL), earnings still look OK to me. Time will tell.

    Starting 10/9/2012 the 10 year numbers for equity investing (barring a meltdown) will show 75%ish Dow/S&P, 130-155% Naz/Russell trailing 10 year returns. Wondering if this will have a behavioral response from other investors/savers, as the two bubble deflator years get lopped off.


    BR: Sold to you

  15. philipat says:

    Barry, the Business Insider chart through 1997 shows a pretty good correlation. Again, it does seem that the Fed Model works fine in “Normal” conditions when the markets operate freely.


    BR: Fee free to invest YOUR money based on the Fed Model

  16. end game says:

    I never understood why anyone would use expected earnings, consensus or otherwise.

  17. philipat says:


    As I am trying to state, NOT in the present circumstances. I thought this was a discussion here? My point being that, historically, there is some evidence that the Fed Model AMY help. AMongst many other factors. BUT, when interest rates are artificially low, and/or during balance sheet recessions, it certainly does NOT work.

    My money? All in Real Estate in London, Singapore and Indonesia (No complaints so far) so no, will not be employing the Fed Model any time soon. Just keen to consider with an open mind. Thanks.

  18. toddtdf says:

    I wouldn’t say the Fed Model is entirely useless. There should be a correlation between earnings yields on stocks and treasuries. But, there are other assumptions that need to be looked at closely.

    I would disagree that market expectations are the undoing of the Fed Model. I think that’s a minor flaw.

    The Fed Model makes the interesting assumption that the risk premium and earnings growth rate cancel out. It’s definitely an approximation, but not the entirely unreasonable, for most situations.

    I think we find ourselves now in a market where the risk premium is higher than normal, and expected growth is lower than normal. So, the Fed Model doesn’t work quite as well now because of this.

    Another interesting alternative (capital structure substitution theory) is to use after tax corporate bond yields instead of 10 year treasury yields. I think this way you have a better adjustment for risk premium, as baa corporate bond yields will reflect a significant margin for risk.

    That still leaves the growth assumption ill-considered. I think economic growth assumptions are bleaker going forward than what we have experienced over the last few decades.

    So, I think the market is close to fairly valued, but I’ll defer to Barry with regard to which way the market will swing.