Critics of the Fed Model

Yesterday' we took note of the Bloomberg story, Cheapest Stocks in Two Decades Signal Bull Market, which quoted several large and successful fund managers.  This story stimulated a post from Barry Ritholtz at the Big Picture, where he said that the Bloomberg story was based upon the “flawed Fed Model.”

This is a timely occasion to summarize our series on valuation, including some of the prominent criticisms.

Our Perspective

Our readers should understand that like them, we are consumers of models.  My own model-developing days ended about ten years ago.  (At least for stocks.  I still dabble in "recreational" modeling.)  Since then I have tested and reviewed many suggested stock market models, and provided feedback for my partner, Vince.

We did not develop the Fed Model, nor have we published any academic reviews.  We have no allegiance to those who created it.  We are perfectly willing to abandon it in favor of some other approach if we could find one that was better.

Some have suggested that the Fed Model is used by bullish pundits and managers.  We find the causation in this statement to be backwards.  The Fed Model helped us and our investors to avoid losses in the bubble era.  Methods precede conclusions, at least for those who have intellectual integrity.

Briefly put, we are just like the average reader seeking methods to improve investment performance.  The only difference is that we bring to the job more than thirty years of experience in developing and evaluating models of all types.  Also, unlike many of those citing critics of the Fed Model, we have actually read the papers and articles involved.

Our work at "A Dash" is not intended as an article in an economic journal, so we are not going to initiate some scholarly debate that few in our audience will understand.  Our conclusions and reasons are simply and clearly stated.  We intend our work to benefit individual investors and some thoughtful traders.

Background

As background, let us say what it is that we like most about the Fed Model.

• It captures the right level of complexity.  It is much, much better than valuation models that look at trailing earnings and/or ignore interest rates.  It is superior to models that add more variables without much more explanatory power.
• The model is based upon a plausible process.  The investor in a particular stock looks at expected earnings for the company and compares that return with the return of a bond.  The summation of millions of such decisions for the 500 S&P stocks is that the equity market as a whole has an expected asset return of bonds.

No other approach shares these virtues.  Anyone with experience in building systems and models will appreciate why these elements are important.

The Criticisms

First, from Barry Ritholtz's column we have the following.

The Fed Model is Imperfect.  This criticism comes from two sources:

1. Those who have never developed a model.  If they had, they would realize that all models are imperfect.  If you read such a criticism, you might ask whether the critic can do any better?
2. Those who are trying to write a journal article or book.  The relevant knowledge quickly gets to the point of practical value and beyond.  A good test is to ask whether a proposed "tweak" to a model really corresponds to the reality of the investment process.

The Fed Model "double counts" interest rates. The idea is that low interest rates are good for corporate profits and also suggest that stock P/E ratios should be higher.

We do not see the point.  The investor has a choice between two assets.  If interest rates rise, the gap narrows.  Maybe the narrowing occurs more quickly if rates rise.  So what?

The Fed Model "assumes" facts about earnings and interest rates.  Barry writes as follows:

The biggest problem with the so-called Fed model is that its built on two assumptions: 1) That profits will stay high, despite being a cyclical peak and decellerating; and 2) that interest rates will stay low.

This is quite incorrect.  The model makes no assumptions.  It takes current data about earnings projections and interest rates.  If earnings fall or interest rates rise, those following the model should adjust their behavior.

It is Barry who is making the assumption.  He believes that earnings are at a cyclical peak, a question that is very much in doubt.  He believes that interest rates are going higher.  Unlike the model, which is based upon current (and forward-looking) data, Barry is saying that he knows better than all of the analysts doing earnings estimates and better than the deep and liquid bond market.  Wow!

One might wish to note that earnings estimates have beaten the published expectation for several consecutive years, not missing a quarter.  In fact, the standard Wall Street line is that companies have attempted to lower the bar so that they can beat estimates.  If this is true, why should we believe that the sum of these forward estimates is overstated?

Valuation models are poor timing tools.  We agree!  We see valuation as a gauge of long-term sentiment.  Sentiment was euphoric in the bubble era, clearly identified by the model.  The current era shows that sentiment is very negative.  Investors have clearly accepted many of the bearish arguments about recession chances, housing problems, oil prices, the Fed, and a variety of misleading charts and anecdotal evidence.

The current valuation gap is useful in showing how much negativity is already reflected in the market.  The risks to earnings and interest rates are not the private knowledge of bloggers and bearish pundits.  Their widespread public recognition has had a major impact on mainstream thinking.  What if things are not as bad as they suggest?  The Fed Model shows what can happen in a return to normalcy -- not roaring good times, just normalcy.

A valuation model imposes discipline on one's thinking.  It provides a way of measuring negative effects.  It is not only the Fed Model.  Various other approaches that include forward earnings and interest rates give similar results.  Without the discipline of a quantitative model, one is free to speculate about each data point.  This is useful for those writing a daily blog, who want freedom to cherry-pick evidence.  It is less useful for understanding what is already "baked in" the market.

The Wall Street Journal Article

Barry cites a breezy two-year old article where Jonathan Clements interviewed some Fed Model critics.  Let us consider those criticisms.

The Fed Model uses operating earnings, not actual earnings.  Right!  And correct for modeling!  Investors making individual stock decisions look forward.  They tend to dismiss one-time events.  Those paying attention have noted that the serial "one-time" charges of the 2000 era are less prevalent.  It is done on a stock-by-stock basis, and the result is a market of stocks.

Bond returns are known and stock returns are not.  This statement is not correct if the investor plans to rebalance asset allocations every year or two, as the model changes.  It is possible to have major capital losses in bonds if interest rates increase.  Meanwhile, the return from stocks has more volatility, but also more upside.  This occurs both with higher inflation or with better than forecast economic conditions.  When Ed Yardeni studied this problem in 2003, he chose not to include a special risk premium for stocks.  It was not that there was no risk -- just that the risk/reward calculation did not favor bonds.  The available data supports his conclusion.  One might also note that the earnings return from stocks was dramatically under-estimated at the time Clements wrote the article.  Maybe it is time for the critics to freshen their data sets.

Summary of Our Work

We believe that an individual investor could spend some time wisely by reviewing our rather extensive material on this topic.

We showed the Fed Model in three time frames.  Look at the data and make your own decision.

We showed why valuation models revealed long-term sentiment, and the factors behind the current cycle of negativity.

We showed the importance of forward earnings.  Many of the critics of the model do not really use forward earnings.  They use some assumed trend, an approach that is frequently erroneous.

We discussed the merits of "tweaking models" like this one, and the pitfalls in trying to over-fit data to recent conditions.

We showed how a leading analyst from a bulge-bracket firm could get it completely wrong by trying to tweak the Fed model.

Finally, we summarized what this all means for the current market.  Understanding this topic shows why the current market is so resilient to negative factors, and why there is so much upside potential.  It is the kindling for what Gary D. Smith has called the "mother of all short squeezes."

There are some other specific criticisms of the Fed Model, including our own.  We shall revisit the topic for these questions, but this summary article captures the key points for an investor who is alert to opportunity.

Comments

[carmelus]

Why so much noise for a model that is useless as it is comparing two assets that are overvalued. US bonds are too expensive considering that us economy resembles to that of an emerging market and us stocks are at the peak of a cycle with unsustainable earnings. Besides one should compare dividend yields vs. yield bonds, not earnings. That model is senseless.

[RB]

I believe CXO compares earnings yield with the inflation rate, no adjustment to earnings yield.
http://www.cxoadvisory.com/REY-details/

[Bill a.k.a. NO DooDahs!]

A word or several about "real" interest rates. Let's not confuse the cause with the effect. Inflation is a monetary phenomenon, and price index increases are *an* effect of inflation, and not "inflation." Of course, now we need to spend hours debating what exactly is "money." Easy Al doesn't even know. Is credit "money?" Well, if I'm competing in the market to buy a house, it sure the frick is. Does (did!) M3 capture all credit? I don't think so. Don't forget that money supply, whatever "money" is, is increasingly global.

I do think it's possible that the Fed model might be improved if interest rates were calculated relative to something that proxied for the negative opportunity cost of holding fiat currency, but I will argue about what that proxy should be.

;-0)

Also, I haven't checked the algebra on this, but if the adjustment is made to both earnings yield and interest rates, does it become irrelevant?

[oldprof]

RB -

Thanks for your usual thoughtful comments and the links. I invite readers to check out the various alternative models. I have tried to write -- in general terms -- why I prefer the simple approach posted here to some of the alternatives. If you apply the criteria I have suggested, you will see some problems.

CXO has several different approaches. Their research is highly professional, and the supporting evidence is generally sound. It is a good source for finding various viewpoints on the Fed Model.

A drawback of a blog like this one is that any research review requires many hours of preparation and careful writing. We do not charge in blindly. We expect to have our errors pointed out.

To summarize, I am not going to do a "partial" review of individual models. If readers have a favorite alternative to the Fed Model, I might put it on the (growing) list of future topics.

Thanks again.

Jeff

[oldprof]

Thanks to Barry for stopping by and commenting, and to those helping to respond. Barry's point seemed to deserve more space, so I featured it in tonight's post.

Even though the conclusion is favorable for the Fed Model, I agree with RW that this is more of a strategic tool.

[RB]

Based on this approach using standard deviation bands, I do not see the Fed model failing in 1982 and which also indicates that stocks are currently significantly undervalued.
http://www.cxoadvisory.com/blog/external/blog1-25-07/

Based on a model using real earnings yield using the Fed's favorite core inflation number, the market today is slightly overvalued.
http://www.cxoadvisory.com/status/

Based on cyclically adjusted measures, the market is currently 35% overvalued.
http://www.smithers.co.uk/page.php?id=33

At this point, the real experts should step in.

[RW]

Gian Gravina once famously remarked that "[T]he bore deprives you of solitude without offering companionship." The online equivalent being one who intrudes upon a conversation and is neither constructive nor civil.

I'm sure the host here will have something substantive and useful to offer and, should a real conversation occur we may all profit.

From my own perspective the Fed model is less a predictor (even though it does use forward-looking data) and more of a stance, a "what has greater likelihood of profit now" matter, at least when it comes to allocation between stocks and bonds (the investment universe is much larger than that of course so further work will need to be done naturally).

[Nova Law]

Barry, interest rates are not historically very low, not when you adjust for the current rate of inflation. 10 years ago interest rates were higher because inflation was higher.

Oh wait, let me do your rejoinder for you to save you time - "inflation is higher now because the Administration is lying about the "real" inflation number, while 10 years ago they did no such thing."

That Damn Bush. He's responsible for everything bad in the world, including that bad case of athlete's foot I got last week.

[Barry Ritholtz]

As I specifically quoted in the post, according to the Fed Model, in 1981 stocks with depressed earnings and sky high interest rates were crazy expensive -- just as the greatest bull market in US history began.

That is a VERY significant flaw in the Fed Model. Missing what may very well be the greatest buying opportunity in a generation raises some very serious questions about any model that proclaims to be able to determine valuation.

In the pre-1982 version, interest rates were inordinately high and profits inordinately low, thus stocks looked extremely expensive. A bit of mean reversion on both of those, and stock valuations suddenly became much more reasonable.

Today, we have the exact opposite -- the 10-year interest rate level is still historically very low, while profits are very high. A little mean reversion on each and sudddenly stocks will look pricey.