Want to Measure Sentiment? Here is a New Idea

Market observers all want to measure sentiment.  Here is the general point of agreement:

When everyone is in the market, there is no one left to buy.

and of course we have the opposite —

When most are on the sidelines, there is no one left to sell.

The general interpretation of sentiment is contrarian.  The indicators include (but are certainly not limited to) the following:

  • Individual investors measures, like the American Association of Individual Investors (AAII);
  • The put/call ratio (interpreted as "public money" buying one or the other);
  • Insider transactions;
  • Who is featured on CNBC;
  • Magazine covers;
  • NYSE specialist positions; and
  • Various polls, including those of supposed experts.

I am sure that I have left out a few, but you get the idea.  Occasionally someone advertises a poll as seeking the opinion of "smart money" but everyone interprets it as contrarian anyway.  It is a tough crowd and no poll gets respect!

Current Readings

Most of the sentiment measures are not objective, especially the interpretation of CNBC guests.  The bears think that it is a parade of market cheerleaders.  The bulls see a bunch of worrywarts.  What is needed is something that you can actually measure.  It would be especially good if the measure had some predictive value.

Looking for a fresh idea?  Here are two new  candidates:

  • US versus GermanyCramer thinks that the market is "too negative" and points out that the S&P has declined more than Germany, despite less exposure.
  • Earnings reactions.  In case you were not watching, the earnings reports were excellent, beating expectations and with generally strong guidance. ( I understand that some feel this was already anticipated by stock prices).  Meanwhile, the results are very dramatic.  The excellent team at Bespoke Investment Group highlights in one of their typically excellent graphs that this was the absolute worst earnings season in a decade.  I am not going to copy their graph, but you should visit their site for a look.

 Both of these ideas are measurable and objective indicators of negative sentiment.

The Very Best Sentiment Measure

There is one indicator that is the absolute best for identifying sentiment.  It is the flash point between the bullish and bearish communities:  Forward Earnings.

For those who view the sell-side analyst community as a work force, helping us to identify trends in future earnings, it is a dramatic and appealing prospect.  The current yield on forward earnings is 7.85%, a rate close to that offered at the market lows of February to March of 2009.  (Brian Gilmartin of Trinity Asset Management covers this for his client base and for TheStreet.com's Real Money site.)

By this measure, the market is just as cheap as it was at the time of the 2009 lows.

The bears believe that analysts are captives of sell-side firms and the companies they cover.  It is a debate that can and should be resolved by data.

Some Comparisons

For those who want some data, I took up this topic many years ago, highlighting the error made by a prominent sell-side analyst.  This article (which I think is one of my best) is still worth reading.

Here (from Tom Armistead at Seeking Alpha)  is another approach, splitting the data set instead of using a quadratic term.   The basic idea is that if there is a great fear of deflation, the PE multiple for stocks is low.  If deflation is off of the table, stocks enjoy a multiple that competes with the higher bond yields.

Tom is getting at the same concept I found with a slightly different method.

Our Take

I think that the forward earnings yield is an excellent sentiment indicator.  If we actually achieve anything close to the projected earnings, the market is cheap — very cheap.

The conclusion is obvious.  For the moment, the marginal market participant is very skeptical about future earnings.

Some big players disagree.  Nick Thakore, manager of the $4 billion Putnam Voyager Fund has this take:

People have been so burned that they're missing the cyclical recovery
and an epic earnings recover.  It looks like 2011 could be an
all-time earnings record for the S&P 500.


It is all about earnings.  We know what to watch for and also how to place our bets.

You may also like


  • JohnF May 25, 2010  

    It’s not really surprising that Germany is down less than the USA. The lower Euro will return them to the number one spot of exporters, which is in a sense what Angela Merkel wanted.
    Of course, the problem is playing with fire. How much will it burn, baby?

  • Mike C May 25, 2010  

    The bears believe that analysts are captives of sell-side firms and the companies they cover. It is a debate that can and should be resolved by data.
    Is there a data set that shows otherwise?
    You’ve made this point before, about forward earnings estimates being a sentiment tool. I’m really trying to understand your point/perspective here but I am not getting it.
    Tell me where I am going wrong here. The conventional view is you’ve got fundamentals and valuation, technicals, and sentiment. Valuation is how are stocks priced relative to earnings, cash flow, dividends, book value, etc. Technicals is price action and price patterns. Sentiment refers to “The Crowd”. Are most market participants bullish or are they bearish? It would appear to me that forward earnings yield is a valuation metric, not a sentiment one.
    I mean forward earnings yields are the same today and yesterday as they were 4-5 weeks ago when the S&P hit 1220. By that metric, sentiment today is unchanged from 5 weeks ago, but doesn’t that seem off?
    5 weeks ago you had record call buying only matched on like 3 other occasions in the past several years which incidentally was a great marker the short-term top. Now you’ve got very elevated put-call ratios.
    Anyways, today looked like a classic reversal day and from my vantage point sentiment is very pessimistic:
    I added stocks aggressively yesterday and today, and might buy more tomorrow.
    FWIW, I’d love to hear your detailed bull thesis on MSFT. It does look very cheap here both on trailing and forward earnings with a rock solid balance sheet.

  • Jeff Miller May 25, 2010  

    Mike C — You have cited Paul’s post describing an evaluation of analyst forecasts.
    Did you read the work cited? If so, please explain all three charts.
    One of the problems on the Internet is a rather loose interpretation of data, so let us discuss this one — in detail.
    Thanks for pointing out something that everyone is talking about, and I await your analysis, especially of chart 3.

  • Mike C May 25, 2010  

    If so, please explain all three charts.
    I hope I get an A. ๐Ÿ™‚ Or at least pass. ๐Ÿ™‚
    Chart 1 appears to show persuasively that for the majority of years analyst forecasts start too high and gradually come down to meet what the actual turns out to be. In a few years, they seem to overshoot to the downside after initially being way too optimistic and then have to revise back up. One exception appears to be the 2003-2006 time period where they were too pessimistic. What is the anomaly here? Those 4 years or the other 20+. Perhaps not directly relevant to the chart interpretation but I think one can ponder that the 2003-2006 profit growth was an abnormal period of profit growth due to a generational credit bubble and profit margins above long-term sustainable levels. Looking at 2003, it appears analysts are too pessimistic at the turn coming out of a recession. Recent experience would appear to confirm that. This would seem to confirm that analyst estimates make the same mistake as many. They are too optimistic at the top and too pessimistic at the bottom. The market of course is a leading indicator. The market turned down in late 2007 well before it became clear just how bad profits would turn down and the market recovered in spring 2009 well before it became clear how robust profit recovery would be. The million dollar question is where are we at in the earnings cycle? I don’t know. If I knew with absolute certainty the SPX was going to do $100 in 2011 and $120 in 2012 I’d be margined to the max long stocks here. Are you absolutely sure about those numbers for those years? Or is it possible this earnings cycle could be a much shorter, compressed one?
    Chart 2 appears to show that analysts predict 10-12% EPS growth for 5 years and then must fall asleep as they appear to never change that forecast except for the late nineties, early 00s where they got even more optimistic. In contrast, actual earnings appear to be quite volatile on a 5-year basis bouncing between 0 and 14%. Interestingly, the average is 6-7% which is the number I’ve seen others use for long-term SPX profit growth. Should we value the market based on a 6% long-term growth rate? Or should we use a single year that could be above trend?
    Chart 3 seems to indicate that except for the late 90s bubble the actual market P/E tends to appropriately trade at a discount to the implied P/E from analysts forecast.
    Still hoping you’ll share your bull thesis on MSFT. I know you’ve mentioned the corporate PC upgrade cycle which makes a ton of sense to me since they’ve put this off.

  • Jeff Miller May 25, 2010  

    Mike C — Thanks for your help and your patience. I am still trying to figure this out.
    Chart 1 — if you look at forward predictions for one year ahead (what I usually do) these seem pretty close for nearly all years. You can try to guess when there will be a recession, of course.
    Chart 2. Do you really think that analysts “fall asleep”? What is the point of five-year forecasts? Who really believes that these are relevant? Or three-year forecasts?
    Chart 3. Please explain “the implied P/E from analysts forecast.” I have NO IDEA what this means, nor do I think that any readers do.
    Meanwhile, the real question is whether the forward earnings for one year out are better than trailing earnings. This research tells us nothing about that question, yet it is mindlessly quoted by many. I don’t think anyone even understands the charts.
    Mostly, you are just quoting what they say in the notes. If they had written a good article they would explain better and also provide access to data for other researchers to validate. I am also suspicious of their correlation analysis, but there is no way to check.
    Just my thoughts —

  • Mike C May 26, 2010  

    Chart 1 — if you look at forward predictions for one year ahead (what I usually do) these seem pretty close for nearly all years. You can try to guess when there will be a recession, of course.
    You lost me here. The curves, path, and slope of the green lines are different and the green lines show how analyst estimates changed as the year progressed. So in 2001 and 2002 estimates started out way too high and as the year progressed they kept ratcheting them down. Eventually, the estimates converge to the actual. 2005 and 2006 were the exact opposite with estimates starting way too low and the analysts playing catch-up to where the actual was.
    I’d prefer not to guess on anything. Hypothetically, let’s assume we were ominscient. Obviously, the correct action would be to dump stocks right before a profits downturn and buy them back before a profits recovery. The data clearly shows forward sell-side estimates will most certainly be wrong at turning points like the profit peak in 2006 or profit trough in 2009. I don’t know what the solution is but my objective is continual improvement and refinement of my process.
    Chart 2. Do you really think that analysts “fall asleep”? What is the point of five-year forecasts? Who really believes that these are relevant? Or three-year forecasts?
    I was being sarcastic. Well, standard finance theory 101 would have a 3-5 year forecast as being more relevant then a single year ahead. Maybe less accurate, but certainly more relevant. The present value of a stock, any stock is the discounted value of ALL future cash flows, not just the cash flows of 1-year ahead. Now we can argue the practical application and pragmatics, but that is the absolutely correct from a theoretical foundation. How do you value a bond? You don’t look at the interest payment one year ahead. You discount all the future interest payments. So the present value is going to change quite a bit if we plug in 14% growth or 7% growth or no growth.
    Chart 3. Please explain “the implied P/E from analysts forecast.” I have NO IDEA what this means, nor do I think that any readers do.
    Not sure what to say. They define it in their footnote quite clearly:
    1 P/E ratio based on 1-year-forward earnings-per-share (EPS) estimate and estimated value of S&P 500. Estimated value
    assumes: for first 5 years, EPS growth rate matches analystsโ€˜ estimates then drops smoothly over next 10 years
    to long-term continuing-value growth rate; continuing value based on growth rate of 6%; return on equity is 13.5%
    (long-term historical median for S&P 500), and cost of equity is 9.5% in all periods.
    They tell you what assumptions they are using to come up with the estimated value which as I stated above they are using the analyst’s own 5-year estimates and then discounting those earnings back to the present at a discount rate of 9.5% (cost of equity).
    Generally, I am not a fan of appeals to authority, and analysis needs to stand on its own merit. That said, this is McKinsey, and McKinsey is one of the top consulting firms in the world recruiting from top 5 MBA schools so odds are the guys who wrote this article and did the research are not dopes.
    Anyways, time for bed. Good exchange.
    P.S. Over/under bet. I’ll take the over on Dow 20K for 2018 ๐Ÿ™‚

  • Jeff Miller May 26, 2010  

    Mike C — I have made a careful note of your comments, and I am working on this question.
    We both know that the people and the firm are excellent, so that is not the question. Meanwhile, I can tell you stories of academic research where there were errors by top people, sometimes unnoticed. I have frequently found some pretty lightweight conclusions from highly respected Wall Street sources.
    In particular, the analysts who are the focus of the article often have similar qualifications and background to the McKinsey authors.
    I definitely do not agree with your interpretation, which seems merely to quote their footnote.
    Here is the test. If I gave you raw data, could you replicate their work. I bet that you could not.
    If not, their presentation (at the minimum) is lacking.
    There is a facile acceptance that this article proves something about data related to a one-year forward earnings forecast. The data in chart one seems to show that the forecast is good, especially as the time gets close.
    I urge you to take a closer look.