The August CFP CE quiz is now available in the Members Section!

As the popularity of tactical asset allocation and using market valuation to inform investment decisions rises, so too do the criticisms to such methodologies. In the long run, this is part of a healthy dialogue that shapes the ongoing evolution of how we invest. But much of the recent criticism to being tactical in particular seems to suggest that if we can't get the timing exactly right, or calculate a valuation that works precisely to predict returns in all environments, that it should be rejected. In reality, though, even just participating in a few booms, or avoiding a handle of extreme busts, can still create significant long-term benefits for achieving client goals. Which raises the question - if we're really focused on the long term for clients, are we expecting too much from market valuation in the short term?

The inspiration for today's blog post comes from an article recently sent to me by another planner. The article, from the AAII Journal, was entitled "A Cautionary Note About Robert Shiller's CAPE" by Stephen Wilcox, and discussed a series of criticisms to the increasingly popular use of Shiller P/E10 (a P/E ratio where the denominator uses a 10-year trailing average of inflation-adjusted as-reported earnings, also known as the Cyclically Adjusted Price/Earnings ratio {CAPE}). Given our recent article in the Journal of Financial Planning on using market valuation to implement tactical asset allocation, he was curious for my thoughts about the AAII article's criticism of CAPE.

The AAII article went into some good depth regarding challenges in how P/E10 is calculated. For instance, earnings are adjusted for inflation so that the P/E ratio will be measured using real earnings, but the methodology the government uses to calculate inflation has changed over the recent decades. Consequently, today's inflation-adjusted earnings may not quite be analogous to real earnings in more distant historical time periods. In addition, accounting standards have changed over time - especially relative to how earnings and balance sheets were reported in the late 1800s when Shiller's time series begins - so our calculation of the P/E ratio over time has not been entirely consistent with respect to the earnings (which is further exacerbated by the fact that corporations were not taxed for over 1/3rd of Shiller's historical data, and it's entirely possible and plausible that taxation has changed some of the incentives for whether/how companies report earnings over time). Furthermore, Shiller's P/E10 uses, as the name implies, a 10-year average of inflation-adjusted earnings, which is done to smooth out the fluctuations of the business cycle. However, the average business cycle is closer to 6 years than 10 years; consequently, Shiller's P/E10 may be measuring something closer to 1.5 business cycles than just 1, which could cause it to become somewhat distorted over time (as it will be tiled in one direction in the midst of a boom-bust-boom 1.5 cycles, but another if it's in the bust-boom-bust phase).

So what's the outcome of all these criticisms? It seems the primary implication is that the long-term historical average of P/E10 - which we use to evaluate if the market is over- or under-valued - might not really be the 16.4 that the current historical data implies. It might be a bit higher or lower. In addition, we might arrive at a different conclusion regarding the P/E10 in today's marketplace if we calculate it differently - for instance, using 6 years of trailing earnings instead of 10. In fact, the AAII article notes that if we calculated P/E6 (i.e., using 6 years of trailing real earnings) as being representative of cyclically-adjusted price-earnings (CAPE) instead of P/E10, the long term average P/E would be 15.78 (instead of 16.41), and the current valuation (as of July 2011 when the article was being written) would be 21.26 instead of 23.35. The end result of this adjustment alone? Instead of suggesting that the market is 42.3% overvalued (using P/E10), it would only be 34.7% overvalued using P/E6.

To which all I can reply is... who cares!? Even with the author’s own adjustments, the CAPE is screaming that the market is overvalued. Perhaps after the market falls 35%, where the author’s CAPE suggests the market is “fair value” and Shiller’s P/E10-based CAPE suggests the market is still 7.6% overvalued, we can debate whether the market really has to fall another 7% (after the first 35%), or if it’s done. Either way, both methodologies “predict” at least a nearly 35% overvaluation, implying severely substandard returns in the coming years! Similarly, while the author suggests that the long-term historical average for CAPE might be adjusted for changes in inflation or accounting standards, nothing in the author’s discussion ever suggests that CAPE could be off by anywhere near 42.3%. In other words, depending on what adjustments you want to apply, maybe the market is 42% overvalued, or maybe it’s 35%, or maybe it’s 30%, or maybe there’s an unforeseen factor in the other direction and it’s 50%. Regardless, all of these results imply huge overvalued and the elevated risk of a severe market decline; do we really care who's most precisely right in the midst of such a meager prediction of forward returns. Are we really going to debate the last 7% of a 42% market decline? Are we completing missing a raging forest fire by focusing on the health of the individual trees?

So the bottom line is that while CAPE may have some minor flaws that make a pinpoint precision of valuation difficult, nothing the AAII author writes suggests that it is in any way incorrect at such extremes as 40%+ overvalued. In point of fact, that is the primary reason why we only focused on making tactical trades to adjust equity exposure at (good or bad) extreme valuation levels in our recent Journal article.

In other words, the purpose of using valuation is not to debate the last few percentage points of potential market movement in an effort to pinpoint the precise expected return and potential risk of equities to the last decimal place. The purpose is to know when stocks are priced for screaming gains that we can participate in, and to know when stocks are priced for screaming losses that we can reduce our exposure to. The rest of the time, perhaps, we can just go along for the ride and take what the market gives us (although when such a valuation approach is applied across all asset classes, inevitably there are always undervalued opportunities to capitalize upon and/or overvalued pitfalls to avoid). Even just participating in the rare opportunities and avoiding the major pitfalls have a significant enhancement to long-term returns and risk, not to mention the success in achieving client goals.

So what do you think? Is valuation-informed investing only relevant at market extremes? If it IS only relevant at market extremes, is that still acceptable? Does valuation have to be "right" more often, or are we expecting too much? Are we missing the forest for the trees?

  • Dick Purcell

    Michael, I think the PE-indicator described in your article is valuable, and agree with your point in this blog about forest v. trees.

    But I’d like to suggest a refinement in how to respond to high or low PE indicator. Rather than go directly from PE indicator to allocation, I’d prefer to respond to high or low PE indicator by adjusting the “expected” stocks return rate down or up for the next X years, then feed that into simulation to see what it means for my particular plan and goals.

    I like this approach for two reasons:

    1. For the world at large, higher stocks “expected” calls for more stock allocation and vice versa — but for a particular plan and goals it can work the other way. If the stocks “expected” is lower, I may need more stocks for the growth required for best prospects for my goals. And similarly, if stocks “expected” is higher, I may be able to pursue best prospects for my goals with less of stocks risk.

    2. From the PE-indicator effect spotlighted in your article, the resulting rise or decline in stocks “expected” for the years ahead may mean it’s best to adjust not only allocation but also my cash flow plan. It may be that the change called for in cash flow is more important than that for allocation.

    If I feed an adjustment of stocks “expected” into my simulation, I can see what it means for my particular plan and goals, in both allocation and cash flow.

    Dick Purcell

    • Michael Kitces

      You have a valid point here. Strictly speaking, I would advocate the same – the valuation should influence the expected return you use in your own models, from determining your allocation to projecting retirement or other goals, and then let that expected return adjustment lead to its own conclusions.

      Given the kind of changes in expected return distributions we see, though, I’m skeptical that lower expected returns would lead to higher allocations (simply because it would likely violate a client risk tolerance constraint along the way), and instead would lead to a lower allocation and an adjustment of saving/spending goals.

      But I do concur with you that strictly speaking, the expected return should be adjusted in your models/projections/etc., and then draw whatever conclusions are appropriate on that basis.

      – Michael

  • Ken Faulkenberry

    I use a tactical asset allocation for the Arbor Asset Allocation Model Portfolio (AAAMP) and believe it is the only risk adverse strategy that makes sense. I don’t believe in using ANY one valuation factor though. They all have flaws.
    PE10 can be skewed by one year (such as 2008) that is far from the mean and has a low probability of being repeated. Net cash flow of companies is one of the best valuation methods in my opinion but should be considered with many other factors.
    I keep my net equity exposure between 25 and 65% 90% of the time. Some of it has to be intuitive. There is no secret formuala that always works. That’s why it is important, as a portfolio manager, to have sound valuation priciples and a diciplined methodology.

    • Michael Kitces

      In point of fact, we are similar internally at Pinnacle – we look at a number of valuation measures to implement a tactical process. There isn’t going to be one simple methodology that works universally (if it was quite THAT simple, it really would be arbitraged away quickly).

      Nonetheless, the underlying point of the post is that I continue to have trouble with the valuation criticism that suggests that if it can’t perfectly time every short-term turn in the market, it must be useless. Ironically, from that perspective, it actually feels like the critics are the ones focusing on short-term market timing, while the tactical folks are long-term oriented, even though they’re often characterized in the opposite manner!

      – Michael

  • Bill Baldwin

    Schiller’s PE10 concerns me on a few levels. First, its backward looking. If the current PE s 13 and forward PE 10.3, shouldn’t we analyze the quality of the earnings, rather than driving it up because the market was previously overvalued, or, as is the case, earnings went to nearly zero in ’09? Secondly, if you read his work and some of the analysis, there is an indication that you should only invest when the PE10 is below historic norms. I woulds suggest you might sit on the sidelines for many years, losing all your clients (or your nerve) during a period such as ’05 – ’08.

  • Kay Conheady

    Hi Micheal,
    I find your response persuasive.

    And, while I realize there simply isn’t enough data for us to draw statistically valid conclusions, as my 12/20/2012 AP/Dshort web article proposes, I do also think we need to be mindful of the long term *trends* of the P/E10, which Wilcox doesn’t touch on. Though things could go differently in the future, thus far we have never seen the P/E10 stop declining when it hits “fair value”. It seems to want to go to extreme lows of undervaluation before the next risk-adjusted worth-investing-in market cycle starts.


  • Joe Pitzl

    If you notice a very large man across the room, it doesn’t matter if he actually weighs 400 pounds or 350. He is overweight.

    That is an oversimplified example, but isn’t that part of the point here with the tactical disucssion? Whether we use PE10 or PE6 is irrelevant. Whether 16.4 or 14.4 is normal is irrelevant when the market is trading at 21.

    At the end of the day, whether we are talking about market valuations or monte carlo simulations, I’d rather be approximately right than precisely wrong.

    Stocks become less risky when they decline in value and they become more risky when they increase in value. Regardless of any change in standard deviation, that is a mathematical fact.

    But let’s not forget that standard deviation also includes the upside, so adjusting it downward or skewing it one direction when markets are undervalued makes no sense. Instead, if we make any adjustment based on market valuation, we ought to change the mean and revise it upward…and vise-versa.

    We all know monte carlo simulations are not perfect, but we also have to acknowledge that its effectiveness is completely a function of the assumptions. Mechanically monte carlo assumes markets are fairly distributed around a specified mean at all times.

    If we simply adjust the mean to a lower number when markets are overvalued, we can continue to simulate the normal distribution, which effectively shifts the range of possible returns more negatively. There is no need to skew the standard deviation (and this simulation is more realistic anyway).

    Without changing the mean, the monte carlo will always assume the market is fairly priced. And therein lies a problem with using it in conjunction with tactical portfolios.

    Valuation is the closest thing we have to gravity in financial markets. Eventually, prices revert to the true intrisic value, but the timing is impossible to gauge. It may take 5 weeks to revert or it may take 5 years, but inevitably, it will happen.

    The problem is that we aren’t dealing with rational systems here, we are dealing with human beings. The market can stay over / under valued much longer than a person (advisor or client) has patience for. As such, the most important element of a tactical approach is simply having the wherewithal (and having clients that have the wherewithal) to patiently wait for reversions to occur, regardless of whether you timed them perfectly.

    Joe Pitzl, CFP®
    Intelligent Financial Strategies

  • Steve Smith


    Presumably there are operational costs absorbed by a firm implementing a tactical overlay, over and above what it would cost the same firm to implement a simple strategic asset allocation strategy for its clients. And a firm that successfully adds value by doing this should be compensated for the application of this skill to the investment process.

    So approximately how much more should a client be charged for this benefit?

  • Kal Salama


    You bring up some excellent discussion topics in your initial blog. After reading it and the responses, I would like to make a few points on the use of valuation analysis and expected returns:


    Valuation should always use forward looking analysis and factor in where markets currently are and expected to go in their earnings, cash flow and dividend cycles.

    Starting point has an enormous influence on realized returns in risky assets and markets, and can produce easily up to 15 plus year returns that are significantly different from long-term expectations. Since valuation (or implied expected return) analysis has no information about the timing of convergence between price and value, it should be used very cautiously in the near term. As a prudent portfolio fiduciary, one should determine formally to what degree, if any, of a client’s long–term risk exposure will be adjusted in the near-term in response to market price/value discrepancies. The near-term adjusting of client risk exposure has the potential to both greatly benefit and greatly detract from client returns.

    As advisors recommending risky asset exposure to clients, we have a responsibility to prepare them for the potential near-term returns they may experience on the way to being compensated for the risks they take. This is true whether we are: buying and holding, rebalancing and adjusting exposure in response to valuation opportunities.

    Expected returns and distribution assumptions

    I would suggest to those using quantitative/simulation models and methods that, particularly given the experience of the last 15 years, they consider techniques that incorporate fat tailed distributions. As an industry, we need to do a better job explaining to clients that the normal distribution, common used to illustrate their expected range of potential outcomes, understates the likelihood of extreme market events.


  • Pingback:

  • Michael Kitces

    P/E10 is backward looking with respect to earnings, although it does look currently with respect to prices, so I wouldn’t entirely call it a backward looking measure overall.

    From a theoretical perspective, I would love to project markets using forward earnings, and not trailing earnings. From a practical perspective, though, I find it totally ineffective to do so, for the simple reason that work from folks like Yardeni has shown that forward analysts are pretty darn bad at accurately projecting forward earnings. Analyst forward projections exhibit numerous blatant behavioral economics heuristics, from the tendency to move with the herd (see how long your career lasts by boldly making a projection that’s materially different from the pack), to projecting the current environment as though it will last forever. Despite the fact that the business cycle produces at least one recession a decade, I don’t believe forward earnings analysts have EVER uniformly predicted a single earnings recession in advance, whereas even trailing earnings data has given warnings signs for at least some recessions and market declines.

    So the bottom line is that I simply don’t see any kind of consistent forward earnings data that actually WORKS, while I consistently see CAPE-style trailing earnings data that DOES work, albeit with some wide error bars that suggests you should only rely upon it at relative extremes (although as our recent Journal article showed, “just” making shifts at extremes still adds material value).

    That aside, I do fully acknowledge that at least when you “just” use something like P/E10, it may be months or even years before the market manifests the returns predicted by the valuation. It consistently works over 10-15+ year time horizons – which are imminently relevant to our clients – but I’ll grant that it creates pressure in the meantime.

    Nonetheless, if the reality is that the track record for predicting a decade of returns based on P/E10 is radically better than the track record for just predicting stocks will always earn their long-term average return, then I’d love to see more conversations focused on how best to get clients to stay the course IN LIGHT OF valuation-based return expectations that may deviate in the short term.

    Or as I’ve written recently (, if we are going to stress over getting clients out of the markets in overvalued, risky environments, let’s at least acknowledge that what we’re really concerned about is not the long-term success of our clients, but the short-term risk to our businesses. That is a real risk, too, but there are ways to handle it.

    – Michael

  • Wade Pfau


    I’ve been thinking more about your point… but something I’m concerned about for the Monte Carlo is what to do with the standard deviation assumption? It probably isn’t safe to assume a normal distribution in cases when valuations are at extremes. Michael (et al.)’s Table 1 from his article shows lower standard deviations when valuations are at extremes, but that is not something I’d be comfortable with including into the Monte Carlo. What do you think?

  • Michael Kitces

    At some point I’m hoping to get at least a blog post up to look at the distribution of those returns in the upper valuation range. I haven’t actually looked in detail at whether they’re still looking like a normal distribution, just shifted to the left with a tighter dispersion, or if something else is going on there as well.

    But offhand, I don’t think it’s necessarily counter-intuitive for the high valuation range to have a lower standard deviation, given that it ALSO has a dramatically lower mean in the first place. That still represents a far uglier distribution of prospective returns; in fact, it has a dramatically greater likelihood of producing losses than the average valuation normal distribution (even with the lower standard deviation).
    – Michael

  • Dick Purcell

    Wade and Michael–

    Interesting issues. But I would not either reduce or de-Normalize the Stocks return-rate distribution, for these reasons:

    1. We know its standard deviation is bigger than calculated from history, so I don’t want to do any reducing of it. That would imply I know more, when in fact I know less than those calculated measures.

    The reason I say bigger than calculated from history is: due to inadequate sample size, our calculated standard deviation has a margin of error – its own standard deviation, so to speak. That is additional uncertainty, suggesting a standard deviation bigger than calculated.

    2. For the extreme PE-indicator highs and lows spotlighted in Michael’s article, we have even less data. So about them, there is more uncertainty, suggesting even bigger standard deviation.

    3. About de-Normalizing, we have the same problem of inadequate data, especially for the high and low PE-indicator extremes. In the face of that I would consider the data adequate for refining the assumed nature of the distribution.

    4. For these multi-year plans, we have our good friend the Central Limit Theorem working with us, saying that for any amount invested for n years, as n grows, Ln(Result) approaches Normal and Result approaches Lognormal regardless of the single-year distribution you are addressing (except for extreme distributions).

    I think the dominant challenge is getting a “best” way of translating PE indicator into an adjustment for the all-important Stocks mean or “expected.” (a) How steep is that slope of “expected” v. PE indicator? (b) Based on what the PE indicator is right now, for how many years ahead do you apply the adjustment? (c) Should we also be considering “expected” for Bonds, since presumably that’s where we put the money taken out of stocks or get the money put into stocks?

    Dick Purcell

One Tweet / Trackback

Michael E. Kitces

I write about financial planning strategies and practice management ideas, and have created several businesses to help people implement them.

For ConsumersFor Advisors

Blog Updates by Email

Nerd’s Eye View Praise

@MichaelKitces Twitter

Out and About

Wednesday, September 9th, 2015

*Future of Financial Planning in the Digital Age *Modern Portfolio Theory 2.0 @ FPA San Diego

Thursday, September 17th, 2015

*Future of Financial Planning in the Digital Age *Social Media for Financial Planners *Understanding Longevity Annuities and their Potential Role in Retirement Income @ FPA Colorado

Monday, September 21st, 2015

*Cutting Edge Tax Planning Developments & Opportunities @ MetLife