Friday, June 12, 2015

Behavioral Finance is the wrong framing for equities

Here is a new paper, on VoxEU (HT: EV).  A key section of the VoxEU summary:
Various studies, beginning with Shiller (1981) have concluded that the volatility in the stock market is too great to represent forecasts of future dividends or other measures of cash flows of corporations. As memorably described by Shiller, the stock market appears to exhibit ‘excess’ volatility, namely volatility that cannot be attributed to rational factors and rather reflects (in the words of Keynes) the ‘animal spirits’ of investors. 
Rare disaster models offer an alternative way to understand excess volatility. Rather than reflecting the day-to-day whims of investors, stock market fluctuations could reflect investors' changing views of the probability of a rare disaster. An increased probability of a disaster implies that future earnings are likely to be both lower and more risky. These effects combine to lower equity prices, even if a disaster itself does not take place. Thus, stock returns, which incorporate these probabilities, can be far more volatile than dividends or consumption, which reflect (primarily) the disaster itself.

I think this is a much more helpful way of imagining equity behavior than behavioral explanations of cognitive biases, fickle moods of fear and greed, and wildly fluctuating required return expectations.  I also happened to see this at an interesting blog called Spontaneous Finance, written by Julien Noizet (the post I am excerpting is a guest post by Justin Merrill.):
The natural rate of interest is equal to the return on assets for corporations. Most economists that try to model the natural rate mistakenly do it as the risk free rate or the policy rate. This is a misreading of Wicksell since he identified the “market rate” as the rate which banks charge for loans, and the important thing was the difference between the market rate and the natural rate.

This all corroborates with my intuition - to begin with the required return on corporate assets and to discount from that to get to low risk securities, instead of starting with a risk free rate and adding risk premiums.  As the Vox paper points out, there are many separate issues going on with equity valuations through a volatile episode, but, the net result appears to create a quite stable level of required returns on corporate assets.  We can model equities based on (1) earnings, (2) growth expectations, and (3) the discount rate.  If the sorts of risks about future changes in income in the Vox paper are manifest in growth expectations, equity valuations become kind of boring.

The discount rate appears to be quite stable over a long period of time - around 6-8%, in real terms, depending on the range of corporations included.  And, there appear to be countervailing influences on growth expectations through the business cycle.  There is a natural tendency for mean reversion, because equity owners are the residual claimants on national income, they tend to experience extreme income fluctuations through business corrections, which are basically disequilibrium episodes.  If the economy does recovery, that disequilibrium will dissipate, and corporate income will return to its natural level as a portion of national income (which is also very stable over time).  But, as Jerry Tsai and Jessica Wachter argue in the Vox paper, risks about the reliability of recovery are especially high during these contractions.  These risks include the possibility of outlier events, and the ability of firms to handle them, that create a drag on probabilistic growth forecasts.  In practice, the added risks related to contractions appear to generally mitigate the expectation of mean reversion, so that growth rates also tend to remain fairly stable over time.


This leaves earnings as the primary source of volatility in equity valuations, and, helpfully, this is a variable that is widely measured, tracked, and forecasted.  As the chart above shows, corporate valuations and earnings move together most of the time.  Even the large swings in valuations since 2003 have largely been in proportion to changes in earnings.  There are two distinct periods where valuations were untethered from earnings.  These periods coincide with unusually low real growth expectations in the 1970s and unusually high growth expectations in the late 1990s.*  In other words, even in the cases where valuations fluctuated, a fluctuating natural interest rate (on corporate returns) is not the likely explanation.

Now, one could argue that this is simply a semantic distinction - that I am just taking "animal spirits" that Robert Shiller would identify as a fickle discount rate and re-categorizing them as deviations in the growth rate.  But, even to the extent that that is the case, this framing makes equity markets much more simple and conceptually manageable.  There is no need to endlessly argue about unidentifiable investor sentiments.  The vast majority of relative equity valuations simply comes down to earnings.  And, where there has been a persistent deviation from the expected valuation, there have been reasonably identifiable sources of deviating growth expectations.  If you take a tactical position, you don't need to put a mood ring on the marginal investor.  You just need to justify a different growth expectation.  You don't even need to think about Treasury Rates, Equity Yields, or Equity Risk Premiums.

Equity valuations, compared to earnings, are roughly at the level trend that, with a little noise on either side and two distinct deviations, has been in effect for 50 years, and corporate growth expectations, which include significant foreign revenues, are 5 1/2% - a bit less than long term NGDP growth.  A bearish position here based on "bubbles" seems wrong.  A bearish position needs to depend on extremely low growth rates or a contractionary shock.


* I admit that the very low valuations of the 1970s are a bit of a mystery to me.  An explanation that I don't quite trust, because it fits my political priors, is that this was the result of the pro-consumption public policy at the time.  High inflation, together with policies such as high minimum wage levels and new public transfer programs, were geared toward the sort of pro-consumption goals that are still associated with a Keynesian paradigm.  Possibly the result of those pro-consumption policies came at the expense of growth oriented investment.

But, what's interesting is that the low real growth expectations caused valuations to fall below relative earnings as much by pushing earnings up as by pulling valuations down.  When expected growth is low, corporations require a larger portion of current income to satisfy investors.  Future corporate growth expectations don't just create higher future incomes.  They create higher relative compensation today because corporate owners substitute expected future cash flows for current cash flows.

And, look at what happened when the Reagan era supply side policies replaced the demand side policies of the 1970s.  I think most people would be surprised to learn that nonfinancial corporate profits didn't top the 1979 level until 1992, after 12 years of the Reagan and Bush presidencies, during a decade when Democrats sponsored tax cuts and the New York Times was against the minimum wage.  And, these are nominal figures while inflation was still high during this period.  From 1Q 1979 to 3Q 1986, nonfinancial corporate earnings fell 60%, in real terms.

Part of this was due to the high inflation premium going to debt, and I have argued that when considering returns to corporate assets as a portion of national income, operating profits to Enterprise Value is more appropriate.  So, here is a graph that adds interest expense and debt to the data.  I have also adjusted the numbers with the GDP deflator.  And, even with high interest payments included, real income on corporate capital declined from 1979 to 1986 and was just above the 1979 level in 1992.

In terms of the main topic of this post, using operating profit and enterprise value still tends to show valuations and profits moving together over time, but here the lag in valuation persists a little longer into the 1980s than it did when we just looked at profits and equity values.

And, we see the same result in the opposite directions in the late 1990s.  There, the high growth expectations caused valuations to soar.  And, since so much of the value of equities was based on future cash flows, corporate owners did not require current income to justify their investments.  So, by any measure, profits were falling during the boom years of the late 1990s, well before the 2000 recession.

In addition to suggesting the downward influence that growth expectations have on current capital income, this also belies the cynical myth that financial markets shortsightedly chase the next quarterly earnings report at the expense of long term value.  First, there is a consistent baseline of boring valuations that simply don't change that much relative to earnings.  But, when they do deviate, they deviate in the opposite direction from this myth.  When current profits were high at the expense of long term growth, equity values plummeted, and when current profits were low while firms plowed investment into highly uncertain long-term growth, equity values soared.

This is like one of those contradictions in consensus ideas that Marc Andreessen likes to tweet.  Everyone simultaneously knows that (1) financial markets are obsessed with short term earnings and (2) financial markets push us into recessions by throwing billions of dollars at outrageous tech. businesses that have no prayer of ever making decent profits.

10 comments:

  1. I like that you take a different approach than just modeling a moody discount factor, but I'm not sure about some of the details. For example: The "low profit growth" of the 1990s can alternatively be thought of an accounting issue related to how to best model amortization of intangibles and the BEA's attempted application of the matching principle, with or without IVA/CCA adjustments. S&P operating profits rose from $21 in 1992 to $56 in 2000. You are talking about investors requiring lower current profits when growth expectations are high, but that doesn't really compute if the accounting is economic. Asset values might rise and or equilibrium returns might rise or decline (again I think you are skipping out on the jointly determined equilibria of time preference, risk, growth from increased investment, and growth from increased productivity when you focus only on "growth" expectations and their interplay with the shape of corporate profits) but the idea that profits should slow or decline today and rise tomorrow seems like more of an accounting mistake or a very strangely shaped TFP curve.

    I don't think you can wish the level or distribution (ERP) out of the picture and I think the difficulty of accounting for profits makes it very tough to tell a NIPA-first empirical story here.

    ReplyDelete
    Replies
    1. Thanks for the thoughtful reply, dlr. I can't explain the full difference between my measure here and your reference to S&P operating profits, so you may have something there. There were a lot of complicated things going on regarding the massive influx of intangibles and entrepreneurial profits. But, are you sure about the effects of amortization? If anything, it seems to me that there would have been a lot of unrecorded intangibles that weren't being amortized, so that this would inflate profits as much as decrease them. Do you have any links with details on that issue. It's worth thinking about, but I'm not sure where to start.

      I thought the distortions of inflation could also change the way the 1970s look. But, I think, there too, correcting for distortions pushes in the other direction. So, in the 1970s economic profits were probably higher than they appear.

      And, just looking at the shape of the time series, profits and valuations grew together until about 1997, and when growth expectations shot up, there was a sharp divergence between profit and valuations. I think issues like amortization would have more gradual effects.

      In terms of the mix of factors regarding risk and growth, I think that is what is potentially fruitful about this framing. If at-risk expected returns are stable, then animal spirits are acting on the discounted rate for low-risk securities. There is probably causal density no matter how we look at it, but this framing makes the higher equity premium a function of changing risk attitudes and expectations, and those changes themselves would change the shape of investments. There would be a higher hurdle rate for long term or high risk investments. Risk free rates don't rise because future growth and productivity rise. Growth and productivity rise because risk free rates rose and changed the shape of investment.

      But, I'm not an expert on BEA data. I don't see an obvious connection between corporate profit share of GDI and future GDI growth, etc. But, there are complexities of global corporate profits vs. domestic value added, unexpected economic shocks that move actual growth away from expected growth, etc.

      These factors don't necessarily come into play in the relationships I show in the post between valuations, expected growth, and current earnings. If you accept my framing, then that relationship has a sort of elegance because it only depends on expectations. But they make it hard to confirm the idea by looking at broader data that brings in these complications.

      I feel challenged by your comment, but I don't quite know how to tackle it.

      Delete
  2. I think stock options were probably the biggest differentiator. The S&P ignored stock options at the time, but the BEA "mistake" was even worse: It counted options as expensed when exercised, which is a very ex-post matching attempt that seriously biases trend earnings downward when stock prices (and thus exercise) are rising. The amortization issue is more complex, because the switch from pooling to purchase accounting in the S&P doesn't have first-derivative implications for either S&P operator profits or NIPA profits. But purchase accounting along with the M&A boom likely allowed S&P firms to transfer more post acquisition amortization to an excluded expense, something that doesn't show up in the BEA data. I think an interesting case can be made that the technology boom led to a lot of startup companies generating P&L losses that were closer to investment (for better or worse) than current expenses. This investments-expensed bias shows up twice in the BEA/S&P profit analysis. First, anytime you have an accounting method that doesn't match well (e.g. expenses R&D & marketing rather than capitalizes and amortizes), trend profits will be biased downward during acceleration of those investments if there is no creative way to write them off via something like purchase accounting amortization. Second, when one index (S&P) has large, public company limitation and its profits exclude goodwill amortization from acquisitions of smaller investing firms, its results will naturally exclude the part of the bias that resides in the P&Ls of those startups and is littered in the tax returns counted by the BEA.

    Backing up from the weeds, I agree with you that the "causal density" seems inescapable, and the usefulness of the framing (i.e. what we are temporarily pretending is the exogenous variable) can even depend on the era we're talking about. It's this lack of flexibility from the animal spirits camp where I think we are probably most in sync. To some people, every market moment is either a bubble or a depression, and all relevant information about intertemporal prices resides in the mass delusions of lemmings.

    If I had to describe the late 90s, I would think first about a higher expected TFP shock. Holding risk equal for the moment, this fattens future consumption and raises the discount rate across the risk spectrum, right down to the RFR. The equilibrium interest rate would depend on how the new shape of the expected consumption curve interacted with people's time preference and marginal utilities, but you would presumably expect a higher quantity of investment at a high equilibria real rate curve. But this wouldn't lead to current corporate profits trends weakening if the accounting was economic, as you would expect higher profit growth along the new equilibrium curve. Asset prices themselves would be buoyed by higher growth/return expectations and stymied by higher discount factors. Obviously that isn't the only equilibrium possible, and when you throw in declining risk premiums as the reframed bubble-monster, you can search for equilibria with higher average discount factors, lower risky discount factors, and a change in the shape of expected growth (higher EBIT growth but lower earnings growth perhaps). But when you actually try to tell a nuance story like that quantitatively in the late 1990s I don't think you get anywhere near a case where weakening near corporate profits, economically accounted, fits a satisfying narrative.

    As a general matter, I think we would probably both agree that the RFR-centrism may be as common a cheat here as animal spirit-centrism. Combining the shape of the risk premia curve with the average-risk discount factor leads to a lot of muddled thinking and communicating.

    ReplyDelete
    Replies
    1. Ah. The options issue is a good point. That could explain the divergence of BEA corporate income and valuations in the late 1990s. And good point on amortization, too. Hell, everything there is a good point. Do you have any links to work trying to quantify all of these issues?

      "I don't think you get anywhere near a case where weakening near corporate profits, economically accounted, fits a satisfying narrative."

      Accepting all of your excellent points, compressing risk premiums would lead to an extension of expected payouts on risky corporate investments. And, the late nineties were defined by IPOs of firms with explosive valuations who were positioning for strategic wins in network effects in internet commerce, with some vague hope of long-future monopolist earnings. Even today, we have Facebook, Amazon, Twitter, etc. More than anything else, the thing that distinguishes the new economy since the 1990s is the prevalence of firms with high valuations that have no near-term earnings potential. Considering the shape of risk premiums, it's surprising that we still see as much of this as we do. Maybe it's the difference between firms today that appear to have one foot in a dominant position, compared to firms in the late 90s that were more of a hope and a prayer. In any case, I think qualitatively, this seems like an easy description of the period.

      Delete
    2. You know, thinking about the options issue some more, it's kind of an interesting factor here. I am arguing that the negative correlation between growth and profit is an emergent phenomenon. But, options are sort of a conscious version of the same thing. Firm owners are prospectively trading profit for growth when they utilize option compensation.

      Maybe my proposed framing of equity valuations is an antidote for the cognitive dissonance we seem to have about compensation with options. The idea is to encourage productive risk taking and growth. Maybe options themselves are an emergent phenomenon, and even though there tends to be some conscious discomfort with them, the market "knows" that growth naturally leads to lower current income, so compensation paradigms keep emerging with a more growth oriented equilibrium than individual owners consciously favor.

      Delete
  3. Re the 1970s and even mood rings: if you lived through the 1970s, you saw a rising inflation, horrific and relentlessy rising urban crime and a long losing war in Vietnam. Even worse, the AMC Gremlins. Drop Dead New York.
    Many wondered if a nation could survive with burned out cities.
    I hate to say it, but single-digit PE seemed appropriate back then. I had no vision.

    ReplyDelete
    Replies
    1. Benjamin, you have a point. This is basically what I am saying. If what we see is a fairly stationary level of expected returns on at-risk capital, the context which you describe may be related to lower growth rates, higher capital income, and lower compensation, until a regime shift changes the shape of investment toward more growth. That happened with Reagan and Volcker, and corporate income fell. I think I might revisit this idea again, regarding NGDPLT and this idea of a stable equity return, in an upcoming post.

      Delete
  4. My take is that you're probably conflating orthogonal phenomena. Overall corporate returns on capital and margins are not low. The investment golf ball shows up a little differently inside the python these days, not because of overall risk aversion, time preference or returns on capital, but because of the accelerated scale and winner take all effects that have exploded in some corners of the world. There were more profit-less IPOs in 2010 than in 1994 or 1988. I don't think that tells you anything about risk aversion, time preference or overall expected marginal returns on investment from those years. It is more about structural geography than it is about economy wide discount rates and that geography is ultimately eliminated in the giant consolidated portfolio segment in the sky.

    Same goes for your employee stock option theory. Even if there is no offsetting adjustment to additional options (or, more recently, RSUs as proportionate option issuance has declined) issued by owners relative to cash comp, it still has nothing to do with the market "knowing" anything about current versus future profits. Employees are market agents too and the overall market still has the same equity ownership when the sun sets in the west.



    ReplyDelete
    Replies
    1. You're right. Taking this idea very far at all or trying to confirm it is difficult. And, your points about the current economy are totally right.

      On the options issue, though, I disagree. Regarding the return to equity holders, it is their future income and gains that matter. An option is a pre-commitment to pay employees based on future valuation changes. If this results in future dilution, that is the product of capital allocation decisions. Firms could simply buy shares at market when options are exercised to sterilize the effect on shares outstanding. I think it is best to think of that as the baseline scenario, in order to keep clear exactly what the transaction is.

      But, I do wonder how much of the current risk premium picture is a product of your points about the winner take all effects, which would create an incentive for high levels of investment in long term earnings potential, even though the prevalence of this factor probably feeds the high risk premium that I would otherwise associate with high earnings. So the current economy is like a mix of 1970s and 1990s. Although, I would expect to see more of a bifurcation among firms of very high PEs and very low PEs. There are a decent number of high PE firms, but I don't see many firms with PEs in the range that was typical in the 1970s.

      Maybe that's because firms incorporate many of these factors within one package - like Apple, which has huge intangibles, a large cash holding, and a low P/E ratio after adjusting for cash.

      Delete
  5. Thank you so much... your blog is giving very useful knowledge for all.i didn’t have the knowledge in this now i get an idea about this..
    thks a lot:-)To know more spot cash for card swipping

    ReplyDelete