Morning Coffee Videocast: Stocks Have Been an Extraordinarily Good Investment: Why?

In which I drink my morning coffee, hog bandwidth, and talk briefly about what I am going to do today--largely think hard about the equity premium, and about why stocks have been such a good investment

Morning Coffee Videocast: Stocks Have Been an Extraordinarily Good Investment: Why?

Marty Weitzman is coming to give a seminar today.

Comments

[STS]

Brad,

Given Weitzman's visit, you might want to take a fresh look at the work of the Cambridge University finance folks at:

http://www.statslab.cam.ac.uk/~chris/papers/EPP070306.pdf

Their Bayesian approach is similar in flavor to Weitzman's. Essentially uncertainty about the parameters of the model used for returns dominates when you try to do the statistics in a fully Bayesian manner.

[I Love Economics]

I love economics because it has all of the scientific glitter and cachet of a hard science-- say physics or higher-order mathematics-- while ultimately resting on a foundation which is no sturdier than that of Sociology or any of the other softer sciences. And yet, try to obtain funding for Sociology research and you'll understand why that scientific legitimacy is so important.

To see what I'm talking about, take the equity premuim, for example. Economists can unfurl all of the most powerful statistical techniques to tell us how an world _should_ behave if it were composed of rational actors, and yet this morning Brad is debating whether any of those calculations have any relevance to the _actual_ world we live in. It's possible that the problem is in the way we're calculating, but more likely it's just that our calculations have nothing to do with reality.

[wcw]

Physics and math are qualitatively different animals. The former models and explains reality, while the latter is pure philosophy, entirely and provably true on its own terms, but resting ultimately not on observation and experiment, but assumption and tautology.

Thanks for the paper. Good stuff.

[marcel]

I think that this is the appropiate venue to mention something along the lines of "Weitzman's work is horses**t, and I don't mean his economics."

[Tracy]

Ah. Perhaps it isn't the shape of the distribution, but the concepts used to formulate it. When I looked at Brad's talk his morning (3 cheers for the idea, btw), I was struck yet again by how much so many of his concerns track those put forward by George Shackle 60 yeas ago.
Suppose we turn the distribution Weitzman is talking about on its head. Instead of measuring uncertainty - you mean risk, actually - we look at the degree of surprise one would feel for certain results. You then get a more managable framework; most results from different mind experiments (we are talking about expectations here, after all) will have a similar result - we wouldn't be surprised, one way or the other. Some results are so far-fetched that we discount them completely. A few results, however, are just plausible enough to be concerned about and constitute "what we have to gain" and "what we have to lose". It is those results - Shackle calls them "focus outcomes" - that tend to dominate our expectations about future events. It is our inherent uncertainty about the future, caused by our capacity to undermine future expectations by present choices, that fuels these considerations.

Soooo .. perhaps the equity premium could be better modeled using different tools. Shackle thought so, if I recall correctly; there are several pieces on investment in his work. The standard refs are Expectations in Economics and Uncertainty in Economics.

[im1dc]

Great Comments, just amazingly great. I had begun to believe I was the last remaining outspoken skeptic on the Economics fora. Phew--glad I'm not and that others far better qualified and knowledgeable than I have chimed in with their appraisals and solutions.

Wonderful, truly wonderful to see that you cannot fool eveyone all the time.

That said, the work of Economists like Weitzman and DeLong do provide value, great value, because they rationalize what they study.

It is not their fault that which they rationalize is not rational, subject to vagaries and perversity.

Indeed, imo and hope, their work might give us a better understanding of intrinsic value as contrasted with nominal of goods, services and most especially commodities.

BTW, "equity premium" is just Wall Street talk for what they can get out of buyers.

It varies with preceived popularity, which on WS means the nearby probability or potential for any given security to go up in price.

Watch Jim Cramer's Mad Money to get a clue. Traders pay more for "earnings growth" a lot more. Check the PE's of 'hot' growth stocks. Actually, they pay up for the possibility of fast growth of earnings. Investors hold on for the ride and hope the company's B-plan actually works out. In short, there are many different interests at work each with their own idea of what a given stock's "equity premium" should be and as long the company in question delivers on the needed earnings numbers "ahead", "in line", but never "lower" the bait and switch continues on. Once the company is mature the equity premium shrinks and the stock has to trade on its real return, making it less attractive for traders, speculators but more so for investors. Of course, other games then come into play such as new products introductions, off patent products, selling off assets, M&A's, etc.,

In other words, the money game is huge and "equity premium" is Street Speak for a temporary phenomenon of the Bulls, i.e., the buy side trade.

The study of equity premium is like the study of chaos and that butterfly in the Amazon flapping his wings--it may be connected but there is no way to find out in what cases, when or how.

[anne]

Brad, for some mysterious reason, I cannot read or hear this post on my office computer or a divinity school library computer. What could the reason be :)

[monkyboy]

I think the equity premium is caused by people forgetting they own stock in a company...but they know when their bonds mature.

Just a theory...

[Ellen1910]

"It is not [Weitzman and De Long]'s fault that which they rationalize is not rational, subject to vagaries and perversity." im1do

Sort of like being a priest explaining the supernatural, eh wot?

[Colin]

Brad, why not just post your script for those of us who are not set up for video?

[Ellen1910]

Do you think the dog's moaning in the background of Brad's blogtv tells us something?

[Ellen1910]

Who says stock returns have been better than the returns on bonds? The S & P 500 has returned 3.84% while long-term treasuries have returned 6.30%. Doesn't sound to me like there's an equity premium puzzle.

Oh, I'll bet you all were talking about "long, long term," right? Not the last five years.

Well, when are you going to start? How about 1926. The time of "blue sky" and "watered" stock, the time before inflation, the time before indexing?

So, how'd our typical investor do between 1926 and 1940?

Or to bring the question up to date, who says there's an equity premium, at all?

[civan93]

First, Brad, did you break down and buy the MacBook Pro? If you did congrats. After weeks of finding myself walking idlly through the Apple section at my local store (we do not yet have an Apple store in NM) I couldn't help myself.

The second comment pertains to risk modelling. While I tend to agree with the basic tenets of Weitzman's criticism, is it fair to say that this risk modelling extends beyond equity returns and can be found in any risk analysis situation? Can any technique, even say game theory, be used to examine the countless numbers of factors that come into play? Are actor's behaviors modelled upon the known or unknown factors? Is this going to save me 15% on my car insurance?

[BobT]

I followed the arguments, but The dog! The Dog! Just walk the dog!!

Nice comic touch!

[jt]

Really enjoying the "morning coffee". One small suggestion: maybe you can experiment with different ways of making eye contact with the webcam.

[Doug]

VIDEO GOOD, BUT CHECK COMPOSITION: Ignore all the naysayers; they are plentiful and need not watch the video. This is a useful and different format for presenting your ideas in a compelling way.

Anyone using a web browser with the QuickTime plug-in should be able to both see and hear the clip. It worked perfectly for me (Windows, Firefox).

This morning's clip has better composition; now just bring the camera up closer to nose or eye level, so we aren't looking up your nostrils. See links [1] and [2] for simple tips on talking head photo composition.

Keep it up!

[1] http://www.pbs4549.org/producer/vmvidang.htm
[2] http://www.hu.mtu.edu/~ciwic/2005/aic_assign4b.htm

[piotr]

Does anyone remember the glory that was Nikkei? What was it -- 40,000? Today, 17,000, and it is more than a decade later.

In its day, Japan was one gigantic growth stock.

[Alex Tolley]

I have to agree with the skeptics.

When you strip the model down, isn't is just describing a distribution with non-independent returns?

It appears to me that the whole argument rests on the distribution of returns, not the posited model. Therefore I would be interested if the same result can be derived from other fat tail distributions. Mandelbrot and others have shown that many market return distributions are power laws with fractional exponents and not random.

If other distributions show the same effect, then this particular model is not required - it just generates a distribution that can describe the effects observed. Someone else has described this as backwards thinking - using data to confirm a pre-existing theoretical model, rather than using data to build models.

[Arun Khanna]

First, thanks for your video. It made me feel part (of the audience) of the active research community.
Second, one could examine Weitzman's explanation of the equity premium puzzle further.
If Weitzman's fat tails and rational investors hypothesis is correct then one should see a negative and significant difference between the equity premium effect in emerging (new technology) industries versus declining industries.
If explanations based on irrational investors are correct then one should see a positive and significant difference between the equity premium effect in emerging (new technology) industries versus declining industries.
If the null hypothesis of a true equity premium puzzle is correct then one should see no significant difference between the equity premium effect in emerging (new technology) industries versus declining industries.

[Matt]

The first reason amounts to saying that we know less about equities than we thought-- the second is that going broke isn't so bad. I.e., equities involve more risk and more of a moral hazard-- hmmm. Well, sounds good to me-- I'll go out and buy me some.

[jim]

I think the "equity premium puzzle" which is really an investor preference for bonds is a threshold effect. Suppose I have an investment strategy using bonds which will provide me $300K at my retirement age. A strategy which involved stocks might provide $500K at that age; it might provide only $100K. $500K is better than $300K. It provides a more comfortable retirement. But I can retire on either. But I can't retire on $100K. In this case $300K is not three times better than $100K. The two are incommensurable. One permits retirement; the other doesn't.

In a sense, the preference for bonds is similar to the desire to play the lottery. In the case of the lottery, two outcomes are possible: (1) I will be rich beyond the dreams of avarice, (2) I'm out a buck. Again, they're incommensurable. Again, there's a threshold. People talk of "The Number", the wealth threshold above which you can consider yourself free of the demands of society. A lottery win would put you above it.

In the nineteenth century, there were income threshold effects. One either could keep a servant or not, for example (this is a threshold which is behind Schumpeter's remark, I think). In modern America there aren't income thresholds: utility is a continuously differentiable monotonically increasing function of income. But there are still wealth thresholds. And if investing in bonds guarantees you stay above a threshold and investing in stocks risks falling below that threshold, it's rational to prefer bonds.

[Ralph]

The current stock market "resurgence" is based on very shaky grounds IMHO. As in early 2000, when you recall the dot com revolution was surging ahead with full abandon, the signs of growing riskiness are everywhere. The Fed continues to raise short term rates incrementally and crude oil prices continue their skyscraping movement after a short pause last year. Housing prices have slowed from their record breaking pace of the past year or two and the US Federal deficit shows no sign of coming to an end with pressure on Congress for more tax reduction for the rich on the march once again while the Iraq war shows no sign of relief and the generals are in full throated cry for Rumsfeld's resignation showing the public how hopeless the situation looks from the military side. All in all not the sort of atmosphere in which to feel confidence in investments IMHO.

[Glen Tomkins]

I thought it was the pyramid

Stable systems achieve homeostasis because they have dominant negative feedback mechanisms for parameters that start to deviate from their set points. While any market has the built-in positive feedbacks of greed and fear that tend to magnify any move in prices up or down, a secondary market like the equities market seems unusually naked in its vulnerability to these destabilizing forces, since the prices of stocks lack much tether to underlying values.

The greed factor has dominated the fear factor in the US equities market for the past century, pushing prices upward in defiance of underlying valuations, because of this self-reinforcing belief in the equities premium provides positive feedback. But how did such a belief get started? Well, the last century has been the US century. It has seen the US rise from being a second tier economic, military and diplomatic power, to being the world's unchallenged hyperpower. How could this rise, whose persistence and size lends it an air of inevitability by now, not lend to the idea of investing in the US an attractiveness even in excess of the true underlying risks and benefits? Why would the US equities market benefit more than the US bond market? Because equities are more speculative, less tethered to the underlying valuations of the commercial and industrial enterprises that issue them. And once the difference in ROI is established in the popular imagination, it becomes a self-fulfilling prophecy.

At some point, of course, negative feedback mechanisms will reassert themselves. Under even the best management, the US could hardly repeat in the upcoming century the spectacular improvement in its relative position it achieved in the last. The curve has to, at least, flatten. More likely, other parts of the globe will at least catch up, and the US will decline, relatively.

With really bad management, of course, there could be an absolute decline. This last consideration seems especially to the point just now in our political life. In this last, American, century, our country has been an attractive investment destination not only because of the success of US commercial and industrial enterpises, but also because we have been the polar opposite of a banana republic. What the current admninistration has done already to move us towards banana republicanism may already have switched on a long-term crisis of confidence in the US that will replace in the upcoming century the underlying confidence in the US that dominated the last. A nuclear attack on Iran, of course, would seal the change in how we and our markets are perceived, and perhaps speed it to a route. The US equities market was probably due, over the long term, for at least a slow deflation in prices even before Bush took office. Now? I got out of US equities four years ago. The point of decision now is whether to get out of the US.

[Alan]

This is great stuff but can you reposition the camera to look down on you?

[atolley]

Regarding the video:

1. "Gauss got it wrong". Gauss didn't get it wrong. Users picking the gaussian or normal distribution got it wrong. This distribution is the almost the default (or even the null hypothesis) distribution that almost everyone picks if the data "looks normally distributed" because its mean and variance is easily calculated and a host of programs and code libraries will do the calculations for you.

2. Risk of bankruptcy. The issue of bankruptcy, aka gambler's ruin was well known and the issues surrounding that on valueing risk taking is quite well studied.

Maybe I am missing something here, Weitzman's model may be good or insightful, or not (see earlier post), but I don't see that it is particularly ground breaking. If these 2 ideas are being used together for the first time, then I could agree it is novel, but again, the model itself may be only one of many that derive the same result based on the distribution charateristics of the returns.

[Blissex]

Jim, your comment:

«I think the "equity premium puzzle" which is really an investor preference for bonds is a threshold effect. Suppose I have an investment strategy using bonds which will provide me $300K at my retirement age. A strategy which involved stocks might provide $500K at that age; it might provide only $100K.»

is completely off the mark.

The assumption here is that stocks are riskier than bonds; but the equity premium puzzle is based on the observation that IN THE LONG TERM stocks are not riskier than bonds.

So, if investing the same amount of money for 40 years has the SAME PROBABILITY of growing to 500,000 if invested in stock as of growing to 300,000 if invested in bonds, why are the multiples on stocks lower than those on bonds? Why isn't everybody investing just in stocks, driving their price up to the point where they no longer outperform stocks?

That is what needs explaining.

The explanations come around to:

* The risk may be the same, but there is no outperformance for stocks (the long term return has been mismeasured, e.g. survivor bias).

* The outperformance may be there, but the risk is higher for stocks (the probability of the outcome has been mismeasured, e.g. by selecting long but abnormally positive periods).

* The _volatility_ of stocks matters.

* Long term locked investments are unrealistic.

I reckon that the single greatest problem is the volatility...

The point with volatility is indeed threshold, and the relevant threshold is zero: once your investment in stocks goes to zero, it stays zero, and cannot revert to the mean.

That is, suppose that the expected value of an investment is 400,000 with stocks and 300,000 with bonds, but the range for stocks is +/- 500,000 and that for bonds is +/ 100,000.

In such a case, there is a chance that the stock investment will become zero at some point, and then _it cannot revert to the mean_.

In other words, (900 + -100)/2 and (500 + 300)/2, but in real life, if your investment goes to zero, it stays zero.

That is there is a fundamental asymmetry between upwards and downwards volatility, so that the lower the volatility the smaller the chances that all will be lost.

Another good point is that there is another fundamental asymmetry, that occurs in practice, even if the ''equity premium puzzle'' statement obscures it...

The true statement of the puzzle is that one invests for the same long period of time in stocks or bonds, like 40 years, and one cannot pull out or switch.

But this is totally unrealistic. Now imagine that we still have the same investment that after 40 years is expected (with the same probability) to 300,000 in stocks or bonds, but the volatility of stocks in +/- 200,000 and that of bonds is +/- 100,000.

The suppose that after 20 years of investing in stocks one finds that the value of the investment is now 500,000: what does one do: cash in the 500,000 and switch into bonds, or leave the money into stocks so that odds are the total will go down to the expected 300,000?

This means that demand for bonds will always be higher than demand for stocks (because half of the time people will switch from stocks into bonds instead of holding stock for the full term), and thus the multiple of bonds will be higher.

Conversely, how many investors are sure of being able to hold an investment for 40 years? Suppose most have to liquidate after 20 years: half of them will liquidate at a bigger loss than if they had invested in bonds.

Therefore most will invest at least part of their finds in bonds, which is an asymmetry that again means that demand for stocks is bound to be greater than demand for bonds.

So suppose you have 100,000 to invest for 40 years, and you expect bonds to return 300,000 (almost 3% compound) over those 40 years, and stocks to return 500,000 over those 40 years (a bit over 4% compound), but with a +/- 500,000 range.

You will still invest say 20,000 in bonds because there is a chance your stock investment will hit the sticky zero threshold, another 30,000 in bonds to guard against the risk of having to liquidate when the stocks are underperforming; and then if after 20 years your stocks have outperformed, you will convert them into bonds to lock in the outperformance.

All these things mean that the demand for bonds, which are less volatile, will be greater than for bonds, thus driving up their multiple even if they return the same or less as stocks.

To recapitulate: volatility is asymmetric, and zero and below are sticky; also, people may choose or be forced to cash in their investment before the 40 year period is over, at time when the investment is either above or below its expected value.

[Blissex]

«because of this self-reinforcing belief in the equities premium provides positive feedback. But how did such a belief get started? Well, the last century has been the US century.»

Note that during the parts of the last century in which the USA grew the most and was most powerful stocks had MUCH LOWER multiples than now.

The great bull market, where stocks have reached amazing multiples, has started in the late 80s, because of two once-only and very important factors:

* The defeat of the USSR has meant that western governments no longer need to care at all about the home front by giving everybody a stake in the success of the corporatist system (poor suckers become poorer? we can ignore that now). Thus the massive transfer of wealth from poorer to richer social strata on the most recent 15 years.

* The ageing of the baby boomers, who are all investing for their retirement according to the same formulas, which prescribe to be long of stocks for most of their life until retirement, has created vast and automatic demand for stocks of any sort, especially those vaguely perceived to be blue chips.

[atolley]

Blissex:

Good explanation.

Minor niggles:- returns are log normally distributed (well at least for most purposes) so that returns never really go to zero and put the investor out of the game.

Secondly, in practice, people tend not to start with lump sums, but continue to invest in the market, adding to the sum over time.

Both these factors may mitigate the preference for bonds (for non-pension fund investors who have other concerns).

Jim's threshold model is not so bad - after all, you could liquidate your relatively riskless securities and invest in equities if equity returns fall dramatically (in fact you should and vice versa).

However, I understood the puzzle to be the degree of the premium, not that it exists. I think any explanation of the size of the premium must also take into account changes in market structure and participants over time. For example, mutual funds are a relatively recent phenomenon and have displaced the predominance of wealthy individuals over the last century.

[atolley]

Bissex:

Re your last post. These factors were not operating in the 1920's bull market (at least not the same ones). You might want to consider the effects of the massive 1970's inflation on returns and risk premia.

I also recall the risk premium to be generally much higher in almost any random time period (sorry I don't have a reference)- so I don't buy that it is the result of the current secular(?) bull market that started in 1982. (I studied finance in the early 1980's when this leg of the market was just getting started and most text books were still puzzling over the previous couple of decades)

[meno]

Blissex,

No, you're wrong.

You're wrong because you were writing in response to someone who was planning for his retirement. And your response is talking about a 40-year investor who is sure they won't want to pull money out early.

Most people save for their retirement in their late 30s, 40s and 50s. The bulk of the saving is in the late 40s and 50s.

So I see no equity premium puzzle as far as retirement savings go. There it is simply that most people are paying the premium of not being a 40-year investor.

[Blissex]

«You're wrong because you were writing in response to someone who was planning for his retirement.»

Not at all, not at all: I am writing to comment about the equity premium puzzle, which is NOT the retirement premium puzzle any more than it is the risk premium puzzle.

The puzzle arose because somebody measured the riskiness over many decades of both stocks and bonds, and found that they were the same, yet stocks were selling at lower multiples than bonds.

So the crucial assumption is that over a long period of time stocks are no more risky than bonds, yet their return is better. Then companies or governments could just sell long term bonds and invest in stocks in the long term and make FREE MONEY, thus eventually arbitraging the difference away.

Obviously this has not happened for over a hundred years, so what's going on?

This has nothing to do with retirement, even if as you point out that retirement savings start late means that a lot of people can't stay invested for many decades...

I chose 40 years as an example of ''a long period of time'', because for example in the past 100 years there have been several 20 year subperiods in which stocks were hideously expensive (that is, their returns were terrible).

[Blissex]

«risk premium to be generally much higher in almost any random time period»

Again, that a risk premium would exist is not a puzzle: if an asset class is riskier it is obvious that it should be cheaper.

The puzzle is the EQUITY premium, not the risk premium: given that in the long term, over several decades, equities apparently are NOT RISKIER than bonds, why are they cheaper? What makes equities cheaper (that is, returns are higher) than bonds given the same riskiness? That's the puzzle.

[STS]

I just checked in again on a different PC to see if I could get the video working. At last it did. Still not sure why.

I'm commenting again because hearing the phrase "Gauss got it wrong" a couple more times annoyed me enough to want to make the useless gesture of typing a comment into this little text box.

Many economists have "got it wrong" for decades because they knew just enough math to convince themselves they were computing something they call an "equity premium" without knowing enough to realize their computation was misleadingly oversimplified or to avoid really silly, misdirected, anachronistic remarks like "Gauss got it wrong"!

Now if Donald Luskin or Victor Davis Hanson said something that bassackwards, we'd hear about it for days. ;)

[Don't go to econ grad school]

Yes, when I learned about the equity premium puzzle, I thought: 'the only puzzle here is "why is my professor puzzled by equities"' The resolution to the so called equity premium puzzle is this: the standard model is wrong. Throw it out, stop teaching it, and come up with a better one. The profession has had decades of unimaginable brainpower and statistical techniques at its disposal to do so and it hasn't - instead it has trodden down the same unproductive path and keeps wondering dully why nothing ever works out. The same rubbish is taught to the next generation and the cycle repeats.

[STS]

To be fair to Brad and his colleagues, identifying the exact problem with the equity premium computations *is* subtle and I appreciate Weitzman's efforts (tho Jobert, Planaria & Rogers are better statisticians -- see the link in the first comment to this thread).

But blaming a long-dead mathematician who never even thought about the "equity premium" for persistent problems with an economic model is really lame. It probably wasn't even seriously intended as a slight -- just Brad trying to popularize. But it is such a silly "blame the other guy" tactic, I'd like to think Brad is honest enough to recognize the absurdity.

[anne]

Heck, stocks are important in a portfolio so figure out how and what to efficiently buy and hopefully wonder at the important gains after. Index, index, index, at Vanguard costs, and do look internationally :) Index, and think thereafter.