Bonds beat stocks

by John Quiggin on March 14, 2009

This Bloomberg story gets the headline (“Bonds beat stocks in earth-shattering reversal”) right., but the lead (or lede) wrong. The intro “Buying 30-year Treasuries is returning more than stocks for the first time since Jimmy Carter was president. ” is wrong – bonds have beaten stocks in quite a few years since then.

Bonds beat stocks (Bloomberg)

Bonds beat stocks (Bloomberg)

The finding in the chart is much more dramatic, to the point that “earth-shattering” is justifiable hyperbole. What it shows is that, over the entire period since 1979, a strategy of buying 30-bonds (trading so that the portfolio always holds the most recently issued bond) has outperformed the strategy of buying stocks and reinvesting the dividends.

This is earth-shattering (or, at least, potentially finance-sector-shattering) because it refutes one of the central assumptions of nearly all investment advice: that, provided you are in for the long haul, stocks always beat bonds. Robert Shiller in Irrational Exuberance pointed out that this was historically true for the US (for periods over 20 years) but not for some other markets. Now it’s no longer true for the US.

Over longer periods, there is still a substantial equity premium; that is, the return to holding stocks exceeds that for bonds. But so there should be. Returns to stocks are much more variable than those to bonds, and, as the latest evidence shows, you can’t eliminate that variability by holding stocks for ten or twenty years.

The big puzzle is, why are stocks so volatile. They are more volatile than profits which, in turn, are much more volatile than aggregate consumption (unless the current stock markets are accurately forecasting a new Great Depression).

(via Brad DeLong and Paul Kedrosky)



Martin Bento 03.14.09 at 8:39 am

I’ve always thought the stock market, and particularly the high-end portions of it that are represented in the Dow and other major indices, tracked not just the overall economy but the size of the portion represented by those companies. In other words, 30 years ago much more of retail was small local stores that would not be represented in Standard & Poor; now much of that economic activity is taken up by Walmart, Home Depot, and the like, which are. 20 years ago, housekeeping for pay would have been an economic activity invisible to “the (stock) market”; still largely true, but there is Merrie Maids and possibly others (I don’t know if there are others). So the tendency of the stock market to grow more than the economy reflected this. Is this idea heterodox? Is it something the economists have considered and rejected?

But I don’t suppose that has any bearing on why stocks are underperforming so. Or maybe it does. Doesn’t the historical concentration of economic activity in large corporate hands seem to be reaching diminishing returns, at least in the US? Retail and restaurants have been largely so concentrated for ten or fifteen years. Perhaps this is one of the factors that drove so much money into derivatives, synthetic bonds, etc., especially based on residential real estate loans: residential real estate was still not concentrated in corporate hands, and mortgage commodification provided a way to treat it as though it were.


Kevin Donoghue 03.14.09 at 8:41 am

The big puzzle is, why are stocks so volatile.

Because we really don’t know what they are worth. They are more volatile than profits because an accountant can make the profit be whatever the CFO wants it to be, within reason; and the CFO usually wants a steadily growing profit. The problem now is that accountants have lost what credibility they had. They are now seen to be the servants of prodigals and projectors and the assurances of the auditors are worth about as much as the tips of Damon Runyon’s Hot Horse Herbie, whose hot horses usually turned out to be colder than a landlord’s heart.


Ginger Yellow 03.14.09 at 9:59 am

I’ve never really understood the great equity risk premium mystery. Of course a broad basket of equities return more than what is considered to be the single safest investment in the world given a long enough time frame. That’s the nature of human psychology and the default instrument of a reserve currency, which keeps yields on T-bills artificially low. And the one time T-bill yields become attractive relative to other bonds is precisely the time that the stock market is booming. Now you can certainly argue the premium should be smaller, although it never seemed all that large to me, and for now it doesn’t exist.

As for the volatility, is it really a surprise that a highly liquid retail investment subject to enormous amounts of day trading, rumour mongering, pump and dump scams, algorithmic trading and insider trading is volatile? Really?

It’s all rather reminiscent of the economics profession’s periodic rediscovery that humans aren’t entirely rational utility maximising automata.


Zamfir 03.14.09 at 1:05 pm

But Ginger,isn’t the point here exactly the opposite, namely that for nearly any starting point since 1979 you would have been richer today if you had invested in bonds than in stocks? I guess this might turn out to be an anomaly, but than again it might not.


Walt 03.14.09 at 2:43 pm

Ginger: The mystery is why it’s so large. Economists expect it to be positive.


F 03.14.09 at 6:24 pm

The big mystery to me is that there are only three periods in US history when bonds have earned substantially more than inflation: 1865-1885, 1920-1934, and 1982-2003. Basically the rest of the time, bonds yield almost nothing in real terms. Why are these the exceptions?


F 03.14.09 at 6:27 pm

I should add that these periods also correspond to the 20 year periods when bonds and stocks have nearly equivalent returns.


John Quiggin 03.14.09 at 8:02 pm

F, this is the risk-free rate puzzle, the overshadowed twin of the equity premium puzzle. Over the whole of US history, the real rate of return on bonds has averaged about 1 per cent, which is lower than economists would expect.


F 03.14.09 at 9:46 pm

Thanks, I’m going through that paper now, but I guess my question was “What is special about the times when bonds do have positive real returns?”, rather than “Why are real returns for bonds nearly zero?”


radish 03.14.09 at 11:59 pm

The big puzzle is, why are stocks so volatile.

I have to agree that it’s puzzling from some angles but perfectly sensible from others, and it’s not just the “info asymmetry” issue Kevin mentions. We think of stock exchanges as “normal” markets in which the products being traded just happen to be securities, but the real purpose of an exchange is to “amplify” signals from other markets rather than to discover new information. The idea behind the institution is to minimize the friction of reallocating capital, in the hopes that capital owners will quickly identify, exploit, and magnify tiny marginal differences in real productivity. So both excess volatility and a certain amount of noise (which offers the benefit of stochastic resonance) are really features, not bugs.

Cf. the Boydians, who very vocal about how performance always costs you in terms of stability and vice versa. Stock markets are supposed to overreact in preference to underreacting, just as sports cars are supposed to overreact rather than underreact, and for more or less the same reasons.


Martin Bento 03.15.09 at 3:49 am

Will “go Galt” for food.


Martin Bento 03.15.09 at 3:50 am

Oops. Meant to post that joke in the other thread. Feel free to delete here.


Martin Bento 03.15.09 at 3:52 am

“So both excess volatility and a certain amount of noise (which offers the benefit of stochastic resonance) are really features, not bugs.”

Hence the affinity between market trading and cocaine.


Kevin Donoghue 03.15.09 at 7:26 am

F: The big mystery to me is that there are only three periods in US history when bonds have earned substantially more than inflation: 1865-1885, 1920-1934, and 1982-2003.

I can’t say I’ve studied this, but aren’t those all periods when high nominal interest rates were being used to bring inflation under control, at least at the beginning of each period? So maybe what you are seeing is a result of adaptive (as opposed to rational) expectations of inflation, so that bond-holders got higher real returns than they anticipated. At least that’s likely to be part of the story. I’d be surprised if there isn’t quite an extensive literature on this.


notsneaky 03.15.09 at 9:08 am

“So both excess volatility and a certain amount of noise (which offers the benefit of stochastic resonance) are really features, not bugs.”

But, except for the word “excess” in that sentence, this just explains (reiterates) why investment in equity is more risky than investment in bonds. It still doesn’t explain why the return is so much higher.

Part of the reason why the EPP is such a big issue is because it also ties into a lot of other “unsolved mysteries” in economics. A bit more specifically it goes right to the big discrepancies between micro and macro data and what they say about how people behave in all kinds of different ways. EPP, taken at face value implies crazy high levels of risk aversion at the macro level but micro studies estimate much lower levels of risk aversion, based on a multitude approaches. Basically, if you’re that risk averse that you need the 4% extra return on a risky investment, you wouldn’t get out of bed in the morning in the fear of the big bad world.
But it also matters for things like how costly are economic fluctuations. Again, micro studies estimates of risk aversion imply that the cost of fluctuations around a trend are nothing compared to the level of that trend. But, taking the EPP at face value – with the implausibly high levels of risk aversion – would imply that recessions and booms matter a lot more. So in a way it’s hard to say how much one should worry about fluctuations vs. how much one should worry about long term average growth, without solving the EPP first.
Another area where it becomes relevant is how much we value future, uncertain benefits. This blog has had some excellent discussions (thanks in large part to JQ) on the Stern Report and the costs of global warming. The central issue there is what is the proper social discount rate to use in estimating these costs. And a key parameter there is the intertemporal elasticity of substitution – which in this context is the same degree of risk aversion that makes the EPP the mess it is. Again, in a way, if we don’t know the answer to the EPP then we don’t know how to properly calculate (more precisely, how people in the economy calculate) these costs.

And another way that micro and macro data disagree with each other in a way that is relevant here is on the elasticity of labor supply – how much do desired hours of work change when the wage rate changes. It turns out to be the same parameter that is key to the EPP (essentially because both phenomenon have the same “substitution effect” vs. “income effect” aspect to them). Basically micro studies find that labor supply is elastic while (traditional) macro studies find that labor supply is not. This has implications for – well lots of stuff. For one if you’re gonna argue that most fluctuations in economic activity are due to productivity shocks (which a policy maker like the central bank or fiscal policy cannot do much about) then to make the explanation coherent you have to have an elastic labor supply. So you’re happy with the micro data but not the macro. But if you think it’s mostly Keynesian style demand side shocks, then that explanation works better with an inelastic labor supply. And what determines which one it is appears to be the same thing which causes the EPP.

The EPP also has implications for social choice theory (how much income redistribution is desirable, given that it has some cost), for consumption externalities (does “Keeping Up With the Joneses” make you save more or less?), for how saving behavior changes as an economy develops and for many other areas. It’s pretty much a big holy grail of a lot of economic research.


Hoover 03.15.09 at 9:42 am

“The big puzzle is, why are stocks so volatile”

John, I’m guessing you’re on the trail of something related to your earlier post entitled “the end of the cash nexus”.

I await with considerable interest your further findings.


Steven 03.15.09 at 2:43 pm

By my quick calculation, this implies that the return on the 30 year was 9.9% annually (assuming moving from 100 to 1800 over 30 years). Isn’t the lesson not that this 30-year period represents a dangerous period for the practice of buying equities (hey, I’d love a 8.9% return right about now, which is what 100 to 1400 over 30 years implies), but rather an extraordinarily fortunate time to buy (and, presumably, sell) long bonds?


Steven 03.15.09 at 3:35 pm

ah, I see I skimmed the article too quickly, they actually give the numbers–I think it’s 9.4% for bonds and 8.8% for equities. Doesn’t change my point, though.


Ginger Yellow 03.16.09 at 2:13 pm

Long government bonds. Another of my frustrations about this sort of question is that Treasuries are used as a proxy for all bonds, when they quite obviously are not. The graph for long dated corporate bonds (and especially high yield bonds) would look very different, especially over the last year and a half. Surely a more appropriate comparator to equities would be a basket of corporate bonds. Of course, the T-bill data is interesting in itself, but it leaves two variables open (fixed income versus dividends, risk) instead of one.


John Quiggin 03.16.09 at 8:05 pm

GY, corporate bonds are like a mixture of government bonds and equity. Most of the time, returns to a portfolio of corporate bonds will lie somewhere between the return to equity and the return on government bonds.

Also, the difference between dividends and fixed interest (on government bonds) is a difference between risky and riskless income. It might be better to state your point as saying that there are two kinds of risk here – fluctuations in earnings (dividends + capital gains) and bankruptcy/default. For an individual security, these have very different effects on the distribution of returns. But for a portfolio, the effects are not so different.


Bob Bronson 03.17.09 at 3:12 am

We solved this puzzle years ago and have successfully predicted the last decade’s results.

See the schematic on page 32 here:

also explained in more detail here:


James B. Shearer 03.17.09 at 6:09 pm

The chart appears to be comparing stocks world wide to US government bonds which is not exactly apples to apples.


Ginger Yellow 03.19.09 at 12:14 pm

John, I grasp all that, but my point is that if you’re trying to analyse the equity risk premium, as opposed to risk premium in general, it makes much more sense to compare directly to other forms of corporate risk, not (the very safest form of) government bond risk.

Comments on this entry are closed.