Felix Salmon

The myth that risk goes away over time

By Felix Salmon
July 9, 2010

Rodney Sullivan has a great column which is nominally about “risk parity” strategies, but which in fact applies much more broadly — to anybody, in fact, who buys in to the idea that if you invest in riskier assets, you’ll end up with higher returns.

Of course, high risk sometimes means that returns are much higher than expected. But it can also mean that returns are far lower, perhaps crippled beyond hope of redemption. Given the reality of fat tails, the likelihood of large negative events is also higher than normal. And poor results can persist for long periods. By now, the myth that risk “goes away” over time should be well put away — risk accumulates over time. In short, investors cannot expect to “get” the expected return, but rather a draw from a very wide distribution. And in reaching for high returns by using leverage, investors dwell in the extremes of the return distribution.

The idea behind a “risk parity” strategy is simple. Let’s say you’re 100% invested in equities, because you’re happy with that level of risk. The problem is that you have no diversification: bonds can outperform equities for very long periods of time. So you rotate some of your holdings out of stocks and into bonds. But bonds don’t return as much as stocks, and that can hurt your total returns. So you leverage your bond-market investments, to bring them up to the same riskiness as the stocks.

Sullivan is not impressed. “Using leverage simply to increase the expected return is almost always a bad idea,” he writes, credibly enough. Which makes sense to me: the person lending you the money you’re borrowing expects to make a profit on the deal, which is only going to make it that much harder for you to make a profit as well.

What’s more, you could hardly pick a worse time to start levering up your bond investments than now:

That this leveraged-bond strategy is emerging after a long bond bull run and poor equity performance is curious. This likely hindsight-driven idea runs strikingly close to the frequent overweighting of equities during the late 1990s and the more recent affinity for uncorrelated assets during the early years of this century. Asset allocation decisions should never be made by simply extrapolating the future from the past.

This, of course, doesn’t just apply to bond-market rallies: it can be applied equally to the stocks-for-the-long-run crew. If you’ve got a coherent fundamental reason why stocks should outperform other asset classes, that’s great. But “they always did in the past” is much less convincing.

5 comments so far | RSS Comments RSS

>>If you’ve got a coherent fundamental reason why stocks should outperform other asset classes, that’s great.>>

They should outperform because they are riskier. They are lower on the capital structure. Bonds have scheduled payouts of known amounts. Stocks don’t.

Investors wouldn’t take on the additional risk unless they got a better price, which leads to a higher expected return. The higher return is just a hope, not a guarantee, else there wouldn’t be any additional risk to be rewarded.

Posted by jabberwocky | Report as abusive

You and Sullivan are both making valid points, but they are not the same point and to a substantial extent they are in opposition – you have misstated Sullivan’s argument.

Sullivan: “The risk-parity idea is that fixed-income investments somehow have a better Sharpe ratio than the diversified market portfolio.”

Sullivan is saying that risk-parity is founded on the assumption that a leveraged bond portfolio has a higher expected return (i.e. is a better value) than an equity portfolio with the same risk, but this is not necessarily (and in fact, probably not) true. His overall prescription is for a reduction in risk, not a different risk allocation.

You: “Let’s say you’re 100% invested in equities, because you’re happy with that level of risk. The problem is that you have no diversification.”

If you’re happy with your risk level, you don’t need any more diversification – diversification is about reducing risk. Anyway, Sullivan is not talking about diversification.

You again: “If you’ve got a coherent fundamental reason why stocks should outperform other asset classes, that’s great. But “they always did in the past” is much less convincing.”

This observation, the complement of the problem seen by Sullivan, is also true: the “real” expected return of equities is as unknowable as that of bonds. But the same applies to the higher moments of asset return “distributions” (aka “risk.”)

The problem is that this observation is of no practical use because there is no possible alternative to taking a view on these matters whether one has grounds to do so or not. Whether one chooses to invest one’s cash, save it, spend it, or borrow more, no alternative is guaranteed to dominate another either for return or risk.

As soon you start to use words like “expectation”, “distribution” and “risk”, you are dealing with a model of the future rather than the future itself. Your model means different things according to different definitions of probability. But the actual future value of an asset does not have a distribution, only a single value that is censored to us.

Posted by Greycap | Report as abusive


I agree that levering up your bond allocation probably won’t ever make sense for retail investors. Essentially you’re attempting to be a one-man bank and profit from the spread, but as an individual your cost of money will be way too high.

I’m a little annoyed at how you come at the equity premium from the side here. Just as I agree it’s not sufficient to say that past performance will absolutely dictate future performance, I think it’s also not sufficient to simply say that past performance is irrelevant. You can’t just invoke “fat tails” and discard 100+ years of market history. You haven’t convinced me that over a 25-year holding period or more, stocks aren’t sure to outperform bonds with a 99% confidence. See this paper at SSRN that attempts to model future returns:

Asset Allocation and Long-Term Returns: An Empirical Approach, Coggeshall/Wu
http://papers.ssrn.com/sol3/papers.cfm?a bstract_id=873184

And I agree with everything Greycap said above about models. But models are necessary, at least a mental model, otherwise you’ll end up with a nihilistic investment approach: Gold, guns, ammo. Even that requires a model of the future.

Posted by nedofbaker | Report as abusive

” But models are necessary, at least a mental model”

That is the point I was trying to make – not well, it seems! One could as easily forgo breathing as the construction of models.

Further, it is meaningless to say that the distribution of asset returns over a particular future period may differ from the distribution of returns over historical periods of the same length because *there is no future distribution*. There is only a distribution in your *estimate* of possible future returns.

If you rule out crystal balls and talking to God, then historical information is the only information on which to form your estimate. This doesn’t have to be as mechanical as measuring the statistics of a historical time series, but even a profound analysis of underlying causal factors is based on historical information and must ultimately assume that some relationships that held in the past will hold in the future.

A good dose of David Hume might be instructive here.

Posted by Greycap | Report as abusive

***If you’ve got a coherent fundamental reason why stocks should outperform other asset classes, that’s great. But “they always did in the past” is much less convincing.***

Excellent point, Felix! Of course, this argument ought to be applied to all investments, not merely stocks. (For that matter, it ought to be applied to individual securities WITHIN asset classes.)

Before you invest in something, you’d better have a coherent reason why that investment makes sense. Historical results are heavily dependent on the time period being considered and also tend to be cyclical.

Posted by TFF | Report as abusive

Post Your Comment

We welcome comments that advance the story through relevant opinion, anecdotes, links and data. If you see a comment that you believe is irrelevant or inappropriate, you can flag it to our editors by using the report abuse links. Views expressed in the comments do not represent those of Reuters. For more information on our comment policy, see http://blogs.reuters.com/fulldisclosure/2010/09/27/toward-a-more-thoughtful-conversation-on-stories/