Buy & Hold, Long Term Returns and All That, from Kim Zussman

February 12, 2008 |

Kim In checking historical US stock returns, the probability of loss declines as the holding period increases. Twenty years is commonly touted as safe, but there have only been 4 such (non-overlapping) periods since the Depression so it's hard to feel secure.

(There is also the problem of whether this came by "luck": e.g., look what happened to German and Japanese markets when they lost WWII)

There were four non-overlapping 238 month (2 short of 20 years) periods in DJIA monthly returns 1928-2008. The compounded return of these (w/o div) shows only one which was down (with ending dates):

Date 20Y cpfactor
2/1/2008    5.994
4/4/1988    2.264
6/3/1968    4.941
8/2/1948    0.721

(Dividends formerly a bigger part of total return, so exclusion under-estimates final compounded return)

Randomly re-ordering the same empirical monthly returns into 100 simulated 80 year series, I calculated compounded 238 month returns and checked for up and down periods. Of the 400 simulated 238 month periods, 47/400 were declines (12%). This is about half as often as actually occurred, suggesting that the negative market momentum around the Depression may not have occurred as result of random ordering of monthly returns.

Kevin Bryant counters:

In the grand span of economic history, 100 years of stock market data is barely a drop in the bucket. This is why I derive little comfort from this kind of analysis particularly during the current period which is quickly proving to be well outside normative experience.

Kim Zussman replies:

Just because long-term stock returns are positive, it doesn't mean they continue into the future, but begs the question whether there are better indicators than history. And a related but very different question is the feasibility of deploying insights/leverage to beat buy-and-hold without increased risk of ruin.

That 3 out of 4 twenty year periods in stocks since 1928 were up should make young people with 401K's feel better, but seems dangerously irrelevant for day traders using leverage.

Riz Din adds:

'In checking historical US stock returns, the probability of loss declines as the holding period increases.' - Kim

My favourite chart to illustrate this important point is Figure 76 in Chapter 7 of the Barclay's Equity-Gilt Study. Limited observations, international examples, and changing times provide good reason to be cautious, but it is all to easy to get lost in the month-to-month or year-to-year volatility and lose track of the extent to which downside risk (negative real returns) have rapidly disappeared over time. Indeed, when looking at the UK data (1899-2005) the study finds that 'For holding periods of five years or longer, the incidence of losses greater than 5% or 10% is the same for equities and gilts.'
Over the long haul, the real returns to UK assets have been 5.3% for equities, 1.1% for gilts and 1.0% for cash. For the US since 1925, the numbers are 7.1%, 2.3% and 0.7% respectively.

To quote Christopher Walken in Wedding Crashers: "We have no way of knowing what lays ahead for us in the future. All we can do is use the information at hand to make the best decision possible."

Phil McDonnell writes:

Phil There can be no guarantee that history will repeat.

Those words, in one form or another, are found in virtually every prospectus ever offered by the financial industry. The main reason is that they are true. There really is no guarantee. But to the speculator the real question is how should one bet?

The converse of the history repeats proposition is that it does not repeat. Should one bet on something that has never happened before? Clearly betting on something which has happened frequently in the past is the better choice than something which has not happened. The best of all worlds is to combine a frequentist approach based on counting, tempered with a modicum of judgment and reason based on any changes in the contemporaneous financial landscape.

J.T. Holley comments:

I couldn't agree more, the art w/ the science. I've often thought in reference to Monsieur Le Cygne Noir why one would bet with such conviction on Sisyphus not to roll the rock up the hill, but furthermore that the rock wouldn't come right back down for ole' Sis to push it back up again? It wouldn't take me too long watchin' that rock n roll to place a bet, I'd be there taken the scrapes from those that thought otherwise as well, but not denying them their fair attempt.

Jim Sogi concludes:

The proper questions to ask are: How are things changing, and how does the trading strategy need to evolve to adapt. A dogmatic approach will not lead to good analysis and will lead to mistakes. Things are changing from the 2003-06 regime.

1. Volatility is up.
2. Global influences are greater
3. Governmental influences are increasing.
4. The industry is consolidating and shifting to electronic.

Time series sample selection in data becomes more important since last year. The idea of regimes being helpful in cycle analysis.

Comments

4 Comments so far

gabe on February 12, 2008 7:38 pm
“young people with 401K’s” don’t invest a lump sum at t0 and cash in after 20 years. that’s why you should use overlapping periods and see what you get. maybe your findings will be published here. I sent my back of the envelope study to the specs but for some reason didn’t make it through.

anyway, for the S&P, 1/3 of the 20 year periods from 1955 to present actually lost money. the expectation was still positive, but considering the size of the bet is not something I would entertain. and keep in mind we know it was the winning horse. the next 50 years who knows?
George Parkanyi on February 12, 2008 10:04 pm
Hmmm … if averages, with all the really crappy stocks that are included in them, still go up on average 10% a year - well, more than 0% anyway - how would the numbers look if you were, like … discriminating?

A couple of guys from the mid-west, Warren Buffett and Charles Munger, may have clued into something like this - but the jury’s still out on whether it’s going to amount to anything.

Cheers,
GP
Martin Johnson on February 13, 2008 9:13 am
“I derive little comfort from this kind of analysis particularly during the current period which is quickly proving to be well outside normative experience.”

The normative experience of the equity market has been 10% yearly drift with a 20% std. deviation. Based on the current period, what is “outside of this normative” experience?
Cecil Purvis on February 13, 2008 3:59 pm
Good work. My only argument with this approach is the use of non-overlapping periods. Since investors are constantly beginning their investment careers, it seems to me that consecutive 20 yr periods would be more meaningful.

Also, I wouldn’t be too concerned about whether or not historical analysis is relevant to day traders and other speculators, because theirs is another world entirely. Speculation is nothing like investing, and it requires a completely different set of analytical tools.

Recent Posts

List of Authors

Feb

12