No. of Recommendations: 19
The chances of a random US stock having a positive real total return in a one year period since 1960 is 56.5%.
This is pretty far from the 4% in the study that got so much press, and a much more useful and meaningful number.
Consequently a diversified portfolio of randomly selected stocks which is reconstituted from time to time almost always does just fine even without the few long run superstars.
In fact your chances of beating the S&P 500 this way over the long run are remarkably high, not low.
...
I think you skipped over some salient details here. For example:
How much positive? Enough to beat S&P or total US stock market weighed my market cap?
What's the magic about one year? How about same stock held for three months or five years or some other random period? Still beats broad index?
When do you rebalance (or reconstitute as you more accurately termed it?) You seem to imply one year (I am guessing). Why sell part or all of winners after a year instead of letting them run?
I concede to your statistics but as stated they are wholly inadequate to good old "buy and hold a market cap weighted index".
The last point is probably sensible for a gut feel--but it simply doesn't seem to be true. It doesn't fit the data.
When you try it with a research database, a periodically reconstituted monkey-with-a-dartboard portfolio has over a 90% chance of beating the S&P 500 over time.
(maybe only 60% or 80%, depending on precisely how the test is structured)
Not that dartboard portfolios are particularly special.
There are LOTS of ways to have a skewed and seemingly sensible subset or weighting of a broad portfolio, sometimes called smart beta or just smart indexing.
It's not so much that any of those strategies are particularly brilliant, but that capitalization weight is such a strong outlier from the pack...to the downside.
Almost any other weighting/sampling approach is better over the long run because of the inherent drags built into cap weight.
The best reading I would recommend is Bob Arnott's research paper, "The Surprising Alpha from Malkiel's Monkey and Upside-Down Strategies"
It's easy to find with a search, but hard to post a link to it.
The shortest possible summary: a broad index portfolio works well, but make sure it's anything but cap weight.
And low in fees, of course.
One specific example:
Buy all S&P 500 constituents which happen to be covered by the SI Pro database. Statistically that's all of them, but I mention it for completeness.
Test 1:
Buy equal dollar amounts of all of them. Hold for a month (21 trading days), check the list again, and repeat.
Once each three months (three holding periods), rebalance all positions to equal weight. These are pretty liquid stocks, so I allow 0.2% round trip trading costs.
(the reason for the one month hold is that any cash from buyouts will get reallocated at the end of the next month, not waiting for the end of the quarter)
Total return 2000-01-03 through 2023-03-10: 9.1952%/year, average across the 21 possible sets of trading dates in the first month.
Test 2:
Buy capitalization weighted dollar amounts of all of them. Hold for a month, check the list again, and repeat.
Once each three months (three holding periods), rebalance all positions to capitalization weight. Same 0.2% round trip trading costs.
Total return 2000-01-03 through 2023-03-10: 6.5696%/year, average across the 21 possible sets of trading dates in the first month.
The difference is 2.6256%/year.
This result might be a bit better than typical because some very large caps did quite badly in the first couple of years I chose, but the equal weight version did better in most of the individual calendar years.
Normally I'd expect a gap in the 1% to 2% range.
But of course, that's also the whole point of the test: the very very largest caps often do quite poorly.
(an equally weighted portfolio of the 5 largest stocks by market cap, reconstituted monthly, underperformed the S&P 500 by 2.86%/year since 1986)
Management fees and trading costs are very important.
But a "monkey with a dartboard" portfolio is a sampling from test 1, so, excluding those costs it will have returns which asymptotically approach the test 1 returns as the number of positions rises towards 500.
(the sampling result does depend on how you handle carryovers from one period to the next...do you throw the dart again or merely rebalance the current pick?...the trading costs will differ a bit, but the range of pre-trading-cost returns will be otherwise the same)
The performance of any given dartboard portfolio will certainly diverge from the all-500-equally-weighted test, but half of them will be higher and half lower.
With the bell curve centred 2.6256% higher than the cap weight return, the chances of beating the S&P are quite high, especially with larger numbers of stocks which narrows the dispersion.
Another rather obvious example:
Even with the drag of very much higher management fees, RSP (S&P 500 equal weight) has beat SPY (float-adjusted cap weight) since its inception 20 years ago.
From inception to end 2022, RSP led by 1.12%/year.
Another test:
I tested a slate of 150 randomly selected stocks from the Value Line 1700 database, requiring them to be stocks with the "Timeliness" criteria populated (some financial history and not currently with a merger/bankruptcy pending).
On average 1493 eligible stocks. Essentially all large caps, a good fraction of midcaps, and a smattering of small stocks.
Hold for a quarter and repeat, no allowance for trading costs in this test.
Out of 20 random runs 2003-2022, the 150-stock random portfolio beat the S&P 500 20 runs out of 20.
Out of 20 random runs 1986-2022, the 150-stock random portfolio beat the S&P 500 19 runs out of 20.
As to your other question, no, there's nothing magical about rebalancing once a year.
I often mention it because a lot of Americans read the board, and there is apparently a tax advantage for longer holds unlike most other jurisdictions.
Yes, generally all the other rebalance periods also beat cap weight. Shorter holding periods generally work better if your trading costs are well constrained, but it's not a big thing.
Based on the evidence one should avoid cap-weight, provided you maintain good control of the fees+trading costs.
If nothing else, equal weight is vastly lower risk.
Apple is 0.2% of RSP but 7.7% of SPY.
Measured by peak capital at risk in a single company, $100 of SPY has the same risk as $3845 worth of RSP.
Jim
