1. In a random portfolio that ended up with 10 stocks going down in price (some of these were positive when dividends were added) 1 that stayed flat(positive with divs) and 9 that went up in price, when all dividends were added to the pool during rebalancing, yet just added to the total for buy and hold, the buy and hold finished 27% ahead, before commissions and taxes were added to the cost of rebalancing.
Rebalancing quarterly adding dividends into the mix for the rebalance, finished at $570,541, before subtracting costs for the rebalance, roughly $400/Qr for commissions. As you would be selling winners only (usually), there is a considerable tax payable as well. There was no need to work this out as it subtracts from rebalancing even more. For example the selling of winning shares in the first quarter alone was $25,898, despite the portfolio being down $6000 in total at that point. You would be creating a current tax liability while losing money on the entire portfolio. Not smart!
2. By the time you added commissions and taxes to rebalancing, while allowing buy and hold to go into Cap gain lower taxes territory,
3. Rebalancing, which is really stealing from winners to add to losers clearly does not work in the real world where stocks go down and some stay down.
4. If the universe of stocks included only the stocks that went up in price, say the current ASX20 from 10 years ago, then yes it probably works.
5. Is anyone prepared to say the current ASX20 will be the same in 10 years time? I'm certainly not.
The stocks used were STO, BSA, OKN, GAP, NBL, MND, AGK, BHP, CAB, NWH, GUD, SHV, ROC, CYG, TGR, RDM, AGO, MGX, NCM, NST. The start date was 9/2/2012, finish 10/2/2014. Closing prices were used.
6. I was skeptical of the gains claimed, now I'm absolutely certain.
7. Thankyou RY for making me do the exercise, it was enlightening.
Wow. I certainly am impressed.
Let's examine the results. I have used the same universe as you have conducted the analysis from. Very good of you to allow your tests to be examined. Research dates are 1/1/2012 to 31/3/2014 to allow clean quarterly rebalancing but otherwise are a representation of the same investment period. I trust they are satisfactory from a survivorship or whatever bias given you were prepared to use it. The data included dividends reinvested as per research convention. They are easily obtained from the ASX or your favoured data vendor.
1. I mentioned earlier, that every portfolio history is just one journey from a garden of forking (got to check my spelling) paths. Just one. And this assertion is a probabilistic one. This message hasn't come through. So let's step it out.
I have taken the selection universe of 20 stocks. We form portfolios in 10% blocks. So each portfolio we are testing contains 10 blocks of 10% initial weight. We then truly randomly build portfolios consisting of 10% blocks taken from the universe of 20. One path in a garden of forking paths.
This simple Monte-Carlo selection method for building portfolios allows for several paths...each of which is at least as valid as the one random portfolio reported. Feel free to question it...it just is. But just in case this helps, I have run the numbers for a 20 unit portfolio, each of 5% which allows up to 20 stocks in the portfolio and the results are stronger than the ones I'll show below.
How many paths are there for the really simple 10 unit portfolio? So many that, if we were to process a portfolio per millisecond (one-thousandth of a second), and assuming the age of the universe is 13.8 bn years as is generally assumed by astronomers and physicists at present, it would need 630 million times as long to check every one. That's a lot of forking paths to explore.
So, given I can't spend all day, let alone 630 million universe years (sure makes dog years pretty trivial), doing this stuff, I'm going to do what everyone else does who understands stochastic processes and has not been able to communicate closed form solutions despite copious efforts - we're going brute force. This is called a Monte-Carlo simulation. It just tries to bring to life what is hard to comprehend behind some squiggly lines full of symbolic stuff.
So, I'm going to make a horrific assumption that a sample of 10,000 portfolios consisting of up to 10 stocks will do and we'll run it through. Oh, if you checked the trivial example of equal weighted stocks in the entire universe at commencement to see where that went, a portfolio of equal validity to any that could be chosen, you'd find that the cumulative return of the unbalanced portfolio was 35.12% vs the rebalanced result of 39.17. Not to shabby. Anyone out there can verify it. Now we have one random portfolio vs another. Reds-1 vs Blues-1.
So, what do the results reveal? From 10,000 (since I ran the 20-unit portfolio too, it is actually more like 20k vs 1 - and Themistokles thought he was having a bad day) draws vs 1. Well, blow me down...the average unrebalanced portfolio return was 34.99% over the period. This compares unfavourably with its MATCHING rebalanced portfolio counterpart which outperformed by 39.21%. Again, not too shabby over a short period like we are examining.
The chances that this is due to luck as developed by 2-sample t-test on a two-tailed basis (given rebal is being contested, we test for 2-sided instead of one. But what the heck, check the p-tests if you want to quibble) is less likely than drawing 4 ROYAL flushes in a row from a clean deck of cards. That's pretty special. $1 per draw down at Monte-Carlo and four hands later you'll have a couple of million bucks. Keep your winnings in play, assuming anyone could stake you and you'd be richer than all the universe....and then some. Better than Rain Man. Also, the likelihood of the rebal portfolio achieving a superior outcome to the unrebalanced portfolio was 60%. That's a whopping edge when sample sizes are this big, but I am talking to Rockstars (not you, unless you voluntarily wish to be classified as such) who think in terms of certainty - so we'll move on.
There is overwhelming statistical evidence that rebal works in the universe which has been specified and the time period used. This result overwhelms any inference that can reasonably be drawn from a single random portfolio. Four thousand portfolios created in this sample underperformed as well. But six thousand outperformed. This is a probabilistic process. Like everything else in investments except pure arbitrage. It is quite outlandish to draw an inference from a single draw in a stochastic process. It is like using Newtonian Laws in the world of relativity and quantum mechanics. Whatever world that is, it's not the real world. It's the stuff of Laplace. And it's so 17th century.
The longer the time period, the higher the probability of success.
2. Let's see. Didn't we already go through this on Commissions? $500k portfolio. Minimum expenses $20 per trade. Trade quarterly. - which is actually stupid because the people who do this don't just get smacked in the head by expenses but think about it. If we rebal all 10 stocks each quarter, brokerage is approximately 0.5% per annum. Tax? Let's say the stock market goes up 10% per annum or 7.5% real before franking. Feel free to muck about up to +2% before you start being put on CNBC as entertainment fodder. 4% is dividend. 6% is cap gain. This is rebal...it doesn't generate full portfolio turnover every tick of the clock. They are small movements. Pensions are not taxed - but let's make that explicit...NOT TAXED. Superannuation taxes cap gain at 15% worse case. Top Tax rate at around 50%. So let's be nuts. Full turnover in a year will generate 3% tax cost for top tax bracket. Full turnover in a year for Pension will generate 0.9% cap gains. If you know anything about tax at all, you'd know that you have discretion around tax parceling and can use modified HIFO standard, you will also reinvest dividends which creates 4% fresh parcels per year, and you might even have savings contribution going in if you happen to be in anything other than pension - and even then. Turnover is frequently reverse as might be expected. All this keeps your parceling fresh and allows low tax cost turnover for anything that is vaguely in the zone of the real world. Then, if you still want to hold out on this argument, you can lock out stocks with no fresh parcels from any rebal. Please see Ves correspondence for how it is actually done.
Woosh...hurdle cleared. If not, then why are you trading at all? Any trading activity outside of rebal freshens parcels. It is a valid concern for embedded capital gains in place. But parcels freshen with cashflows. If you have negative cashflow, you are probably in pension phase. If you happen to be in the situation of drawdown at max tax rate, there is a balancing act, but the basics apply.
3. Yes it does, welcome to the real real world. Take the Blue pill Neo.
4. In the universe of stocks provided, there were 7 that recorded negative returns over the period inclusive of dividends: BSA, MND, NWH, GUD, AGO, MGX, NCM. Building a trivial portfolio of equally weighting these and comparing the unbalanced vs rebalanced portfolios yields...are you getting tired of this yet (?)...unbalanced -31.57% vs rebalanced -27.37%. Rebalancing adds probable value to any portfolio, including those consisting of stocks that go down - including portfolios consisting entirely of stocks that go down - so long as they wiggle on their merry way to oblivion.
5. Per discussion with MichaelD, this process works if a monkey or astrologer, or monkey who is an astrologer, with a low IQ threw 20 darts and randomly selected stock names. New dart throws can take place every period. It doesn't matter if the monkey was trained by NASA. It doesn't matter if the stocks come from the top, bottom or inside out 20. They can be random numbers as long as they wiggle. Please go ahead and run 10,000 tests and verify this for yourself (you are not the monkey - whether astrologer or other). But in tests, they actually used monkeys to prove the point.
6. I wouldn't be so sure. But that's me. This is not bread and butter. Further, it is probabilistic. So there is always the chance that a single draw won't add value. There is no certainty in a world where Heisenberg exists. To reject rebal with certainty seems...somewhat unsupported.
7. Likewise. It's been fun. All the best with it.