Maximize portfolio diversification with a minimum amount of stocks

20th Aug '15 by AnonymousUser192157

37 comments 23573 reads

Dear all,

As investors, we are all concerned about portfolio diversification. Personally, I follow a highly concentrated strategy (5-10 stocks), but that does not mean I do not investigate portfolio diversification.

But what is the best way to achieve diversification? The academic literature is filled with articles that suggest a minimum number of stocks needed to lower positive correlation among stocks in the portfolio. Typically, investors are told to hold at least 25-30 stocks to achieve some form of diversification.

However, what is often left out of the discussion is which stocks to pick and how this influences the diversification benefits. I believe that a good amount of diversification can be achieved by picking stocks across 5 well-chosen industry groups.

More concretely, I have found that over the last 10, 5, 3 and 1 year, consumer staples, health technology and health services have been the best performing industries, trumping the overall market on both return and risk-adjusted return basis by quite a bit (a Sharpe Ratio that is at least 0.5 higher). At the same time I find that these industries are not very highly correlated with each other (especially health services vs. the other two). Implementing K-Means clustering algorithms on the industry portfolios, I find a very good result for 5 industry groups:

1) Technology (both electronic and health)
2) Health Services
3) Retail/Consumer
4) 'Heavy': energy, process industries
5) 'Boring': finance, communication, utilities

My finding is that if you pick stocks across these 5 industry groups, you can already achieve a great deal of diversification with a minimum amount of stocks. Hence, the benefits of a concentrated high-conviction portfolio can be combined with the risk-reducing benefits of diversification. I also see this working quite well in practice.

But now I am wondering. What does the Stockopedia community think of this? Just a research curiosity, or actual investable research?

Best,

Tim

Unlock with Email

Unlock with Google

Already have an account?
Login here

Graham Neary

21st Aug '15

Thanks Jane!

mrwhits

Hi Ed,

was wondering if you had read any of these?

http://www.newtraderu.com/2013/08/06/twelve-of-the-biggest-trading-losses-in-history/

I'm not saying this will happen to you, but how many "pairs" would have to go the wrong way before you would be wiped out?

Regards

Edward Croft

Tim, from a top-down perspective you are right. Many of the best performers in the market are indeed from those sectors. I think Morningstar have done analyses of the number of economic moats in each sector and certainly there are proportionally more in healthcare, defensives, consumer & technology sectors.

Saying that though some of the best StockRank performers in UK portfolios this year have come from the sectors you are shunning... e.g. Adept Telecom up 85% year to date, Dart Group and Wizz Air, airline stocks both up more than 50%, International Greetings up 100% which is in the basic materials sector. And in the financials sector you'll find lots of fund managers which can be great moat businesses.

I'm not saying that your approach isn't right, it's just that I personally believe there is more power in a bottom up selection approach that diversifies sector biases, than a top down sector based one that outright shuns whole categories of stocks.

Saying that though I'll probably ditch my utility exposure next year as it really isn't adding much right now.

Joel Greenblatt in "You Can be a Stock Market Genius" describes how he used LEAPS options to devastating effect.

Lifecyle Investing by Ayres and Nalebuff explains why people should use leverage to access stock market exposure in the decades prior to retirement, and shows how to implement this with call options.

But if trading stocks on margin is the only route to leverage, then rolling over single-stock futures tends to be much more cost-efficient than the daily bets.

As an aside, I think it's funny that banks won't extend personal loans for direct investment in stocks. But they will extend plenty of credit to (and refinance) massively leveraged homeowners, even if those homeowners are choosing to spread bet instead of pay down the huge principal on their mortgage!!

AnonymousUser192157

Dear Graham,

Thanks for the excellent comments! Obviously, I agree with everyone in this discussion that different things work for different people. And thank god! The Stockopedia forum would not have been as interesting otherwise.

I just wanted to come back on two things you mentioned. First is the graph on portfolio diversification. Those are indeed the kind of graphs you see all over the place, but I generally do not like them. Why? Because they simply look at individual stock return volatility as a measure of risk, which is quite unsophisticated. As we all know from introductory financial markets courses, portfolio risk is stemming from correlation rather than volatility. That's also why I revert to simple correlation charts and more advanced correlation clustering algorithms to get a sense of how concentrated my portfolio really is.

This also brings me to the second point, where you mention that risk and volatility aren't the same thing, which is obviously correct. But as an academic myself, researching market efficiency among other things, I just wanted to point out that academics typically think in the factor-bet paradigm. They are not saying higher volatility will yield higher returns. It has just been shown that certain factor bets have generated returns in the past (size, momentum, value, ...), and they assume that if you outperform the market you are just overexposing yourself to these factors. And in fact, just looking at the StockRanks system this should be very true. Although, it's up to us to assess whether this is really increasing risk, or just a very common sense thing to do.

Thanks again for the interesting comments!

Dear Ed,

Thanks again for the insightful remarks!

Just discussing the idea here has already made me think further and deeper to improve the overall strategy. Let's see what the results are in a couple of years.

ExpectingValue

Tell me about it - and personal loans are available at a 3.6% rate if you want to spend on a depreciating asset like a car, or even on a holiday with no tangible residual value (not that I am denigrating holidays!), but they're not keen on lending to invest.

I find it bizarre. Mainstream banks are about as stubborn and inflexible as can be.

pka

22nd Aug '15

Ed,

This is just an idea I've had on your long-short leveraged spread betting portfolio. I know it's worked very well for you so far, but I think there is a risk of it performing badly when there are major changes in market sentiment that cause some sectors of the market to suddenly come back into favour after being out of favour in the recent past. I wonder whether you could reduce that risk by matching long and short positions within each market sector, or even within each 'industry group'. For example, you might take a long position in a company with a high StockRank in the 'Speciality Retailers' industry group, and match that with a short position in a company with a low StockRank in the same industry group. Then do the same in a different industry group, and so on. That would protect you against a sudden change in market sentiment causing one industry group to start doing better than another. But you would still have the benefit that high StockRank companies will hopefully do better than low StockRank companies in the same industry group or sector.

Pka - that's essentially what I do, but I'm doing it at the sector level (of which we have 10) rather than at the industry group level (of which we have 54). Each sector is thus matched equally long short. I've been considering doubling the size of the portfolio to 50 longs and 50 shorts and doing the matching at the more granular industry level. There's a lot of academic research that shows you can improve factor returns through industry level long shorts - so it's been on my mind to test it in the wild. At the industry level though you often can't find such extreme ranking pairs.

The thing to remember is that if you are leveraged and trade a lot the spreads eat you up. The average bid ask spread on the stocks I'm trading is over 1% - so lever that up 5x and the round trip cost of a trade is 5%... Clearly I don't want to be churning the portfolio at that cost too frequently. I've found at rebalancing that being more forgiving on sell rules keeps the costs of trading lower. So I'm selling outright when the long position StockRanks fall under 90 and closing shorts if it rises above 50. Over the last 6 months less than 20% of positions have thus been completely closed - the rest being shaved.

I'm well aware of the risks here. What I'm doing is seeking the spread in returns between high and low StockRanks. As long as the high ranks outperform then all is good, but if there's a regime change and the factor returns invert then yes it could cause problems. In 2007 there was a famous 'Quant Meltdown' where all the Hedge Funds doing this kind of factor arbitrage trading saw big reversals. All the quality, value and momentum premiums inverted at once - the trades had become crowded and there were forced liquidations. Some funds went bust.

This historically has been rare but worth remembering. I'm managing my liquidity risk and not betting the farm - I can exit these position sizes easily in case of trouble and I'm actively controlling the exposures in quite a dispassionate way. Rebalancing feels like a harvest and resowing procedure rather than the heart in the mouth feeling you often get if trading a 5 or 10 stock folio. A single stock can't hurt me too much nor can the indices, it's other subtleties that could. Will keep you posted.

Mechanical Bull

Peter Lynch talks about the importance of looking at how strategies work over very long time periods. So I would argue that ten years is not really long enough to make these kind of conclusions and doing this analysis at different times over the past 10 to 15 years would have yielded different results. Tech stocks looked great in the late nineties and then crashed. Mining stocks were looking good four or five years ago. Energy stocks were a good bet until about a year ago. I would be careful in assuming the future to be a continuation of the recent past (i.e. within the last 10 years), because a) these expectations are usually built into the share price, and b) something else happens.

In any case it would be useful - if you have the data - to replicate your analysis if you had done it say, 5 years ago.

mpat89

23rd Aug '15

Ed you wrote "Often I have to try several shorts before I can find one that will stick."
Please can you elaborate?

Have you considered weighting your position sizes based on Beta to give a portfolio with hypothetically zero market risk?

To illustrate 'expensive cost of leverage', Ed's strategy of using 5x gross leverage will incur financing costs of 12.5% per annum using IG Index - undeniably a substantial chunk of change!
Relatively speaking this is only 2.5% of the gross position value but may cause discomfort during an extended drawdown.

Novice Investor

mpat89,

Did you include the rollover cost every 9 months?

NI

It's based on DFB's so there would be no rollover cost. I can't find any transparent details on financing costs included in long dated bets.

underscored

Would IG levy a financing fee on the entire bet, or just the exposure not covered by your deposit?

justinian

24th Aug '15

It seems the entire position has to be financed. Shorts can cost too. This is in addition to the spreads which they add 0.1% either side of the market buy and sell prices (i.e. 0.2% widening of the market available spread).

https://www.ig.com/uk/spread-betting-cfds-charges

The example of the costs is:

Longs - you pay 2.5% on base rate i.e. you pay 3% on long positions currently.
Shorts - you receive base rate less 2.5% i.e. you pay 2% on short positions currently.

I've looked at this strategy in the past because on the surface it seems appealing. However, much of market returns are either at the entire market, industry or sector level and only a fraction at the intra company level - certainly not sufficient to my mind to cover the sorts of costs of holding that you would bear with ig (basically 5% for each long short pairing plus the cost of the spread). However, it's great if you can negotiate with a friendly broker (typically in America) to reduce these costs to make it work better.

The other thing that concerned me, particularly with the high leverage that is required to boost those potentially small differential intra company returns, is what happens if the behaviours of stocks alter such that they are no longer inversely correlated.

For instance, I've just had a quick check of the various ranking systems to see what sorts of ranks are currently having the biggest corrections using the performance charts. Somewhat counter intuitively, the high stockranks and value ranks seem to have experienced the biggest corrections in the last few days. I can only think this is because the lower stock ranks have already been on a downward trajectory and are a little washed out with not so much more to give up. However, the point is that under stressed conditions they are all moving the same way unlike the long run divergent path since inception where high ranks were positive and low ranks were negative. No doubt this will resume, but in the meantime at 5x leverage that experience can be painful. I'm not entirely sure what 5x leverage means either - is it 2.5x long and 2.5x short, or 5x either side?

I would be interested to hear from Ed about the costs he finds he's bearing and what % they represent of the overall return and also how the long short portfolio has behaved of late. Finally, in terms of uniform drawdowns in correlated corrections, what sort of drawdown could be tolerated before the dreaded margin calls come in?

You could not be more right, and I have not been considering this enough. Below the two snapshots (2010-2015 vs. 2005-2010).

As you mentioned, energy/utilities was looking far better back then.

What would be extremely helpful here, is a good aggregate stockrank metric for industries, so to spot these kind of changes as we are looking at them (not through a backtest). For now, I will still overweigh consumer/health/tech, but I will be more cautious in not considering energy/utilities/'boring' stocks because today they seem like a bad idea.

Thanks for the great feedback!

http://www.stockopedia.com/sectors/ ?

Certainly a good starting point, but there's obviously a lot of limitations to a simple median value calculation. I think I saw a more sophisticated aggregate market/sector approach pass by in the suggestion box.

rahulstockopedia

23rd Dec '15

Re: post 26

Edward

1) Do you use individual stops when running this portfolio, or do you just use QVM deterioration at the time of rebalancing?
2) How often are you rebalancing?

Thanks.

Maximize portfolio diversification with a minimum amount of stocks

Unlock the rest of this article with a 14 day trial

About the Author

AnonymousUser192157

37 comments

Unlock the metrics for all these stocks and more...

Unlock the metrics for all these stocks and more...