A Change of Strategy

Well it's safe to say it's been a long time since my last post just over a year ago. Plenty has been happening during that time - I've been too busy enjoying life and learning to put my thoughts to paper, but for some reason I felt like writing up a new piece today. A warning, it's fairly lengthy. For any of you diehards who are still following my blog, I wouldn't hold out for another post soon, but I definitely appreciate your continued patronage and will try to write up any noteworthy thoughts I have. 

One of the best things that happened over the past year was my trip to the Berkshire Hathaway Annual Meeting in late April and early May. It was an fantastic trip and something I've wanted to do for a long time. Yes, the opportunity cost of the travel expenses is probably worth millions by the time I'm old, but to me the opportunity to hear from Warren Buffett and Charlie Munger in person (and even get a selfie with Warren!) is priceless. To the right is one of the photos I took of Buffett. You can read a little bit more about my trip in a little piece I wrote for Forager. 

The main topic I wanted to write about was my change in investing approach. Don't worry, it's nothing drastic like becoming a day trader - if I do, you'll know I've lost my marbles - but it's a fairly substantial change. I now have a much greater appreciation of the power of mechanical investing strategies. This involves using a fairly simple rule or rules to pick stocks. One of the most well-known mechanical investing strategies was first outlined by Joel Greenblatt in The Little Book That Beats the Market, which he named the 'magic formula' - yes, a dodgy sounding name but the formula does actually work very well. Using just two ratios, for value of the stock and quality of the business, Greenblatt reported that a portfolio of the top 30 stocks ranked by his magic formula would have returned 23.8% p.a. from 1988 to 2009, compared to the market return of 9.6% p.a. Subsequent research and performance casts a little suspicion on the veracity of his backtesting but whichever way you cut it, the formula outperforms the market. 

Other very simple strategies that implement value investing systematically include buying stocks that have a low price in relation to book value, earnings, dividends, revenue etc. While some work better than others, it's rather astonishing to see how well such a naive strategy performs. If you're interested in reading more about this I highly suggest you take a look at Tweedy Browne's excellent paper, What Has Worked In Investing, Wesley Gray and Tobias Carlisle's book Quantitative Value, and Tobias Carlisle's book Deep Value: Why Activists and Other Contrarians Battle for Control of Losing Corporations. 

You may be wondering why these mechanical investing strategies work so well and why you don't see every investor/fund manager adopting them. They work so well because unlike us humans, these strategies are perfectly unbiased and ignore irrelevant information that often clouds our judgement. It doesn't matter how terrible or attractive a particular stock may look to a human, these strategies weigh up the critical information (the price in relation to value) and stick with it, allowing them to purchase really cheap businesses and stay away from expensive ones. It's value investing implemented systematically, protecting us from our own biases and allowing us to take advantage of other people's biases. As for why it isn't widely adopted, I suspect that it's hard to convince people to put their savings (and pay fees) to a fund manager that lets their computer do all the hard (or rather easy) work. Perhaps more important is what Charlie Munger refers to as Excessive Self-Regard Tendency - we think we're all above average drivers (90% of Swedish drivers think so), we tend to buy more lottery tickets if we can pick the numbers, surely we can beat a dumb formula at stock picking. Personally speaking, I can also attest that it's hard to admit that a formula is likely to outperform you, despite all the hard work I put into analysing a business. 

However, I've decided to put my wallet ahead of my pride and largely stick to a mechanical investing strategy. This is not to say that researching businesses and using expert judgement to pick stocks isn't a good approach - one only needs to look at some of the most famous investors to see that it can work extremely well - but I believe that far too many investors are fooling themselves into thinking they can pick stocks better than some of the strategies mentioned in the Tweedy Browne paper. I'll still be open to exceptional ideas that don't qualify for my mechanical investing approach, indeed I've bought one recently that I might detail in another post sometime. As for what particular approach I've decided to go with, the net-net strategy is the most appealing to me as I'll try to explain next. Many of you reading this who are value investors would have heard of the net-net strategy, but for those of you who don't, I'll try my best to explain it. 

Way back in 1934, Benjamin Graham and David Dodd described the net-net strategy in Security Analysis. The idea was that if you could buy a business for a price substantially below its liquidation value, you would do pretty well as an investor. As a rough proxy for this liquidation value, they calculated a 'net current asset value' (NCAV) as follows: current assets minus total liabilities and preferred stock. While the current assets (typically cash, receivables and inventory) would likely not be worth as much as stated on the balance sheet in an actual liquidation, the idea was that the sale of the rest of the non-current assets (mostly property, plant and equipment) would usually make up the difference. Most businesses have more total liabilities than they have in current assets so have negative NCAV and can be ignored for this strategy. Graham and Dodd's recommendation was to purchase a stock when the price of the stock (or market capitalisation) was less than two-thirds of this NCAV, or alternatively stating that, when the NCAV to price ratio was 1.5 or greater. For example, if the NCAV of a business is $30 million, you'd buy the stock if its market capitalisation is $20 million or less. This buffer between the NCAV and the price of the business is the margin of safety that value investors require. 

It's a very simple, straightforward mechanical investing strategy described over 80 years ago. I had assumed until recently that these net-nets were largely extinct due to the ability of computers to easily ferret them out in recent decades. It turns out that my assumption was wrong and surprisingly there are still hundreds of net-nets in developed markets today. In addition, I believe I had fallen under the illusion that many other value investors would have had - nowadays Warren Buffett tends to purchase high quality businesses hence many people want to emulate his current approach, but his highest returns were many decades ago when he bought these net-nets and similarly neglected, poor-performing businesses. He's even said in recent years that if he were young and working with much less money than he is today, he'd do pretty much what he did in the past again.  

So how well does this strategy perform? Benjamin Graham implemented it his Graham-Newman fund from 1930 to 1956 and reported an average return of 20% per year from buying these net-nets. That's a pretty juicy return, but how does the strategy stack up with more rigorous testing under different time periods and different markets? Luckily, a number of academic studies have fairly comprehensively answered that question. It performs unbelievably well. I'll give an overview of the findings of these papers (feel free to skip over it if you take my word for it), but for a more comprehensive review I suggest you read the actual papers. 

In one of the earlier studies, Ben Graham's Net Current Asset Values: A Performance Update, Henry Oppenheimer looked at the performance of net-nets in the US stock market over the 1970 to 1982 period. Once per year, a portfolio of net-nets was formed - that is, the NCAV to price ratio of the stocks chosen was at least 1.5. One year later, the portfolio was sold and replaced with a new one. He found these net-net portfolios compounded at 28% p.a., whereas the market as measured by the NYSE-AMEX benchmark returned just under 11% p.a. He found that the greater the ratio of NCAV to price (i.e. the cheaper the stock), the higher the return. Interestingly, he found that the stocks that had positive earnings in the last year actually underperformed those that had lost money in the previous year. In addition, he found that of those stocks that had positive earnings, those that paid a dividend significantly underperformed those that didn't. Those last two findings are pretty counterintuitive - most people, including myself, would have expected that a profitable, dividend paying net-net is more attractive than an unprofitable net-net that doesn't pay dividends, but in fact it's the exact opposite. 

In their paper, Ben Graham's Net Nets: Seventy-Five Years Old and Outperforming, Tobias Carlisle, Sunhil Mohanty and Jeffery Oxman essentially extended Oppenheimer's study to the 1984-2008 period. The results were even more impressive, net-nets overall compounded at 35% p.a. compared to the market at 11% p.a. In addition, they largely confirmed the other findings of Oppenheimer with a small exception. While the higher NCAV to price stocks tended to perform better than the more expensive ones, the very cheapest quintile of net-nets recorded a relatively disappointing 23% p.a. compared to the second cheapest quintile of 53% p.a. There's probably something not quite right about those net-nets that appear to be the very cheapest according to the net-net criterion - in my experience I've found that many of these are either in bankruptcy, delisting, their assets have disappeared since the last financial statement date, or there are substantial non-controlling interests so not all the assets belong to shareholders. Putting this aside, the researchers found confirming evidence that profitable net-nets underperform their unprofitable peers - 27% p.a. compared to 49% p.a. Likewise, the profitable, dividend paying stocks underperformed the profitable, non-dividend paying stocks 19% p.a. to 33% p.a. This is pretty powerful evidence for the performance of the net-net strategy - with 1362 net-nets over 26 years, there is strong statistical significance. Furthermore, digging down into the yearly returns shows that despite the prevalence of computers in recent years that should make such a strategy much easier to implement, net-nets have continued to outperform in the 21st century. 

What about outside of US markets you might ask? Fortunately, James Montier has us covered here in Graham’s Net Net’s: Outdated or Outstanding? where he analysed the performance of net-nets in three regions, the US, Europe and Japan. Between 1985 to 2007, he found that the global portfolio of net-nets returned 35% p.a. In the US, the result was over 40% p.a., in Europe just under 20% p.a. and in Japan 20% p.a. Keep in mind that during this period the US share market performed much more strongly than the European and Japanese markets, so this likely explains the absolute performance differential. I don't see any reason why net-nets wouldn't perform well in other developed share markets. 

I could go on and on for much longer about other studies corroborating these results but hopefully you get the idea by now. It's worth noting that the net-net strategy won't outperform the market every single year and is still subject to some gut-wrenching declines like those which occurred in the GFC. Moreover, actual returns from implementing it would likely be a few percentage points lower than stated in these studies due to transaction costs and difficulties in purchasing illiquid stocks. I'm willing to put up with that but of course it's not for everybody. On the plus side, this is a relatively simple and time efficient way to invest - you don't need an IQ of 200 nor spend all night researching stocks. All you have to do is run a stock screen to find these net-nets and purchase them - it can be done in less than an hour, although I spend much longer checking out each business. If you're really keen and would like to look at the results of some other studies, this is a good place to start (keep in mind the methodology varies a little bit across studies but the conclusion is still the same, net-nets drastically outperform). So we've established that this simple strategy works across markets and across different time periods, with some results indicating roughly 50% p.a can be achieved. But why stop there, is it possible to do even better? At risk of stuffing up a very good strategy, I think we can. 

In addition to the NCAV calculation described above, Benjamin Graham also suggested that you could be even more conservative in your calculation of liquidation value. Given that in a liquidation, you might not get all the receivables owed to you and you may have to sell inventory at a big discount, Graham came up with an alternative, which I like to call 'discounted NCAV'. You take cash + receivables x 0.75 + inventory x 0.5 - total liabilities - preferred stock. You're assuming that receivables are only worth 75% as much as stated on the balance sheet and inventories 50% of their balance sheet value. This tends to favour cash rich businesses compared to the previous formula. Once again, you purchase if the discounted NCAV to price ratio is 1.5 or greater. It's a pretty blunt approach and some people prefer to use different discount factors for receivables and inventory, but intuitively it should work better than the previous formula as you're demanding a greater margin of safety. Even fewer stocks meet this discounted NCAV criterion than the previous one but it appears that this greater conservatism leads to greater rewards.

For this we turn to a relatively hard to find paper, Dissecting the Returns on Deep Value Investing, once again conducted by Tobias Carlisle, Sunhil Mohanty and Jeffery Oxman. They investigated the performance of this 'discounted NCAV' approach over the 1975 to 2010 period. To qualify, stocks had to have a minimum price of $3 per share, and another portfolio tested performance only for stocks trading for at least $5 per share. Instead of requiring a discounted NCAV to price ratio of 1.5, they only required a ratio of 1.0, a less stringent requirement but perhaps they did it to obtain a greater sample of stocks. You'd expect if they tested it on a ratio of 1.5 and ignored the arbitrary minimum price requirement, the performance would be even better. Here's the unbelievable bit: net-nets in the minimum $5 price portfolio returned a compound 75% p.a. and those in the minimum $3 price portfolio returned 85% p.a! I've never seen anything like that before, and in the back of my mind I still doubt the validity of those numbers as they appear too good to be true. Nevertheless, at the very least I would expect such a discounted NCAV approach to do as well as the traditional NCAV approach that most papers have studied. Therefore, this is the strategy that I'm going to stick with. 

It's tempting to say, "Well that's fantastic, I'll just take that list of net-nets and pick a few that look good to me". However, this kind of cherry picking is very tricky to get right as more often than not, we tend to hamper the performance of mechanical investing strategies by using our own judgement. See James Montier's Painting By Numbers: An Ode To Quant for an excellent description of the applications of simple statistical models to improve results outside of investing and evidence that we should try to avoid interfering with the models. In addition, I highly recommend Daniel Kahneman's book, Thinking, Fast and Slow to better understand how our brains work and read about more evidence detailing the usefulness of simple statistical models for decision making. 

Even going back to Joel Greenblatt's magic formula, Greenblatt set up a website that either allowed subscribers to access the list of magic formula stocks and choose to invest from that list, or he managed the portfolios for investors by strictly adhering to the formula. From 2009 to 2011, those that self-managed their portfolios had a cumulative return of 59.4% compared to the S&P 500 of 62.7% and the professionally managed portfolios of 84.1%. Greenblatt says those that cherry picked from the magic formula list systematically avoided the best performing stocks because they looked the scariest and most ugly. In addition, they also timed the market poorly, selling stocks after poor performance and buying after good performance. This goes to show that cherry picking can eliminate all of the outperformance of a winning strategy and even lead to underperformance relative to the share market. Nevertheless, I think that very limited human intervention can improve results. In the case of this net-net strategy for example, simply checking that the stocks aren't about to go bankrupt and confirming the NCAV calculation is correct are no-brainers that I believe will improve performance.

In my personal net-net approach I've added additional filter rules based on empirical evidence, for example, as mentioned above, the unprofitable net-nets perform far better than the profitable ones so I'm filtering out any profitable net-nets. In addition, I have some extra quantitative factors that I can apply objectively and have also been proven to improve performance. For instance, if a net-net is buying back lots of shares I'll give it a few bonus points to its NCAV to price ratio, and if it's issuing tons of shares, I'll subtract a few. I won't go into detail about these extra factors as this post is already getting quite lengthy, but they generally don't make much of an impact on my decisions - the NCAV to price ratio is all important. I've decided to pick the top five stocks each quarter to eventually come up with a portfolio of 20 net-nets which should provide more than enough diversification. My first five picks were made a few weeks ago - these averaged a NCAV to price ratio at the time of purchase of 2.55, which is well above the minimum requirement of 1.5, so my margin of safety should be much higher than the studies listed above. There's no guarantee they'll do as well as the academic studies have demonstrated - I've made some adjustments with the goal of improving performance but none of the studies have explicitly tested my exact rules. While I certainly don't expect 75% p.a. returns, I'm confident the results will be good, so I'm putting my money where my mouth is. I'll keep you posted on the performance. 

The five stocks I've purchased so far are largely listed in Canada, have lots of cash and very few liabilities. The names are S i2i Limited, Africo Resources Limited, Petrofrontier Corp, INV Metals Inc and Black Iron Inc - I'll let you research them yourselves if you're interested. It seems hard to imagine that you can buy these stocks that are trading at around half their cash (after all liabilities have been deducted). There is no way in the world someone would sell you their entire business for prices like these - they'd call you crazy and tell you to get stuffed. But in the share market, small ownership stakes in companies can and do trade at ridiculously cheap prices. Proponents of the strong-form efficient market hypothesis ought to take a hard look at these examples as they're clearly inconsistent with the notion that all securities are priced rationally. 

After all this, you may be wondering why I'm willing to lay out this strategy for everyone to read and possibly implement - after all, shouldn't it create more buying and selling competition for these net-nets? Well I don't think it'll cause much, if any, competition. The obvious reason is that barely anyone reads this blog! But even assuming I had a large audience, only a very select group of people would be comfortable with such a strategy and follow through with it. First off, the vast majority of net-nets are small, illiquid stocks so it is near impossible for anyone with millions of dollars to buy them up in decent quantities. That eliminates virtually all of the smart professional investors that would no doubt love to adopt such a strategy. Once you get to a portfolio of a million dollars or so, I suspect that this net-net strategy will be difficult to implement, but it's a good problem to have too much money. Secondly, as previously discussed it goes against human nature to cede control to a simple formula or rule even though it's proven to do better than human judgement, thus this net-net strategy will appear too scary to most people. Thirdly, you need to be a value investor that understands why this method of investing works - of which we are in a minority. Finally, some people will look at these net-nets and view them as garbage as they've often been terrible businesses - but that's fine with me, as they say, one man's trash is another man's treasure. Much of the reading I've done has helped me realise that it's the ugliest, scariest businesses that outperform the most, largely due to mean reversion as business performance improves and people realise the stock isn't as crappy as they thought - the typical net-net being a good example. But that's a topic for another time. In summary, the net-net secret has been out for more than 80 years, I doubt people are suddenly going to start paying attention to it now.