Random thoughts about net nets

[rukawa]

I have been running backtests regularly on net-nets and related screens. An interesting thing I found was that the following screen returns about 80%  a year according to EQBT on bloomberg:

 

1) market cap - net cash < -0.3 * market cap

2) and price/book < 0.6

 

I define net cash as cash - financial liabilities. Note I don't include operational liabilities like trade payables etc. Including operational liabilities reduces the performance of the screen.

 

My reasoning about net-nets is that the reason they do well is that they are able to last long enough that some good thing happens and the market revalues their business. They avoid the typical problem with deep value which is that the debt destroys the company and the assets end up being worthless. So I started looking for companies with a good debt situation which could last a long time. Hence the first criterion.

 

The idea of the second criterion is that you want a stock that is cheap. This is a net net like screen. My view is that it excludes the net-nets which typically do badly which are those which have very little cash, plenty of accounts receivables and even worse inventory and also a lot of financial liabilities. A typical example of this is SHOS (Sears Hometown). The reason these net-nets are not great is that the inventory often ends up worthless and the financial liabilities destroy the company. I've run screens for SHOS like companies and they don't do well though the appear to be very cheap when viewed as net-nets.

 

A typical net net screen does well but it includes shit and sugar. I think as investors we aught to think more critically about what assets are cash-like and what liabilities are truly dangerous vs non-dangerous. And why exactly net-nets tend to outperform vs the ones that don't

 

I've begin to consider a way of analyzing which assets are truly cash like using time series of balance sheet data. Often with assets that are convertible easily to cash you will observe that one asset will decrease at the same time as the other increases. Their increases/decreases tend to be very strongly negatively correlated. A sum of such variables would tend to be relatively constant over time. Over time there must be a statistical/mathematical way to figure out which assets are cash-like. In some sense they are not linearly independent...they are linearly dependent. There are non-current assets that I also think are cash like...an example would be long-term security investments.

[frommi]

Thanks! My observation over the last 2 years has resulted in similar conclusions. A large cash balance and/or share buybacks are the things you want to see in a netnet for huge returns. (Still hold SHOS, but my patience is getting smaller every quarter)

 

I have some questions:

 

Was this a backtest with international data?

What was the rebalance timeframe?

Are the results stable when you exclude really low liquidity stocks like price < 0.1 and daily volume < 10k USD?

When you exclude 2009 and 2003 are the returns still that high?

[rukawa]

Thanks! My observation over the last 2 years has resulted in similar conclusions. A large cash balance and/or share buybacks are the things you want to see in a netnet for huge returns. (Still hold SHOS, but my patience is getting smaller every quarter)

 

I have some questions:

 

Was this a backtest with international data?

International. But excludes developing countries and China/HK. I'll extract my exact screening parameters and post them later on. Generally I've found that ex-financial improves performance..and ex China/HK does as well. But not a lot. My own practice mostly excludes China, Financials and even Hong Kong.

 

What was the rebalance timeframe?

Quarterly is my standard

 

 

Are the results stable when you exclude really low liquidity stocks like price < 0.1 and daily volume < 10k USD?

I only include stocks with market cap > 1m market cap which helps a bit with liquidity. I'm not sure whether I excluded dark companies. I'll have to try the other things you mentioned.

 

When you exclude 2009 and 2003 are the returns still that high?

Don't know and its fairly difficult for me to answer that so I won't. My screen was from 01/01/2000 to 01/01/2017 so it includes both the years you mentioned. But as far as I can see there are always some monster  years. E.g. investing from Dec 31, 1993 to Dec 31, 1996 would result in a 10 bagger based on a conventional NCAV screen. In fact look at this paper it appears that the killer time for net-nets was actually the 1990's (see Exhibit 3). I don't really see the point in excluding years like this.

https://www.valuewalk.com/wp-content/uploads/2014/10/benjamin-grahams-net-nets-seventy-five-years-old-and-outperforming-full-tables1.pdf

 

 

[frommi]

Don't know and its fairly difficult for me to answer that so I won't. My screen was from 01/01/2000 to 01/01/2017 so it includes both the years you mentioned. But as far as I can see there are always some monster  years. E.g. investing from Dec 31, 1993 to Dec 31, 1996 would result in a 10 bagger based on a conventional NCAV screen. In fact look at this paper it appears that the killer time for net-nets was actually the 1990's (see Exhibit 3). I don't really see the point in excluding years like this.

 

Thanks for your answers. I asked about the years because i am under the impression that these type of returns come mostly from the years after a larger market correction, so its probably not something to start after an 8 year long bullmarket. Thats the reason i reduced my NCAV portfolio (from 80%->20% of the whole portfolio right now) and put more money into low beta/high dividend stocks. Maybe at the end of the year i will realize that that was a mistake, but i don`t really find a lot of good netnets right now and i really don`t want to hold crappy cash burning companies in a bear market. My plan is to switch gradually back into NCAV stocks when some good ones re-appear on my screens.

[frog03]

1m market cap seems way too low to be meanindgful.  What results do you get if you bump this to say 50 mn?

Thx

[rukawa]

1m market cap seems way too low to be meanindgful.  What results do you get if you bump this to say 50 mn?

Thx

 

What is your objective or reason for increasing the threshold to 50m? And why is 50m more meaningful than 1m?

 

I'm not going to bother to check because I basically know it would result in much lower returns, more volatility since it would reduce the quality and number of net-nets...maybe to a handful. I would guess that I could find 50 net-nets right now...but maybe <10 satisfying your criterion,

 

I've invested in sub-1m companies. Its harder but its possible. >2.5 m I generally have no problems as long as the company isn't dark. To me >50m is a ridiculously high threshold that would exclude a lot of very good net-nets. Including many very high quality, high liquidity net-nets in Japan. Not sure what it accomplishes except to lower returns and make the strategy completely untenable.

 

 

[frog03]

1M market cap this gives you what 5K daily volume, maybe much less.  Unless you have a very small portfolio, this is not going to work. 

[Hielko]

1m market cap seems way too low to be meanindgful.  What results do you get if you bump this to say 50 mn?

Thx

 

What is your objective or reason for increasing the threshold to 50m? And why is 50m more meaningful than 1m?

If you include companies with $1 million market caps you will have stuff with almost zero liquidity and super wide bid ask spreads. If your results for example include a company where someone sells 100 shares at $0.0001 and next year there is some random trade at $0.01/share for 100 shares as well it looks like a 10,000% return that will totally skew any result. But even if you managed to get 100% of all traded volume you would have made a profit of less than a dollar. That's why if you want to run a meaningful backtest you have to add liquidity constraints and/or take into account trading inpact, bid/ask spread etc. Otherwise you just filter out this kind of noise that cannot be traded.

[frommi]

If you include companies with $1 million market caps you will have stuff with almost zero liquidity and super wide bid ask spreads. If your results for example include a company where someone sells 100 shares at $0.0001 and next year there is some random trade at $0.01/share for 100 shares as well it looks like a 10,000% return that will totally skew any result. But even if you managed to get 100% of all traded volume you would have made a profit of less than a dollar. That's why if you want to run a meaningful backtest you have to add liquidity constraints and/or take into account trading inpact, bid/ask spread etc. Otherwise you just filter out this kind of noise that cannot be traded.

 

Yes, but rukawa is right, 50m is way to high. The average NCAV stock in my portfolio is between 5 and 50 million and i won`t include anything above 150m because these tend to have lower returns. But adding a filter like >10k daily volume and price > 0.1 should help with the noise, at least in my backtests it did.

[Sharad]

If you include companies with $1 million market caps you will have stuff with almost zero liquidity and super wide bid ask spreads. If your results for example include a company where someone sells 100 shares at $0.0001 and next year there is some random trade at $0.01/share for 100 shares as well it looks like a 10,000% return that will totally skew any result. But even if you managed to get 100% of all traded volume you would have made a profit of less than a dollar. That's why if you want to run a meaningful backtest you have to add liquidity constraints and/or take into account trading inpact, bid/ask spread etc. Otherwise you just filter out this kind of noise that cannot be traded.

 

Yes, but rukawa is right, 50m is way to high. The average NCAV stock in my portfolio is between 5 and 50 million and i won`t include anything above 150m because these tend to have lower returns. But adding a filter like >10k daily volume and price > 0.1 should help with the noise, at least in my backtests it did.

 

In completing backtesting, is there a way for a screener to take the "ask" on purchase, and sell at the "bid", or is that too much for a Bloomberg screener to do? This would clean the noise that many backtests have with stocks that are illiquid, and reduce the credibility of the performance indicated.

 

I don't use a Bloomberg terminal, and likely will never experience the joy of using one (sarcasm).