Obviously one should not invest based solely on a stock screen, but I find that it is a good first step.
Regardless of one’s investment style, how can it not be a central element of an objective and consistent stock-picking process ?
Is there a more efficient way to narrow-down the universe of investable securities into a list of securities that match one’s requirements ?
Is there a better way to quickly and efficiently benchmark an issue ?
I think that I would rather put my money on the guy who selects his stocks based solely on stock screens, than on the guy who selects his stocks solely on some 20 pages individual DCF model with 10 decimals WACC.
Obviously both above investors are probably bad investors, but my point is that you need to know your opportunity costs and relative pricings and risk profiles.
I think the only criticism is for simple screens like low P/E with no further analysis, and this criticism is less about the value of the screen itself and more about how easy it is to do, and how therefore filled with more value traps than one would otherwise think.
If it’s an industry that you’re not too familiar with and you’re not putting too much cash into it, (i) forward P/E, (ii) forward EPS growth rate (Bloomberg consensus estimates) and (iii) true cash flow generation are good places to start. I hate to say it b/c it’s simplistic but stocks I’ve bought based on those three screens have been some of my best investments. You can’t really screen for cash flow well so it takes some knowledge on analyzing CF statements but sometimes being simple is OK
There people that invest using mainly screens, some of them do quite well. A lot of people, now that I think about it do well with mechanical type investing in the small/micro space. They hold very diversified portfolios.
I personally love low P/E screens… for other people to give me their money. When I look at some of my best shorts of 2014, several were low P/E with high but unsustainable trailing growth numbers. When you look pro forma, a few of these were trading at 50-60x pro forma EPS. There are indeed lots of people who invest exclusively based on P/E ratios – of course if you ask them, they probably did some “analysis” but it’s hard to see how much they could have done and yet been so wrong. The market systematically overvalues trailing growth as a general rule.
The problem with low P/E screens is that in 2009 you don’t need them and in 2014 all that’s left on them is garbage. And everyone and their mom runs them now. Low P/E screens were a lot more effective circa 1995 and before. Now it is a pure commodity.
PE’s though are worse cuz mgmt can control so I prefer ebitda. Also ppl who uses juss PE typically use 1 tiny stat. I feel like using 1 year trailing or even ntm are both faulty. earnings are too volatile. you really gotta look at minimum 10 years of financial statements. to see an entire mini credit cycle IMO. Also id beware of value traps at those with PEs below 10x. they are down there for a reason. and i agree with everyone on cash flow its always best. i prefer FCF.
Did you guys know many Chinese retail investors pick stocks based on price? Like if it is only 3RMB it’s definetly a good deal, and if it’s 100RMB it’s probably a bad deal “because it has a greater distance to fall”. Never mind if the 3RMB stock is trading at 100X revenues, and the 100RMB is 3X. Not making this up.
I don’t have an issue with screens, as a tool to generate ideas for fundamental research. I do have an issue with systematic investment strategies that use them. It’s very easy to do things like create a screen and track the performance of a strategy that follows the screen. Then you revise and work on creating a screen that produces the best returns. This is a very dangerous approach, for what should be obvious reasons.
I would add that I don’t particularly care for the double sort methodology in factor investing either, though I don’t think it’s as bad as using screens. I really don’t like the way the double-sort approach throws away information.
Can you clarify what you mean by the double-sort method? You mean breaking things down into quintiles by factor and then buying/selling the stocks that fall into a specific cross section of quintles (e.g. lowest quintile of PE + highest quintile of momentum)?
If so, i’ve always found that approach suspicious, although I don’t have many better alternatives, since it does tend to keep portfolios at least minimally diversified.
^I would describe it basically as you have. On Ken French’s website he has a number of double sorts based on a variety of different variables. So you can look at the returns of portfolios that are small cap value/growth, etc.
I would focus on the difference between estimating a factor model and using the double sort approach to create a portfolio. The goal of estimating the factor model is to try to figure out how certain stocks with similar characteristics have common components (means, variances, correlations). There isn’t anything inherently wrong with the Fama-French/double-sort approach, though the statistical motivation is undermined here.
I prefer to re-interprept the double sort as a multilevel/hierarchical model that is estimated in each period. Once you start thinking about it as a cross-sectional regression in each period, then it becomes a little more obvious how weird it is to arbitrarily create groups instead of just running the regression on the relevant variables (P/B, e.g.).
However, I see no reason to continue to think in terms of double sorts when going from the factor model to portfolio construction. What about turnover? What about objectives? The double sort portfolio is, at best, a hypothetical trading strategy that you would never invest in (and really never really want to invest in).