Skip to main content

Soon Window-Dressing Won't Work

  • Author:
  • Updated:
    Original:

Will it? I don't know. The Wall Street Journal has a good article today about creepy stuff going on around window-dressing, where investment managers bid up the illiquid stocks they own on the last day of a quarter to make their quarterly numbers look good and increase their fee income and stuff. Sadly the Journal leads with its own study, which is kind of like statistics for people who don't like statistics:

The Journal's analysis compared the performance of those 10,000 stocks to the one-day return of the Standard & Poor's 500-stock index. On days that didn't end the quarter, an average of 217 stocks beat that index by at least 5 percentage points then trailed it by at least three the next day. But on the final trading days of quarters, an average of 280 stocks did.

Umm! There's a chart of that, too, but ... umm! Would you trade on that?1 Who cares, though; the Journal also cites forthcoming statistics from the Journal of Finance, which is like by, for and about people who love statistics:

[Let's talk about] Rabih Moussawi, a finance researcher at the Wharton School at the University of Pennsylvania who has studied window dressing. In a study slated for publication next year in the Journal of Finance, he and his colleagues found that the stocks most heavily owned by hedge funds outperformed the market by an average of 0.3 percentage points on the final day of the quarter — and underperformed by 0.25 points on the following trading day.

You could trade on that! Here's what you do. First thing in the morning of the last day of each quarter, you buy a basket of $X worth hedge-fund-favored stocks constructed however they construct it in their paper - you can do it, I have faith in you, look at 13-Fs, etc. - and short an equal-weighted2 amount of S&P 500. Then, sell $2X of your basket at the close that day (and buy in 2x whatever you sold of S&P), leaving you short $X. Then, the first trading day of the next quarter, close out your short (and your S&P long). You've now made on average 0.55% of $X in two days. You can only do this once a quarter, for just 2.2% a year, but lever it a bit - why not, it's market-neutral, etc. - or, more importantly, focus on small-cap and high-hedge-fund-concentration stocks that are more likely to get the most manipulated, and you start getting close to hedge fund returns. (I mean, low bar.)

GO FORTH AND GET RICH! I don't know. We talked a while back about this amusing paper where two academics, David McLean and Jeffrey Pontiff, found that, when their fellow spoilsports published studies of trading anomalies, the value of those anomalies dissipate by about 35%. So I guess now this strategy should only be worth 1.4% a year. Because hey: now that I've told you about it, you're going to start moving the market up earlier, and someone else is going to figure out that they should start buying the day before quarter end, and then someone else will jump ahead of them, and in a frictionless universe everyone will backwards-induct until you end up meeting yourself buying and selling on the first day of the quarter. Or something.

The McLean & Pontiff study notes that some anomalies are just, like, anomalies - dumb technicals - where everyone should rush in to correct them, while others have "strong theoretical motivation ..., are a rational response to risk and costs, and thus sophisticated traders that observe the findings of these papers should not be induced to trade.” Sadly it didn't break down whether the dumb anomalies dissipate faster than the rational ones; it's worth wondering whether it's true.

Take window dressing. Everybody knows about this and has known about it forever - the studies are new but the issue is as old as quarterly performance reporting. And yet here we are, with constant smallish outperformance. It's hard to argue that that's because quarter-end outperformance is "a rational response to risk and costs," unless "costs" include "agency costs" like "I want to make more fees by inflating the value of my holdings."3 But how would it dissipate? I mean, you can trade against it in your personal account. But your PA is tiny. And you could build a hedge fund that trades on the anti-window-dressing strategy, and you might make money, but that has the problem of leaving you short a bunch of stocks at quarter-end, with those stocks all in the green. How attractive is that to investors? The agency costs may protect the anomaly. I'm starting an anti-window-dressing fund with a November fiscal year.

Fund Managers Lift Results With Timely Trading Sprees [WSJ]
Do Hedge Funds Manipulate Stock Prices? [SSRN]

1.How would you identify your 280 stocks? Also don't the arbitrary thresholds of "up by 5% down by 3%" smack of data-mining? Like, what percentage of stocks are up by >=4% at quarter-end and down by < =1% the next day, or whatever? Is it less than the average of non-quarter-end days? Etc.

2.As an exercisea I am writing this without reading the paper; if you want to make money off of this you gotta read it yourself. But if I know finance professors - and I don't - they're calculating "outperformance" in some sort of complex CAPM-y way, so it's not like "equal dollar amounts of SPX" or "equal betas of SPX" but like "equal risks of SPX determined by some multifactor regression," so there'll be math involved. Again though that's your problem, you're the one making money off this foolproof trading strategy, you gotta do the math.

a.This exercises both my "predicting how academic finance papers work" and my "laziness" muscles.

3.Or, I guess ... I mean I'm being blasé about your making actual money off a 55bps anomaly. That could be eaten by your E*Trade commissions.

Related

So Maybe Greek CDS Won't Be Fine, Who Knows, I Give Up

ISDA decided today that there has been no credit event for purposes of Greek CDS. Obvs! And by "obvs!" I mean what I said the other day, which is that with 100% certainty there's been no credit event yet, but with 100% certainty there will be, so everyone should just chill out. Except that it seems like that last part may be wrong. So go ahead and panic. I used to make convertible bonds and some of my time was spent answering questions about what happened to things upon Events. The most popular was: what happens after a merger? If you have a convertible that converts into 10 shares of XYZ stock, but now XYZ is being acquired and each share of XYZ is being acquired for $30 in cash and 4.5 shares of PQR stock and a pony - what happens to the convertible? And the answer I would give usually started with "don't trouble your pretty little head about it." Like, it's fine: you have a convertible that converts into 10 Things, and before the merger each Thing was an XYZ share, and after each Thing is exactly what an XYZ share transformed into, so you convert into $300 and 45 PQR shares and 10 ponies. It just works because it has to work. Economic interests follow without interruption from changes in form; derivative securities poof into derivatives of things that the underlying poofs into. There is no arbitrage! That assumption is central to doing any sort of derivative work, and it spoiled me a bit. Sometimes people would come up with more complicated scenarios involving dividends, multiple-step transactions, weird splits and spinoffs and sales, etc. etc. And I would generally start from the bias "it has to work, so I am sure the document written in the way that works." Where "works" means "the economics and intent of the trade are preserved after the change in form." But of course the document was written by humans, often specifically me, and those humans, often including me, are fallible. So there may well be documents from my former line of work that don't "work" in the sense that an issuer could do some structural tricks that would screw holders out of their economics - where the derivative doesn't follow the underlying everywhere it might go. These tricks are unlikely enough that I don't lose sleep over them. You can't predict everything. I sort of assumed that Greek CDS also had to just work but here is Felix Salmon at Reuters saying no. Lisa Pollack at FT Alphaville said something similar a week ago but I could not fathom that she meant it so I read it to mean something else. But she means it, and Felix does too. Go read it but the basic gist of this theory is: