Undressing the window-dressers

Painting the tape: An old issue, getting worse

The Wall Street Journal’s Jesse Eisinger yesterday reported on the mysterious correlation between late-month moves [subscription required] in several microcap stocks prominent in the portfolio of Jeff Feinberg’s JLF Asset Management, a $500 million hedge fund.

But by focusing on a single hedge fund, and suggesting that “the proliferation of hedge funds” may be responsible for “strange movements appearing at the end of the month,” Eisinger missed the bigger window-dressing story. It took a few hours with a spreadsheet and some some datapoints freely available on Russell’s website to demonstrate the substantial, and growing, impact of window-dressing on on small- and micro-cap indices, where the more easy-to-manipulate stocks reside.

Eisinger said that “the practice was never widespread, but was too common for comfort.” The practice is, in fact, widespread, obvious, and appears to be spreading. It is of such long-standing, and so obvious, that hedge funds – leverage notwithstanding – cannot be alone in abetting the manipulation.

The thesis of Eisinger’s article is essentially that window-dressing, “historically focused on the [end] of the quarter, since mutual-fund managers focus intently on quarterly performance,” has moved to a monthly focus because “Many hedge-fund investors get monthly updates. A down month sets off red flags for hot-money investors and can cause an unwanted exodus of cash.” That “blame the hedge funds” position does rather ignore the reality that mutual-fund managers have become equally focused on monthly results for various reasons, including the facts that:

• Investors – after their 2000-2002 beating – have become more willing to move away from poorly-performing products, and
• Technology allows those investors to more easily, more closely, monitor the performance of their investments against peer funds, and other investments.

But this isn't intended as a swipe at Eisinger, for an otherwise fascinating article based on an array of facts indicating either an excellent source, or many hours digging through SEC filings, or some combination thereof. So let’s move on, and fuel the fire he ignited with some facts gleaned from some simple - and some will doubtless argue, simplistic - analysis.

Methodology

From 10 years of daily total-return data for the large-cap Russell 1000, and small-cap Russell 2000, indexes, along with the five years of available data for the newly-introduced Russell Micro-Cap Index, I generated average returns for three ‘strategies’ over four periods: The full decade, the five-year bubble of 1995-2000, the most-recent five-year period, and the last 12 months.

The spreadsheet detailing the analysis, with formulae intact, can be downloaded here. [I have no reason to believe that it is in anyway contaminated, but if you choose to open the file, the usual common-sense caveats about viruses etc applies].

The data was analyzed against three different strategies:

Strategy 1: Each month, buy the index at the close of the 4th-last trading day of the month, and sell the index at the close of the last trading day. This essentially means holding the index over the last three trading days of the month, the period in which window-dressing would be expected.
Strategy 2: Each quarter, buy the index at the close of the 4th-last trading day of the quarter, and sell the index at the close of the last trading day. This holds the index over the last three trading days of the quarter, and is intended to provide an indication as to whether window-dressing is more prevalent at quarter-, rather than month-, end.
Strategy 3: Each quarter, buy the index at the close of the 10th-last trading day of the quarter, and sell the index at the close of the last trading day. This holds the index over the last nine trading days of the quarter, or a period equal to that analyzed in Strategy 1. Again, it is intended to provide an indication as to whether window-dressing is more prevalent at quarter-, rather than month-, end.

Findings

The analysis shows that an investor in the Russell small- and micro-cap indexes could, over any time period and especially over the last five years, collect much of the index gains by simply buying at the close of the fourth-last trading day of the month, and selling at the close of the last-day of the month. A summary showing the percentage of total index move captured by each strategy, over each time period, is available here.

It demonstrates clearly that window-dressing – or some other phenomenon giving rise to an equivalent result – contributes substantially to the performance of the Russell 2000 and Russell Micro-Cap indices. And while the limited period creates a substantial caveat, it appears to have become a factor in the performance of the large-cap Russell 1000 over the last 12 months.

Without beating every calculation to death:

Over the five- and 10-year periods analyzed, Strategy 1 generated at least half the return of the Russell 2000, and for the five years available, 75% of the micro-cap index return. Over the last 12 months, the strategy out-performed both indices, almost doubling the return of the micro-cap index. By contrast, the strategy captured 80% of the return of the large-cap index over the last 12 months, but much less over the longer time-frames.

Strategy 2 captured significantly less of the return in all three indices, which is to be expected given its very short holding periods; just 120 trading days in the last 10 years. Again, however, the two smaller-cap indexes captured a much larger proportion of the total move than the large-cap index.

Benefiting from its longer hold period, Strategy 3 outperformed Strategy 2 in most time periods for the smaller-cap indexes, while producing mixed results against the larger-cap index. However, over the last 12 months the strategy spectacularly outperformed the index, making 25% more than the Russell 1000, 75% more than the Russell 2000, and almost five times the return of the Russell Micro-Cap index.

The strategies appear to work as well as they do in the Russell 2000 and Russell Micro-Cap by eliminating the vast majority of the losses, and sometimes accruing profits, in index down months; for example, over the five years to June 30 2005, the strategy increases the number of profitable months in the Russell 2000 from 34 to 39, and in the Russell Micro-Cap from 34 to 42. By comparison, it increases the number of Russell 1000 profitable months from 33 to 35.

The strategy does, however, forfeit most – but not all – of the gains in up months, a fact that would seemingly confirm the expectation that window-dressers work much harder at the end of down months, compared with up months.

Conclusions

Well, I kind of led with those. Window-dressing – or some other phenomenon giving rise to an equivalent result – contributes substantially to the performance of the Russell 2000 and Russell Micro-Cap indices, and appears to have had some impact in the Russell 1000 over the last 12 months. The index performance discrepancy appears to be big enough to be tradeable, although further analysis of the performance of the Russell 2000 ETF (IWM) would be needed to confirm the results based on “buy at close” and “sell at close” orders, and any slippage.

I do not believe that the larger discrepancy in the Russell Micro-Cap index can be traded effectively, as the iShares ETF (IWC) covering the index is relatively thinly traded and often at exorbitant buy-sell spreads.

Eisinger correctly reported in his original article that “Regulators have rarely done much about the curious movements of certain stocks in the final days of a quarter or a month,” but regulators, of course, rarely see an issue until it either smacks them over the back of the head – Enron, et al – or is drawn to their attention by Eliot Spitzer.

It is clear from this analysis, however, that the window-dressing phenomenon – or some other similar observable fact giving rise to an equivalent result – in underlying stocks is materially distorting the performance of indexes to which many hundreds of billions of dollars of investment capital are benchmarked, and in which – through ETFs and many other instruments – many more billions are invested.

Noted with interest: The US Securities & Exchange Commission does not currently have a director of market regulation, who would normally be responsible for investigating market manipulation-related questions. Annette Nazareth, the most recent incumbent, was recently appointed to the commission itself.

NASD? No, I won't go there. You really, really, don't want to get me started on NASD. Think FEMA, but without the excuse of communication and logistical challenges.

Disclaimers
Positions: Long PZI (PowerShares Zacks Micro-Cap ETF). Positions may change at any time, without notice, and probably will. Past performance is not indicative of future results. I hope.

The usual comments on conspiracy theories apply.

The opinions expressed are solely those of its author. They cannot be bought, although they can be rented for the weekend. The discussion of specific investments is in no way to be construed as investment advice (unless you're looking for a quicker and cleaner way of getting rid of money than burning it).

Comments

[David Jackson]

Greg,

This is a *terrific* article. I've posted it (consistent with your prior permission) on no less than three Seeking Alpha sites: US Market Blog, ETF Investor, and the soon-to-be launched Micro Cap Stock Blog.

Best Wishes,
David Jackson