07.01.2015 Views

Fund liquidation, self-selection and look-ahead bias in the hedge ...

Fund liquidation, self-selection and look-ahead bias in the hedge ...

Fund liquidation, self-selection and look-ahead bias in the hedge ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

has an average of 1.6%. The underwater <strong>in</strong>dicator is equal to one if a<br />

<strong>and</strong><br />

has a negative cumulative return over <strong>the</strong> past eight quarters 6 ,which<br />

fund<br />

<strong>in</strong> 14% of <strong>the</strong> cases. The age of <strong>the</strong> funds varies between 13 months<br />

occurs<br />

275 months (about 23 years), while <strong>the</strong> average age is about 45 months.<br />

<strong>and</strong><br />

average size of <strong>the</strong> <strong>hedge</strong> funds, measured by <strong>the</strong>ir log net asset value<br />

The<br />

to about 18 million US$. Total risk is measured by <strong>the</strong> st<strong>and</strong>ard<br />

corresponds<br />

of <strong>the</strong> previous six quarterly returns (StDev).<br />

deviation<br />

size (NAV) is not available for each quarter for all funds <strong>in</strong> our<br />

<strong>Fund</strong><br />

<strong>the</strong> <strong>liquidation</strong> <strong>and</strong> <strong>self</strong>-<strong>selection</strong> models us<strong>in</strong>g two specifications,<br />

estimated<br />

<strong>in</strong>clud<strong>in</strong>g size (based on 20138 fund/period observations) <strong>and</strong> one ex-<br />

one<br />

size (based on 21297 fund/period observations). Follow<strong>in</strong>g Baquero,<br />

clud<strong>in</strong>g<br />

Horst <strong>and</strong> Verbeek (2005) we employ <strong>the</strong> TASS <strong>in</strong>vestment style clas-<br />

ter<br />

which is closely related to <strong>the</strong> 9 commonly used Tremont <strong>hedge</strong><br />

sification,<br />

style <strong>in</strong>dices 7 . In Table 3 we report <strong>the</strong> estimation results based on<br />

fund<br />

20138 fund/period observations for <strong>the</strong> probit specification for <strong>liquidation</strong><br />

that non-<strong>liquidation</strong> means<br />

Note<br />

it is still possible that <strong>the</strong> fund <strong>self</strong>-selected dur<strong>in</strong>g <strong>the</strong> sample period.<br />

that<br />

we subsequently remove all <strong>the</strong> fund/period observations where<br />

Therefore,<br />

fund liquidated (417 fund/period observations) <strong>and</strong> estimate <strong>the</strong> probit<br />

a<br />

to expla<strong>in</strong> <strong>self</strong>-<strong>selection</strong> versus survival (<strong>in</strong>clud<strong>in</strong>g size), where<br />

specification<br />

implies that <strong>the</strong> fund did not liquidate <strong>and</strong> still prefers to report<br />

survival<br />

performance to <strong>the</strong> data vendor. The estimation results are reported<br />

<strong>the</strong>ir<br />

Table 4. The estimation results for both specifications exclud<strong>in</strong>g size are<br />

<strong>in</strong><br />

<strong>in</strong> <strong>the</strong> appendix (Table 9 <strong>and</strong> Table 10). All models <strong>in</strong>clude <strong>in</strong>vestment<br />

reported<br />

style dummies, while time dummies are <strong>in</strong>cluded to capture aggregate<br />

to <strong>the</strong> probabilities of <strong>liquidation</strong> <strong>and</strong> <strong>self</strong>-<strong>selection</strong>. The <strong>in</strong>vestment<br />

shocks<br />

“convertible arbitrage” <strong>and</strong> “dedicated short <strong>bias</strong>” conta<strong>in</strong> very lim-<br />

styles<br />

6 The cumulative return is determ<strong>in</strong>ed over at least five quarters with a maximum of<br />

quarters.<br />

7 eight<br />

has <strong>the</strong> follow<strong>in</strong>g n<strong>in</strong>e <strong>hedge</strong> fund <strong>in</strong>vestment styles: Convertible Arbitrage,<br />

Tremont<br />

Short Bias, Emerg<strong>in</strong>g Markets, Equity Market Neutral, Event Driven, Fixed<br />

Dedicated<br />

Arbitrage, Global Macro, Long/Short Equity, Managed Futures, <strong>and</strong> Hedge <strong>Fund</strong><br />

Income<br />

Therefore, we use <strong>the</strong> most recent observation of net asset value<br />

sample.<br />

from <strong>the</strong> TASS database. However, <strong>the</strong>re rema<strong>in</strong> some observa-<br />

available<br />

for which NAV is miss<strong>in</strong>g <strong>and</strong> cannot be imputed. This occurs <strong>in</strong> 7%<br />

tions<br />

<strong>the</strong> cases. Because we do not want to elim<strong>in</strong>ate <strong>the</strong>se observations, we<br />

of<br />

versus non-<strong>liquidation</strong> (<strong>in</strong>clud<strong>in</strong>g size).<br />

Index.<br />

10

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!