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

Daniel and Naik, 2003, Baquero and Verbeek, 2006), this may
(Agarwal,
us to better identify funds that are expected to liquidate in the future
allow
of large negative flows after their reporting period). Compared
(because
the previous classification, 19 cases move from the self-selection to the
to
bin.
liquidation
the classification of cases is based on past returns, this obviously
Because
hassomeimpactuponthecoefficientestimatesforthelaggedreturnsin
binary probit models explaining liquidation and self-selection. For the
the
model the cumulative estimated impact of past returns increases
liquidation
from 3.394 (Table 3) to 3.775, not taking into account the underwater
slightly
For the self-selection model, the cumulative impact reduces from
dummy.
(Table 4) to 3.364. Apart from this, the estimation results for these two
4.386
with and without net asset value (available upon request), are not
models,
different from before. Note that, even with this new classification,
noticeably
probability of self-selection is negatively (and statistically significantly)
the
to past performance.
related
use these new models to correct the persistence analysis for lookahead
We
bias due to liquidation and self-selection. As before, look-ahead bias
limited impact at the two-quarter horizon, reported in Table 7, but substantial
has
impact at the four-quarter horizon, reported in Table 8. Given that
past returns are used to classify a case as liquidation, it is not surprising
poor
see that the bias due to fund liquidation has increased in magnitude, at
to
cost of the self-selection bias. However, the expected returns for each of
the
deciles, after correcting for both sources of bias, are virtually the same as
the
reported in Tables 5 and 6. To some extent this reflects the ambiguity
those
classifying a case as either “liquidation” or “self-selection”. Two conclu-
in
are clear, however. First, no matter how we define it, self-selection has
sions
offsetting impact upon the liquidation bias. Second, it is inappropriate to
no
correct for liquidation bias while at the same time restricting attention
only
those cases that are classified by TASS as “liquidated”.
to
analyzing hedge fund performance and its persistence, a multi-period
When
bias or look-ahead bias may arise if funds attrite from the available
sampling
6 Concluding remarks
databases due to reasons that relate to their performance. In this paper, we
24