英语论文网

find that using a reference portfolio approach yields biased test statistics.
They argue that benchmarking performance by a matched firm approach yields well-
specified test statistics for detecting long-run excess performance.Hence,we use the
matched sample long horizon methodology to capture the firm specific revaluation and
cost of capital effects around inter-listing.The matched firm approach also allows us to
abstract from the influences of economic and market reforms.This is because the impact
of any reforms on a matched pair of firms should be similar.11 12
We use buy-and-hold returns to measure long-run stock price performance.
Conrad and Kaul(1993)and Barber and Lyon(1996a,1996b)show that buy-and-hold
11
Note that,although this concern is greater for EMs that have liberalized their economies and markets in
recent years,it is tempered by the greater intra-market return correlations of EM firms.
12
Further,this approach permits us to risk adjust by using returns in excess of a matched-sample of home-
market firms.This is especially important given that data beyond three years is unavailable for most of
our sample.This data restriction precludes us from capturing fundamental changes in expected returns
and risk in the vein of Bekaert and Harvey(1995)or Errunza,Hogan and Hung(1999).We also augment
this analysis using a pooled time-series cross-section regression framework in the spirit of Bekaert and
Harvey(1998).12
returns are appropriate when conducting long-run analyses.We examine returns within
various windows around the liberalization,starting 36 months before to 36 months after
the date the liberalization is announced.For sample firms which have less than 36 months
of pre or post-announcement data,we truncate the buy-and-hold return of those firms and
their associated control firms.13
We benchmark performance by selecting a size and country control(match)firm
for each sample firm.14 The procedure for selecting control firms is as follows:At the
beginning of each sample firm’s pre-announcement window,we rank all firms in the
Datastream International database that do not have a DR program by country and then by
market value of equity.We then choose four firms with a market value closest to(two
smaller,two larger)the sample firm within the same country.We randomly choose either
the larger or smaller firm as the match firm.When a matched firm delists or issues an
ADR,we substitute the next firm from the market value pair.Of the 126 firms in our
sample,116 were paired with only one control firm while 10 required two control firms.
C.Defining Test Periods
The introduction of ADRs for firms that are not priced in a completely integrated
market would result in a lowering of expected returns through reduction in segmentation
premia.In a well functioning market,the price effect should occur on announcement
followed by price stabilization.Unfortunately,there is information dissemination(leakage)
prior to announcement.For example,Bhattacharya et.al(1998)find that in Mexico,
unrestricted insider trading causes some corporate information to be fully incorporated
into prices prior to announcement.In a number of emerging markets,the access to global
capital markets is controlled by the government concerned about the impact of foreign
investments on the economy,markets,balance of payments,and socio-political climate
whi