95 /304
95
臺大管理論叢
第
28
卷第
1
期
3. Empirical Methodology
3.1 Data Description and Variable Construction
Our study sample comprises of all firms listed on the Taiwan Stock Exchange
(TWSE) and GreTai Securities Market (GTSM) which had, at some time between 2009
and 2012, received a corporate social responsibility award which was duly announced in
the Global Views or CommonWealth Magazines. Those firms that were in receipt of CSR
awards are defined in this study as ‘CSR winners’. We refer to Lee, Liu, and Yang (2011)
for our selection of the paired samples (non-CSR firms) since candidates for CSR awards
are not made public, and as such, we do not have access to lists of the firms that failed to
receive such awards.
In an attempt to reduce the potential error caused by the over-sampling of CSR
winners, robust firms were matched from the full sample of listed firms and the number of
matching samples was increased to 1:3, with the following criteria being applied:
12
(i)
firms in receipt of CSR awards must have had positive net margins over the past three
years; (ii) matching firms must be located in the same industry as the CSR winners; and
(iii) matching firms must have had total assets as close as possible to those of the CSR
winners in the one-year period prior to the CSR award announcements, with the difference
in assets being within 50%.
Any listed firms located in the financial sector were excluded from the sample since
the measures of financial performance in the financial sector differ from those of other
firms. The data descriptions are represented in Table 1, which shows the CSR award
announcement dates announced by the Global Views and CommonWealth Magazines
from 2009 to 2012. The exclusion process resulted in final samples of 157 CSR winners
and 498 non-CSR firms.
12 Another selection of the paired samples (with the incorporation of the industry, profit, size of assets
and P/B ratio) is examined in this study; the results were then compared as a robustness test. The
results using the new criteria (with the incorporation of the P/B ratio) are consistent with those using
our original criteria. We only present the results applying the original sample selection criterion due to
space consideration. The other results from the authors are available upon request.