42 measurement of variables csr ratings 4 this used

observations, and the post-SOX sample contains 1,645 firms and 8,702 firm-year observations.
4.2 Measurement of Variables
CSR Ratings
4
This used to be called KLD STATS. We use interchangeably ESG STATS and KLD STATS.

19
The MSCI ESG STATS (Statistical Tool for Analyzing Trends in Social and
Environmental Performance)
5
is an annual data set of environmental, social, and governance
(ESG) ratings for publicly traded companies conducted by MSCI ESG Research. The MSCI ESG
STATS ratings model includes over 50 ESG indicators in seven ESG categories for the three
topic areas:
Seven ESG categories
• Environment
• Social:
o Community
o Human Rights
o Employee Relations
o Diversity
o Customers
• Governance
6
The spread of companies covered by STATS in 2010 is the largest 3,000 United States
companies as defined by market capitalization. The STATS product incorporates a binary
representation of ESG ratings: “if a company does satisfy the criteria established for a rating, this
is indicated with a ‘1.’ If a company does not meet the criteria established for a rating, this is
indicated with a ‘0.’ If an ESG indicator has not been researched for a particular company, this is
indicated by ‘NR’ (Not rated).”
7
Each of the seven ESG categories has “strength” and “concern”
variables; when a company meets the criteria for “concern” variables this is transformed into a “-
1.” For analysis in this study we aggregate the ratings into total strengths and sum the concerns
over the seven ESG categories. We follow common practice in the academic literature, of adding
5
MSCI ESG Research builds on the expertise and achievements of sustainability pioneers KLD, Innovest, and
IRRC acquired via MSCI’s acquisition of RiskMetrics.
6
We remove out this corporate governance scores in the KLD data because we analyze corporate governance as an
endogenous variable, one of the key factors we are considering, in the model. We use this KLD rating of
governance as a proxy for corporate governance variable in our model.
7
We excerpt the description of the rating system from MSCI ESG STATS User Guide & ESG Ratings Definition,
MSCI ESG Research (2011).

20
the concerns to the strengths in order to arrive at a single net total score (i.e., Kim, Park, and
Wier, 2012; Griffin and Mahon, 1997; Johnson and Greening, 1999; Waddock and Graves,
1997).
Discretionary Accruals
We estimate discretionary accruals as earnings management proxy variables by utilizing
the annual cross-sectional industry regression with the modified Jones model based on DeFond
and Subramanyam (1998) and Kothari, Leone, and Wasley (2005). The equation to measure
discretionary accruals is as follows:
!"
!"
!
!" ! !
= ࠵?
!
+ ࠵?
!
!
!
!" ! !
+ ࠵?
!
∆ !" !
!"
! ∆ !" !
!"
!
!" ! !
+ ࠵?
!
!! !
!"
!
!" ! !
+ ࠵?
!
!"# !
!" ! !
!
!" ! !
+ ࠵?
!"
,
(2)
where TA
it
is total accruals for a firm i at year t, ∆ REV
it
is the change in net revenues for firm i at
year t, ∆ REC
it
is the change in net receivables for firm i at year t, PPE
it
is gross property, plant,
and equipment for firm i at year t, IBEI
it
is income before extraordinary items for firm i at year t,
and A
it-1
is total assets for firm i at year t-1. Total accurals ( TA
it
) are calculated by deducting cash
flow from operating activities (CFO
it
) from income before extraordinary items (IBEIT
it
). We
obtain all items in the equations from the Compustat database. The residuals from the modified