

259
臺大管理論叢
第
28
卷第
2
期
(3)
Where Δ
ln(Cost)
is the log-change in costs for a firm from year
t-1
to year
t
. We use
three measures for Δ
ln(Cost)
: selling, general, and administrative costs (
SGA
), cost of
goods sold (
COGS
), and total operating costs (
OC
). First, we follow Anderson et al.
(2003) to use selling, general, and administrative (
SG&A
) costs in relation to revenue
activity because sales volume will affect many of the elements of SG&A cost. SG&A cost
can be meaningfully studied in relation to revenue activity because sales volume drives
many of the components of SG&A costs (Cooper and Kaplan, 1998). In addition, we
follow Banker, Byzalov, Ciftci, and Mashruwala (2014) to use cost of goods sold. Costs of
goods sold (
COGS
) by a manufacturing firm usually includes direct material, direct labor,
and allocated overhead that needed to produce goods for sale. Finally, following Kama
and Weiss (2013), we use operating costs (
OC
), defined as revenue minus operating
income, as a summary measure of total firm costs. Operating costs (
OC
) are the expenses
which are related to the operation of a firm. For a manufacturing company, operating costs
include not only SG&A costs, but also research and development expenses, which is also
related to production. Therefore, the change in OC to change in sales should be more
sensitive than the change in SG&A to the change in sales. However, there is no theory
suggesting which categories of costs are more “sticky”.
In addition, Δ
ln(Sales)
is defined as the log-change in sales for a firm between year
t-1
and year
t
.
CC
is the measure of customer concentration, calculated as in Eq. (2).
Dec
is an indicator variable which equals to 1 if sales decreased in year
t
relative to year
t-1
,
and 0 otherwise.
GDPGrowth
refers to the log-change in Gross Domestic Product from
year
t-1
to year
t
.
Size
is defined as the natural log of sales for a firm in year
t
.
RD
is
defined as the research and development expenditures in year
t
scaled by total assets;
RD
is set to zero if value of the research and development expenditures is missing. We include
GDPGrowth
,
Size
, and
RD
as control variables. We expect the coefficient on
GDPGrowth
and
Size
to be positive because costs are more sensitive to change in sales and because
larger firms have lower adjustment costs. We include
RD
because the high intensity of