**Unformatted text preview: **equations to
ensure identiﬁcation of the system. The results of estimating this model,
ˆ
ˆ
β 1 = 0.4671 (t -statistic = 11.49) and β 2 = −0.2207 (t -statistic = −4.93), support the sticky costs hypothesis and the reversal of stickiness in subseˆ
quent periods (β 4 = 0.0839, t -statistic = 3.54). The coefﬁcients on the contemporaneous and lagged changes in SG&A costs in the second equation
ˆ
ˆ
are signiﬁcantly positive, γ1 = 1.5285 (t -statistic = 131.53) and γ2 = 0.2284
(t -statistic = 18.27), suggesting SG&A costs positively inﬂuence sales.
Advertising is a speciﬁc discretionary expenditure included with SG&A
costs that inﬂuences the level of revenue activity. For ﬁrms that reported advertising costs separately, we estimated a model that related changes in nonadvertising SG&A costs to changes in revenue in one equation and changes
in revenue to changes in advertising costs in a second equation. Results of
ˆ
estimating this model support the sticky costs hypothesis, β 1 = 0.6298 > 0
ˆ2 = −0.1232 < 0 (t -statistic = −6.56), and the re(t -statistic = 68.77) and β
ˆ
versal of stickiness in subsequent periods (β 4 = 0.1142, t -statistic = 6.11).
Signiﬁcant and positive coefﬁcients on the contemporaneous and lagged
ˆ
ˆ
advertising change terms, γ1 = 0.2214 (t -statistic = 51.69) and γ2 = 0.1007
(t -statistic = 24.05) support the effect advertising has on current and future
sales. 4. Variation in the Degree of Stickiness
Hypotheses 3a through 4b describe conditions and circumstances that
would affect the degree of stickiness across ﬁrms and over time under the STICKY COSTS 59 alternative model of cost behavior. The coefﬁcient on the sticky costs term,
β2 in model (I), may be expanded to include the various economic factors
described in hypotheses 3a through 4b as follows:
β2 = γ0 + γ1 ∗ Successive Decrease i,t + γ2 ∗ Growth i,t + γ3 ∗ log
+ γ4 ∗ log Employees i,t
Revenue i,t Assets i,t
Revenue i,t . The Successive Decrease i,t dummy is activated for ﬁrm-year observations when
revenue declined in the preceding period. The Growth i,t variable is the percentage growth in real gross national product (GNP) during year t . Substituting this relation into model (I) gives:
log SG&Ai,t
Revenue i,t
= β0 + β1 log
+ γ0 + γ1 ∗ Successive Decrease i,t
SG&Ai,t −1
Revenue i,t −1
Assets i,t
Revenue i,t + γ2 ∗ Growth i,t + γ3 ∗ log
+ γ4 ∗ log
∗ log Employees i,t
Revenue i,t ∗ Decrease Dummyi,t Revenue i,t
+ εi,t .
Revenue i,t −1 This is restated as model (III), where βk = γk −2 in the expanded version of
model (I), k = 2, 3, 4, 5, and 6.
Model (III):
log SG&Ai,t
SG&Ai,t −1 = β0 + β1 log Revenue i,t
Revenue i,t
+ β2 ∗ Decrease Dummyi,t ∗ log
Revenue i,t −1
Revenue i,t −1 + β3 ∗ Decrease Dummyi,t ∗ log Revenue i,t
∗ Successive Decrease i,t
Revenue i,t −1 + β4 ∗ Decrease Dummyi,t ∗ log Revenue i,t
∗ Growth i,t
Revenue i,...

View
Full
Document