18
CHAPTER III
MEASURES, RESEARCH DESIGN,
AND SAMPLE SELECTION
3.1 Measures of Over-investment
To measure the increase or decrease in over-investment, I first examine whether a
firm is more likely to over-invest following the approach in Biddle et al. (2009). I use two
firm-specific characteristics that affect the likelihood a firm will over-invest. Specifically,
I use the firm’s cash level (
CASH
) and free-cash flow (
FCF
) as two partitioning variables
based on the argument that firms with low
CASH
and low
FCF
are more likely to be
financially constrained. Alternatively,
firms with high
CASH
and high
FCF
are more
likely to face agency problems and to over-invest (e.g., Jensen 1986; Harford 1999; Lie
2000; Richardson 2006).
FCF
is measured as cash flow from operating activities less predicted capital
expenditures (
CAPEX
). Following Biddle et al. (2009), I estimate predicted
CAPEX
as a
function of sales growth based on the following model:
CAPEX
t
= γ
0
+ γ
1
SALES_GROWTH
t-1
+ υ
t
(1)
where
CAPEX
t
is the natural logarithm of capital expenditures
scaled by lagged total
assets in year
t
.
SALES_GROWTH
t-1
is the percentage change in net sales in year
t-1
.
I
estimate equation (1) by year for each 2-digit industry with at least 10 observations in a
given year. I then estimate the predicted capital expenditures for each firm
i
using the
estimated coefficients from equation (1):
PREDICTED_CAPEX
it
= γ
0
+ γ
1
SALES_GROWTH
it-1
(2)
where
PREDICTED_CAPEX
it
is predicted capital expenditures for firm
i
in year
t
.
Hence, the free-cash flow (
FCF
it
)
for each firm
i
in year
t
is measured as follows:
FCF
it
=
CFO
it
–
PREDICTED_CAPEX
it
(3)
where
CFO
it
is cash flow from operating activities scaled by lagged total assets.
19
My second proxy for the likelihood of over-investment is cash level (
CASH
). For
each firm
i
, I measure
CASH
it
as cash and cash equivalents at end of year
t
scaled by
lagged total assets.
After measuring
FCF
and
CASH
for each
firm-year, I rank firm-years into terciles
based on
FCF
and
CASH
. I re-scale the ranked values to range between zero and one.
Following the approach in Biddle et al. (2009), I then create
a composite score measure,
OVER_INV
, which is computed as the average of ranked values of the two partitioning
variables. I do so because each variable is likely to measure the
likelihood of over-
investment with error and aggregating these two variables reduces the measurement error
in the individual variables. Thus,
OVER_INV
measures the likelihood
of over-investment
based on
CASH
and
FCF
. As
OVER_INV
increases, the likelihood of over-investment
increases.
Dostları ilə paylaş: