|
2007 Annual International CHRIE Conference & Exposition
130
All cash inflows stemming from an unexpected change in marketing expenditures are captured either in
SFEPS or in Z. If the marketing shock, with its related effects on future cash flow, sales, and other expenses, is the
only economic event affecting the firm in the period ending at time t, the change in per-share market price, which
will be reflected in the periodic return, is:
Pi,t – Pi,t-1 = SFEPSi,t – MFEPSi,t – OEXFEPSi,t + Zi,t
(1)
If marketing yields benefits only in the current period, then Z will have a value of zero. Any marketing
forecast error should be accompanied by contemporaneous, unexpected cash flows for sales and other expenses.
Consider the following linear regression model:
Pi,t – Pi,t-1 = A0 +B1SFEPSi,t – B2MFEPSi,t – B3OEXFEPSi,t + ei,t
(2)
The coefficients for SFEPS and OEXFEPS should be positive and negative, respectively, and the
coefficient for MFEPS should be negative and the same as the coefficient for other expenses. In other words, an
unexpected increase in marketing outlays should be associated with a negative price response when controlling for
contemporaneous changes in sales and other expenses.If, instead, benefits from marketing occur only in future
periods, then an unexpected increase in marketing for a given period will not be associated with same-period shocks
in sales and other expenses. However,
Zi,t
will now be positive and will reflected in the periodic return. If Z could
be observed, its coefficient should be positive while the estimated coefficient for MFEPS would be negative.
Because Z is unobservable, however, it constitutes an omitted variable that is positively correlated with MFEPS in a
cross-sectional test. The estimated efficient for MFEPS will, therefore, be positively biased toward zero. In fact, if
all benefits from marketing occurred in future periods and if marketing were a zero net-present-value project at the
margin, Z would be perfectly and positively correlated with MFEPS, and the slope coefficient of MFEPS would be
zero in the absence of a Z-variable.
The discussion above suggests that the estimated coefficient of association between marketing forecast
errors and returns will depend on the longevity of benefits arising from these activities. If marketing outlays
constitute expenses, then a negative coefficient of association with CAR should be observed, similar to the negative
partial correlation coefficient Hopwood and Mckeown (1985) found for total expense forecast errors. If marketing
outlays produce future-period cash flows, the present value of the cash flows makes a positively correlated omitted
variable which would bias the estimated coefficient for marketing toward zero. Relaxing the assumption that
accounting revenues and expenses correspond to cash flows should not invalidate the analysis.
Estimated coefficients for returns with MFEPS and RDFEPS can, therefore, be compared against two
benchmarks following Bublitz and Ettredge (1989). First, there is the “zero” measure aforementioned. Unexpected
investments in zero net-present value projects which are long-lived should have no association with CAR.
Therefore, we have, in null form,
H1: the coefficient of association between MFEPS and CAR is greater than or equal to zero.
Failure to show significance in a directional test can be interpreted as an indication that unexpected outlays
for marketing constitutes long-lived, zero net-present-value investments.
As a second benchmark, the estimated coefficients for MFEPS can be compared with the estimated
coefficient for Other Expense forecast errors. Thus, we have, in null form,
H2: the coefficient of association between MFEPS and CAR is less than or equal to the coefficient of
association between OEXFEPS and CAR.
If the model is correctly specified, its inability to show significance in a directional test would suggest that
marketing outlays are treated similar as ordinary expenses in the valuation process. Alternatively, exhibition of
significance in a directional test would suggest that marketing has future benefits. However, caveats apply. Inability
to show significance may be due to lack of power in the test. For example, measurement errors might affect our
analysis, especially considering our data is collected annually.
Because the focus of this study is on the marketing outlays, which are typically available on the annual
basis, we use annual rather than quarterly data. We draw all publicly traded hotel firms (SIC code 7000, 7010 and
7011) from COMPUSTAT Industry Annual and CRSP from 1996 to 2005. Outliers of variables that lie outside three
standard deviations of sample means are trimmed, which cause our firm-year observations drop to 64. Following
Do'stlaringiz bilan baham: |
