per acre was estimated annually and for the five-year
period for each of the 36 combinations (i.e. cultivars
X end-of-season harvest options X insecticide treat-
ments X herbicide treatments).
Table 1 shows total adjusted returns per acre for
the five-year period. Total adjusted returns ranged
from $2,841 to $1,968 per acre. The 31 percent
difference in adjusted returns ($174.60 per acre per
year) suggests a need to understand what manage-
ment practices explain the wide difference in annual
per acre returns.
MODELS SPECIFIED
Tomek discussed the application of zero-one (bi-
nary) regression variables in time series analyses.
Binary variables (also called dummy variables) are
frequently used in price analysis to account for
within-period variation (such as seasonal prices) or
between-period variation (such as annual price
level). Binary variables are also applied in hedonic
pricing models to account for quality attributes or
other discrete characteristics of the dependent vari-
able.
Many regression models which include binary
variables include one or more continuous inde-
pendent variables. However, there are applications
both of time series and cross section regression
analyses using multiple binary variables alone,
referred to here as binary variable regression (BVR).
Madsen and Liu used BVR to study price differences
for feeder cattle (incorporating independent vari-
ables for grade, weight and sex, market location, and
lot size classes). Sersland applied BVR to a cost
analysis of meatpacking plants (incorporating inde-
pendent variables for plant size, hours worked per
shift, shifts per day, days per week, and percent of
capacity utilized). Regression analyses when all in-
dependent variables are binary, such as BVR, yield
results similar to those of an analysis of variance
(ANOVA) approach.
In this study, two models were specified and es-
timated for each of the five years and for the five-
year period combined. Model A assumed no
interaction among the four independent variables, as
in a main effects ANOVA model. Model A was:
3 2 2
Y = α + X βli Xli + X β2i ¾ + X β3i ×3i
i=l i=l i=l
3
+ X β4i X4i
i=l
where
Y was the adjusted value ($) of alfalfa per acre,
Xii was end-of-season harvest option (i= 1 -3,
I=Fall cut, 2=Winter grazed, 3=Unhar-
vested),
X2i was herbicide treatment (i= 1-2, I=No her-
bicide, 2=Annual herbicide treatment),
X3i was insecticide treatment (i= 1-2, I=No insec-
ticide, 2=Annual insecticide treatment, and
X⅜i was alfalfa cultivar (i=l-3,1=WL318,2=Arc,
3=OK08).
Of specific interest in this research was the pos-
sible three-way interaction between end-of-season
harvest option, insect control, and weed control. In
model B, end-of-season harvest option (Xii), weed
control (X2i), and insect control (X3i), were com-
bined into a single variable (X5i), thereby replacing
three variables in Model A (Xii, Xa, and X3i). The
combined variable in Model B was Xsi, i = 1 - 12
{i.e. 3 end-of-season harvest options X 2 herbicide
treatments X 2 insecticide treatments).
ESTIMATION RESULTS AND
IMPLICATIONS
Results from each model are reported here, and
groups of variables associated with alternative
management options are discussed separately.
ModelA
Regression results for Model A are shown in Table
2. One independent variable from each variable
group was left out and is referred to as the base
variable (Suits). Thus, the intercept for Model A,
1983, can be interpreted as follows. The mean ad-
justed value for unharvested OK08 without insec-
ticide and herbicide treatments and with no fall
cutting was $474.50 per acre. Beta coefficients are
interpreted as differences from the base variable
within each variable group. For example, the mean
adjusted value for fall-cut alfalfa in 1983 was $35.34
per acre more than for unharvested alfalfa. Winter
grazing increased returns an additional $7.32 per
acre ($42.66 - $35.34) compared with fall-cut alfal-
fa. If the beta coefficient was not significantly dif-
ferent from zero, then adjusted returns for that
variable were not statistically different from ad-
justed returns for the base variable.
End-Of-Season Harvest Options
Winter grazing increased adjusted alfalfa returns
each year and for the five-year period in relation to
other end-of-season harvest options. Based on pre-
vious research (Senst and Berberet; Dowdy),
removal of fall alfalfa growth by grazing reduced
stress on alfalfa plants resulting from insect and
weed infestations. Consequently, increased yields
111