29
ratio of first-term members on the SFRC with respect to all first-termers prior to the
reform. Indeed, a glance back at Figure 1 indicates that the ratio is slightly greater than
one at that time. For the model of internal prestige, therefore, the observations run from
1949 to 2002. For the external prestige models, the regression covers the period from
1947-2001. The loss of observations does not appear to have a detrimental effect on our
results.
Finally, we used a variety of diagnostic techniques in testing the models above.
Given the small size of the data set, we were particularly concerned about the impact of
outliers on the estimates, but we also looked for effects of collinearity, heteroskedasticity,
and autocorrelation.
12
In addition, we considered the effects of other independent
variables, including dummies for divided government and the pre/post 1974 period
(based on the Grosewart Index findings)
13
, as well as the number of U.S. military
conflicts, among others. We also tested for interaction effects. The diagnostics revealed
t
ha
t
t
he
i
nc
l
us
i
on
of
t
he
s
e
a
nd
ot
he
r
va
r
i
a
bl
e
s
i
mpa
i
r
e
d
t
he
qua
l
i
t
y
of
t
he
mode
l
s
’
f
i
t
,
usually in terms of outliers and heteroskedasticity.
V. Results
12
For
out
l
i
e
r
s
,
we
a
na
l
yz
e
d
me
a
s
ur
e
s
of
Cook’
s
Di
s
t
a
nc
e
a
nd
l
e
ve
r
a
ge
s
t
o
unc
ove
r
t
he
r
a
r
e
suspect years and then ran the analysis with and without their inclusion to test for changes in the
results. For collinearity we looked at correlations and the plot matrix between independent
variables. For heteroskedasticity we employed externally studentized residual plots versus
predicted values and q-q plots. Finally, for autocorrelation we used both the Durbin-Watson and
LM tests.
13
Based
on
Gr
os
e
c
l
os
e
a
nd
St
e
wa
r
t
’
s
f
i
ndi
ngs
i
n
t
he
i
r
e
xa
mi
na
t
i
on
of
c
ommi
t
t
e
e
pr
e
s
t
i
ge
,
we
investigated using a dummy variable for the period before and after 1974, the end of the 93rd
Congr
e
s
s
.
As
t
he
y
e
xpl
a
i
n,
t
hi
s
ye
a
r
“
ha
ppe
ns
t
o
be
t
he
t
i
me
of
s
ome
ma
j
or
reforms in Congress
t
o
t
he
c
ommi
t
t
e
e
s
ys
t
e
m…t
he
c
ha
nge
of
[Grosewart index] value for the Foreign Relations and
Appropriations agrees respectively with the research of Deering and Smith (1997, 79-8
0)
.
”
(Stewart 1999)
Convention