14
more conservative legislators are near the top. The particular scale of the scores is
arbitrary, but the relative distances and placement are substantively important. Each of
the y-axis comparison figures are coded differently so we will expect to see negative and
positive patterns, and the main issues are consistency and the tightness of the data around
the regression lines.
In figure 1, our scores are compared to the interest group ratings provided by the
Arizona League of Conservation Voters, and one can see that conservative legislators in
our estimates score low on the environmental scorecard and liberals score higher. The fit
is strong with an R-square of .61, and most of the diffusion appears to be among the more
liberal legislators (according to our estimates). Though positive in direction because of
the coding, the relationship between our estimates and the interest group scores from the
National Federation of Independent Businesses are also quite consistent with a strong R-
square of .64 in Figure 2. We would not expect perfect congruence with single issue
interest group scores, but the strong pattern provides face validity.
The x-axis comparison variable in figure 3 is a composite scale developed by
Richardson, Russell, and Cooper (2004), and it uses interest group endorsements. Four
conservative groups and four liberal groups that endorsed candidates in the 2000 election
were examined, and each conservative endorsement received a negative value of one and
each liberal endorsement received a positive one. As figure 3 shows, the estimated scores
are strongly related to these endorsement scores with an R-square of .66, and the scores
are particularly well linked for conservative legislators.
The comparison variables in figures 4 and 5 do not reflect legislator ideology, but
they provide information about the minority composition of the district and partisan
Convention