8
In order to test the predictions of the valence theories, I develop a measure where senators’ ideal
points are estimated on a common scale with state median voter’s ideal points. These estimates are
COnstituency-REpresenative ideal point estimates, or CORE scores. The logic of the median voter
theorem, though by adding an assumption that some deviation by elected officials will occur, is employed
to estimate state ideal points on a common scale with senators.
I assume that a state’s median voter supports policies when both senators and the president agree
(if the state voted for the president). Alternatively, I assume that a state’s median voter supports policies
if both senators agree, but the president does not agree (if the state voted for the president’s opponent). It
is highly unlikely that three actors that were previously supported by the state would disagree if the
median voter of the state did not agree with them. Especially given that the president typically takes
positions only on high-profile roll-call votes, these are policy positions that are likely known to or easily
determined by voters in a state. Thus, elected officials will rarely agree on high-profile issues unless their
constituents also agree.
3
TABLE 1 ABOUT HERE
To follow this logic, consider the following hypothetical example. Maine is currently represented
by two senators who are considered moderate Republicans, Olympia Snowe and Susan Collins. The state
has voted for the Democratic candidate for president in the last three presidential elections covered in the
time studied in this paper (1992, 1996, and 2000). Thus, in the 106
th
Congress (1999-2000), for instance,
we assume that the median voter of Maine is in favor of any issues that both Clinton, Snowe, and Collins
agree on. See Table 1 for more details. In Table 1, five roll-call votes are displayed. The three actors, all
supported by the state of Maine, agree on roll-call 1 and roll-call 4. Additionally, Collins and Snowe
agree on roll-calls 2 and 3, though Clinton disagrees or does not take a position. Thus, we do not count
these roll-calls in the estimation of Maine’s ideal point. Perhaps Clinton has taken the state’s median
position, and Snowe and Collins have diverged due to party pressure. Alternatively, perhaps Snowe and
3
Of course, these estimates have potential pitfalls as they are based only on elite behavior and require some fairly
strong assumptions. However, given that no existing measure captures the concept of constituency median, these
scores are advances.
Convention