17
per state divided by the normal state party vote.
7
Campaign spending captures candidate
mobilization efforts and therefore represents one way to measure competitiveness. We expect
competitiveness to have a positive effect on turnout as it does in general election campaigns. We
show our results with and without this variable for three reasons. First, reporting rules only apply
to candidate who choose to take federal matching funds. Thus, in 1996 spending data for Steve
Forbes and in 2000 Steve Forbes and George W. Bush are not available. So, in these years,
spending data under reports actual spending. Second, the FEC rules cap spending in states based
upon an Electoral College formula. These caps force candidates to show spending meant for one
state as spending in another state. This happens when a candidate buys TV ads in the Boston
media market for the New Hampshire primary or a candidate uses a survey firm in Nebraska to
assess voter attitudes in New Hampshire. If the candidate is worried about reaching spending
limits imposed by the FEC he will place those costs into Massachusetts and Nebraska instead of
New Hampshire. And, third, reporting requirements only began in 1976 and therefore 1972
becomes a missing case. Thus, this data has some amount of error, but still represents an
adequate measure of spending.
8
Because our unit of analysis is party elections in primary states we also control for party.
The variable is coded 0 for Democrats and 1 for Republicans. We also control for year of race as
each contest is ultimately centered on a particular general election contest that is important to the
general dynamic of the nomination contest.
9
Finally we also control for demographic characteristics of states including percent black,
7
Data on candidate spending comes from FEC reports. We used constant 2000 dollars. The formula for constant
dollars was obtained from the Bureau of Labor Statistics.
8
When you look at the data, spending levels makes sense with expectations about primary sequence. Most of the
money is placed in the early primaries with the latest primaries often indicating no money spent.
9
An F test actually showed that including the dummies did not provide any increase in predictive power. Models
without election year dummies show slightly stronger effects for our two main independent variables of interest.
Convention