9
As with the primary variables, missing values also pose a problem for contributions data.
Candidates, for example, are not required to report expenditures or contributions if the total amount of
each is under $5,000. Following Jacobson (1990), one can assume that candidates spend at least $5,000
during the campaign, unless otherwise reported. This assumption implies that the candidate has at least
$5,000 to spend. For the other two resource measures, however, I set missing values to zero; if
unreported, I assume a challenger received no money from the party and received no contributions from
individual contributors.
[Figure 1 here]
I collected data on the above indicators for major party House challengers in the 2000, 2002, and
2004 elections and scaled them using a principal component factors analysis; the eigenvalues are
presented in Figure 1. By the Kaiser Criterion (as described in Kim and Mueller 1978), which suggests
the retention of components with eigenvalues greater than one, one principal component is retained. The
remarkable conclusion to draw from the analysis is that indicators relating to experience and resources,
which are usually included separately in analyses, as well as popularity, all load on one factor. This result
is advantageous to the study of incumbent deterrence for two reasons. First, the retention of one factor
means that there is one dimension with which we need to be concerned, a dimension I have called
challenger strength. Second, in examining how incumbent activities can affect the strength of the
challenger’s campaign, only one equation is needed given that there is only one dependent variable.
[Table 1 here]
Table 1 contains the factor loadings of the five indicators described above. Not surprisingly, the
indicators that have the highest loadings are the ones that relate to monetary resources. The loadings also
suggest that electoral experience, while important to determining campaign strength, is not a strong
enough proxy to warrant exclusion of the other variables in future analyses. This can also be seen in the
correlations between electoral experience and the other indicators included in the analysis. These
correlations are presented in Table 2.
[Table 2 here]
Convention