V. Endogenous Lobbying? An IV Reestimation
Our discussion to this point has raised the possibility of an endogenous relationship
between contributions and access, but we have not directly confronted it. It is plausible that the
causal arrow runs from access to money rather than the other way around, though the logic of this
is tricky. Groups may give money in response to (in compensation for) meeting with a legislator
or after being contacted by a legislator. The structure of our data does not allow us to observe
whether contacts precede donations or vice-versa. Fortunately, the inferential problems that
reverse causality would cause are limited if the exchange is still a quid pro quo understood by
both parties -- the member is opening his doors because of the money, given or promised.
More troublesome is the possibility that there is a correlation between the amount
organizations donate to legislators and their contact with these legislators, simply because they
both give to their friends and meet with their friends. One possible cause of this problem would
be the constituency bases of lobbying groups. A group or firm might, say, have a large
membership or high employment in a particular state which would independently increase the
access to the state’s senators and, if Wright’s (1985) story about the House holds for the Senate, it
would increase group contributions to senators. In this case, we might spuriously infer that the
donations are buying access, when they are proxy for or signal of friendship—the underlying
cause of access in our account. Hojnacki and Kimball (2001), for instance, control for the
constituency connection between groups and legislators clearly, and find that the control renders
insignificant the correlation between donations and access.
The models we have presented to this point deal with this concern by including measures
of constituency support and friendliness of legislators to groups that respectively favor or oppose
their position. However, the controls we use (i.e. apha score, party, sites, medicare eligible
population) might not adequately capture the non-monetary connections between groups and
senators. The preferred solution is to estimate a richer model that addresses the inferential issues
that complicate our interpretations thus far. We thus investigate a series of instrumental variables
that have the potential to better isolate the causal relationship between donations and access.
27
Convention