I am trying to compare a database of campaign finance contributions to candidates with a database of congressional votes and see if I can unearth any causal relationships between campaign finance sources and legislative action.
Causal relationships are difficult to establish of course, because it is not necessarily obvious whether a donor is supporting a candidate who agrees with their policy objectives, or is offering money to change a candidate's policy objectives. There are many many practical difficulties with determining the sources of campaign financing, which I will ignore for the purposes of this question.
Therefore, my primary assumption is that campaign finance contributions from a given source would have a strong correlation to a certain voting block for a piece of legislation relevant to their interests.
Suppose we take a list of campaign contributions from the following PACs to federal congressmen and senators:
- American Healthcare Association PAC
- Boeing PAC
- AFLAC PAC
- GlaxoSmithKline PAC
- American Academy of Family Physicians PAC (FamMedPAC)
Eli Lilly and Company PAC
Lockheed Martin Employees PAC And we run an analysis of all congressmen and senators' votes on only one piece of legislation, something healthcare-related like the Affordable Care Act. We would expect that Lockheed Martin and Boeing's contributions to have a comparatively little causal relationship on a representatives' vote, and the rest of the PACs (which are healthcare-related) would have a strong relationship, either with a vote for, or a vote against, the legislation.
How would you run this analysis? Linear Regression seems silly because a vote is a binary yes/no value.