Data
We obtained the data from the official FEC files. Here we focus on donations from political action committees to candidates. We also obtained win-loss information for 157 Senate races between 1980-2006, 69 of them open-seat races (no incumbent running). In 2008, there will be 5 open-seat races, in Colorado, Idaho, Nebraska, New Mexico, and Virginia, which we are tracking.
For/against donations to Republican and Democratic candidates
The FEC data lists both donations for and against a candidate. For instance, if the DNC spends $1M on a marketing campaign against George W. Bush, this is listed as a specific type of transaction between the DNC and Bush. The following plots represent where different committees send their money both for and against candidates. We annotated the 20 committees that donated the most money to Senate campaigns overall.
Returned Contributions
For various reasons, candidates sometimes have to return contributions due to certain regulations. Here we plot the amount of contributions returned, versus the number of checks that were returned. Notice that the line follows a linear trend, with notable exceptions for most (but not all) presidential candidates. Another outlier is Rick Lazio, who was charged with excessive contributions (MUR #4821 from FEC).
Donations and Outcomes of Open-Seat Senate Races
The following plot represents the amount of money donated to winners and losers of open-seat Senate elections. The horizontal axis is the total amount donated to the losing candidate, and the vertical axis is the total amount donated to the winning candidate. Blue x's represent past races. Note that in most races, the winning candidate has more money.
We also put points in for the 2008 races, based on donations up until early September 2008. The green dots represent points if the candidates with more money were to win, with the "projected winning" candidate listed on top and the "projected losing" candidate annotated next. Blue circles represent the points that would occur if the candidate with less money were to win. Notice that for the Virginia and New Mexico races were to be awarded to the candidates with less money, they would be outliers. (The only point that far away from the 1:1 line is the point at ($1M, 100K), which is the 1988 Wisconsin seat where Herb Kohl beat Susan Engeleiter. Herb Kohl was already well-known in that race, being the heir to a chain of department stores.)
Geographic Distribution of Donations
We chose to test the hypothesis that links occur with probability related to the distance between sender and recipient. We took distances for all donations (including that to presidential, senatorial, and congresional seats), based on the zip code listed for each committee and candidate. Taking raw distance, the pattern was not clear. However, the explanation for this is the different density of population (thereby PAC's).
Instead, we chose to use a measurement of rank (proposed by Tomkins et al)-- the rank between node u and node v is the number of nodes that are geographically closer to u than v is. That is, if the National Education Association is 100km from Barack Obama, and there are 5000 committees that are inside a 100km-radius of Barack Obama, then the rank of the NEA is 5000. Using this rank, a much clearer power-law emerges:
