More on networks and donations
Andrew Gelman responds to my previous post on this topic.
it’s important to separate two aspects of the above research: network analysis as a general statistical/social-science research method as applied to American politics, and the analysis of political contributions in particular. … In social science as a whole, networks have become very trendy—and I pretty much think that’s a good trend. There are some roadblocks in applying these ideas to the study of public opinion and voting, however, since we’re talking about a network of 250 million adults where the average person knows only 750 other Americans. You can get this sort of data from surveys but it’s hard to know what to make of it. Tian Zheng, Tom DiPrete, Julien Teitler, and I have been involved in a research project estimating the segregation of Democrats and Republicans in social networks … the analysis is difficult, just at the technical level of building a statistical model for what we’ve got. It’s no surprise that a lot more work has been done on networks in Congress. Moving to research on political contribution networks, I wonder if one reason you don’t hear much about it is that this sort of work is politically marginalized, as it’s associated with left-wing critiques of the political system, rather than more traditional representations of American politics as being generally representative of public opinion. … I agree with Henry that political donations would be a natural place for network analysis, since many of the major contributors have clear enough links that sparseness is less of an issue.
Thinking about this a bit more, it seems obvious to me that the difficulties of putting together data on political contributions is not only a problem for political science, but for public accountability more generally. If the FEC data were presented on the Internet in a way that allowed researchers (and non-profits) to easily scrape it and make it usable, this would make it much easier to figure out who is giving what to who.
Comments
I agree that transparency for organizational political donations is a good thing, but analysis, especially like that on the web site you gave as an example, is going to stumble over buying influence vs. paying protection. Microsoft is a classic example of a company that did little political work until its adversaries used politics as a weapon instead of competing in the marketplace. Microsoft has to pay huge amounts to politicians just to be left alone, never mind trying to buy influence. As government power continues to expand, all organizations have to spend money on politicians just to make sure some other adversarial group doesn’t buy influence to screw them over. If you are a large US company you cannot afford not to pay off politicians in the US and, increasingly, in the EU as well.
Posted by: Robert L. | April 5, 2008 04:22 PM
>If the FEC data were presented on the Internet in a way that allowed researchers (and non-profits) to easily scrape it and make it usable, this would make it much easier to figure out who is giving what to who.
The issues is not so much getting the data from FEC. You can download the contribution records from the FEC website. The site is easy to scrape, and a number of non-profits have done it. The problem is that there seems to be very little validation of the data submitted to the FEC. Transactions are not cross-referenced, the antiquated DB design (think COBOL) means some data is overwritten, and the coding scheme is opaque to those of us not initiated to the minutia of campaign finance law. The data collection seems to have been designed to aid the FEC in fulfilling its bureaucratic duties, not so much to help researchers or citizens with investigations.
Another part of the challenge is the entity resolution problem: politicians have ids for each campaign, but there are no ids for other individuals or companies included in the data, so all the alternate spellings must be matched up somehow.
Clearly we should have higher standards for election data. (For some reason the IRS doesn’t seem to make nearly as many errors as the FEC..) But considering that the FEC is currently several members short of a quorum needed to make any enforcement decisions in the middle of an election year of record-breaking fund raising, DB design may be the least of their problems. Congress lacks the political will to hold themselves accountable.
I think it would be great if more social science researches would get involved in this area - and I think we are going to see a burst of papers in the next year or two. From a DB perspective, the data is somewhat dirty. But cleaning enough data to answer a specific question is still much less work than trying to go out and interview each candidate.
Posted by: Skye Bender-deMoll | April 8, 2008 02:54 PM
Wendy K. Tam Cho (2003) wrote an excellent article in AJPS about spatial contagion effects in contribution networks among Asian Americans.
Posted by: Jason McDaniel | April 9, 2008 05:01 PM