4
$\begingroup$

Does anyone know, where I can learn about applying data science to win a political campaign? I know the Obama campaign had 12 data scientists in 2008 and 165 data scientists in 2012. In 2012, they ran over 65,000 simulations every night, for 14 months. They correctly predicted every state within 0.5% and Florida within 0.05%. How did they do this? And where can I find the data they used?

$\endgroup$
4
  • 1
    $\begingroup$ Their data is almost certainly not public. $\endgroup$ – Danica Apr 3 '15 at 22:06
  • $\begingroup$ datafloq.com/read/big-data-obama-campaign/516 $\endgroup$ – Tyrion Lannister Apr 3 '15 at 23:40
  • $\begingroup$ @TyrionLannister if anything is public, it could be worth asking on opendata.se. $\endgroup$ – aeroNotAuto Apr 6 '15 at 14:04
  • $\begingroup$ You can find some insights in Nate Silver's popular science book "The Signal and the Noise: Why So Many Predictions Fail--but Some Don't". However, if I remember correctly, his predictions were not perfect for the Scottish referendum. $\endgroup$ – Valentas Apr 14 '16 at 13:43
5
$\begingroup$

This is an interesting and relevant question. I think that from data science perspective, it should not be, in principle, any different from any other similar data science tasks, such as prediction, forecasting or other analyses. Similarly to any data science work, the quality of applying data science to politics very much depends on understanding not only data science approaches, methods and tools, but, first and foremost, the domain being analyzed, that is politics domain.

Rapidly rising popularity of data science and machine learning (ML), in general, certainly has a significant impact on particular verticals and politics is not an exception. This impact can be seen not only in increased research interest in applying data science and ML to political science (for example, see this presentation, this paper, this overview paper and this whole virtual/open issue in a prominent Oxford journal), but in practical applications. Moreover, a new term - political informatics or poliInformatics or poli-informatics - has been coined to name an interdisciplinary field, which stated goal is to study and use data science, big data and ML in the government and politics domains. As I've said earlier, the interest in applying data science to politics goes beyond research and often results in politics-focused startups, such as PoliticIt or Para Bellum Labs. Following the unfortunate, but established trend in startup ecosystem, many of those ventures fail. For example, read the story of one of such startups.

I am pretty sure that you will be able to find neither proprietary algorithms that political startups or election data science teams used and use, nor the their data sets. However, I am rather positive that you can get some understanding about typical data sets as well as data collection and analysis methods via the resources that I have referenced above. Hope this helps.

$\endgroup$
3
  • $\begingroup$ Thanks Alexsandr. I'll take a look. I think the most fascinating part of the Obama campaign and big data, in general, is the ability to personalize. Because they had such rich data sets about us, the President was able to give you a different campaign than he gave me. A campaign volunteer probably read a different script to my neighbor than to me. $\endgroup$ – Tyrion Lannister Apr 8 '15 at 2:32
  • $\begingroup$ I think the hardest part is to generate or find these data. The analysis is the easy part. $\endgroup$ – Tyrion Lannister Apr 8 '15 at 2:36
  • $\begingroup$ @TyrionLannister: You're welcome. While personalization can certainly benefit from data science, I doubt it's used to such level of granularity - IMHO it is simply not feasible. I also don't think that analysis in political informatics is easy (I mean here a comprehensive analysis, not some trivial one). I partially agree with you about data collection, though - it is definitely not an easy part, but doable and rather simple (especially, considering big budgets that can be thrown at such task); analysis, on the other hand, is challenging and of critical importance (to the stakeholders). $\endgroup$ – Aleksandr Blekh Apr 8 '15 at 2:55
2
$\begingroup$
  1. Voter registration files are public data and are available at the state or county level (since the passage of the Help America Vote Act states are required to maintain lists of eligible voters).
  2. A popular account of the Obama data analytics is in the book, "The Victory Lab" by Sash Issenberg http://www.thevictorylab.com/
  3. Academic paper, "Political Campaigns and Big Data" https://www3.nd.edu/~dnickers/files/papers/nickerson_rogers.2014.pdf
$\endgroup$
0
$\begingroup$

Erin Hartmann who just finished up a PhD in Political Science at UC-Berkley has a paper that uses Obama for America data from 2012. She is using the paper on the job-market for an American Politics sub-field position Might want to check out her work, though I'm not sure it will answer your question. She has a public site here: http://www.erinhartman.com

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.