Blog 2.0: “Open-Source” Economic Research – Time-Series Analysis of the 2012 Presidential Election
Last night I set the tone for the new incarnation of my blog. This morning I realized what the blog’s main purpose should be. I actually really should have thought of this months ago, but the idea just occurred to me.
Over the course of the year, as I’ve started doing my own original research, I’ve found that some projects grow into bigger projects that could potentially become publishable research and other times you hit a point where you’re just not sure where to go with the project so you turn to the ones that are progressing faster. In the latter case, you often have some interesting results but because of the uncertainty about how to develop the project, the results just languish on your computer. What I realized this morning is that this is a perfect forum for me to put up research that falls into the second category so that maybe someone else will find a good use for the research idea that I haven’t thought of.
So with that in mind I have created a page for a small article I wrote entitled “A Time-Series Analysis of the Factors that Influence Presidential Elections.” You can find the page with the article in full (7 pages) above. The paper contains all of the relevant technical econometric details and the statistical results that I’ve consider thus far. For those of you not interested in those details, I will just provide the abstract and the conclusion here along with a quick, informal summary. Also, my plan is to put my code and the data online so that if someone, someday wants to move forward with the research in some way, shape, or form, everything is there for the taking. So without further adieu, I bring you:
Open-Source Economics Research Project#1: A Time-Series Analysis of the Factors that Influence Presidential Elections
In this research I wanted to know how changes in different factors like economic performance (captured by unemployment) and voter sentiment (captured by polls) effect the probability that a President is re-elected. Before Intrade was effectively shut-down (an issue I will blog about at some point in the future) I acquired data on the probability that President Obama was re-elected over the course of the election cycle where this probability is captured by the price of an Obama Futures contract on Intrade. I then combined this with Bureau of Labor Statistics unemployment data and Gallup Poll measures of the President’s disapproval rating to represent the economic and the sentiment factor.
From a technical perspective, which is discussed more in the paper, when working with observational time-series data (non-experimental data that varies with time) there are a number of issues that the researcher needs to be careful to deal with appropriately. To try and summarize these issues as painlessly as possible, the basic issue is that in any normal statistics class you only work with data that was assumed to be independent and identically distributed. However, the President’s Intrade price today is not independent of the President’s Intrade price yesterday. Thus, in this research I make a first attempt at dealing with some of the statistical issues that one encounters when dealing with time-series data of this nature.
If your eyes haven’t already glazed over, then you’re a brave soul and I welcome you to continue the adventure by downloading the article which has its own page at the top of the blog. For everyone else, here is the introduction and my tentative conclusion.
A number of papers have argued that the prices listed on Intrade for presidential election futures contracts can be interpreted as the probability that a candidate will be elected president in an upcoming election (Wolfers & Zitzewitz 2004, 2006). This paper uses this premise as a foundation for analyzing the economic and non-economic factors that drive outcomes in presidential elections. The idea that economic factors drive presidential elections has a long history in economic research (Fair 1978). However, in the absence of real-time, observable variation in the probability that a candidate is elected, much of the justification for this proposition has either been theoretical or predicated on the notion that since economic factors effectively predict presidential outcomes such factors must have a direct impact on election outcomes. The idea that non-economic factors like voter sentiment drive election outcomes most likely dates back to the invention of Democracy. However, it seems that most often the relationship between non-economic factors and outcomes is presumed rather than based on any sort of rigorous analysis. Despite the fact that Intrade has existed for the 2004, 2008, and 2012 American elections it seems that no one has yet to use the real-time variation in Intrade prices to assess the factors that drive presidential election outcomes. The purpose of this paper is to demonstrate an initial attempt perform such an analysis.
Both economic underperformance and negative voter sentiment seem to have been associated with reductions in the probability that the President was re-elected. The second panel results indicate that polling results may be subject to momentum effects. Also just as one would expect, increases in unemployment are associated with increases in in the President’s disapproval rate. Finally, none of the lagged-differences affect unemployment which is consistent with the intuitive premise that large-scale macroeconomic phenomena like unemployment are not determined by the forces that determine polling results and the President’s price on Intrade.