Running on Empty: Data-Driven Analysis is Hard Without Data

[Image courtesy of polyglotprogramminginc]

My friend and colleague Paul Gronke of the Early Voting Information Center recently posted the following on Facebook:

I have been training my new RA for the past HOUR how to collect early voting statistics for the 2014 election, a data product that people apparently expect me (and earlyvoting.net) to collect. Here’s the rub–we went through four states–Alabama, Arizona, Alaska, and North Carolina, and successfully found the statistics we wanted for ONE.

Now she understands why I am slow to get this information.

Aside from the fact that this is a departure for Paul – who usually uses Facebook to gripe about air travel or post pictures of vegetables from his garden – his complaint is significant and worth noting. In recent years, academics and election officials alike have been coming around to the notion that data-driven analysis is increasingly important in the drive toward better election administration nationwide.

But, as Paul notes, that can be difficult without data.

The reasons why the kind of information Paul is seeking is so hard to find are actually common to many “missing data” problems in the field and usually include:

1. State or local law doesn’t require collection or reporting of such data;
2. Varying terminology and practice makes it difficult to identify what’s being sought; or
3. Existing systems like election management software aren’t built to handle it.

An even bigger problem can be the perception that some data is of more interest to researchers like Paul than it is to election administrators. But in this case, the explosion of voting opportunities suggests that getting a handle on who is casting ballots before Election Day or otherwise outside of the polling place should be a priority for the field. This information can help in allocating resources and preparing to handle the flow of voters and ballots across a single election. The fact that it’s still hard to find is probably more a symptom of election offices scrambling to adapt existing data collection procedures than evidence of unwillingness to collect and share the data.

Until then, we are all heavily reliant on – and grateful to – researchers like Paul for their “archeological” work in state election data to make these figures available. Keeping up with his chantarelle mushroom harvest or his latest airport SNAFU is a small price to pay for his continued good work on documenting the growth of non-precinct place voting in America.

Thanks, Paul. [We kid because we care. And because it’s fun.]