Update on FL Early Voting with Saturday Data

Here is the update on in-person early voting statistics, in three graphs.

First is the day-to-day in-person early voting turnout, compared to 2010. Because all counties have now begun early voting, I don’t account for when counties started.  (Over twenty counties opened their in-person early voting operations yesterday.)Unlike 2010, 2014 was the biggest day of in-person early voting during the first week.  (Click on this, and all other graphs, to see the full picture.)

day_to_day_gross2Because of the big surge of in-person early voting yesterday, more voters have now cast a ballot via in-person early voting in 2014 than in 2010, despite the fact that over 20 counties only began early voting yesterday:

day_to_day_cumulative_grossFinally, the in-person early voting electorate has been much more Democratic in 2014 than it was in 2010:


Florida Early Voting Surges Ahead

Florida has begun in-person early voting under the new regime that allows, though doesn’t require, two weeks of early voting.  Because counties may start early voting on different days, it is not straightforward to compare early voting trends in 2014 with those in 2010.

Using the early voting turnout files available from the state, it is easy to see that early voting in 2014 is well ahead of 2010.  Here are the two basic graphs.

First, just leaving aside the fact that over 20 counties did not start in-person early voting until today, the day-to-day early voting turnout has matched 2010 totals, as the following graph shows. (Click on the graph to get the full picture.  Unfortunately, the thumbnail cuts off values from the y-axis.)

day_to_day_grossSo, for instance, despite the fact that only 35 counties opened early voting on Monday (15 days before the election), 48,214 people voted early in those counties, compared to 65,276 early voters in all of Florida’s counties in 2010.  The total number of early voters in 2014 in the 45 counties that were conducting early voting that day exceeded total early voter turnout in all 67 Florida counties in 2010.

A more apples-to-apples comparison is possible.  Let us look at the counties that began early voting in the period when early voting is optional, and compare turnout in those counties with 2010.  On Monday, 45 counties reported early voting totals.  As I’ve already noted, 48,214 people voted early in those counties.  If we look at those same 45 counties in 2010, we see that only 37,297 individuals voted the first Monday in 2010.  If we continue this comparison across the first five days of early voting, the pattern is stark: (Again, click on the thumbnail for the full graph)


When we do this apples-to-apples to comparison, we see that turnout has vastly exceeded 2010 totals.  Thursday’s 77,365 early voters was 43% greater than the number who voted on the comparable day in these counties in 2010.  Overall, early voting is up 37% compared to 2010, when we confine ourselves to counties that have started early voting.

More can (and will) be said about these numbers.  The most obvious initial conclusion is that early voting continues to rise in popularity in the Sunshine State.  One thing to watch is whether the counties that have opened their early voting centers as late as possible experience the type of congestion that causes long lines to form at the polls.  With lower turnout overall in a midterm election, we shouldn’t expect problems to emerge like we saw in 2012.  However, the initial evidence is that interest in in-person early voting continues to grow in Florida, so there is a real possibility that the late-opening counties will have to struggle with unanticipated congestion.

I have yet to run these numbers by party, so it’s unclear whether in-person early voting (as distinct from absentee voting, which is also a form of early voting) continues to skew towards the Democrats.  I also have to remind the reader that I have not been able to examine parallel figures with absentee ballots, though there is some evidence that interest in absentee voting has also grow in 2014 (compared to 2010).

To quote my friend, Doug Chapin, stay tuned.


Thoughts on the GAO report on wait times

On Tuesday the Government Accountability Office issued its long-awaited report on wait times at polling places.  I recommend it to all who are interested in this topic.

It is no criticism of the report to say that much of what is contained within it has appeared elsewhere.  The report provides one-stop shopping for those interested in the established research on the topic. More importantly, the independent verification of existing research — in the way that only the meticulous, scrupulously nonpartisan GAO can do it — underscores that certain facts about long waiting times are actually facts.

Most importantly, long lines are not universal.  They are concentrated in particular places — certain states, cities, and areas with large minority populations.

While this is bad news for these particular places, it is good news for doing something about long wait times.  Assuming that the jurisdictions beset by wait times are dedicated to doing something about the problem, the policy response can be focused on a half-dozen states and a relatively limited number of large jurisdictions.  (The flip side of this conclusion is that a scatter-shot approach to long wait times would be a mistake.)

The report contains one new finding that deserves attention.  This finding is contained on the very first page of the report:

Estimates from our nationwide survey of local election jurisdictions indicate that most jurisdictions did not collect data that would allow them to calculate voter wait times at individual polling places on the November 2012 General Election Day.

The GAO research team conducted a survey of local election officials, and asked them about the data they did collect that might help with the management of wait times.  This is what they found:

  • 36% of jurisdictions recorded “observations by election officials of voter wait times at polling places”
  • 31% recorded “the number of votes cast at a precinct during a specific time period”
  • 18% recorded the “length of time polling places remained open after designated closing times”
  • 17% recorded the “time individuals checked into a polling place, recorded by an electronic poll book”
  • 16% recorded “voter complaints about wait times at polling places”

My only criticism of the report is that it credits too readily the utility of these data gathering efforts.  Data such as voter complaints and after-hours closings are better than nothing.  They are indicators that local election officials are taking the problem seriously.  But, they are still blunt instruments for helping to manage problems of polling place congestion.  (Imagine, for instance, if the only statistic a Walmart manager had to judge whether to add another cashier line was how long it took to check-out the last customer when the store closed at night.)

Both queuing theory and the application of line-management techniques in retail and manufacturing teach us that specific types of data are needed to manage lines effectively.  Mostly importantly, we need to know when people arrive (not when they get to the front of the line to check in) and how long it takes to complete all the tasks required of them.  The percentage of election jurisdictions gathering this data is effectively zero.

The report does mention an example of two jurisdictions that have taken it upon themselves to gather the type of data that is needed for the proper management of the polls using standard techniques that are common in the private sector:

In at least one election, 1 of these jurisdictions distributed time-stamped cards to every 15th voter upon arrival. Poll workers then recorded the time on each card at various stages of the voting process and collected the cards when voting was complete. In the other jurisdiction, officials stated that they began measuring wait times from arrival to check-in in the August 2014 election by distributing cards to voters upon arrival and then collecting those cards at the check-in station, where they recorded the time of check-in in an electronic poll book.

This is exactly what needs to be done.  It probably doesn’t need to be done in every election and at every precinct.  But, if managers of the nation’s largest jurisdictions began conducting these exercises in representative precincts on a regular basis, they would reap great dividends.

The final question — not covered in the GAO report — is what to do with this data?  I close with some shameless self-promotion.  Earlier this year, at the request of the Presidential Commission on Election Administration, the VTP posted three election management tools that can take input that comes from data-gathering exercises and convert it into output to help guide decisions about the allocation of resources (poll books, privacy booths, etc.) in polling places.  With the support of the Democracy Fund, we are working hard to fine-tune these tools.  If you haven’t checked out the tools, please do.  I am looking forward to sharing the results of our R&D efforts in the coming months.

Voter Identification and Discretion

We have a blog post on our Voter ID and Discretion article out on the LSEUSA blog site.  Poll workers often are influenced by their own biases when implementing voter identification laws but this problem can be mitigated in part by having better educated poll workers.

Improving survey quality — and implications for research on election administration

I did a Q&A recently with Lonna Atkeson, which is now available on the OUPblog, “Improving Survey Methodology: a Q&A with Lonna Atkeson.” This Q&A builds off of a recent Symposium on Advances in Survey Methodology that Lonna and I co-edited in Political Analysis.

New research on Voter ID

A paper by Lonna Atkeson, Yann Kerevel, Thad Hall and myself, “Who Asks for Voter Identification? Explaining Poll-Worker Discretion” is now available in Journal of Politics Early View. Here is the abstract:

As street-level bureaucrats, poll workers bear the primary responsibility for implementing voter identification requirements. Voter identification requirements are not implemented equally across groups of voters, and poll workers exercise substantial discretion in how they apply election law. In states with minimal and varying identification requirements, poll workers appear to treat especially minority voters differently, requesting more stringent voter identification. We explain why poll workers are different from other street-level bureaucrats and how traditional mechanisms of control have little impact on limiting poll-worker discretion. We test why many poll workers appear not to follow the law using a post-election survey of New Mexico poll workers. We find little evidence that race, training, or partisanship matters. Instead, poll worker attitudes toward photo-identification policies and their educational attainment influences implementation of voter-identification laws.

Blueprint to Implementation: Election Administration Reform for 2014, 2016, and Beyond

On Monday, May 19, this event will take place in Chicago, and a number of VTP folks will be there — including Charles Stewart, Steve Graves and myself. Looks like it will be an interesting event, and I’ll try to write more about it on Monday!

VTP’s Charles Stewart Testifies Before US Senate Rules and Administration

The headline says it all — Charles testified at a hearing of the US Senate Rules and Administration committee earlier this week. This link will take you to his written testimony and the webcast.

Improving the quality of surveys

Here’s a Q&A that I recently did with Daniel Oberski on the OUPblog, who has recently developed a helpful software package (Survey Quality Prediction) that is getting an award at AAPOR this week.

Auditing Risks

There is a great story in the NYTimes today about new British rules related to auditing.  Specifically, under the new rules:

Auditors are supposed to comment on the particular risks that companies face and to say what they did to deal with those risks.

They are supposed to discuss how much of the company they actually audited, to disclose what figure they deemed to be the lower limit for materiality [the importance/significance of an amount, transaction, or discrepancy], and to explain how they arrived at that number.

Imagine if we did this in elections!  What if, in every election, we knew the particular risks that were evident in each jurisdiction — based on an audit of the election, processes, and procedures in the jurisdiction — and what the jurisdiction had done to mitigate the risk?  It would provide excellent data on management and allow people to know how well a jurisdiction is working to minimize problems, reduce the possibility of malfeasance, and ensure elections are of the highest quality.