Archive

Posts Tagged ‘VBM’

Vote by Mail and Election Results

An emerging theme from the 2012 elections is the impact of vote by mail (VBM) and other convenience voting reforms, such as provisional ballots, on the speed with which we know the results. John Wildermouth, for example, argued yesterday that the prevalence of n0-fault, permanent VBM and early voting in California means that it’s taking longer than it should to know who won on November 6. He writes:

It’s taking longer and longer to get a final count of a statewide election and the problem only is going to get worse.

The growing number of vote-by-mail ballots turned in at the polls, combined with more and more provisional ballots that need to be hand-checked, means that election night is becoming election week. Or election month.

The relationship between the use of VBM and other convenience voting reforms and the speed with which we know the results of an election is an interesting question, but it is one that we do not have a lot of data on at this point.

As the following graph shows, voters in California’s counties vary in their use of VBM. The graph shows the percentage of voters casting their ballot through the mail in the June 2012 election. I did a quick and dirty analysis exploiting this variation to see if there is a relationship between the prevalence of VBM in a county and whether or not we know its results by now. If greater VBM usage leads to less certainty about the election outcomes, then counties at the top of the chart should be done with their counts while counties at the bottom should still be counting. The analysis calls into question this emerging theme.

There are two dependent variables for the analysis: First, has a county sent in its county canvass complete (CCC) numbers to the state, thereby signaling it has counted all its ballots? Second, and conversely, has a county still just reported its final election night update (FENU)? The data for county reporting status come from here. Since these are binary outcomes (yes or no), logistic regression is appropriate here.

I use three independent variables in each model: (1) The percentage of VBM ballots in a county in the June 2012 election, (2) the total number of registered voters in a county, and (3) the total number of ballots cast in the county. I use the prevalence of VBM in a county from the June election as the numbers are not yet available for the November election. VBM usage is generally higher in the June election, however, so it should give us a good idea of how many people were likely to use VBM.

The results (shown in Table 1) are suggestive of a relationship but not encouraging for the VBM causes delay hypothesis. The coefficient for the percentage of voters using VBM in the CCC status model is -0.052 with a z of -0.97 (p=0.334). While the estimated effect is negative, meaning that the greater the percentage of VBM ballots the less likely it is a county will have moved to CCC status, given the z-score we cannot conclude that the results is due to anything other than random chance. The coefficient for the percent VBM in the FENU model is 0.023 with a z of 0.89 (p=0.373). Again, the estimated effect is in the right direction–greater VBM usage leads to a higher likelihood of a county still being in FENU status–but the z-score is too small to let us conclude the relationship is real. Substantively the signs are in the right direction, but statistically we can’t say there is a relationship on the basis of these results.

Caveats: (1) The data are for this year only. There may a change due to VBM over time. (2) The data are for California only. There may be differences due to VBM across states. (3) The data are only to date. There may be differences due to VBM that emerge once all of the counties have reported their final counts.

Follow

Get every new post delivered to your Inbox.

Join 516 other followers

%d bloggers like this: