A close-up of a USPS box on a city street

Analysis: Vote By Mail in Colorado and Washington

Vote-by-mail (VBM) policies continue to be adopted by more and more states as the pandemic changes the way American elections are held. Yet, there is a lot about vote-by-mail that we don’t know. Luckily for us, the Elections Performance Index can shed some light on the policies and its potential effects. 

If you’re unfamiliar with the Elections Performance Index, we suggest you check out this explainer piece! In it, we explain what the EPI is and the questions it can and can’t answer. In this article, we’ll focus our analysis on Colorado and Washington because of the time bounds of the EPI (which currently has data from 2008 to 2018). Washington transitioned to an all VBM system in 2011, and Colorado made the switch in 2013. To try and glean how VBM impacted various aspects of these states’ elections, we will compare EPI indicators from before VBM was adopted to after. The results are complex and require more analysis to make sense of,  but they’re a good example of how the EPI can spur research and help point data-informed students of elections in the right direction. 

What exactly are we going to analyze with the EPI? The index has many indicators to choose from. In this article, we’ll look at mail ballots rejected, provisional ballots, and the residual vote rate.


Mail ballots rejected

This indicator measures what percentage of mail and absentee ballots are rejected out of all votes cast. States that have adopted universal VBM should anticipate that their mail ballot rate rises just because more of their ballots are cast by mail than other states. 

Provisional ballots

This indicator tracks the percentage of provisional ballots cast out of all votes cast. Provisional ballots are often offered to voters if their eligibility is suspect. If the voter is deemed able to vote (after for example, resolving a discrepancy between their listed address of residence and the address they provided on their absentee ballot), then the provisional ballot is counted. 

Residual vote rate

The MIT Election Lab’s Charles Stewart explains the residual vote rate in great detail here. The  short explanation is that the residual vote rate is an indicator that measures the number of over- and undervotes for president.


The following figure depicts Washington’s EPI scores in the 2008 and 2012 elections. Washington adopted VBM in 2011.

As expected, the percentage of mail ballots rejected (out of all ballots cast) increased by a small amount in Washington. The gold bar to the right illustrates the national average for that EPI indicator in this specific electoral comparison, meaning the average state saw a decrease in mail ballots rejected. However, it is important to remember that Washington adopted VBM in 2011, greatly increasing the number of mail ballots cast in their state. This means that there were more opportunities for mail ballot rejections than before, especially as the percentage of mail ballots cast  as a percentage of all ballots cast increases. There are also other contributing factors to this increase.

From an administrative perspective, this means that Washington election officials suddenly dealt with an (expected but notable) surge in mail ballots that may have impacted the rejection rate. Washington voters also needed to learn how to apply to VBM and fill out a mail ballot correctly, and the novelty of the experience may have also increased the rejection rate.

The percentage of provisional ballots cast decreased and the residual vote rate increased slightly. 

For completeness, let’s also compare two election years post-VBM to see whether the effects analyzed above are part of a larger trend.  Let’s compare Washington’s EPI scores in the 2012 and 2016 elections. Mail-in-voting would have been in full effect in both years.

The first important thing to note from the figure above is how small the changes are compared to the previous electoral comparison. It seems like the changes we found comparing pre-VBM to post-VBM Washington were not part of a larger trend endemic to the state. The mail ballot rejection rate even decreased by a small margin.


The below figure compares Colorado’s EPI results from the 2012  (pre-VBM) to the 2016 election (3 years into VBM). 

Like Washington, Colorado experienced an increase in the rate of mail ballots rejected after implementing VBM. The same possible explanations for this increase discussed in the Washington section of this article apply here. Colorado also saw a decrease in the rate of provisional ballots cast and an increase in the residual vote rate.

Just like Washington, let’s take a look at other years to make sure the changes we’re seeing aren’t part of a larger trend. The below figure compares Colorado’s EPI scores from the 2008 to the 2012 election (both pre-VBM).

Interestingly, even before VBM was adopted by Colorado, the state was still experiencing an increase in mail ballots rejected and a (smaller) increase in the residual vote rate.


The changes in the indicators we analyze are quite small. However, just a percentage increase in the residual vote rate could have big consequences across the state. The changes seen here need more analysis to fully understand, but their smaller magnitude should not be interpreted as a low significance result.  

After the introduction of VBM, both Washington and Colorado experienced an increase in the residual vote rate. Many of the changes Washington experienced after the implementation of VBM did not seem like they were part of a larger trend, but Colorado saw an increase in mail ballots rejected prior to their adoption of VBM.  

These findings and the complexity surrounding them is why any analysis of vote by mail, including ongoing and future analysis of the increases in voters casting ballots by mail in the 2020 election, needs to start with a core understanding of the specific state in question and its electoral practices. Especially when research focuses on analyzing indicators directly related to VBM, a broader context for any changes must be provided. Otherwise, states may be penalized just because they have more mail ballots cast, and that isn’t useful to anyone. 

What is useful is looking at the EPI in context when considering how states can prepare for the upcoming November election. Both Washington and Colorado experienced an uptick in the residual vote rate after implementing VBM. The cause of this increase can never be fully explained (it may in part be due to VBM or other things entirely), but high residual vote rates are associated with things like increased levels of voter confusion due to confusing paper ballot instructions or procedural errors during the counting process. States can work to combat this in November by ensuring that their paper ballots are as clear as possible and, if they are able, try to bolster their capacity for counting many mail-in ballots in the coming months. 

Up next: we’ll look at what the EPI reveals about in-person voting. Though political scientists are (and should) be paying attention to voting by mail in light of the pandemic, in-person voting remains a method that many Americans use and prefer. See you there! 

A photo of Pia Deshpande

Pia Deshpande is a writer with the MIT Election Data + Science Lab.

Topics Vote by mail and absentee voting