To: Interested Parties

Re: Latest AVR Data From Oregon and California

From: Sean McElwee, Executive Director of AVR NOW

Date: March 10, 2019

Summary

Official data, made available recently from California and Oregon, allow a more thorough look at how automatic voter registration (AVR) has affected overall rates of registration. These data points are helpful, not just in comparison to states without an AVR system, but also to understand the difference between two different and distinct automatic voter registration systems. Both models significantly improved the rate at which individuals are being registered, but the Oregon model is accomplishing a larger reach, much of it by keeping the voter rolls more up-to-date through automated address updates. This newest data further reinforces the lessons of previously available information showing far higher rates of successful transactions in Oregon than in California and in states that have implemented model similar to California. Previous analysis indicates that 94 percent of eligible but unregistered Oregonians who visit the DMV end up registered to vote. This new data and analysis continues to indicate that Oregon’s model is the proven method to have complete and up-to-date voter rolls.

Lessons Learned:

Lesson #1: All Automation Models Improve Voter Registration Outcomes

Lesson #2: Different Models Achieve Different Registration Outcomes

Lesson #3: Automating Updates Improves Registration Rates

Lesson #4: Different States Will Have Different Baselines

Lesson #5: Functional DMVs May Make a Big Difference

Our Goals

As advocates, we closely monitor the outcomes of this work in order to be sure that AVR is achieving the promise of creating a system that is accurate, complete, and registers every eligible voter, except for individuals who genuinely don't want to be registered. Realizing this goal of creating the most accurate and complete voter file possible results in three key benefits:

  • Greater Integrity - Accurate and up-to-date rolls are a meaningful safeguard to ensure that votes are being cast only by eligible voters and to create greater trust among the public. 

  • Greater Access - Confusion about voter registration remains one of the greatest impediments to voting for eligible voters. Out-of-date registrations pose burdens on voters who believe that registration is already portable. Finally, as more states move to vote-from-home systems, it's even more important that governments have the correct mailing address for voters.

  • Greater Efficiency - Processing electronic registration information costs significantly less than processing paper forms. Additionally, well-designed automated interfaces can reduce demands on front-line employees’ time to as little as 0-15 seconds per registration compared to traditional voter registration systems that require roughly 90 seconds per interaction.

Lesson #1: All Automation Models Improve Voter Registration Outcomes

The latest data from California and Oregon indicates that both systems -- despite operating very differently -- continue to outperform traditional paper forms and manual transmission of data. In Oregon, the AVR system alone processed 926,550 new registrations and updates in 2017 and 2018.[1] That is significant growth for the state. In the 2014 election cycle, prior to adoption of AVR, Oregon’s DMV processed 230,295 registrations,[2] meaning registration rates at the DMV have roughly quadrupled.

California does not allow for an apples-to-apples comparison as the state’s automated system was not fully implemented for the entire election cycle, but we can extrapolate from existing data. In its first six full months of operation, California’s new system processed 759,173 new registrations and updates;[3] multiplying by four to estimate for a two-year cycle produces a total of roughly 3,000,000 new and updated registrations. The comparison with previous cycles is even more pronounced. In the 2016 cycle before AVR had been implemented, California’s DMV processed only 694,209 registrations over the two-year period.[4]

State Registrations and
Updates Pre-AVR
Registrations and
Updates w/AVR
CA 694,209 3,000,000* (est)
OR 230,295 926,550

 Lesson #2: Different Models Achieve Different Outcomes

It is worth noting that while these two systems achieved similar improvements from their states’ historic baselines, the overall impact is quite different for the two systems. One could look at the growth in registration rates in the two states and see a greater improvement in California. One could look at absolute numbers and see that California is registering more than three times as many voters as Oregon. But these analyses overlook that California has a voting eligible population nearly eight times larger than Oregon and is building from a baseline of a far weaker registration regime at the DMV.

Rather than looking at raw numbers or changes in rates, we can also compare the number of applications processed in a state to the size of the voting eligible population, giving us a rough way of measuring the reach of the AVR system. By this tally, the impact is not close. Oregon’s DMV in the 2017-2018 election cycle processed new or updated application registrations for 29.8 percent of the eligible voters in the state.[5] By comparison, California’s system reached the equivalent of 11.8 percent of the eligible voters in the state.[6]

StateRegistrations & Updates Pre-AVRRegistrations & Updates w/ AVRVoting Eligible Population - 2016Percent of State Reached
CA694,2093,000,000 (est)25,635,13911.8 percent*
OR230,295926,5503,113,17829.8 percent

This data, showing significantly greater reach for the Oregon model relative to the voting eligible population, strengthens the conclusions from previous analyses looking at opt-out rates by state. Oregon’s 94 percent registration rate is significantly higher than Colorado’s 30 percent registration rate among currently unregistered voters. Because California does not distinguish decline rates by eligibility or registration status, we do not know what the equivalent registration rate is for that state.  However, these numbers indicate that the system is almost certainly not achieving the registration rates of Oregon.

 Lesson #3: Automating Updates Matters

Notably, by far the biggest difference between California and Oregon outlined in lesson #2 is not in the processing of new registrations but in processing updates. Only a new registration will change the total number of registered voters in a state but address updates are crucial to ensure that transient communities are engaged in our democratic process. California is on track for 2,356,844 new registrations over an election cycle, approximately 9.2 percent of the state’s eligible voters. Oregon processed 303,362 new registrations over the cycle, approximately 9.7 percent of the state.[7] But the rates of updates in the two systems are very different, so different in fact that California only updated a few more registrations than Oregon despite eight times as many eligible voters.

StateRegistrations & Updates w/ AVRNew RegistrationsUpdated RegistrationsPercent of Applications that are Updates
CA3,000,000 (est)2,356,844679,84822.4 percent
OR926,550303,362623,18867.3 percent

Notably, Colorado’s former secretary of state implemented a front-end automatic voter registration system similar to California’s and according to that office saw only 30 percent percent of unregistered eligible voters become registered. Colorado also started automating all address updates leading to significant improvement in quality of data. The state’s new secretary of state is looking to further reform and improve the system, including moving to a back-end system for new registrations as well. With California moving toward replicating Colorado’s universal mail ballot system, it will become even more important to ensure that addresses of transient communities are kept up-to-date in the elections system.

 Lesson #4: Different States Will Have Different Baselines

While California is on track to register nearly as many new voters through their AVR system as Oregon’s in terms of share of the eligible population, at least some of the gap between the two states in terms of new registrations is likely a reflection of Oregon’s higher level of voter registration. (Following several years of successful AVR implementation, Oregon has fewer eligible unregistered voters.) But we do not yet know what ideal rates of registration growth should look like. Colorado’s share of registered eligible voters is currently higher than Oregon’s but Colorado is still pursuing reforms to their AVR system because agency-level data indicates too high of a declination rate. California’s new registration rate looks promising but we simply do not know how many eligible voters in the state are being denied access to elections under the current AVR system. As AVR continues to propagate throughout the country, we will improve our ability to evaluate registration rates at various agencies. For now, this data merits more in-state discussion among lawmakers, advocates, and civic engagement practitioners to identify mechanisms to close the gaps.

 Lesson #5: Functional DMVs May Make a Big Difference

Even if California completely eliminated declinations at the DMV, they would be moving fewer new and updated registration than Oregon as a share of voting eligible population. This may be an indication that a big issue in California is lower-than-ideal utilization rates of DMV services because of widely reported issues with long waits and other problems.

 Conclusion

Early public reports from state governments indicate that automating registration creates significant improvements over traditional agency registration approaches that were reliant on paper and manual transfer of data. But there are notable variations between states in terms of accuracy and completeness of voter rolls under different models of automatic voter registration. By comparing rates of new and updated registrations from DMVs to the voting eligible populations for states, we can adjust for other variations in state systems and data to allow for a greater comparison.

 This comparison reveals that Oregon’s system is processing far more applications than California with the vast majority of those occurring as updates. As policymakers, administrators, and advocates work to ensure that we have the most accurate and complete voter rolls possible, these comparisons can provide real insight into how different models operate and where we may find additional improvements for all of them.


[1] Oregon Secretary of State, “2017 Oregon Motor Voter Registrations by County” (2/21/19) and “2018 Oregon Motor Voter Registrations by County” (2/15/19). Accessed 3/3/19.

[2] Table 2a of US Election Assistance Commission, “The 2014 EAC Election Administration and Voting Survey Comprehensive Report,” June 30, 2015.

[3] Calculated from California Secretary of State, “DMV New Motor Voter Registration Transactions April 2018 - Current” previous version with data from March-November 2018.

The newer version of the reports aggregate all 2018 data into a single line, making extrapolation to other time periods more difficult.

[4] Table 2a of US Election Assistance Commission, “The Election Administration and Voting Survey 2016 Comprehensive report,” 2017.

* Projection based on semi-annual data extrapolated and averaged for a two-year estimate

[5] Voting Eligible Population (VEP) corrects US Census estimates of Voting Age Population (VAP) to account for non-citizens and justice-involved Americans to estimate only the number of individuals in a state who are eligible to vote. Eligible voter estimates from US Elections Project, “2018 November General Election Turnout Rates,” 12/14/18. Accessed 3/3/2019.

[6] California’s eligible voter estimates also from US Elections Project. Evaluating registrations as a share of voting eligible population is a useful adjustment in California, where published data from the SOS includes ineligible individuals as an unknown portion of the declination universe. This presents a challenge to other forms of comparisons to Oregon. Because the voting eligible population estimates also exclude ineligible individuals, this approach should appropriately correct for these individuals.

[7] While these rates are similar, it is worth keeping in mind that Oregon started the 2017 election cycle from a much higher registration rate, meaning there are significantly fewer eligible unregistered voters to find. That implies that it should be possible to get a higher new registration rate in California.