Honest Elections Seattle InitiativeDemocracy Voucher Usage and Low-Turnout Voter EngagementEvaluation in 2017 and 2019 Prepared for
Piper Fund, a Proteus Fund initiative.
15 Research Drive, Suite BAmherst, MA 01002
Prepared by
Keshavan Sridhar, Data and Analytics [email protected] Chris Langeler, Program [email protected] Win|Win Network1402 3rd Ave, Ste 201Seattle, WA 98101
August 14, 2021
Summary
Participation in the City of Seattle’s Democracy VoucherProgram through using at least oneDemocracy Voucher in either 2017 or 2019 was associatedwith a higher likelihood of voting in the 2017and 2019 Seattle elections among low-turnout voters.The three categories of low-turnout votersexamined are “First-Time Eligible Voters”: new votersthat can register and vote for the first time,“Non-Voters”: voters who never voted in the previouseight eligible odd-year Seattle elections, and “LowPropensity Voters”: voters who had up to a 25% votingrecord in the previous eight eligible odd-yearSeattle elections. “Odd-year Seattle elections” refersto both primary and general Seattle municipalelections held during each odd-year. The comparisonof likelihood to vote is between that of low-turnoutvoters who used at least one voucher and that of low-turnoutvoters who did not use a voucher. ● First-Time Eligible Voters who used at least one voucherin 2017 were 11.75 times more likely tovote in 2017. ● First-Time Eligible Voters who used at least one voucherin 2019 were 6.02 times more likely tovote in 2019. ● Non-Voters who used at least one voucher in 2017 were10.2 times more likely to vote in 2017. ● Non-Voters who used at least one voucher in 2019 were6.4 times more likely to vote in 2019. ● Low Propensity Voters who used at least one voucherin 2017 were 7.4 times more likely to votein 2017. ● Low Propensity Voters who used at least one voucherin 2019 were 4.5 times more likely to votein 2019.
When analyzing combined 2017 and 2019 voucher programparticipation and voting, it is clearthe more elections in which low-turnout voters participatedin the voucher program, the higher theirpropensity to vote was through those years. The fourelections in 2017 and 2019 were the 2017 Primary,2017 General, 2019 Primary, and 2019 General elections. ● Low-turnout voters from all three voter categorieswho used at least one voucher in both 2017and 2019 on average voted in 3 more elections thana low-turnout voter who did not use avoucher in any election. ● Low-turnout voters from all three voter categorieswho used at least one voucher in exactly oneelection year on average voted in 2 more electionsthan a low-turnout voter who did not use avoucher in any election. 2 1 Introduction
During the last decade, several different forms oflocal public financing options 1 have beenpiloted or established in municipalities throughoutthe U.S. The City of Seattle’s ‘Democracy Voucher’Program, created via the “Honest Elections Seattle” 2 (HES) initiative in 2015, represents one such modelfully implemented within a jurisdiction via the SeattleEthics & Elections Commission (SEEC). Dataanalysis on Democracy Voucher usage during the program’sfirst six years of implementation yieldedsome encouraging civic participation outcomes.
During the 2017 election, which was the first electionDemocracy Vouchers were distributed andaccepted as contributions to Seattle City Councilcandidates, nearly 9 out of 10 voucher users had neverpreviously contributed to a candidate for local officein Seattle. 3 The percentage of Seattle residentsdonating to any candidate also increased from 1.3%of all eligible adults in 2015 to 3.4% in 2017. 4 84% ofDemocracy Voucher users in 2017 were “new donors,”meaning they had not previously made a cashdonation to a political candidate according to campaignfinance records. 4 These results suggest progresstoward a primary goal of the original HES initiative:enabling constituents who are not traditional donorsto contribute to candidates that best reflect theirvalues and priorities.
Another core goal of the original HES initiative wasto increase voter participation amongunderserved communities and infrequent voters by removingbarriers that encumbered them fromcontributing to supported candidates during the electioncycle. 5 Preliminary analysis from 2017 showsthat among Seattle residents who voted in fewer thanhalf of the previous elections for which they wereeligible, voucher users were four times as likelyto vote in the 2017 election than non-voucher userswere. 6 The impact of voucher usage on voting among low-turnoutvoters is the main focus of this report.Our working hypothesis is that participation in theDemocracy Voucher Program positively impactedvoting rates among low-turnout voters and this impacton voting behavior persisted beyond a singleelection cycle. While our analytical methods will not result in causal conclusions, we predict findingswillsuggest a positive relationship between the DemocracyVoucher program and voting behavior. ________________________________________________________________ 1 “Faces of Small Donor Public Financing 2021.”BrennanCenter for Justice. March 11, 2021. 2 “Honest Elections Seattle Initiative (I-122).”Cityof Seattle. 3 “Seattle Democracy Program Evaluation.”BERK. April25, 2018. Page 2. 4 “Seattle Democracy Program Evaluation.”BERK. April25, 2018. Page 18. 5 Liz Dupee and Alissa Haslam. “The Honest ElectionsSeattle Story: The Path to Democracy Vouchers, theCampaign, and WhatWe Learned Along the Way.”Win|Win Network and WashingtonDemocracy Hub. Page 3. 6 Jennifer Heerwig and Brian J. McCabe. “ExpandingParticipation in Municipal Elections: Assessing theImpact of Seattle’sDemocracy Voucher Program.”University of Washington.Page 5. 3 2 Research Scope, Methods, and Data Sources
2.1 Voter Categories
Voters are segmented into six different categoriesbased on their prior vote history. For everyvoter category that has a percentage range, the lowerpercentage is not included in the voter category,while the upper percentage is included in that votercategory.
Super Voter75% - 100% vote history for past eight eligible odd-yearprimary and general elections Frequent Voter50% - 75% vote history for past eight eligible odd-yearprimary and general elections Infrequent Voter25% - 50% vote history for past eight eligible odd-yearprimary and general elections Low Propensity Voter0% - 25% vote history for past eight eligible odd-yearprimary and general elections Non-Voter (never voted)0% vote history for past eight eligible odd-year primaryand general elections First-Time Eligible Voters (new voter)This is the first election in which the individualwas eligible to vote When determining voter categories, only prior odd-yearelections are used. This is to avoidcomparing odd-year elections to even-year elections,which are different in nature. Someone who may bea low-turnout voter in odd-year elections may havea much better voting record in even-year elections,which often include statewide or federal races. Givenone of the goals of the voucher program is toincrease civic engagement in municipal elections,the focus on only odd-year elections is an appropriateapproach.
Low-turnout voters examined in this report encompassthree voter categories: Low PropensityVoters, Non-Voters, and First-Time Eligible Voters.
2.2 Research Scope and Methods
We first quantify the likelihood of low-turnout votersvoting in an election year when at least onevoucher was used that same year, compared to low-turnoutvoters who did not use a voucher that year.Specifically, we investigate whether low-turnout votersusing at least one voucher in Seattle municipalelections in 2017 and 2019 increased the likelihoodof a low-turnout voter participating in those elections 4 by either voting in the primary or general elections of those years. This first question looks at thepresence or absence of voting in an election yearand looks at 2017 and 2019 separately.
Bar graphs are used to visualize the likelihood increaseof voting in an election between voucherusers and non-voucher users for each year and eachof the three low-turnout voter categories of interest(Fig. 4-9). Likelihood is measured in terms of theodds ratio (OR) of voting with voucher usage andvoting without voucher usage.
Calculating Odds Ratio (OR):LetVc
= voucher userswho voted,Vu
= voucher users who didnot vote,Nc
= non-voucher users who voted, andNu
= non-voucher users who did not vote. Then,OR= (Vc
/Vu
)/(Nc
/Nu
). In our analysis, the ORgives how much more likely a voucher userwas to vote than a non-voucher user in a given year,for each of the three low-turnout votercategories.
Our second analysis considers the combined numberof votes cast by low-turnout voters in the2017 and 2019 Primary and General elections comparedto each group’s participation in the 2017 and2019 voucher programs. This looks at 2017 and 2019combined and compares either zero, one, or twoelection years of voucher program participation tothe exact number of votes that were cast spanningthetwo years, which was between 0-4. The average (median)value is reported, and is compared to the levelsof voucher program participation over the two electionyears (0, 1 or 2). The population of voters in thisanalysis that is used are those who were eligibleto both vote and use a voucher in 2017 and 2019. Hence2017 voter categories were used for the combined analysis.
Some of the voters after 2017 moved into higher-turnoutvoter categories in 2019; thus, theaverage change to individuals’ past vote histories between 2017 and 2019 as a percentage is also givenforeach of the three categories. Boxplots are createdfor each voter category to illustrate the correlationbetween the number of elections a voter used at leastone voucher in and the increase in the amount ofvotes cast (and the related increase in vote historypercentage) (Fig. 11-16).
Unregistered or ineligible voters who were voucherusers are not considered in this analysis.
2.3 Data Sources
The voter data for analysis in this report was obtainedfrom the 2017 and 2019 Seattle voter files 7 and then appended to expanded vote history data sourcedfrom Catalist. 8 The 2017 and 2019 DemocracyVoucher program data was sourced from the SeattleEthics and Elections Commission (SEEC) 9 , whichadministers the Democracy Voucher Program.
________________________________________________________________ 7 ”Research and Past results - Elections and Voting.”WA Secretary of State.8 “Progressive data.”Catalist.9
“Seattle Ethics and Elections Commission.”City ofSeattle. 5
2.4 Why Low Turnout Voters?
To examine low-turnout voter behavior, we are focusingspecifically on registered voters who areLow Propensity Voters, Non-Voters, and First-TimeEligible Voters. The correlation between voucherusage and vote propensity with these voter categoriesis more relevant, given that higher propensity votersare more likely to vote regardless of any externalfactors.
Critical in shaping the focus of our analysis is anunderstanding of the makeup of registered,voucher-using voters. Voucher users who were registeredvoters in 2017 were overwhelmingly SuperVoters (44.5%), followed by Frequent Voters (15.9%), Infrequent Voters (13.7%), Low Propensity Voters(12.8%), Non-Voters (12.5%), and finally First-TimeEligible Voters (0.5%) (Fig. 1).
Figure 1. 2017 Voucher Usage by Voter Type
Fig. 1. Breakdown of Voucher Usage (%) by Voter Typein the 2017 Primary and General Elections in Seattle(N =19426). 7-9
6 The same pattern is seen when looking at 2019 voucher usage, during which voucher users whowere registered voters were again predominantly SuperVoters (41.7%), followed by Frequent Voters(17.4%), Infrequent Voters (14.6%), Low PropensityVoters (14.3%), Non-Voters (11.3%), and First-TimeEligible Voters (0.7%) (Fig. 2).
Figure 2. 2019 Voucher Usage by Voter Type
Fig. 2. Breakdown of Voucher Usage (%) by Voter Typein the 2019 Primary and General Elections in Seattle(N =37941). 7-9
7 Figure 3shows the distribution of voucher users who were registered voters by vote percentage in both2017 and 2019, with the black dot indicating the meanvalue. Noticeably, the distributions are identicalforboth years. The average (median) voucher user hada prior vote history of 75% in both 2017 and 2019.Inother words, the average voucher user was a FrequentVoter.
Figure 3. Voucher Users Vote History Distributionin 2017 and 2019 Fig. 3. Boxplot of vote history distribution for voucherusers in the 2017 and 2019 Primary and General Electionsin Seattle. For2017, N = 19426, Mean (black dot) = 62.7% , Median(vertical black line) = 75%.. For 2019, N = 37941,Mean (black dot) =61.9%, Median (vertical black line) = 75%. 7-9
Examining any differences in likelihood of votingwhen comparing voucher and non-voucherusage among high propensity voters, while interesting,is not particularly salient. What is noteworthy toexamine is if voucher usage increases the likelihoodof low-turnout voters participation. The voterscharacterized by the “whisker” portion of the boxplotsdepicted inFigure 3are the focus of our analysis,namely Low Propensity or Non-Voters, along with First-TimeEligible Voters, who are not represented onthe plot due to having no prior voting ability. 3 Results: Single Election Comparison
In this section, for each low-turnout voter category,we examine how much more likely it was for avoucher user to vote compared to a non-voucher userfor both the 2017 and 2019 Seattle elections(summarized inFigures 4-9). The calculation for thislikelihood is the OR calculation, which is foundinSection 2.2of this report. 8 3.1 First-Time Eligible Voters: 2017
Figure 4. First-Time Eligible Voters: 2017 Vote History Fig. 4. Percentage of registered First-Time EligibleVoters, Vote History, and Voucher Usage in the 2017Primary andGeneral Elections in Seattle. For Voucher Users, N= 106. For Non-Voucher Users, N = 5493. 7-9
First-Time Eligible Voter Voucher Users were 11.75times more likely to vote at least once in 2017compared to First-Time Eligible Voter Non-VoucherUsers. 10 3.2 First-Time Eligible Voters: 2019
Figure 5. First-Time Eligible Voters: 2019 Vote History Fig. 5. Percentage of registered First-Time EligibleVoters, Vote History, and Voucher Usage in the 2019Primary andGeneral Elections in Seattle. For Voucher Users, N= 268. For Non-Voucher Users, N = 6072. 7-9
First-Time Eligible Voter Voucher Users were 6.02times more likely to vote at least once in 2019compared to First-Time Eligible Voter Non-VoucherUsers. 11 ________________________________________________________________ 10 The 95% Confidence Interval is from 6.68 to 20.6811
The 95% Confidence Interval is from 4.38 to 8.29 9
3.3 Non-Voters: 2017
Figure 6. Non-Voters: 2017 Vote History
Fig. 6. Registered Non-Voters (%), Vote History, andVoucher Usage in the 2017 Primary and General ElectionsinSeattle. For Voucher Users, N = 2428. For Non-VoucherUsers, N = 185573. 7-9
Non-Voter Voucher Users were 10.2 times more likelyto vote at least once in 2017 compared toNon-Voter Non-Voucher Users. 12 3.4 Non-Voters: 2019
Figure 7. Non-Voters: 2017 Vote History
Fig. 7. Registered Non-Voters (%), Vote History, andVoucher Usage in the 2019 Primary and General ElectionsinSeattle. For Voucher Users, N = 4282. For Non-VoucherUsers, N = 199100. 7-9
Non-Voter Voucher Users were 6.4 times more likelyto vote at least once in 2019 compared to Non-VoterNon-Voucher Users. 13 ________________________________________________________________ 12 The 95% Confidence Interval is from 9.32 to 11.1813
The 95% Confidence Interval is from 6.01 to 6.83
10 3.5 Low Propensity Voters: 2017
Figure 8. Low Propensity Voters: 2017 Vote History
Fig. 8. Registered Low Propensity Voters (%), VoteHistory, and Voucher Usage in the 2017 Primary andGeneralElections in Seattle. For Voucher Users, N = 2494.For Non-Voucher Users, N = 91108. 7-9
Low Propensity Voucher Users were 7.4 times more likelyto vote at least once in 2017 compared to LowPropensity Non-Voucher Users. 14 3.6 Low Propensity Voters: 2019
Figure 9. Low Propensity Voters: 2017 Vote History
Fig. 9. Registered Low Propensity Voters (%), VoteHistory, and Voucher Usage in the 2019 Primary andGeneralElections in Seattle. For Voucher Users, N = 5437.For Non-Voucher Users, N = 102516. 7-9
Low Propensity Voter Voucher Users were 4.5 timesmore likely to vote at least once in 2019 comparedtoLow Propensity Voter Non-Voucher Users. 15 ________________________________________________________________ 14 The 95% Confidence Interval is from 6.55 to 8.2915
The 95% Confidence Interval is from 4.21 to 4.88 11
4 Results: Comparison of Combined 2017 and 2019 Elections In the previous section, analysis is confined to asingle election, examining using at least one voucherandvoting at least once in that same election. In thissection, we additionally examine the impact of usingatleast one voucher in 0, 1, or 2 elections between2017 and 2019 on the exact number of votes castbetween the two years, namely in voting in 0, 1, 2,3, or 4 elections in 2017 and 2019 combined (Primaryand General elections in both years).
This analysis is confined to those individuals whowere able to both use a voucher and cast a vote inboththe 2017 and 2019 elections. Voter categories arebased on 2017 voter groupings.
Figure 10illustrates the pattern we expect: regardlessof the low-turnout voter category, we see anincrease in the number of elections in which individualsvoted with an increase in the number of electionsin which individuals used at least one voucher.
Figure 10. Vouchers Used and Voting in 2017 and 2019
Fig. 10. Average (mean) number of votes cast (outof 4) and number of elections (either 2017, 2019,or both)Democracy Vouchers were used in (0, 1 or 2). For First-TimeEligible Voters, N = 4987. For Non-Voters, N = 153112.For Low Propensity Voters, N = 78833. 7-9
Figure 10depicts mean values as the average.Figures11, 13,and15in the next subsections depictmedian average votes cast for each low-turnout votercategory, and these median values are used foranalysis. 12 4.1 First-Time Eligible Voters: 2017 and 2019 Elections Combined Analysis Figure 11. First-Time Eligible Voters: Vouchers andVoting (2017/2019) Fig. 11. Total votes cast (0-4) and total years voucherprogram participation (0-2) for 2017 and 2019 electionscombined. Median values represented by black solidlines. For First-Time Eligible Voters who did notuse a voucher in2017 or 2019, N = 4690. For First-Time Eligible Voterswho used at least one voucher in 2017 or 2019, butnot in bothyears, N = 267. For First-Time Eligible Voters whoused at least one voucher in both 2017 and 2019,N = 30. 7-9
A First-Time Eligible Voter who didn’t use a voucher in either election had an average (median) voterecord of 0 votes. A First-Time Eligible Voter who used at least one voucher in either 2017 or 2019(butnot both) had an average vote record of 2 votes. AFirst-Time Eligible Voter who used at least onevoucher in both 2017 and 2019 had an average voterecord of 3 votes (Fig. 11).
A First-Time Eligible Voter who used at least one voucher in both 2017 and 2019 on average voted in3more elections than a voter who did not use a voucher.Just using at least one voucher in one electionmade the First-Time Eligible Voter on average votein 2 more elections than if no voucher was used (Fig.11). 13 Figure 12. First-Time Eligible Voters: Vouchers and Vote History Change (2017/2019) Fig. 12. Vote history change (%) and total years voucherprogram participation (0-2) for 2017 and 2019 electionscombined. Median values represented by black solidlines. For First-Time Eligible Voters who did notuse a voucher in2017 or 2019, N = 4690. For First-Time Eligible Voterswho used at least one voucher in 2017 or 2019, butnot in bothyears, N = 267. For First-Time Eligible Voters whoused at least one voucher in both 2017 and 2019,N = 30. 7-9
A First-Time Eligible Voter who didn’t use a voucher in either election had an average (median) votehistory change of 0% by the end of 2019. A First-Time Eligible Voter who used at least one voucher ineither 2017 or 2019 (but not both) had an averagevote history change of +50% by the end of 2019 . AFirst-Time Eligible Voter who used at least one voucherin both 2017 and 2019 had an average votehistory change of +100% by the end of 2019 (Fig. 12). 14 4.2 Non-Voters: 2017 and 2019 Elections Combined Analysis Figure 13. Non-Voters: Vouchers and Voting (2017/2019)
Fig. 13. Total votes cast (0-4) and total years voucherprogram participation (0-2) for 2017 and 2019 electionscombined for Non-Voters. Median values representedby black solid lines. For Non-Voters who did notuse a voucher in2017 or 2019, N = 147260. For Non-Voters who usedat least one voucher in 2017 or 2019, but not inboth years, N =5313. For Non-Voters who used at least one voucherin both 2017 and 2019, N = 539. 7-9
A Non-Voter who didn’t use a voucher in either electionhad an average (median) vote record of 0 votes.A Non-Voter who used at least one voucher in either 2017 or 2019 (but not both) had an average voterecord of 2 votes. A Non-Voter who used at least one voucher in both 2017 and 2019 had an average voterecord of 3 votes (Fig. 13).
A Non-Voter who used at least one voucher in both2017 and 2019 on average voted in 3 more electionsthan a voter who did not use a voucher. Just usingat least one voucher in only one election made theNon-Voter on average vote in 2 more elections thanif no voucher was used (Fig. 13). 15 Figure 14. Non-Voters: Vouchers and Vote History Change (2017/2019) Fig. 14. Vote history change (%) and total years voucherprogram participation (0-2) for 2017 and 2019 electionscombined for Non-Voters. Median values representedby black solid lines. For Non-Voters who did notuse a voucher in2017 or 2019, N = 147260. For Non-Voters who usedat least one voucher in 2017 or 2019, but not inboth years, N =5313. For Non-Voters who used at least one voucherin both 2017 and 2019, N = 539. 7-9
A Non-Voter who didn’t use a voucher in either electionhad an average (median) vote history change of0% by the end of 2019. A Non-Voter who used at least one voucher in either 2017 or 2019 (but not both)had an average vote history change of +12.5% by theend of 2019. A Non-Voter who used at least onevoucher in both 2017 and 2019 had an average votehistory change of +25% by the end of 2019 (Fig. 14). 16 4.3 Low Propensity Voters: 2017 and 2019 Elections Combined Analysis Figure 15. Non-Voters: Vouchers and Voting (2017/2019)
Fig. 15. Total votes cast (0-4) and total years voucherprogram participation (0-2) for 2017 and 2019 electionscombined for Low Propensity Voters. Median valuesrepresented by black solid lines. For Low PropensityVoters whodid not use a voucher in 2017 or 2019, N = 72911.For Low Propensity Voters who used at least one voucherin 2017or 2019, but not in both years, N = 5304. For LowPropensity Voters who used at least one voucher inboth 2017 and2019, N = 618. 7-9
A Low Propensity Voter who didn’t use a voucher ineither election had an average (median) vote recordof 1 vote out of 4. A Low Propensity Voter who used at least one voucher in either 2017 or 2019 (butnotboth) had an average vote record of 3 votes. A LowPropensity Voter who used at least one voucher inboth 2017 and 2019 had an average vote record of 4votes-- the voter participated in every electionin both2017 and 2019 (Fig. 15).
A Low Propensity Voter who used a voucher in both2017 and 2019 will have on average voted in 3 moreelections than a voter who did not use a voucher.Just using at least one voucher in one election madetheLow Propensity Voter on average vote in 2 more electionsthan if no voucher was used (Fig. 15). 17 Figure 16. Low Propensity Voters: Vouchers and Voting (2017/2019) Fig. 16. Vote history change (%) and total years voucherprogram participation (0-2) for 2017 and 2019 electionscombined for Low Propensity Voters. Median valuesrepresented by black solid lines. For Low PropensityVoters whodid not use a voucher in 2017 or 2019, N = 72911.For Low Propensity Voters who used at least one voucherin 2017or 2019, but not in both years, N = 5304. For LowPropensity Voters who used at least one voucher inboth 2017 and2019, N = 618. 7-9
A Low Propensity Voter who doesn’t use a voucher ineither election had an average (median) votehistory change of 0% by the end of 2019. A Low Propensity Voter who used at least one voucher in either2017 or 2019 (but not both) had an average vote historychange of +12.5% by the end of 2019. A LowPropensity Voter who used at least one voucher inboth 2017 and 2019 had an average vote history changeof +25% by the end of 2019 (Fig. 16). 5 Conclusions and Recommendation
The analyses in this report suggest a strong relationshipbetween participation in the DemocracyVoucher Program and voting behavior among low-turnoutSeattle voters. In every category of low-turnoutvoter in both the 2017 and 2019 Seattle elections,it is shown that a voucher user was more likely tovotethan a non-voucher user. It is furthermore shown thatthe number of elections in which a low-turnoutvoter participated using a Democracy Voucher positivelycorrelates with the total number of votes cast inthose elections by that voter.
In this report, we are only able to draw correlationalconclusions that suggest a strong positivelikelihood between participation in the DemocracyVoucher Program and increased voter participation 18 among low-turnout voters in both individual and multiple elections. Additionally, because all analyseswere conducted in 2021, there was no opportunity tocreate control groups or prescribed conditions duringthe 2017 or 2019 cycles to establish more nuanced,causal conclusions. There is a clear need for morecomprehensive research that would allow for prescribedconditions, control groups, and collection andanalysis of data during a local election cycle tofurther understand the nature of these relationships.
Despite the aforementioned limitations, it is clearthere exists a strong potential for greaterparticipation in the voting process by removing financialand bureaucratic barriers to supportingcandidates in local elections through the DemocracyVoucher Program. Expanding the DemocracyVoucher Program in Seattle and other jurisdictionscould be an impactful way to increase equitable votingparticipation in a sustainable manner in local elections. 19