Objectives. To examine trends in partisan polarization of childhood vaccine bills and the impact of polarization on bill passage in the United States.
Methods. We performed content analysis on 1497 US state bills (1995–2020) and obtained voting returns for 228 legislative votes (2011‒2020). We performed descriptive and statistical analyses using 2 measures of polarization.
Results. Vote polarization rose more rapidly for immunization than abortion or veterans’ affairs bills. Bills in 2019–2020 were more than 7 times more likely to be polarized than in 1995–1996 (odds ratio [OR] = 7.04; 95% confidence interval [CI] = 3.54, 13.99). Bills related to public health emergencies were more polarized (OR = 1.76; 95% CI = 1.13, 2.75). Sponsor polarization was associated with 34% lower odds of passage (OR = 0.66; 95% CI = 0.42, 1.03).
Conclusions. State lawmakers were more divided on vaccine policy, but partisan bills were less likely to pass. Bill characteristics associated with lower polarization could signal opportunities for future bipartisanship.
Public Health Implications. Increasing partisan polarization could alter state-level vaccine policies in ways that jeopardize childhood immunization rates or weaken responsiveness during public health emergencies. Authorities should look for areas of bipartisan agreement on how to maintain vaccination rates. (Am J Public Health. 2022;112(10):1471–1479. https://doi.org/10.2105/AJPH.2022.306964)
Responses to the availability and promotion of COVID-19 vaccines have been influenced by political party. Partisanship is a significant predictor of intent to be vaccinated against COVID-19,1 actual rates of vaccination,2 and lawmakers’ support for vaccine mandates.3 Did the unique circumstances of the pandemic give rise to these partisan outcomes, or was immunization already a polarized issue?
We define “polarization” as an alignment of partisan identity with particular policy positions. Scholars have given attention to such partisan alignments on public health policy,4,5 often with warnings that health policymaking is most likely to advance population health when it remains free of partisan politics.6,7 Difficulties related to managing the COVID-19 pandemic support those claims.1,8,9 When it comes to childhood immunizations, evidence suggests state policies have a strong effect on rates of vaccination and disease.6,10,11 Therefore, it is especially important to examine polarization among those making these important policy decisions.
Goldstein et al.12 examined immunization bills proposed in state legislatures from 2011 to 2017 and found that Republicans were more likely to sponsor “antivaccination” bills. Their analysis was limited to exemption bills and does not account for important political and epidemiological occurrences since 2017 (e.g., the majority of Trump’s presidency, measles outbreaks in 2019). We build on their work by examining a broader spectrum of childhood vaccine bills over a much longer period (1995–2020). This allows us to address 3 specific questions: (1) Have childhood vaccination policies become more polarized over time? (2) Are some types of vaccination bills more polarized? (3) What is the relationship between polarization and the likelihood of a bill passing into law?
Our study period encompasses significant episodes in the history of US immunization policy, including the autism‒mercury controversy, the politicized rollout of the human papillomavirus (HPV) vaccine, and the resurgence of measles.6,13,14 Although childhood immunization laws have historically enjoyed widespread support across party lines,15 organized campaigns to “green our vaccines” or to oppose HPV mandates on the basis of concerns about sexual promiscuity14 could resonate more with the ideological commitments of the political left or right. Thus, the emergence and decline of these conflicts could produce changes in polarization over time.
Recent public opinion16 as well as academic research on policymakers,12 health communication,17 and vaccine exemptions18 indicate a correlation between vaccine opposition and political conservativism. Importantly, vaccine opposition is increasingly framed as an issue of civil liberties.17 Thus, lawmakers may take positions based on partisan ideologies concerning state authority and parental rights14,19 rather than on scientific recommendations or safety concerns. Certain bills could be more likely to trigger these ideological differences, such as those that target parents, modify vaccine mandates, or adjust public health authority during outbreaks.
To address these questions, we used sponsorship data on 1497 bills and voting returns for 228 legislative votes to create 2 separate measures of polarization: sponsor polarization and vote polarization. The first allowed us to view changes in polarization over a longer period; the second allowed us to compare polarization levels on immunization with 2 issues commonly considered more and less polarized (abortion and veterans’ affairs).
We searched NexisUni’s (LexisNexis) comprehensive “bill text” database to identify childhood vaccination bills introduced into state legislatures between 1995 and 2020. Because of our focus on childhood immunizations, we excluded bills concerning adults, college students, insurance coverage, funding appropriations, provider education, and professional criteria for administering vaccines. To obtain search results that best matched our inclusion criteria, we used the following Boolean search: “(Vaccin* OR Immuniz*) AND NOT (military OR dog OR veterinar* OR pet OR animal OR rabies OR flu OR influenza OR appropriations OR ‘higher education’ OR college).” We identified 1044 unique bills from 7093 results.
We located an additional 432 bills in a data set of vaccine bills (1995–2012) provided by the authors of a previous study13 and 127 bills in a list compiled by the Association of State and Territorial Health Officials (2011‒2016).20 This yielded 1603 bills. We eliminated 106 cases for missing data on key variables (as noted subsequently), resulting in 1497 bills in our primary data set. Two of these sources cover only a subset of years in our study: the period for the Lillvis et al. study ended in 2012,13 and the Association of State and Territorial Health Officials only tracked immunization bills for 5 years.20 To ensure this did not bias our results, we conducted a sensitivity analysis using only bills from the NexisUni search.
We used yearly partisan composition data from the National Conference of State Legislatures to identify the majority party in the chamber in which each bill was introduced.21 We were also interested in whether polarization is related to how strict or lenient a state’s policy is regarding exemptions. We used National Conference of State Legislatures’ data on current exemption policies22 and Conis’s23 historical analysis of changes to state exemption policies to generate a yearly variable with categories: medical exemptions only, medical and religious exemptions, or medical, religious, and philosophical exemptions.
We created a second data set that captures information on how Republican and Democratic legislators voted on the bills in our primary data set. We obtained voting records from LegiScan (LegiScan LLC, Elkview, WV), which pulls data from state legislature Web sites. We looked up each bill introduced between 2011 (the beginning of the earliest complete legislative session in the LegiScan database) and 2020 and recorded available data for votes that occurred in committees or on the House or Senate floor. Votes on amendments and administrative procedures were not included. When multiple votes occurred in a single committee or chamber, we recorded the final vote. A bill that progressed through 1 committee in each chamber and ultimately passed into law would yield 4 observations: 1 committee vote and 1 floor vote in both the House and Senate. We were able to obtain data for 228 votes.
To benchmark levels of vaccine polarization, we calculated vote polarization for 2 comparison issues to represent the high and low ends of the issue-polarization spectrum. Abortion is considered highly partisan, whereas bipartisanship on veterans’ affairs is common even during times of intense conflict between the major parties.24 Details on the creation of these benchmarks are included in “Methodology for Abortion and Veterans’ Affairs Benchmarks” in the appendix (available as a supplement to the online version of this article at https://ajph.org).
The first author (K. E.) read the text of each bill and performed content analysis. To check for subjectivity in coding, the second author (A. M.) coded 50 randomly selected bills and obtained 95% observed agreement for variables used in this analysis, with a pooled κ of 0.83.25
We defined our primary outcome of interest, partisan polarization, as alignment of vaccine positions with party identity. To operationalize this, we first classified bills as “pro” or “anti.” Consistent with Goldstein et al.,12 we coded a bill as pro if it was interpreted as an effort to sustain or increase vaccinations among the target population, and anti if interpreted as having the potential to decrease vaccination rates. Twenty-five bills were dropped because of insufficient publicly available information to determine pro‒anti valence. We also located the primary sponsor’s party affiliation on state legislature Web sites. We eliminated 81 cases for missing data on this variable (51 sponsored by committees, 21 sponsored by Independent or other party, 9 for which sponsor party could not be found). Anti bills sponsored by Republicans and pro bills sponsored by Democrats were coded 1 for sponsor polarization; anti bills sponsored by Democrats and pro bills sponsored by Republicans were coded 0.
In our content analysis, we also generated the following non‒mutually exclusive dichotomous variables to indicate the goals of each bill: exemption (medical or nonmedical exemptions from school-based requirements); mandate change (adding or removing required immunization or changes to how requirements are established); vaccine contents (materials used in vaccines [e.g., mercury, fetal tissue]); education (provide or require education about risks, benefits, schedules, or exemptions); registry (establishes or modifies immunization tracking systems; guidelines for sharing immunization data); consent (rules for obtaining consent before immunization); and HPV; meningitis; measles, mumps, and rubella; and varicella (related to those specific vaccines).
We created a categorical variable indicating the bill’s primary target (i.e., the constituency whose interests, activities, responsibilities, or authority are most affected). Categories were health care providers, local or state health department, schools, parents or patients, pharmaceutical companies, and other.
Using data from state legislature Web sites, we created a variable to indicate if the bill ultimately passed into law. This variable was used to evaluate the effect of sponsor polarization on a bill’s likelihood of becoming law.
We used our secondary data set consisting of 228 votes on vaccine bills to create a second measure of polarization: vote polarization. We calculated an index of dissimilarity equal to half the sum of the absolute difference between the proportions of Democrats or Republicans voting yes or no. We used the following equation:
where n = number of vote categories (i.e., 2: yes or no); di = number of Democratic votes in vote category i; DT = total number Democratic votes in the floor or committee roll call; ri = number of Republican votes in vote category i; and RT = total number Republican votes in the floor or committee roll call. Vote polarization (P) represents the proportion of Republicans (or Democrats) that would need to change their votes for the bill to be perfectly bipartisan. The index ranges from 0 (bipartisan; equal proportions of Democrats and Republicans voted yes or no) to 1 (completely partisan; Republicans and Democrats segregated into yes‒no vote categories). A unanimous vote would have a value of 0. If all Republicans voted for and all Democrats against a bill (or vice versa), the value would be 1. We multiplied the index by 100 to improve interpretation. We also calculated vote polarization for abortion and veterans’ affairs bills.
Finally, legislative session indicates when the bill was originally introduced. We used 2-year increments because sessions in many states span 2 years.
In our analysis, we used both descriptive and inferential statistical techniques. To examine changes in polarization over time, we first conducted a linear time-trend test using an ordinary least squares regression model predicting sponsor polarization with a continuous measure of legislative session as the only predictor. Next, we plotted the proportion of bills in each session (1995–2020) coded 1 for sponsor polarization. Finally, we plotted the mean for vote polarization in each session (2011–2020), compared with abortion and veterans’ affairs bills, and conducted similar time-trend tests as described previously for each of the 3 issues by regressing vote polarization on our continuous legislative session variable.
To identify characteristics of bills that tend to be more polarized, we estimated a Bayesian generalized linear mixed-effects regression model with sponsor polarization as the dependent variable and various bill characteristics (goals, targets, specific vaccines, political context, legislative session) as independent variables. Finally, using passed as the dependent variable and sponsor polarization as a key independent variable, we estimated a second regression model to assess whether polarized bills are more or less likely to pass into law, while controlling for bill characteristics, political context, and session. Both regression models specified state as a random intercept to account for nonindependence of multiple bills being proposed in the states. The use of a Bayesian framework improved model convergence by allowing us to specify a Gaussian prior on the fixed effects with a mean of 0 and standard deviation of 10 (intercepts) and 2.5 (slopes).
We used Excel Version 16.37 (Microsoft Corp, Redmond, WA) to perform content coding. We used R version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria) to perform statistical analyses.
Results of our time-trend test indicated a statistically significant increase in sponsor polarization between 1995 and 2020 (b = 0.096; P < .01). Figure 1 displays the level of sponsor polarization in each legislative session. Between 1995 and 2014, the alignment of pro or anti with political party was somewhat inconsistent, ranging from a low of 50% in 2005–2006 (i.e., Republicans and Democrats were equally likely to sponsor pro and anti bills) to a high of 65% in 2003–2004. Overall, the degree of sponsor polarization during this period was relatively low—5 of 10 legislative sessions at or below 55%. However, beginning in 2015, sponsor polarization increased rapidly. In 2013–2014, just over half of childhood vaccine bills (55%) were either Republican‒anti or Democrat‒pro bills, but this increased to 85% in 2019–2020.
Figure 2 shows vote polarization for immunization, abortion, and veterans’ bills for 2011–2020. As expected, abortion was the most polarized during this period, while votes related to veterans were far less partisan. Time-trend tests showed vote polarization for immunization bills increased at twice the rate of abortion bills (b = 0.092 and 0.044, respectively), with no statistically significant change for veterans’ bills. In 2011–2012, polarization for immunization bills was somewhat higher than for veterans’ affairs bills (24% Republicans or Democrats required to change vote to achieve bipartisanship, compared with 9%). In 2019–2020, polarization for veterans’ bills was still very low (13%), but polarization for vaccine bills rose to 58%. Abortion remained the most polarized issue, but the gap between abortion and immunization closed substantially.
Figure 3 displays odds ratios (ORs) for the regression model predicting sponsor polarization. Bills proposed in 2015–2016 or 2017–2018 were more than twice as likely to be polarized compared with 1995–1996; bills introduced in 2019–2020 were more than 7 times as likely to be polarized. Because the model adjusts for bill characteristics, this indicates that the trend in Figure 1 is not simply an artifact of different kinds of bills being proposed in this later period.
Figure 3 also reveals several important insights about which kinds of bills were more (or less) likely to follow the polarized Republican‒anti and Democrat‒pro pattern. When we looked at goals of the bill, policies concerning immunizations during public health emergencies had 76% greater odds of being polarized (OR = 1.76; 95% confidence interval [CI] = 1.13, 2.75). Bills related to exemptions (OR = 0.58; 95% CI = 0.40, 0.84) were polarized in the reverse direction, with Democrats more likely to sponsor anti bills. For bills that mentioned specific vaccines, odds of polarization increased 2-fold for HPV bills (OR = 2.1; 95% CI = 1.29, 3.43), whereas meningitis was associated with reverse polarization (OR = 0.49; 95% CI = 0.29, 0.85). Results also suggest that Republican-sponsored bills were less polarized (OR = 0.64; 95% CI = 0.49, 0.83).
Figure 4 addresses our final research question about the effect of polarization on the likelihood of a bill becoming law. Adjusting for bill characteristics, sponsor polarization was associated with a 34% decrease in the odds of passage (OR = 0.66; 95% CI = 0.42, 1.03). Also notable, bills that would increase or decrease access to exemptions had 66% decreased odds of passage (OR = 0.34; 95% CI = 0.18, 0.64).
Our sensitivity analyses restricted to bills obtained only from NexisUni did not meaningfully alter our overall conclusions (Tables A and B, available as supplements to the online version of this article at https://ajph.org).
Some have described the nature of political polarization in the United States as an “oil spill” that gradually taints previously “apolitical” issues.26 Our analysis indicates that political polarization has, in fact, spread to vaccine policy. Among state legislators, positions on childhood immunization policy became substantially more aligned with political party between 2015 and 2020. This was true for 2 separate measures of partisan polarization—sponsorship and voting. Although increases in polarization across the political arena likely contributed to this shift,27 our comparative analysis of voting on immunization, abortion, and veterans’ affairs bills revealed an especially rapid increase in polarization for vaccine bills in recent years. By 2019–2020, nearly 60% of one party would need to “defect” to obtain bipartisanship on immunization bills. To put this in perspective, legislators were more likely to vote along party lines for vaccine bills in 2019–2020 than for abortion bills in 2011–2012, which was the height of polarized conflict between the Tea Party and the Obama administration (Figure 2).
The trends we have identified can be explained, in part, by the rise and fall of controversies during this timeframe. Beginning in 1998, Democratic and Republican lawmakers alike were focused on the possibility that the mercury-containing vaccine preservative thimerosal could be linked to autism. The high frequency of bills to prohibit mercury in vaccines and to expand access to exemptions in the early 2000s were almost certainly motivated by these safety concerns.13 However, a 2004 Institute of Medicine report established the scientific consensus against the autism‒mercury hypotheses, the Lancet retracted the study that first proposed a vaccine‒autism link because of ethical and scientific concerns, and the study’s lead author, Andrew Wakefield, was removed from the United Kingdom’s registry of physicians in 2010.13,14,28
Faced with the loss of legitimate scientific arguments, leaders of vaccine-critical organizations were compelled to innovate new narratives that might resonate with blocks of potential supporters.28 In their analysis of antivaccine Facebook pages from 2009 to 2019, Broniatowski et al.17 found a decreasing prevalence of safety or efficacy concerns and an increasing emphasis on protecting individual liberty from government intrusion. This shift in emphasis from safety to parental freedom would certainly have been felt by policymakers. Legislators might strongly disagree about whether certain vaccines are safe for children, but such disagreements are not inherently partisan. By contrast, as vaccine policies became increasingly framed in terms of individual liberty versus the state, legislators might have faced mounting pressure to choose their side based on adherence to conservative or progressive views on the appropriate role of government.
In addition, the 2015 measles outbreak in California, the publicity generated by the outbreak’s association with Disneyland, and the legislature’s subsequent decision to eliminate nonmedical exemptions brought intense public attention to this issue.17 Sociologists have shown that such “takeoff” conditions bring previously unrecognized disagreements to the surface.29 This, in turn, increases partisan polarization,30 as people interpret the newly contentious issue through the lens of existing partisan identities.31 In this case, California’s identity as a liberal state led by a well-known Democratic governor clearly signaled the Democratic position on vaccine exemptions to legislators in others states who were forming their own policy positions. The measles epidemic of 2019 put vaccines in the spotlight once again,17 forcing state legislatures to weigh calls from public health officials to restrict exemptions against the demands of those who saw this as an infringement on parental rights. Our findings suggest this second round of “takeoff” conditions further solidified party lines.
The partisan divide is narrower for some proposed vaccine policy changes than others. For instance, HPV tended to be more partisan, as others have shown,32 but not bills concerning varicella or measles, mumps, and rubella. Pro‒anti stances on bills related to immunization registries, education, consent, and even proposed changes to vaccine mandates were not associated with political party. Such findings could help lawmakers identify areas of potential bipartisanship.
Importantly, Republicans and Democrats did not see eye-to-eye on how to respond during public health emergencies—such as whether to exclude children with exemptions from school (102 bills) or to allow mercury-containing vaccines (52 bills) during outbreaks. The frequency of these bills peaked in 2005–2006, declined sharply even as cases of pertussis surged in 2012–2014,33 and remained very low during the measles outbreaks of 2015–2020, suggesting epidemiological threat might not have been legislators’ central concern. Here again, partisan convictions about the role of the state could be relevant. Exclusion of nonvaccinated students during an outbreak seems like a reasonable precaution, but such proposals could appear threatening to conservative legislators who are eager to protect parental control over school attendance. This sense of threat could explain why, in response to the actual crisis of a global pandemic, Republican lawmakers in at least 26 states have pushed to limit emergency powers of public health officials.34
Republican-sponsored bills were less likely to be polarized and, though falling just short of statistical significance, anti bills appear to be more polarized than pro bills. In other words, bills that might decrease vaccination coverage were usually sponsored by Republicans, but Republicans were also more willing than Democrats to cross the party line.
Experts suggest partisan polarization hinders effective public health policy.6,7 From this perspective, our findings are mostly disheartening. However, we find some encouragement in the fact that polarization decreased likelihood of a bill becoming law. Put differently, the policy proposals that are most likely to be passed and implemented are those that bridge the partisan divide. By its very design, lawmaking in the United States requires compromise and coalition building,35 which might protect population health from the worst effects of polarization.
We analyzed bills before the COVID-19 pandemic. Our findings shed light on current vaccination policy conflicts, but our data did not include COVID-19‒specific bills. Our analysis focused on childhood immunization, so some caution is warranted when applying these findings to adult vaccination policies. In addition, although we relied on multiple sources to identify bills that met our inclusion criteria, it is possible that some bills were unintentionally omitted. Finally, we relied on publicly available records, which could be incomplete.
This analysis demonstrated a clear increase in partisan polarization of childhood immunization bills in state legislatures between 1995 and 2020, which contradicts popular notions that vaccine polarization began with the COVID-19 vaccines. Fortunately, state-level legislative processes appear to have a protective effect, as polarized bills were less likely to pass into law, thereby lowering the potential for partisan agendas to translate into policies that could jeopardize childhood vaccination rates. Nevertheless, given the current consideration being paid to mandating COVID-19 vaccination among children in schools and previous legislative and legal controversies over mandating COVID-19 vaccination for adults, our analysis might suggest characteristics of bills less likely to engender polarization.
Members of the Creighton University Kingfisher Fellows program provided helpful comments. Jonathan Purtle provided early input on study design. Cara Buckley, Aidan Herrera, and Gracie Peterson provided supplemental research assistance.
CONFLICTS OF INTEREST
The authors have no known conflicts of interest.
HUMAN PARTICIPANT PROTECTION
Institutional review board approval was not required because all data are publicly available, and the study was not human participant research.