This paper studies the impacts of a vaccine program on the vaccine take-up of eligible and ineligible individuals. Even though vaccines are generally considered one of the most cost-effective health interventions and are estimated to prevent millions of deaths each year (Li et al., 2021, Olshansky and Hayflick, 2017, WHO, 2021), widespread hesitancy regarding vaccinations exists and has been deemed one of the greatest threats to future public health (WHO, 2019). To combat vaccine hesitancy, policymakers must understand the social and behavioral forces driving vaccine reluctance. Previous research has successfully identified several factors that impact vaccine take-up (e.g., cost-benefit (Brilli et al., 2020); socioeconomic background (Mullahy, 1999, Schmitz and Wübker, 2011, Shahjahan et al., 2022); information (Anderberg et al., 2011); prevalence of disease (Philipson, 1996, Oster, 2018); insurance and prices of vaccines (Chang, 2016, Churchill, 2021b, Lipton and Decker, 2015); recommendations (Lawler, 2017); and mandates (Carpenter and Lawler, 2019, Churchill, 2021a)). There is also increasing evidence that health interventions more generally affect not only the targeted individuals but also their peer groups (e.g., Breining, 2014, Al-Janabi et al., 2016, Fletcher and Marksteiner, 2017, Alsan, 2017, Black et al., 2020, Fadlon and Nielsen, 2019 and Daysal et al. (2020)). Understanding if and how health interventions may spill over and affect others than the intended individuals is crucial knowledge for the designers of health care policy in their pursuit of herd immunity.

We study the introduction of national vaccination programs that provided the Human Papillomavirus (HPV) vaccine free of charge to Danish adolescents. In 2008, a national HPV vaccination program targeting adolescent girls was introduced.1 While the HPV vaccine was available to adolescents prior to the introduction of the vaccination program, girls who were eligible for the program could get the vaccine for free - as opposed to paying approximately 560 USD for the three recommended doses.2 In 2020, a similar program was implemented for boys which gives us the opportunity to study the program impact in two separate settings.3 The introduction of the vaccination programs allows us to investigate the extent to which the programs affected not only the targeted individuals (direct effects), but also individuals not directly targeted by the programs (spillover effects) and to explore the mechanisms underlying the decision to get vaccinated.

The decision to vaccinate is of particular interest, since vaccinations play a large role in preventive care. This is especially true for children and the elderly, although the COVID-19 pandemic has demonstrated that vaccination can be of great importance for all population groups. While many individuals choose to get themselves or their children vaccinated, a non-negligible part of the population typically are not vaccinated.4 Failure to vaccinate poses a challenge for policymakers in public health since the main objective of childhood vaccination programs is the eradication of diseases or at least the achievement of herd immunity. Our setting allows us to study the effects of two large-scale introductions of national vaccination programs on vaccine take-up. It also allows us to overcome some of the methodological challenges inherent in studies of take-up and spillovers as the programs generate plausibly exogenous variation in eligibility (and take-up) of the targeted adolescents.

We leverage population-level administrative data from Denmark and link several registers to obtain a data set with rich information on socioeconomic background, detailed vaccination history and family linkages. Our study takes advantage of the fact that eligibility for the HPV vaccination programs was determined by date of birth. In October 2008, girls born January 1, 1993, or later were eligible for the program, whereas girls born earlier were not. We exploit the eligibility criterion to estimate the effects of the program in a regression discontinuity framework. The first step in our analysis is to assess the direct effect of the program on the targeted girls. The program increased the HPV vaccine take-up of targeted girls by 53.2 percentage points. Heterogeneity analysis reveals that the impact is greatest among low-income and low-education families, implying that the program reduced socioeconomic inequality in vaccine take-up for the targeted girls.

In the next step of our main analysis, we estimate the spillover effect of the program on the older sisters of the targeted girls. The older sisters are ineligible for the program. We find a robust and precisely estimated spillover effect of 4.5 percentage points, which corresponds to a take-up increase of about 30 percent for the group of older sisters. In comparison, we do not find spillovers to ineligible older brothers. However, in 2020 the HPV vaccination program was implemented for boys. We conduct a similar analysis for boys and their older ineligible brothers and find highly comparable results, namely that the program had a direct effect of 36.0 percentage points on the targeted cohort of boys and a 3.5 percentage points spillover to older brothers. Across both programs, we document a socioeconomic gradient in the spillover effect as we find larger spillover effects in high-SES families. This ultimately implies that while the implementation of the vaccine programs reduced the SES-gap in vaccine take-up among the targeted children, the programs led to increased socioeconomic inequality in vaccine take-up among the older ineligible siblings.

The rich register data allows us to study not only spillovers to siblings but also to extended family networks. While we find spillover effects on older ineligible same-sex siblings, we do not find evidence of spillover effects on ineligible cousins. The data also enables us to examine whether eligibility for HPV vaccine programs spills over to the take-up of other preventive health care services. We find a 1.3 percentage points cross-vaccine spillover to take-up of the MMR booster vaccine for the targeted girls (booster recommended at age 12). We do not detect any sizable cross-vaccine spillovers for boys (booster recommended at age 4). Moreover, spillovers to take-up of cervical cancer screening in adolescence among the targeted girls are close to zero and statistically insignificant.

From the perspective of the household, eligibility for the vaccination program implied a reduced cost of getting the eligible child vaccinated and the receipt of information material. We exploit the richness of the administrative data to conduct a range of analyses to explore the heterogeneity and timing of both the direct effects and the spillover effects. Focusing on the results for girls, estimated spillover effects are generally larger for high-SES families, closely spaced siblings and sibling pairs where the older sister still lives at home. Whether the mother has a post-secondary health education is not associated with differences in the estimated spillover effects. However, baseline HPV vaccine take-up is much higher for older sisters with a mother with a health education emphasizing that having a health professional in the family is associated with higher HPV vaccine take-up among the older sisters in general. When we examine the timing of HPV vaccination, we find that both eligible girls and their older sisters respond swiftly to eligibility. Also, 16 percent of siblings pairs are vaccinated in the same week. This suggests that information is an important mechanism and that parents may to some extent be bundling the physician visits for convenience. Finally, we conduct a set of analyses where we estimate spillovers to eligible siblings, i.e. siblings who are themselves eligible for the free vaccine, and find positive spillover effects on younger same-sex siblings. All in all, these analyses suggest that the monetary cost of the vaccine is not the main driver of the estimated spillover effects.

Our unique combination of high-quality data and several reforms allows us to contribute significantly to a growing literature on within-family spillovers in health behavior. Crucially, we are able to link eligible and ineligible individuals to other family members using administrative data, which enables us to study not only vaccine take-up of the focal individual but also vaccine take-up of siblings and cousins in the relevant age range. Further, we are able to examine the existence of cross-vaccine spillovers as well as spillovers to cervical cancer screening. The ability to link several administrative registers also implies that we can investigate heterogeneity of effects across a number of interesting dimensions including whether the mother has a health education and characteristics of the sibling pairs. There is a smaller number of studies that estimate peer effects or family spillovers related to vaccine take-up. Rao et al. (2011) investigate peer effects in influenza vaccination among students at a U.S. university. They use the random assignment of students to residence halls as an exogenous variation, as some residence halls have a flu clinic while others do not. They find that the share of your friends in treated houses has a positive effect on your vaccine take-up. Using data from Japan, Itaya et al. (2018) find evidence of positive peer effects in influenza vaccine take-up; both based on geographical peer groups and households. They take advantage of an age-dependent eligibility threshold (age 65) to generate exogenous variation in take-up. The study closest related to ours is Bouckaert et al. (2020), who study spillovers in vaccine take-up within families of a population-based influenza vaccination program in the Netherlands. Like (Itaya et al., 2018), they exploit an age-dependent eligibility threshold (age 65) to generate exogenous variation in take-up. They estimate both the direct effect on the targeted individuals and spillover effects on other family members using a regression discontinuity design. The authors find that the program increased vaccine take-up among the targeted individuals and their younger ineligible partners by 10 percentage points. However, they do not find a spillover to older eligible partners and they find a negative spillover to adult children. The eligible individuals are notified about their eligibility and vaccine benefits in a personal invitation. The authors argue that spillover effect sizes are consistent with the partners and children learning about the program target group, the risks of getting influenza, and the costs and benefits of vaccination. Hirani (2021) investigates the spillover effects of a reminder vaccination policy in Denmark on siblings and cousins and finds evidence of a negative effect on younger siblings’ vaccine take-up. In a related study, Hirani and Wüst (2023) find that vaccine reminders do not cause any major spillovers to other health investments. Alsan (2017) finds that the older sisters of vaccine-eligible Turkish children have more favorable schooling outcomes. Also, Sato and Takasaki (2019) estimate peer effects in vaccine take-up among women in Nigeria using randomized conditional cash transfers that increased vaccine take-up. They find that the female friends of women who received the conditional cash transfers were more likely to take up the vaccine. In comparison, Freedman et al. (2022) find no evidence of an effect of COVID-19 vaccination on COVID-19 incidence of schoolmates. In addition to the work on vaccine behavior, our paper also relate to is a growing number of studies of within-family spillovers of health shocks at different ages (e.g., at birth (Daysal et al., 2020, Black et al., 2020, Breining, 2014); early childhood (Alsan, 2017); middle age (Fadlon and Nielsen, 2019, Fletcher and Marksteiner, 2017); and late adulthood (Costa-Font et al., 2021)).

We also contribute to a literature that studies the spillovers of public programs and implications for health equity. Economic research is increasingly focusing not only on the direct effects of a given program, but also derived effects such as spillover or peer effects (e.g., spillovers of medical treatment (Daysal et al., 2020); peer effects in parental leave take-up (Dahl et al., 2014); and spillovers of retirement to spousal health (Zang, 2020)). The availability of information regarding the potential effects of a program is crucial for policymakers to be able to weigh the costs and benefits of the program in question (Fletcher and Marksteiner, 2017). Along the same lines, understanding how and why different public programs may affect other individuals than those originally targeted by the program provides both researchers and policymakers with valuable information concerning the behavior and interaction of individual decision makers (Figlio et al., 2023). Inequalities in health remain substantial in many countries (see, e.g., Dahl et al. (2020), for a recent study of inequality in life expectancy and mortality inequalities in Denmark and the US), and spillover effects may play an important role, as they possibly have different distributional consequences than the direct effects of a given intervention.

Further, we contribute to the existing economic research on vaccine take-up and compliance. This literature has previously tended to focus on either childhood vaccinations or influenza vaccinations while COVID-19 vaccine take-up is the focus of several recent studies (e.g., Karaivanov et al. (2021)). Only a few recent exceptions have considered the take-up of the HPV vaccine. Carpenter and Lawler (2019) investigate the effects of middle school vaccination requirements and document the cross-vaccine spillover effects on HPV vaccine take-up for U.S. adolescents. Using Swedish administrative data, Chen et al. (2019) find that having a doctor or a nurse in the family increases HPV vaccine take-up along with a wide range of other favorable health outcomes. Finally, HPV vaccine take-up has recently been shown to be positively affected by the intensity of opt-out requirements (Churchill, 2021a) and the Affordable Care Act (Churchill, 2021b, Lipton and Decker, 2015). The richness of our data and the magnitude of the reaction to the implementation of the vaccination programs allow us to conduct detailed analysis and characterizations of the group of individuals who get vaccinated due to the introduction of the vaccination programs. While we draw our conclusions based on programs promoting HPV vaccination in specific contexts, our findings provide valuable insights into which demographic groups are prone to adhere to a vaccination program and the potential impact on health disparities as well as potential drivers of these effects. This understanding is crucial for identifying specific populations that could benefit from targeted interventions to enhance vaccine compliance.

The outline of the paper is as follows: In Section 2, we describe the institutional framework surrounding the vaccination program in Denmark, the HPV vaccine, and the epidemiology of the HPV infection. In Section 3, we outline our empirical strategy. In Section 4, we present the data used in the analyses and validity tests. In Section 5, we present our findings and in Section 6, we discuss potential mechanisms driving the spillover effects. Finally, in Section 7, we conclude.