Nudging the trendsetters: Increasing second-dose HPV vaccination in Bogota, Colombia

Martínez, Déborah; Díaz, Lina; Maldonado Zambrano, Stanislao Working Paper Nudging the trendsetters: Increasing second-dose HPV vaccination in Bogota, Colombia IDB Working Paper Series, No. IDB-WP-1547 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Martínez, Déborah; Díaz, Lina; Maldonado Zambrano, Stanislao (2023) : Nudging the trendsetters: Increasing second-dose HPV vaccination in Bogota, Colombia, IDB Working Paper Series, No. IDB-WP-1547, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0005331 This Version is available at: https://hdl.handle.net/10419/299466 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc-nd/3.0/igo/legalcode Nudging the Trendsetters: Increasing Second-dose HPV Vaccination in Bogota, Colombia Déborah Martínez Villarreal Lina Díaz Stanislao Maldonado IDB WORKING PAPER SERIES Nº IDB-WP-1547 December 2023 Department of Research and Chief Economist Inter-American Development Bank December 2023 Nudging the Trendsetters: Increasing Second-dose HPV Vaccination in Bogota, Colombia Déborah Martínez Villarreal* Lina Díaz* Stanislao Maldonado** * Inter-American Development Bank ** Universidad del Rosario Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Martínez V., Deborah. Nudging the trendsetters: increasing second-dose HPV vaccination in Bogota, Colombia / Deborah Martinez, Lina Diaz, Stanislao Maldonado. p. cm. — (IDB Working Paper Series ; 1547) Includes biblioraphical references. 1. Papillomaviruses-Vaccination-Colombia. 2. Medical policy-Colombia. 3. Economics- Psychological aspect-Colombia. I. Diaz, Lina M. II. Maldonado, Stanislao. III. Inter- A merican Development Bank. Department of Research and Chief Economist. IV. Title. V. Series. IDB-WP-1547 Copyright © Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 Attribution- NonCommercial-NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license (http://creativecommons.org/licenses/by-nc-nd/3.0/igo/ legalcode) and may be reproduced with attribution to the IDB and for any non-commercial purpose, as provided below. No derivative work is allowed. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the UNCITRAL rules. The use of the IDB's name for any purpose other than for attribution, and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC-IGO license. Following a peer review process, and with previous written consent by the Inter-American Development Bank (IDB), a revised version of this work may also be reproduced in any academic journal, including those indexed by the American Economic Association's EconLit, provided that the IDB is credited and that the author(s) receive no income from the publication. Therefore, the restriction to receive income from such publication shall only extend to the publication's author(s). With regard to such restriction, in case of any inconsistency between the Creative Commons IGO 3.0 Attribution-NonCommercial-NoDerivatives license and these statements, the latter shall prevail. Note that link provided above includes additional terms and conditions of the license. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. http://www.iadb.org 2023 Abstract1 This study investigates the effectiveness of dynamic norm nudges in promoting second-dose HPV vaccinations among trendsetters—parents who initiated the firstdose HPV vaccine for their daughters between 2017-2020. Utilizing administrative data from Bogota’s Secretariat of Health in a field experiment, we measure the impact of various norm nudges, including trending, qualitative, and quantitative dynamic norms, on actual vaccination rates. Contrary to our hypothesis, dynamic norms alone fail to influence second-dose HPV vaccination rates for these trendsetters. However, the study reveals a 5.22 percent increase attributed to injunctive norms, representing a substantial 34 percent boost compared to the control group’s 15.2 percent average. These findings underscore the importance of tailoring nudge strategies to the unique characteristics and preferences of the target population. This research significantly advances our understanding of norm-based interventions’ efficacy in influencing minority behaviors, offering valuable insights for developing targeted and impactful public health strategies. JEL classifications: C93, D91, I10, I12, I15, I18 Keywords: Nudge, Behavioral economics, Health, Vaccination, HPV, Field experiment, Social norms, Trendsetters 1 The authors gratefully acknowledge financial and program support from the American Cancer Society in addition to the feedback provided throughout the project. This work would not have been possible without the support from Claudia Acosta, Patricia Calderón and José Garzón from Secretaría de Salud de Bogotá, Marta Marin from Ministerio de Salud de Colombia; Carlos Castro, Karen Salinas, Mónica Giraldo and Laura Gil from Liga Colombiana contra el Cáncer; Meenu Anand, Erika Rees-Punia and Itziar Belasteuguigoitia from the American Cancer Society; Julian Peña from the Behavioral Government Lab at Universidad del Rosario; and Carlos Scartascini from Inter-American Development Bank. Author affiliations: Martínez Villarreal (corresponding author, [email protected]), Inter- American Development Bank; Díaz, Inter-American Development Bank; Maldonado, Universidad del Rosario. 2 1. Introduction Findings from an experiment on first-dose HPV vaccinations (Martínez Villarreal, 2023) indicate that dynamic norm nudges increase vaccine uptake among girls and adolescents in Bogota, Colombia, where only a minority of the population is vaccinated against HPV. In such a context, first-dose HPV vaccinations are considered a minority behavior. Since 2016, the World Health Organization health guidelines have recommended two doses of the HPV vaccine for coverage against cervical cancer (WHO, 2022),2 but first-dose and second-dose HPV vaccination have been a minority behavior in Bogota since 2016. In this study, we test the hypothesis that dynamic norm nudges increase second-dose HPV vaccinations for trendsetters. We employ Bicchieri and Funcke’s (2018) definition of trendsetters, i.e., as the initiators of norm abandonment. Norm abandonment occurs when societies replace one social norm for another (Andreoni et al., 2021; Bicchieri, 2017). In this context, trendsetters are the group of parents who vaccinated their daughters with the first-dose HPV vaccine between 2017 and 2020. This is the first study that tests dynamic norm nudges’ effect on trendsetters’ behavior. We conducted a field experiment in Bogota, Colombia, to test our hypothesis. The experiment consists of text messages to parents of daughters 9-12 years old who have received the first-dose HPV vaccine but not the second. The experiment studies the effect of five norm nudge treatments. Three treatments are dynamic norms, one treatment is a descriptive norm, and another treatment is an injunctive norm. This experiment has one control group, one experimental control, and one policy control. Administrative records on vaccination from Bogota’s Secretary of Health allow us to measure the effect of norm nudges on actual HPV vaccinations. The literature most closely related to this study is studies of the effect of dynamic norm nudges on minority behaviors (Aldoh et al., 2021; Cheng et al., 2022; Mortensen et al., 2017; Loschelder et al., 2019; Sparkman and Walton, 2017; Milkman et al., 2022). In these studies, dynamic norm nudges inform experimental participants how other people’s behavior has changed, or is changing, over time (Sparkman and Walton, 2017). The literature also tests different variations of dynamic norms, such as framings that either include or exclude elements like the percentage change in adopting the minority behavior. For example, Mortensen et al. (2017) and 2 Recent studies demonstrate that a single dose of the HPV vaccine is sufficient to provide the same protection as a multidose regimen against HPV (WHO, 2022). However, when the experiment was conducted, the two-dose vaccine schedule was still the public health recommendation in Bogota. Starting September 30, 2023, Colombia has started a single-dose schedule countrywide, except for immunocompromised people. 3 Sparkman and Walton (2017) study the framings of dynamic norm nudges, which is also a focus of this study. This approach helps to determine which dynamic norm nudge has the largest effect on changing minority behaviors. We test the effect of three framings based on the seminal work by Mortensen et al. (2017) and Sparkman and Walton (2017). In the first treatment, the trending norm contains information about the percentage change in the adoption of the minority behavior by the reference population: “Since 2016, the number of parents in your town who got the second dose of the HPV vaccine for their daughters has increased by 83 percent.” In the second treatment, the qualitative dynamic norm communicates the trend in HPV vaccinations without mentioning the percentage change: “More and more parents in your area are giving their daughters their second dose of the HPV vaccine.” In the third treatment, the quantitative dynamic norm adds the descriptive norm, which communicates the prevalence of the minority behavior, to the qualitative dynamic norm message: “Eight percent of parents in your area have already gotten the second dose of the HPV vaccine for their daughters, and more and more are doing it.” The messages in each treatment refer to an increase in the trend of second-dose HPV vaccinations. In addition to the research on the effects of dynamic norms on increasing minority behaviors, this paper draws on several other strands of literature. One of these strands is the research on trendsetters’ behaviors (Bicchieri, 2017; Bicchieri and Funcke, 2018). Trendsetters have also been called positive deviants in the health literature (Herington and van de Fliert, 2018). Spreitzer and Sonenshein (2003) define positive deviants as individuals or groups that depart from the norms of a reference group in honorable ways. The research on positive deviants informs the design of interventions for behavioral change (Herington and van de Fliert, 2018; Bicchieri, 2017). For example, Pascale, Sternin and Sternin (2010) decreased children’s malnutrition in Vietnam by applying the strategies of mothers who belonged to the minority and did not have malnourished children in the community.3 Pascale, Sternin and Sternin (2010) refer to mothers who belong to the minority as positive deviants. Unlike studies such as theirs, however, this study focuses on an intervention directed at trendsetters. 3 The strategies applied by these mothers go against locally accepted wisdom. Some of these strategies are feeding children even when they have diarrhea; feeding children several smaller meals rather than one or two large ones; and adding to children’s rice foods that are associated with low socioeconomic status (Pascale, Sternin and Sternin, 2010). 4 Contrary to our hypothesis, the results indicate that dynamic norms do not increase seconddose HPV vaccination rates of trendsetters. This is also the case for the descriptive norm. Only the quantitative dynamic norm has a marginal statistically significant effect compared to the control group at the 90 percent confidence level. This is a surprising result since dynamic norms effectively increase minority behaviors in other contexts (Aldoh et al., 2021; Cheng et al., 2022; Mortensen et al., 2017; Loschelder et al., 2019; Sparkman and Walton, 2017; Milkman et al., 2022). The injunctive norm has a statistically significant increase in second-dose HPV vaccinations of 5.22 percent compared to the control average of 15.2 percent. This difference is equivalent to a 34 percent difference at a 99 percent confidence level. The most effective message for increasing second-dose HPV vaccination is the experimental control. The experimental control is a personalized reminder signed by the Secretariat of Health of the following form: “Hi [Name of the parent]. Get your daughter the second dose of the HPV vaccine: give her all the protection. Secretariat of Health.” Its effect represents a statistically significant increase of 50 percent compared to the control group. 2. Vaccination Context Cervical cancer (CC) is the fourth most common cancer in women worldwide, and it is one of the three most frequent cancers in women younger than 45 (D’Oria et al., 2022). Almost all cervical cancers are caused by the human papillomavirus, or HPV (Walboomers et al., 1999). In addition to CC, HPV is associated with oropharyngeal, anus, genitals, head, and neck cancer. Estimates show that 75 percent of women and men who are sexually active will acquire HPV in their lifetime (Mavundza et al., 2021). Fortunately, the risk of HPV infection and the development of CC can be significantly reduced through a set of HPV vaccines (WHO, 2017). According to the Colombian Ministry of Health, CC is the leading cause of death from cancer in Colombia's women aged 30 to 59. In 2020, new CC cases represented 7.9 percent of all cancer cases, equivalent to 4,742 cases in that year (Córdoba-Sánchez et al., 2022). In this country, the risk of HPV infection can be reduced with two HPV vaccines administered through the Expanded Program on Immunization (PAI). The country’s health system allows citizens to be vaccinated at any vaccination point regardless of their health provider. These vaccines are free for girls between 9 and 17. The Expanded Immunization Program of Colombia's Ministry of Health and Social Protection prioritizes 9-year-old girls’ HPV vaccinations. 5 In 2012, Colombia was one of the leaders in HPV vaccination coverage in Latin America (Córdoba-Sánchez et al., 2022). After the initial introduction of the vaccine in 2012, it became recommended by the health authority and was administered in schools. However, the success of the country’s vaccination program stopped after an outbreak of unknown etiology in the municipality of Carmen de Bolivar. Although safety studies found no association between the HPV vaccine and Carmen de Bolivar’s events, vaccine coverage rates began to decline steadily, reaching their lowest point in 2016 (Córdoba-Sánchez et al., 2022). Coverage levels of HPV vaccination have been recovering over the past years but are still far from the pre-Carmen de Bolivar levels. Figure 1 shows the vaccination rate of the first and second doses of the HPV vaccine for 9-year-old girls in Colombia. The second-dose vaccination rate is substantially lower than the first dose. Figure 1. HPV Vaccination Rates in Colombia since the Introduction of the Vaccine in 2012 Source: Authors’ compilation based on data from the Information System of the Expanded Immunization Program (PAI) of the Ministry of Health and Social Protection of Colombia. Through a large text message communications campaign, we tested the impact of several behavioral economics principles on first and second-dose HPV vaccinations. To provide recommendations to increase HPV vaccination rates, we partnered with the Health Secretariat of Bogota, Colombia, La Liga Colombiana Contra el Cáncer, and the American Cancer Society to run six experiments. 12 Milkman et al. (2022) refer to the dynamic norm as growing norm. The growing norm reads the following way: “More Americans are getting the flu shot than ever in the last decade. Last year, 45 percent of American adults got one.” Previous studies find dynamic norms to be effective at increasing minority behavior despite informing subjects about the minority behavior (Mortensen et al., 2017; Sparkman and Walton, 2017; Milkman et al., 2022). 4. Empirical Analysis 4.1 Sample Characteristics and Balance Across Groups Table A2 in the Appendix shows the descriptive statistics of available variables in the database, and Table A3 shows that treatments are balanced on the observable characteristics of the sample, the t-tests in this table compare each treatment to the control. Out of 84 comparisons, only three differences are statistically significant at the 95 percent confidence level. The differences are equivalent to less than 2 percent of the comparisons. EPS (name of an insurance provider), contributory insurance, uninsured, subsidized insurance, ethnic group, displaced by the armed conflict, Colombian nationality, and low stratum are binary variables. The low stratum is also binary and is constructed by grouping the two lowest neighborhood levels that the government of Bogota uses to characterize low socioeconomic status. 4.2 Regression Model The impact evaluation is based on a standard intention-to-treat analysis (ITT). The main outcome variable is a binary measure of whether a parent’s daughter gets vaccinated with a second-dose HPV vaccine during the text message campaign window or within three months after the campaign ends. The software we use to send text messages does not allow us to identify who receives or reads the messages. Thus, a treatment-on-the-treated (TOT) analysis is not possible. We estimate three models. The first is 𝑦𝑦𝑖𝑖=𝛼𝛼+ 𝛽𝛽1 𝑇𝑇1𝑖𝑖 +𝛽𝛽2 𝑇𝑇2𝑖𝑖 +𝛽𝛽3 𝑇𝑇3−7𝑖𝑖 +𝛾𝛾𝑋𝑋𝑖𝑖+𝜃𝜃𝑠𝑠+ 𝜇𝜇𝑖𝑖 (1) where 𝑦𝑦𝑖𝑖 is the value of a dependent variable that indicates if the daughter of parent i gets vaccinated with the second-dose HPV vaccine (0 = daughter does not get vaccinated, 1 = daughter gets vaccinated). 𝑇𝑇1 is an indicator variable taking the value of 1 when i is assigned to the policy control, and 𝑇𝑇2 is an indicator variable taking the value of 1 when i is assigned to the experimental control. 𝑇𝑇3−7 is an indicator variable taking the value of 1 when i is assigned to a norm nudge. The 13 reference group for this estimation is the control group. 𝑋𝑋 is a vector of controls that includes all observable characteristics available in the administrative database: insurance company, type of insurance, ethnic group, displaced by the armed conflict, Colombian nationality, and a variable identifying whether the family lives in a low-income area (low stratum). 𝜃𝜃𝑠𝑠 is a vector of randomization strata dummy variables (locality*age), and 𝜇𝜇𝑖𝑖 is the error term. The second model is 𝑦𝑦𝑖𝑖=𝛼𝛼+ 𝛽𝛽1 𝑇𝑇1𝑖𝑖 +𝛽𝛽2 𝑇𝑇2𝑖𝑖 +𝛽𝛽3 𝑇𝑇3𝑖𝑖 +𝛽𝛽4 𝑇𝑇4𝑖𝑖 +𝛽𝛽5 𝑇𝑇5−7𝑖𝑖 +𝛾𝛾𝑋𝑋𝑖𝑖+𝜃𝜃𝑠𝑠+ 𝜇𝜇𝑖𝑖 (2) where 𝑦𝑦𝑖𝑖 is the value of a dependent variable that indicates if the daughter of parent i gets vaccinated with the second-dose HPV vaccine (0 = daughter does not get vaccinated, 1 = daughter gets vaccinated). 𝑇𝑇1 is an indicator variable taking the value of 1 when i is assigned to the policy control. 𝑇𝑇2, 𝑇𝑇3, and 𝑇𝑇4 take the value of 1 when i is assigned to the experimental control, descriptive norm, and injunctive norm treatments, respectively. 𝑇𝑇5−7 is an indicator variable taking the value of 1 when i is assigned to a norm nudge. The reference group for this estimation is the control group. 𝑋𝑋 is a vector of controls that includes all observable characteristics available in the administrative database: insurance company, type of insurance, ethnic group, displaced by the armed conflict, Colombian nationality, and a variable identifying whether the family lives in a low-income area (low stratum). 𝜃𝜃𝑠𝑠 is a vector of randomization strata dummy variables (locality*age), and 𝜇𝜇𝑖𝑖 is the error term. The third and final model is 𝑦𝑦𝑖𝑖=𝛼𝛼+ 𝛽𝛽𝑗𝑗 𝑇𝑇 𝑗𝑗+𝛾𝛾𝑋𝑋𝑖𝑖+𝜃𝜃𝑠𝑠+ 𝑣𝑣𝑖𝑖 (3) Similarly to the previous equations, 𝑦𝑦𝑖𝑖 is the value of a dependent variable that indicates if the daughter of parent i gets vaccinated with the second-dose HPV vaccine (0 = daughter does not get vaccinated, 1 = daughter gets vaccinated), and 𝑇𝑇 𝑗𝑗 are indicator variables for i’s treatment assignments j=1-7. In this case, the coefficients 𝛽𝛽j estimate the average treatment effects of treatment j compared to the reference control group. 𝑋𝑋 is the same vector of controls in equation 1 that includes all observable characteristics available in the administrative database, 𝜃𝜃𝑠𝑠 is a vector of randomization strata dummy variables (locality*age), and 𝑣𝑣𝑖𝑖 is the error term. 14 5. Results Table 1 presents the results of equations (1)-(3) that show the effect of norm nudges on increasing the second-dose HPV vaccinations for trendsetters.6 Columns (1), (3), and (5) display the OLS estimates without controls, and columns (2), (4), and (6) show the OLS estimates controlling for relevant covariates. The control variables include insurance provider, type of insurance, ethnic group, displaced by armed forces, Colombian nationality, and low socioeconomic stratum. All the controls are dummy variables. The average vaccination rate in the control group during the experimental period was 15.2 percent. Column one of Table 1 shows that the average second-dose HPV vaccination rate of girls whose parents received a norm nudge treatment is 2.8 percent higher than the control group’s average. This result is statistically significant at a 95 percent confidence level. Column two shows that this result is robust when we control for covariates. This estimate is equivalent to an 18.4 percent difference between norm nudges and the control group. This result does not support H1, which states that norm nudges do not increase second-dose HPV vaccination rates for trendsetters. However, norm nudges include descriptive, injunctive, and dynamic norms. The subsequent analysis will allow us to identify what elements of norm nudges impact this population. 6 This study focuses on the pure effect of social norms on trendsetters, which requires restricting the original sample, depicted in Figure 3A in the Appendix. All the results presented here exclude the sample of parents of daughters 13- 17 years old to focus on the trendsetters and exclude the parents who received a follow-up message after each weekly message that served as a planning tool. Columns one and two in Table A5 in the Appendix show results for the whole sample. Columns two and three show the effect of the planning tool on the whole sample, while columns four and five show the effect of the planning tool on the trendsetters. 15 Table 1. There is no evidence that dynamic norms effectively increase second-dose HPV vaccinations for trendsetters (1) OLS (2) OLS (3) OLS (4) OLS (5) OLS (6) OLS VARIABLES Applied vaccine Applied vaccine Applied vaccine Applied vaccine Applied vaccine Applied vaccine Policy control -0.0120 -0.0109 -0.0120 -0.0109 -0.0120 -0.0109 (0.0198) (0.0198) (0.0198) (0.0198) (0.0198) (0.0198) Experimental control 0.0749*** 0.0760*** 0.0749*** 0.0760*** 0.0749*** 0.0760*** (0.0198) (0.0198) (0.0198) (0.0198) (0.0198) (0.0198) Norm nudges 0.0280** 0.0283** - - - - (0.0135) (0.0135) Positive descriptive norm 0.0238 0.0244 0.0238 0.0244 (0.0198) (0.0197) (0.0198) (0.0197) Injunctive norm - - 0.0509*** 0.0522*** 0.0509*** 0.0522*** (0.0198) (0.0197) (0.0198) (0.0197) Dynamic norms - - 0.0218 0.0217 - - (0.0148) (0.0147) Quantitative dynamic norm - - - - 0.0383* 0.0373* (0.0198) (0.0198) Qualitative dynamic norm - - - - 0.0105 0.0120 (0.0198) (0.0198) Trending norm - - - - 0.0165 0.0157 (0.0198) (0.0197) Constant 0.152*** 0.111 0.152*** 0.110 0.152*** 0.110 (0.0114) (0.0769) (0.0114) (0.0769) (0.0114) (0.0769) Observations 4,956 4,956 4,956 4,956 4,956 4,956 R-squared 0.004 0.014 0.004 0.015 0.005 0.015 Controls NO YES NO YES NO YES Note: The control variables include insurance provider, type of insurance, ethnic group, displaced by armed forces, Colombian nationality, and low stratum. All the controls are dummy variables. The unreported coefficient values for the Probit model show the same coefficients as the OLS estimation. Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Column three in Table 1 shows the impact of descriptive and injunctive norms on seconddose HPV vaccinations for trendsetters. Column three shows that the average second-dose HPV vaccination rate of girls whose parents received the descriptive norms treatment is 2.4 percent higher than the control group’s average. This result is not statistically significant and remains the same after controlling for covariates. Thus, the result does not support H2, which states that descriptive norms do not increase second-dose HPV vaccination rates for trendsetters. This result does not show evidence of the expected “boomerang effect” of descriptive norms (Cialdini, 1990; Bicchieri and Xiao, 2009; Bicchieri and Dimant, 2022; Kuang et al., 2020; 16 Schultz et al., 2007). The backfire effect might still be present in the population that corrected overstated beliefs of the descriptive norm, as in Schultz et al. (2007). However, our setting limits the strength of our conclusion since beliefs on current vaccination rates held by the participants are not elicited, impeding analysis of heterogenous effects of descriptive norms on HPV vaccinations. Regarding the effect of injunctive norms, the average second-dose HPV vaccination rate of trendsetters in the injunctive norm treatment is 5.1 percent higher than the control group’s average. The result is robust to the specification, including covariates, and statistically significant at the 99 percent confidence level. This is a 33.55 percent difference from the control group, and the result reaches statistical significance after the Bonferroni correction for multiple comparisons. This finding supports H3, which states that injunctive norms increase trendsetters’ seconddose HPV vaccinations. Moreover, the result supports the literature that finds injunctive norms effective at increasing a minority behavior (Allcott, 2011; Bonan et al., 2020; Jachimowicz et al., 2018; Ryo et al., 2021; Schultz et al., 2007). A potential mechanism, as suggested by Bicchieri and Dimant (2022) and Hauser (2018), is that injunctive norms address the underlying motivations of trendsetters. Column three in Table 1 also shows the effects of dynamic norms loosely defined. The estimation shows a marginal coefficient of 2.2 percent, which is not statistically significant compared to the control group. Albeit positive, this result does not support H4, which states that dynamic norms increase second-dose HPV vaccinations. Furthermore, this goes against recent studies which find that dynamic norms effectively increase minority behaviors (Aldon et al., 2021; Cheng et al., 2022; Mortensen et al., 2017; Loschelder et al., 2019; Sparkman and Walton, 2017; Milkman et al., 2022). The results from column five in Table 1 disentangle the effect of each separate dynamic norm treatment on second-dose HPV vaccinations for trendsetters. The marginal coefficients for the trending, qualitative, and quantitative dynamic norms show a positive sign. However, none are statistically significant at a 95 percent confidence level. These results are relevant for H5 and H6, which state that trending and qualitative norms increase second-dose HPV vaccinations. With the 17 caveat that this effect might be due to a lack of power,7 these results do not support hypotheses H5 and H6. The quantitative dynamic norm shows a marginal effect of 3.8 percent statistically significant at the 90 percent confidence level compared to the control group. The coefficient remains the same when we control for covariates. Although there is a large effect equivalent to a 25 percent difference in second-dose HPV vaccinations compared to the control group, the result does not show significant effects with a Bonferroni correction for multiple comparisons. A heterogeneous effects estimation shows that quantitative dynamic norms have a negative effect on the population with subsidized insurance. The marginal coefficient is -11.4 percent, statistically significant at the 95 percent confidence level (See Table A4 in the Appendix). This result is consistent with studies that find boomerang effects of norm nudges in sub-populations (Cialdini, 1990; Bicchieri and Xiao, 2009; Bicchieri and Dimant, 2022; Kuang et al., 2020; Schultz et al., 2007). The result does not show significant effects with a Bonferroni correction for multiple comparisons. Lastly, the results show that the most effective nudge of the intervention to increase seconddose HPV vaccination for trendsetters is the experimental control for increasing second-dose HPV vaccination. This treatment shows a marginal increase of 7.5 percent compared to the control group. The result is robust to including covariates and statistically significant at a 99 percent confidence level. This difference is equivalent to an approximately 50 percent increase compared to the control group’s average and reaches statistical significance after the Bonferroni correction for multiple comparisons. The experimental control is a non-norm nudge containing two elements: the recipient's name, and the sender’s information, in this case, the Secretariat of Health. The content of the experimental control is the following, “Get your daughter the second dose of the HPV vaccine: give her all the protection.” Thus, it can be considered a reminder. This result supports the vast literature on the reminders’ role in increasing vaccination (Briss et al., 2000; Jacobson et al., 2005; Busso, 2015; Busso, 2017; Stockwell, 2012; Szilagyi, 2013). 7 We conducted a power analysis to estimate the number of parents assigned to each treatment. However, we use only a subset of that sample to analyze the trendsetters specifically. See Table A5 in the Appendix for regressions over the entire sample in columns one and two. 18 This intervention based on SMS norm nudges is highly cost-effective in increasing seconddose HPV vaccinations of trendsetters. The cost per additional girl vaccinated is estimated at USD $0.61. This cost considers the cost of all messages bought for the intervention and the marginal vaccination rate per treatment. However, a simple reminder to the same population would cost USD $0.24. This is a cost reduction of 61 percent. 6. Conclusion In this study, we run a field experiment through a text message campaign to increase the minority behavior of second-dose HPV vaccinations for trendsetters in Bogota, Colombia. The target population is parents with daughters between 9 and 12 who already have the first dose of the HPV vaccine. Because this population of parents has acted against social norms in the past, we refer to them as the HPV vaccination trendsetters. The vaccination rate of the first-dose HPV vaccine at the time of the experiment is approximately 30 percent, and the second-dose HPV vaccination rate is 9 percent. We test the effect of five norm nudges, one experimental control, one policy control, and one control group on second-dose HPV vaccinations. The main findings are the following. First, we find a lack of statistically significant evidence of the effect of dynamic norms in increasing second-dose HPV vaccinations for trendsetters. Second, the results show a positive statistically significant effect of injunctive norms on second-dose HPV vaccinations for trendsetters. The difference in the mean of vaccinations for the injunctive norm treatment group and the control group was sizable at 33 percent. Third, the most effective nudge at increasing second-dose HPV vaccination is the experimental control, i.e., a personalized reminder signed by the Secretariat of Health. The experimental control shows a statistically significant increase of 7.5 percent, equivalent to an approximately 50 percent increase compared to the control group’s second-dose HPV vaccination average. The results in this study do not support other studies that find dynamic norms effective at increasing minority behaviors. However, the results support the literature that finds the effect of norm nudges depends on the underlying preferences of the target population.8 The differences in 8 For example, Castro and Scartascini (2015) find that a descriptive nudge does not affect the average population’s behavior; however, it increases tax compliance on previously non-compliers but decreases compliance on previously compliant taxpayers. Unlike this study, norm nudge experiments typically find differential effects of norm nudges by 19 the effect of norm nudges containing the same social norm components on first-dose and seconddose HPV vaccinations illustrate the importance of understanding the underlying characteristics of the population to develop effective nudge interventions. This study’s results allow us to reflect on trendsetters’ underlying preferences. Trendsetters who have gone against social norms may have preferences less influenced by others' behaviors, resulting in the ineffectiveness of dynamic norm nudges. Additionally, these individuals may adhere to a moral rule for behavior that favors their daughters’ health, explaining the significant impact of injunctive norms on second-dose HPV vaccination. Furthermore, a simple reminder, i.e., the experimental control, is highly effective for trendsetters who may have forgotten to administer the second HPV vaccine six months after the first dose. The implications of this study's findings are relevant for developing cost-effective public health nudge interventions. The estimated cost per additional vaccinated girl in this study was approximately USD 0.61. However, had the simple reminder been implemented across all groups, the cost would have decreased to USD 0.24 per additional vaccinated girl. This estimation highlights the importance of experiments that find effective nudges for the target population, as they can help keep the costs low when implemented at scale. When vaccination completion is the problem, important public health goals can be achieved by norm nudges or reminders. Furthermore, given the link between HPV vaccination and reduced risk of cervical cancer, norm nudge interventions and reminders may ultimately lower public resources allocated to cancerrelated medical care. analyzing heterogeneous effects (Allcott, 2011; Beshears et al., 2015; Bicchieri and Dimant, 2022; Castro and Scartascini, 2015; Fellner et al., 2013; Ferraro et al., 2011; Kantorowicz‐Reznichenko, 2021; Peth, 2018; Richter et al., 2018; Schultz et al., 2007). 20 References Aldoh, A., P. Sparks, and P.R. Harris. 2021. “Dynamic Norms and Food Choice: Reflections on a Failure of Minority Norm Information to Influence Motivation to Reduce Meat Consumption.” Sustainability 13(15): 8315. https://doi.org/10.3390/su13158315. Allcott, H. 2011. “Social Norms and Energy Conservation.” Journal of Public Economics 95(9– 10): 1082–95. https://doi.org/10.1016/j.jpubeco.2011.03.003. Andreoni, J., N. Nikiforakis, and S. Siegenthaler. 2021. “Predicting Social Tipping and Norm Change in Controlled Experiments.” Proceedings of the National Academy of Sciences 118(16): e2014893118. https://doi.org/10.1073/pnas.2014893118. Beshears, J. et al. 2015. “The Effect of Providing Peer Information on Retirement Savings Decisions: Peer Information and Retirement Savings Decisions.” Journal of Finance 70(3): 1161–1201. https://doi.org/10.1111/jofi.12258. Bhanot, S.P. 2021. “Isolating the Effect of Injunctive Norms on Conservation Behavior: New Evidence from a Field Experiment in California.” Organizational Behavior and Human Decision Processes 163(March): 30–42. https://doi.org/10.1016/j.obhdp.2018.11.002. Bicchieri, C. 2000. “Words and Deeds: A Focus Theory of Norms.” In: J. Nida-Rümelin and Wolfgang Spohn, editors. Rationality, Rules, and Structure. Dordrecht, The Netherlands: Springer Netherlands. https://doi.org/10.1007/978-94-015-9616-9_10. Bicchieri, C. 2006.. The Grammar of Society: The Nature and Dynamics of Social Norms. Cambridge, United Kingdom: Cambridge University Press. Bicchieri, C. 2017. Norms in the Wild: How to Diagnose, Measure, and Change Social Norms. Oxford, United Kingdom: Oxford University Press. https://doi.org/10.1093/acprof:oso/9780190622046.001.0001. Bicchieri, C, and E. Dimant. 2022. “Nudging with Care: The Risks and Benefits of Social Information.” Public Choice 191(3–4): 443–64. https://doi.org/10.1007/s11127-019- 00684-6. Bicchieri, C., and E. Xiao. 2009. “Do the Right Thing: But Only If Others Do So.” Journal of Behavioral Decision Making 22(2): 191-208. http://dx.doi.org/10.1002/bdm.621. Bicchieri, C., and A. Funcke. 2018. “Norm Change: Trendsetters and Social Structure.” Social Research: An International Quarterly 85(1): 1–21. https://doi.org/10.1353/sor.2018.0002. 21 Bonan, J. et al. 2020. “The Interaction of Descriptive and Injunctive Social Norms in Promoting Energy Conservation.” Nature Energy 5(11): 900–909. https://doi.org/10.1038/s41560- 020-00719-z. Bursztyn, L., A.L. González, and D. Yanagizawa-Drott. 2020. “Misperceived Social Norms: Women Working Outside the Home in Saudi Arabia.” American Economic Review 110(10): 2997–3029. https://doi.org/10.1257/aer.20180975. Castro, E., and C. Scartascini. 2015. “Tax Compliance and Enforcement in the Pampas: Evidence from a Field Experiment.” Journal of Economic Behavior & Organization 116: 65–82. https://doi.org/10.1016/j.jebo.2015.04.002. Cheng, L. et al. 2022). “Join Us: Dynamic Norms Encourage Women to Pursue STEM.” Current Psychology 41(9): 5967–77. https://doi.org/10.1007/s12144-020-01105-4. Cialdini, R.B. 2003. “Crafting Normative Messages to Protect the Environment.” Current Directions in Psychological Science 12(4): 105–9. https://doi.org/10.1111/1467- 8721.01242. Cialdini, R.B., RR. Reno, and C.A. Kallgren. 1990. “A Focus Theory of Normative Conduct: Recycling the Concept of Norms to Reduce Littering in Public Places.” Journal of Personality and Social Psychology 58(6): 1015–1026. https://doi.org/10.1037/0022- 3514.58.6.1015 Constantino, S.M. et al. 2022. “Scaling Up Change: A Critical Review and Practical Guide to Harnessing Social Norms for Climate Action.” Psychological Science in the Public Interest 2 (2): 50–97. https://doi.org/10.1177/15291006221105279. Córdoba-Sánchez, V. et al. 2022. “HPV-Vaccine Hesitancy in Colombia: A Mixed-Methods Study.” Vaccines 10(8): 1187. Dimant, E., G.A. van Kleef, and S. Shalvi. 2020. “Requiem for a Nudge: Framing Effects in Nudging Honesty.” Journal of Economic Behavior & Organization 172: 247–66. https://doi.org/10.1016/j.jebo.2020.02.015. D’Oria, O. et al. 2022. “New Advances in Cervical Cancer: From Bench to Bedside.” International Journal of Environmental Research and Public Health 19(12): 7094. https://doi.org/10.3390/ijerph19127094. 28 Table A4. Quantitative Norms Result in a Negative Heterogeneous Effect on the Subsidized Population (1) Colombian (2) Displaced (3) Ethnic (4) Contributory (5) Subsidized (6) Stratum low (7) Uninsured VARIABLES Applied vaccine Applied vaccine Applied vaccine Applied vaccine Applied vaccine Applied vaccine Applied vaccine Policy control -0.0264 -0.163 -0.154 -0.0530 0.0609 0.0638 -0.0036 (0.111) (0.156) (0.304) (0.0468) (0.0544) (0.0429) (0.112) Experimental control -0.0209 0.0138 -0.248 -0.0425 0.0490 -0.0082 -0.0419 (0.115) (0.161) (0.262) (0.0459) (0.0518) (0.0439) (0.108) Positive descriptive -0.0410 -0.293 0.114 -0.0099 0.0146 0.0141 0.0007 (0.129) (0.197) (0.239) (0.0465) (0.0537) (0.0432) (0.113) Injunctive norm 0.0706 -0.189 -0.142 -0.0619 0.0631 0.0393 0.0519 (0.115) (0.168) (0.305) (0.0463) (0.0531) (0.0436) (0.112) Quantitative dynamic norm -0.175 0.145 -0.194 0.0433 -0.114** -0.0368 0.152 (0.123) (0.286) (0.407) (0.0486) (0.0551) (0.0430) (0.127) Qualitative dynamic norm 0.0803 -0.148 0.191 -0.0132 0.0176 0.0009 -0.0263 (0.121) (0.175) (0.262) (0.0466) (0.0528) (0.0431) (0.110) Trending norm 0.0275 -0.158 -0.153 -0.0357 0.0481 0.0433 -0.0128 (0.121) (0.167) (0.306) (0.0473) (0.0539) (0.0428) (0.118) Observations 4,956 4,956 4,956 4,956 4,956 4,956 4,956 Controls YES YES YES YES YES YES YES Note: Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 29 Table A5. Regression Results of the Full Sample, per Pre-registered Regressions (1) (2) (3) (4) (5) (6) Target Population 9-17 9-17 9-17 9-12 9-12 9-12 VARIABLES Vaccination rate Vaccination rate Vaccination rate Vaccination rate Vaccination rate Vaccination rate 1. Policy control -0.0170 -0.0100 -0.00935 -0.0288 -0.0191 -0.0109 (0.0126) (0.0127) (0.0129) (0.0193) (0.0195) (0.0198) 2. Experimental control 0.0466*** 0.0431*** 0.0465*** 0.0538*** 0.0490*** 0.0760*** (0.00993) (0.00997) (0.0129) (0.0152) (0.0153) (0.0198) 3. Positive descriptive 0.0102 0.00675 0.00930 0.0226 0.0179 0.0244 (0.00992) (0.00996) (0.0129) (0.0152) (0.0153) (0.0197) 4. Injunctive norm 0.0345*** 0.0311*** 0.0445*** 0.0298* 0.0250 0.0522*** (0.00992) (0.00996) (0.0129) (0.0152) (0.0153) (0.0197) 5. Quant dynamic norm 0.00707 0.00361 0.0126 0.0283* 0.0236 0.0373* (0.00992) (0.00996) (0.0129) (0.0152) (0.0153) (0.0198) 6. Qual dynamic norm 0.00266 -0.000796 -0.0113 0.0191 0.0143 0.0120 (0.00992) (0.00996) (0.0129) (0.0153) (0.0153) (0.0198) 7. Trending norm 0.0175* 0.0141 0.00186 0.0317** 0.0270* 0.0157 (0.00992) (0.00996) (0.0129) (0.0152) (0.0153) (0.0197) Planning tool 0.0208*** - 0.0291*** - (0.00567) (0.00871) Constant 0.0990** 0.0920* 0.0709 0.187*** 0.176*** 0.110 (0.0471) (0.0471) (0.0599) (0.0605) (0.0606) (0.0769) Observations 16,398 16,398 9,235 8,807 8,807 4,956 R-squared 0.030 0.031 0.031 0.011 0.012 0.015 Control variables YES YES YES YES YES YES Link group included YES YES NO YES YES NO Link control NO YES - NO YES NO Control mean .1317 .1317 .1238 .1695 .1695 .1520 Note: The control variables are the observable characteristics of the sample, including uninsured, subsidized insurance, contributory insurance, ethnic group, whether the armed conflict displaced the girl, and whether the girl is Colombian. These variables get a value of 1 if it applies to the girl’s record. Low stratum is also a binary variable and was constructed by grouping the two lowest neighborhood levels used by Bogota to characterize low socioeconomic status. See Figure 3A for the graphical representation of the experimental design containing the full sample of 9-17-year-old girls. 30 Figure 3A. Experimental Groups of Full Samples of 9-17-year-old Girls Note: The stratified randomization allows us to study the pure effect of the norm nudges on trendsetters. To this end, we focus on the subsample parents of 9-12-year-old girls who were vaccinated during a minority norm scenario and who only received the norm nudge and not the planning tool. Thus, these parents received 8 messages, one per week, during the field experiment.