Tackling barriers to collective action for effective vaccination campaigns: rabies in rural Africa as an example

The paper addresses how to overcome collective action barriers to achieve effective mass vaccination in resource-limited settings, using canine rabies vaccination in rural Tanzania as a case study. While vaccination confers individual protection, eliminating infectious diseases requires community-wide participation to reach herd immunity thresholds (approximately 70% for rabies). Despite evidence that mass dog vaccination reduces human rabies risk and demand for post-exposure prophylaxis, campaigns often fail to achieve sufficient coverage due to demand-side (awareness, salience, competing priorities) and supply-side (accessibility, logistics) constraints. The study examines whether communication strategies (text messages and leveraging community leaders) and operational adjustments (more vaccination points, longer operating hours, earlier advertising) can increase participation by making rabies risk salient and lowering participation costs, thereby solving the collective action problem underlying public-good provision.

Study setting and design: The study was conducted in Morogoro Rural District, Tanzania. From 93 villages with confirmed mobile network coverage, 56 were randomly selected and randomly assigned to one of four groups: (1) Routine advertising only; (2) Community leaders only (routine advertising plus promotion by religious/Maasai leaders); (3) Text messaging only (routine advertising plus SMS); (4) All advertising types (both community leaders and SMS in addition to routine advertising). District dog vaccination campaigns occurred January 20–21, 2018; sampled villages were vaccinated on January 20. A total of 7,210 dogs were vaccinated district-wide (3,256 in sampled villages). Interventions: - Text messaging: Phone numbers were manually collected and uploaded to a mass-SMS platform (Rasello). Message content, timing, and tone were developed via workshops and focus groups. The SMS emphasized a recent local rabies death, the protective benefit of vaccinating dogs, and provided exact time (8 am–4 pm) and locations, stressing that vaccination was free. Messages were sent 5, 3, and 1 day before campaigns at ~4 pm. - Community leaders: District veterinary officers identified religious (Muslim and Christian) and Maasai leaders. Leaders received letters ≥1 week before campaigns inviting them to inform their communities, with a sample message but freedom regarding timing, frequency, and content; reminders were sent by SMS. Leaders were asked to clearly convey date/time and free vaccination. - Routine advertising: Posters at schools, markets, town halls 1 week prior; loudspeaker announcements by local leaders. Operational changes (2018 comparative analysis): To reduce participation costs and improve access, campaigns implemented more vaccination points (at sub-village level where possible) and kept vaccination points open all day, with earlier and extended advertising to facilitate preparation (e.g., dog restraint). A central-point vaccination strategy was used (up to 3 points in larger villages; total 88 points across sampled villages). Vaccinated dogs received colored collars. Data collection: - Post-vaccination transects: Vaccinators conducted evening transects to estimate village-level coverage (collared vs. total dogs observed). - Surveys: Short exit surveys at vaccination points captured information sources. Village leader questionnaires collected socio-economic and contextual data (used selectively based on AIC contribution). Household questionnaires were administered to 20 randomly selected households per village (total 1,117 households) within 5 weeks to capture participation and information pathways. - Workshops and focus groups: Fifteen workshops (97 participants) from 43 villages and five focus groups (47 participants) informed intervention design and identified barriers. An additional workshop with leaders from outside study villages refined the community-leader intervention. - Collective action experiment: In each village, 20 surveyed households participated in a coordination game. Each participant received TZS 5,000 and could return it (cooperate) or keep it. If ≥14/20 returned cash, all participants would receive TZS 2,000; otherwise, cooperators lost their TZS 5,000. The proportion cooperating served as a village-level collective-action indicator. Ethical approvals and informed consent were obtained; cash was selected over alternatives following ethical and practical considerations. Quantitative analysis: - Primary analysis: Generalized linear mixed models (GLMMs) with logit link modeled estimated vaccination coverage as the dependent variable. Predictors included indicators for SMS coverage (0 if no SMS; otherwise number of phone numbers collected divided by estimated households), community leader advertising (binary), presence of pastoralists (binary), presence of a livestock field officer (binary), and, in extended models, the SMS×community leader interaction and the collective-action indicator. Models weighted by dog counts from transects; village-level fixed effects used for intervention coverage estimates. - Operational changes analysis: Collated estimated coverage from 2014, 2016, and 2018 for available villages (29 in 2014; 33 in 2016; all 56 in 2018). Modeled effects of operational changes (categorical: 1 in 2018 when implemented; 0 otherwise) with year indicators to assess association with coverage changes across years, acknowledging potential temporal confounders. Theoretical framework (Box 1): A large finite population of agents chooses whether to pay attention and then whether to participate. Payoffs incorporate opportunity costs of attention (k) and participation benefits and costs (a−c), with probability of public good production p(L) increasing with the number of participants L. Fractions of agents differ in opportunity costs and participation costs (parameters s, t). Attention and participation occur when p(M)+(a−c) ≥ k, where M is the number informed (proportion y). Interventions act by increasing y (information/salience via SMS and leaders) and/or reducing costs (operational changes), thereby raising participation s* and total participants s*·N.

- Advertising interventions: Neither SMS nor community leader advertising alone produced a statistically significant effect on coverage at the 10% level. Villages receiving both interventions achieved notably higher mean coverage (~70%), compared to Routine Advertising Only (~60%), Community Leaders Only (~65%), and Text Messaging Only (~30%), with reduced variance when both were combined. - GLMM results (advertising): In models without interaction, the community leader coefficient was positive and significant at 5%, while pastoralist presence (negative) and livestock officer presence (positive) were not significant. Including the SMS×community leader interaction yielded a positive and significant interaction (5%), a negative SMS main effect significant at 10%, and loss of significance for the leader main effect—indicating complementarity: the combined interventions outweighed the negative standalone SMS effect. - Qualitative insights: Many SMS recipients sought confirmation via other sources (e.g., loudspeakers, local leaders). Barriers to SMS effectiveness included illiteracy and weak signal. The SMS anecdote about a recent rabies death increased perceived risk; respondents reported heightened motivation to vaccinate. Community leaders were valued for mobilization rather than information per se. - Operational changes: Mean coverage was ~25% in 2014 and 2016, rising to ~60% in 2018 when operational changes (more points, all-day hours, earlier advertising) were implemented. Models indicated operational changes had a positive, statistically significant association with increased coverage in 2018, whereas 2016 effects were not significant, suggesting operational adjustments substantially enhanced participation. - Scale of campaign: 7,210 dogs vaccinated across the district; 3,256 in sampled villages. Overall, results support that increasing salience and lowering participation costs together can overcome collective action barriers, with operational logistics exerting particularly strong effects.

The study demonstrates that overcoming collective action challenges in mass dog vaccination requires both making disease risk salient and reducing participation costs. Advertising strategies alone, especially SMS with limited reach and literacy constraints, were insufficient in this setting. However, SMS and community leader engagement were complementary—leaders can validate and amplify messages while SMS provides timely, salient cues—yielding improved coverage when combined. The strongest gains were associated with operational changes that decreased access costs (more points closer to households, all-day availability, earlier notice), enabling households to act on intentions. Context strongly moderated intervention effectiveness. Compared with Haiti, where near-universal phone contact made SMS the primary channel, Tanzanian villages relied more on routine advertising and word-of-mouth, diminishing marginal benefits of SMS alone. Trust in local leaders helped mobilize participation but was not a standalone determinant of coverage. The theoretical framework clarifies how increasing the informed proportion (y) and reducing opportunity and participation costs shift attention and participation thresholds, jointly increasing s* and total participation. These insights generalize to other public goods requiring coordinated community action, including human vaccination programs aiming for herd immunity thresholds. By aligning communication with trusted channels and addressing logistical barriers, programs can more effectively translate awareness into action, improving vaccination coverage and advancing rabies elimination.

This work combines experimental interventions and theory to show that solving collective action problems in vaccination campaigns requires both salience and accessibility. In rural Tanzania, neither SMS nor community leader advertising alone significantly increased dog vaccination coverage, but their combination did, and operational changes (more vaccination points, extended hours, earlier advertising) were strongly associated with large coverage gains. The findings underscore the need to tailor interventions to local contexts, leveraging trusted leaders and communication norms while minimizing participation costs through practical logistics. Future research should isolate the causal impacts of individual components (advertising versus operations), explore scalability and cost-effectiveness of operational adjustments (e.g., optimal number/location of points, duration), assess equity of access across sub-populations (e.g., pastoralists), and test generalizability to other diseases and settings with varying communication infrastructures and social structures.

- Inability to fully isolate advertising effects from operational changes across years; potential temporal confounders in 2018 analyses. - SMS reach constraints (incomplete phone lists), literacy issues, and weak signal likely attenuated SMS effects. - Vaccination coverage estimates from transects are noisy, especially in low dog-count villages; although models weighted by counts, measurement error remains. - Limited availability and reliability of village-level covariates; many were excluded due to lack of model improvement. - Collective action indicator may exhibit limited between-village variation and may not perfectly map to vaccination contexts. - Historical comparison used partial village samples in 2014/2016 (29 and 33 villages), limiting comparability. - Ethical and contextual considerations around the cash-based coordination game, though approved and locally appropriate, could influence responses. - Generalizability is context-dependent; effectiveness of communication modalities varies with local infrastructure and norms.