Vaccination incentives don’t backfire.

We first study the concern that offering financial incentives may reduce future vaccination uptake and other health behaviours when no payments are offered. Using administrative data on vaccination uptake, Fig. 1 and Table 1 show no evidence that participants in the financial incentives condition were less likely to take an unincentivized second dose of a COVID-19 vaccine (if anything, uptake increased; OLS regression, B = 0.055, standard error (s.e.) = 0.033, P = 0.097) or to delay the uptake of the second dose (OLS regression, B = 0.046, s.e. = 0.033, P = 0.164). We also do not find any effects on second-dose uptake when we restrict the sample to those who took the first dose (OLS regression, B = 0.049, s.e. = 0.035, P = 0.158). Using data from the first survey, we do not find evidence that offering monetary incentives affected the intention to take the third dose (OLS regression, B = −0.026, s.e. = 0.044, P = 0.560) nor the willingness of participants to take a third dose if they were hypothetically offered SEK 100 (OLS regression, B = 0.001, s.e. = 0.043, P = 0.983) or SEK 500 (OLS regression, B = −0.008, s.e. = 0.043, P = 0.850). Using data from the second survey, we do not find negative effects on self-reported actual third-dose uptake (OLS regression, B = −0.007, s.e. = 0.046, P = 0.879), the delay to take the third dose (OLS regression, B = 0.030, s.e. = 0.046, P = 0.524) or third-dose uptake when we restrict the sample to those who took the second dose (OLS regression, B = −0.016, s.e. = 0.049, P = 0.745).

Fig. 1: Regression-estimated effects of offering financial incentives for first-dose uptake on further COVID-19 vaccination, other health behaviours, morals and civic responsibility, perceived safety, efficacy and trust, and other concerns.

The figure is based on RCT data linked to comprehensive survey data and population-wide Swedish administrative data capturing each vaccination in Sweden. The figure shows regression-estimated effects of the financial incentives condition relative to the control condition. All regressions use the pre-registered controls consisting of gender, age, region, interactions between age and region, being in an at-risk group for COVID-19, civil status, having children in the household, employment status, education, parents’ place of birth and income (see Supplementary Information section 1.1 for details, see Supplementary Information section 2.3 and Extended Data Fig. 1 for results without controls). The blue dots indicate the estimated impact in standard deviations on the respective variables; all outcomes are defined as pre-registered. Error bars represent 95% confidence intervals (two-sided CI: mean ± 1.96 s.e.) from OLS regressions with heteroscedasticity-robust standard errors. The dashed grey lines indicate the threshold for small effect sizes of 0.2 standard deviations (Cohen’s d). The sample sizes for the control and incentives conditions across datasets are as follows: administrative data, n incentives = 1,132, n control = 3,888; first survey data, n incentives = 726, n control = 2,512; second survey data, n incentives = 606, n control = 2,100.