An (un)healthy social dilemma: a normative messaging field experiment with flu vaccinations

Health Economics Review

Table 4 Measures of prediction

Probability of getting vaccinated (%)
	(1)	(2)	(3)	(4)	(5)
Baseline (no message)	31	31	31	29	33
Self	39	39	40	38	34
Others	44	43	42	41	45
Both	51	51	51	48	52
Receiver operating characteristic (ROC) analysis: Area Under the Curve value
Cross-validated mean AUC	0.59	0.57	0.59	0.69	0.59
Standard Deviation	0.04	0.03	0.04	0.06	0.03
Bootstrap bias-corrected CI
Lower bound	0.49	0.52	0.54	0.63	0.54
Upper bound	0.59	0.60	0.62	0.71	0.62
Controlling for:
Socio-demographic	No	Yes	Yes	Yes	Yes
Preferences	No	No	Yes	Yes	Yes
Beliefs	No	No	No	Yes	No
Prior vaccination	No	No	No	Yes	No
Heterogeneity:
Gender interactions	No	No	No	No	Yes
N	828	828	828	828	828

Notes: *p = 0.10 **p = 0.05 ***p = 0.01. The probability of getting vaccinated is the predicted probability of getting the flu vaccine if everyone was assigned to a specific treatment, under a linear probability model. The Receiver operating characteristic (ROC) analysis: Area Under the Curve value is used for comparing predictive models in both model selection and model evaluation after fitting a logit regression, and it is widely used in health-based research. It ranges from 0.5 to 1, where 1 is perfect accuracy. Acceptable predictive values start around 0.65. The Area Under the Curve was calculated using the Stata command cvauroc [80]. We use a 10 K-fold cross-validation

ISSN: 2191-1991