Warming up cool cooperators

Abstract

Explaining why someone repeats high-cost cooperation towards non-reciprocating strangers is difficult. Warm glow offers an explanation. We argue that warm glow, as a mechanism to sustain long-term cooperation, cools off over time but can be warmed up with a simple intervention message. We tested our predictions in the context of repeat voluntary blood donation (high-cost helping of a non-reciprocating stranger) across 6 studies: a field-based experiment (n = 5,821) comparing warm-glow and impure-altruism messages; an implementation study comparing a 3-yr pre-implementation period among all first-time donors in Australia (N = 270,353) with a 2-yr post-implementation period (N = 170, 317); and 4 studies (n = 716, 1,124, 932, 1,592) exploring mechanisms. We show that there are relatively warm and cool cooperators, not cooling cooperators. Cooperation among cool cooperators is enhanced by a warm-glow-plus-identity message. Furthermore, the behavioural facilitation of future cooperation, by booking an appointment, is associated with being a warm cooperator. Societal implications are discussed.

Fig. 1. Theoretical models linking warm glow to cooperation.

a, Current knowledge. Summary of the general findings supporting warm-glow theory. Specifically, people experience variation in warm glow from giving (ω) (see main text for factors that influence the extent of experienced warm glow) and this predicts future donations (ρ) based on the reinforcing effects of warm glow (λ) (see main text). Exogenous manipulations of warm glow (χ) predict repeat donations. b, Model to be tested. We extend a by suggesting an additional behavioural mechanism, booking another appointment to donate (β), which enhances commitment to cooperate (Ø) and identity as a donor (π). Overall increased feelings of experienced warm glow increase the probability of booking, with this being stronger for warm cooperators (those with higher levels of experienced warm glow) than cool cooperators (those with lower levels of experienced warm glow). Booking (β) predicts return donation (ρ), which is again stronger for warm cooperators. An exogenous manipulation of warm glow/impure altruism will increase the probability of making another donation if a donor identity prime is present. As cool cooperators are less likely to book and have their donor identity strengthened, the warm-glow/impure-altruism-plus-identity messages should be maximally effective for cool cooperators. Green, behavioural mechanism; peach, psychological mechanism; brown, exogenous manipulation; light blue, overall experienced warm-glow; yellow, effects linked to being a warm-cooperator; dark blue, effects linked to being a cool cooperator; greys, indications of associated behaviours or action.

Fig. 2. Treatment messages used in Studies 1 and 6.

The messages formed when warm glow (warm glow, impure altruism) is crossed with identity (present, absent).

Fig. 3. Aggregate return behaviour pre- and post-implementation of the ‘warm-glow-plus-identity’ message in new donors as a function of booking status (Study 2).

For the pre-implementation, the aggregated point estimate for the percentage return rate for those who booked was 61.47 (95% CI = 61.20, 61.74; n = 75,414/122,681) and for those who did not book, 21.49 (95% CI = 21.28, 21.70; n = 31,738/147,672). For the post-implementation, the aggregated point estimate for the percentage return rate for those who booked was 62.35 (95% CI = 61.96, 62.74; n = 36,472/58,500) and for those who did not book, 30.55 (95% CI = 30.28, 30.82; n = 34,156/111,817). For those who had booked, there was a 0.88% (95% CI 0.402, 1.358) significant (Z = 3.610, P = 0.0003) increase in returning donors. For those who had not booked there was a 9.06% (95% CI = 8.718, 9.402) significant (Z = 51.963, P < 0.001) increase in returning donors. There was a significant interaction with the percentage increase in return rates for those who had not booked, significantly greater than those who had booked (Z = 27.292, P < 0.001), with a percentage difference of 8.18% (95% CI = 7.593, 8.767). Analyses were conducted using procedures for Z tests for proportions detailed in refs. ^, and implemented in ZumaStat 4.0. All analyses were two-tailed and no adjustments were made for multiple comparisons. Error bars are 95% CIs.

Fig. 4. Mediation models specifying the effects of messages to predict ‘afforded warm glow’ directly and indirectly via a focus on the donor, recipient or both (Study 6).

These models specify the effects of messages (X) to predict the outcome (Y) ‘afforded warm glow’ directly and indirectly via focus (‘afforded focus on the donor, recipient or both’ (the mediator, M)). These models were specified in PROCESS 4.0, with estimates based on 5,000 bootstraps with a multinominal X (messages) and BAU as the reference (n = 1,521). Age, gender and supply (‘afforded focus on donating to maintain blood supply’) were specified as confounders of M and Y. Predicting M from X has an R² of 0.086, P < 0.001. Predicting Y from both X and M has an R² of 0.333, P < 0.001. All analyses were two-tailed and no adjustments were made for multiple comparisons.