Table S4.
Results split by past gift acceptance
| Specification | A10 |
| Dependent variable | Chose to donate |
| Regression model | OLS |
| Population | All |
| N | 3,641 |
| R2 | 0.024 |
| Treatment | |
| Swag offereda | −0.038 (0.013)*** |
| Meals offereda | −0.040 (0.013)*** |
| Donor type | |
| Accepted gift in past | −0.012 (0.015) |
| Interactions | |
| Swag × Accepted Gift in Past | 0.039 (0.022)* |
| Meals × Accepted Gift in Past | 0.020 (0.019) |
| Constant | 0.137 (0.019)*** |
| Month fixed effects | Yes |
| Linear combination of coefficientsb | |
| Swag + Swag × Accepted-in-Past | 0.001 (0.017) |
| Meals + Meals × Accepted-in-Past | −0.020 (0.014) |
All SEs are Huber–White robust. *P < 0.10; ***P < 0.01. This regression separately tests for gift effects on donors who have accepted gifts in the past (e.g., during past on-air or online pledge drives that offered gifts) and those who never have done so. Results suggest those who have never accepted gifts before are largely driving the results. This result is consistent with the proposed attention mechanism. Donors who value the gifts more (because they have accepted them before) may show less of a change in net motivation when their attention shifts toward the gift. In addition, it is also possible that these donors (who may be more extrinsically motivated to begin with) may already be using a more cost–benefit mindset, and thus, their mindset is not as influenced by the gift offer.
These coefficients represent the gift effect on those who have never accepted gifts in the past. Results show that these donors do experience crowding out.
The linear combination results test whether those who have accepted gifts in the past show different donation rates in the gift versus control treatment; since these results are not statistically significant, these donors are not affected by the gift offer.