Table 4.
Logistic regressions modelling whether the social learners chose the social learner optimum. Predictors included (a) the centred proportion of demonstrators who chose the demonstrator optimum, (b) dummies for each combination of similarity and reliability information, minus the omitted category of reliably incorrect–similar signals and (c) interactions between each of these dummies and the centred proportion of demonstrators who chose the demonstrator optimum. Robust standard error clustered on social learner. See Appendix 9 for the regressions with control predictors, although the only significant control predictor was that the social learners were more likely to answer optimally on blocks where the % symbol was optimal, suggesting an arbitrary preference to choose this symbol across both games. We also give the 95% confidence interval lower and upper bounds for each estimate.
| Parameter | Estimate (game against nature, all signals) |
Estimate (coordination game, all signals, full data) |
|---|---|---|
| Intercept | 0.210 (0.148) 95% CI [−0.08, 0.50] |
0.047 (0.109) 95% CI [−0.17, 0.26] |
| Centred proportion of demonstrators choosing the demonstrator optimum | 0.166 (0.642) 95% CI [−1.09, 1.42] |
0.613 (0.462) 95% CI [−0.29, 1.52] |
|
Reliably incorrect–different dummy [indicates different and is correct with 0.1 probability] |
−0.136 (0.184) 95% CI [−0.50, 0.22] |
0.184 (0.165) 95% CI [−0.14, 0.51] |
|
Uninformative–same dummy [indicates same and is correct with 0.5 probability] |
−0.193 (0.201) 95% CI [−0.59, 0.20] |
−0.026 (0.159) 95% CI [−0.34, 0.29] |
|
Uninformative–different dummy [indicates different and is correct with 0.5 probability] |
−0.131 (0.193) 95% CI [−0.51, −0.25] |
−0.136 (0.160) 95% CI [−0.45, 0.18] |
|
Reliably correct–same dummy [indicates same and is correct with 0.9 probability] |
−0.278 (0.222) 95% CI [−0.71, −0.16] |
0.254 (0.204) 95% CI [−0.14, 0.65] |
|
Reliably correct–different dummy [indicates different and is correct with 0.9 probability] |
−0.048 (0.192) 95% CI [−0.42, −0.33] |
−0.017 (0.168) 95% CI [−0.35, 0.31] |
| Centred proportion of demonstrators choosing optimum × reliably incorrect–different dummy | 0.374 (0.815) 95% CI [−1.22, 1.97] |
0.786 (0.622) 95% CI [−0.43, 2.00] |
| Centred proportion of demonstrators choosing optimum × uninformative–same dummy | −0.787 (0.819) 95% CI [−2.39, 0.81] |
−1.197 * (0.595) 95% CI [−2.36, −0.03] |
| Centred proportion of demonstrators choosing optimum × uninformative–different dummy | −0.901 (0.805) 95% CI [−2.48, 0.67] |
−0.023 (0.616) 95% CI [−1.23, 1.18] |
| Centred proportion of demonstrators choosing optimum × reliably correct–same dummy | 2.53 ** (0.911) 95% CI [0.75, 4.31] |
1.460. (0.770) 95% CI [−0.05, 2.96] |
| Centred proportion of demonstrators choosing optimum × reliably correct–different dummy | 0.724 (0.785) 95% CI [−0.81, 2.26] |
0.573 (0.605) 95% CI [−0.61, 1.76] |
Asterisks denote the level of significance of our p-values, with the following key: *** p < 0.001; ** p < 0.01; * p < 0.05; trend, p = 0.05–0.10 significance.