Table 2 Population-averaged model tests of hypothesis.

From: Heterogeneous groups cooperate in public good problems despite normative disagreements about individual contribution levels

 

(1)

(2)

(3)

(4)

(5)

(6)

Ind. all rounds

Ind. start rounds

Ind. end rounds

Group. all rounds

Group. start rounds

Group. end rounds

Treatment (disagreement)

.608

.115

− .156

.608

.115

− .156

(.672)

(.764)

(.929)

(1.014)

(.831)

(1.274)

Intercept

13.561***

12.563***

13.188***

13.561***

12.563***

13.187***

(.475)

(.540)

(.657)

(.717)

(.588)

(.901)

N observations

1920

192

192

640

64

64

N participants

192

192

192

N groups

64

64

64

64

64

64

  1. We use population-averaged regression models to statistically examine the hypothesis that the contribution level is higher in the normative agreement condition than in the normative disagreement condition. The contribution decision is the dependent variable and the experimental condition is the predictive factor. Across six models we vary whether the outcome variable is on the individual-level (models 1–3) or group-level (models 4–6), and whether we include all rounds (models 1 and 4), only the first round (models 2 and 5), or only the end round (models 3 and 6) as observations. Regardless of which model is used, we find no significant difference in contribution levels between conditions according to conventional standards.
  2. *p < 0.05, **p < 0.01, ***p < 0.001 (Bonferroni-adjusted p/6, two-tailed tests). Standard errors in parentheses. In model 1 and 4 there are repeated measures on individuals/groups. We take these repeated measures into account with the exchangeable working correlation matrix.
Back to article page