Table 5. Probit Regression Results
Sharing
Variables
Coefficients
Marginal Effects
(1)
(2)
(3)
(4)
(1)
(2)
(3)
(4)
-.29*
-4.97***
-3.22**
-3.96
Const.
(.16)
(1.27)
(1.34)
(2.87)
.33
.13
-6.38***
-10.85***
.13
.05
-.99***
-.99***
Public
(.40)
(.29)
(1.76)
(3.33)
(.16)
(.11)
(.02)
(.00)
-.001
-.02
-.09
-.0005
-.006
-.03
Popularity
(.02)
(.02)
(.19)
(.007)
(.01)
(.07)
.06*
-.41
.02
-.16
Public×Popularity
(.03)
(.25)
(.01)
(.10)
.45***
.30***
.38
.17***
.11***
.14
Age
(.10)
(.08)
(.25)
(.04)
(.03)
(.10)
.56***
1.02***
.21***
.39***
Public×Age
(.18)
(.34)
(.07)
(.14)
.008
.003
Popularity×Age
(.02)
(.008)
Public×Popularity
.05
.02
× Age
(.03)
(.01)
-.23
-.35
-.34
-.08
-.13
-.13
Male
(.21)
(.24)
(.25)
(.08)
(.09)
(.10)
.34
.45
.13
.18
Public×Male
(.41)
(.43)
(.16)
(.17)
.10
.03
.04
.04
.01
.02
Prosocial
(.06)
(.07)
(.08)
(.03)
(.03)
(.03)
.22**
.21*
.08**
.08*
Public×Prosocial
(.11)
(.11)
(.04)
(.04)
N
231
182
182
182
Pseudo R2
0.01
0.19
0.24
0.25
Standard errors are in the parenthesis. Errors are clustered at group level. ***,**,* indicate significant level
at p=.01,.05, .10, respectively .
In regression (3), we focus on the interaction effects of population and age in Public and
Private: popularity has a significant and positive effect on the sharing probability only in
Public (p = .04). However, this effect is not seen in Private (p= .57), which suggests that
more popular kids are more likely to share in Public. Nonetheless, in actuality, they are
no more generous than in Private. On the other hand, less popular kids behave more
similarly between Public and Private. Indeed, children ranked first are 20% more likely to
share than children ranked eleventh in Public, although a first ranked child is no more
likely to share than a child ranked eleventh when sharing occurs in Private. Additionally,
age has a significant and positive impact on both Public and Private treatments (p < .01
and p < .01, respectively). Pro-social tendencies play a significant and positive role in
!
18!