We then proceed with a regression analysis. Results are shown in table 1. We regress the amount of coins appropriated by dictators on the dummy $treated$—that is, equal to 1 for participants in the sample who experienced the land tenure reform—and a set of sociodemographic controls.14 Model 1 implements an OLS regression. Standard errors are clustered at the village level, and we use wild cluster bootstrap with 999 replications to resample with the sixteen clusters (Cameron, Gelbach, & Miller, 2008). The coefficient of the treatment dummy is negative and statistically significant at the 1% level, suggesting that participants in the treated sample take significantly less than those in the control sample. In model 2, we replicate the same specification but for using a censored tobit regression to account for a mass point at zero coins taken and implementing score bootstrap with 999 replications. The result remains significant at the 1% level, confirming that participants who experienced the reform appropriate significantly less of the passive player's endowment than those in the control group.15 Point estimates suggest that experiencing the reform on average determines a 40% to 60% reduction of the passive players' endowment appropriated by the dictators.

Table 1.

Coins Appropriated by the Dictator

Model 1Model 2Model 3Model 4
treated $-$2.046 $-$2.518 $-$2.097 $-$2.525
C.I. [$-$3.53, $-$.700]  [$-$3.651, $-$.691]
P-value 0.004 0.001 0.005 0.001
education   $-$0.130 $-$0.192
[$-$.567, .329]
0.52 0.38
income10k   $-$0.118 $-$0.172
[$-$.565, .191]
0.49 0.38
credit   0.057 0.157
[$-$.681, .697]
.84 0.68
Controls
Constant 3.903 3.643 4.319 4.165
[1.818, 5.987]  [2.499, 6.203]
Number of 127 127 127 127
observations
Model 1Model 2Model 3Model 4
treated $-$2.046 $-$2.518 $-$2.097 $-$2.525
C.I. [$-$3.53, $-$.700]  [$-$3.651, $-$.691]
P-value 0.004 0.001 0.005 0.001
education   $-$0.130 $-$0.192
[$-$.567, .329]
0.52 0.38
income10k   $-$0.118 $-$0.172
[$-$.565, .191]
0.49 0.38
credit   0.057 0.157
[$-$.681, .697]
.84 0.68
Controls
Constant 3.903 3.643 4.319 4.165
[1.818, 5.987]  [2.499, 6.203]
Number of 127 127 127 127
observations

Dependent variable: coins appropriated by the dictator. Models 1 and 3: OLS regression, wild cluster bootstrap with 999 replication. Models 2 and 4: left-censored tobit regression, score bootstrap with 999 replications. Standard errors clustered at the village level. Compared to models 1 and 2, models 3 and 4 additionally control for income (in thousands of XOF), access to credit, and education levels. Controls include age, gender, religion, village distance to paved roads, village population, and estimation of risk preferences. The two lines below the coefficient estimates report the confidence intervals (for OLS regressions) and the $p$-value, respectively.

Close Modal