Table B-3.

Impact of Theft Shocks on Social Preferences

Explanatory variable	Dependent variable	Correlation coefficient	Regression coefficient	# Obs.
Δ log theft 2002 to 2007	Outcome in 2007
	sent as dictator in anonymous game	−0.030	−0.024 (0.076) [0.708]	103
	sent as dictator in chosen non-revealed game	−0.027	0.013 (0.105) [0.919]	103
	sent as dictator in revealed game	0.055	0.073 (0.085) [0.367]	103
	sent as dictator in chosen revealed game	0.001	0.083 (0.104) [0.550]	103
	trust people in the world	−0.116	−0.059** (0.024)+ [0.024]**	123
	trust people in the village	−0.079	−0.064** (0.027)+ [0.121]	123
	trust closest neighbor	0.037	0.007 (0.030) [0.852]	123
	bad to buy something you know is stolen	0.057	0.008 (0.007) [0.426]	123
	would villagemates take advantage if had opportunity	−0.075	−0.007 (0.025) [0.739]	123
Δ log theft 2007 to 2009	Outcome in 2009
	sent as dictator in anonymous game	0.117	0.080 (0.106)	41
	sent as dictator in chosen non-revealed game	0.078	−0.015 (0.095)	33
	sent as dictator in revealed game	0.156	0.070 (0.086)	41
	sent as dictator in chosen revealed game	0.122	0.068 (0.117)	33
	trust people in the world	−0.025	−0.001 (0.049)	49
	trust people in the village	0.035	−0.013 (0.068)	49
	trust closest neighbor	−0.077	−0.083 (0.079)	49
	bad to buy something you know is stolen	0.022	0.005 (0.009)	49
	would villagemates take advantage if had opportunity	0.064	0.066 (0.058)	49
	negative reciprocity	−0.008	−0.002 (0.021)	49

Notes: Heteroskedasticity robust standard errors are in parentheses. Wild cluster bootrap p-values are in brackets.

Per-comparison significance: ***p <0.01,

^**p <0.05,

^*p <0.10.

FDR q-values: +++ q <0.01,

⁺⁺q <0.05,

⁺q <0.10 calculated for 9 or 10 hypotheses within table and column and panel.

In regression column these are based on the heteroskedasticity robust standard errors. Controls in regressions include log income, sex, age, education, and village fixed effects. Each regression additionally controls for social preferences in the previous round. In the top panel, the first four rows control for the amount sent as trustor and the average share returned as trustee in 2002. In the remaining rows of the first panel, the 2002 version of the 2007 outcome variable is included as a control variable. In all rows of the second panel, the 2007 version of the 2009 outcome is included as a control variable.