The Burden of Social Connectedness: Do Escalating Gift Expenditures Make You Happy?

Our previous estimators likely suffer from sample selection bias, i.e., attrition bias, if the subsample of these dropped observations is not randomly selected from the entire population. Firstly, to test this potential bias we use standard Heckman selection correction methods, where we identify the selection term of being repeatedly surveyed using household characteristics (such as gender, age, hukou status, political status, marital status, education, household wealth and number of children). We find that the coefficient of the estimated ρ (i.e., the correlation of the two error terms) is quite small (0.028) and its p value is almost equal to 0.9, indicating a weak evidence of sample selection bias. We also divide the full sample into two subgroups based on whether the respondent participated in all rounds of surveys, and then look at whether there is significant difference in happiness between these two subgroups, holding other thing being equal. Results by regressing happiness on RGI and Repeated, which is an indicator variable that equals to one if the respondent repeatedly participated in all three rounds of surveys and equals to zero otherwise, while controlling for a number of household characteristics, county dummies and year dummies, show that the coefficient of Repeated is not significant (p value = 0.345). This finding reaffirms that attrition poses little threat to the baseline estimators.

In this study, the RGI is measured at the household level, and happiness and some control variables are measured at the individual level. Therefore, such cross-level analysis may induce bias in the main findings. To ease this concern, we restrict the sample used in the analysis to household heads and their spouse, who are mainly responsible for supporting the family (Wang 2012). They would feel more pressure than other family members when the family cannot or struggles to make ends meet.

The total effect refers to the effect of the independent variable RGI on the dependent variable Happiness without the mediating variable Household consumption. The direct effect refers to the effect of the independent variable RGI on the dependent variable Happiness when controlling for the mediating variable Household consumption. The indirect (mediating) effect of Household consumption refers to the effect of the independent variable RGI on the dependent variable Happiness through the mediating variable Household consumption.