Table 8. Heterogeneous effects of COVID-19 on e-commerce: By age of the household head.

	Online shopping after the pandemic
	(1)	(2)	(3)
	2SLS	2SLS	2SLS
Share of coronavirus cases	0.010	0.097	0.824
	(0.221)	(0.301)	(0.772)
	0.969	0.748	0.227
Share of coronavirus cases * household head age below 35	0.547***	0.469**	0.512**
	(0.211)	(0.206)	(0.223)
	[0.008]	[0.021]	[0.018]
Control variables	Yes	Yes	Yes
Regional fixed effects	No	Yes	No
Provincial fixed effects	No	No	Yes
First-stage F-stat	50.84	32.82	33.46
Observations	770	770	770
Test Share of coronavirus cases + Share of coronavirus cases * household head age below 35 = 0	0.557*	0.566*	1.336*
	[0.051]	[0.076]	[0.077]

The dependent variable is a dummy for online shopping after the pandemic. Household head age below 35 is a dummy for household head age less than 35. Above prefecture-level cities include sub-provincial and provincial cities. Below prefecture-level cities include counties and below. The share of COVID-19 cases is calculated as the number of confirmed COVID-19 cases on the survey day/city population. The instrumental variable for the share of COVID-19 cases is the distance between the city and Wuhan, which is transformed using the log function. Control variables include gender, age, education levels, income, household size, share of children and share of the elderly. The region refers to the east, center and west. Robust standard errors clustered at the city level are reported in parentheses. P-values from wild bootstrap clustering are reported in brackets. We use Rademacher weights and 1000 replications.

*** significant at the 1% level

** significant at the 5% level

* significant at the 10% level.