Table 1.

Behavioral impact of social media use.

	Reducing the frequency of:			Increasing the demand for:
	Face-to-face conversation	Dining out: dinner	Dining out: lunch	Books/games	Disinfectant	Face mask	Fermented soybeans	Toilet paper rolls
Encouraged by the government?	Yes	Yes	Yes	No	Yes	Yes	No	No
Panel A:	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Twitter	0.040** (0.018)	0.022 (0.015)	0.028** (0.013)	0.058*** (0.019)	0.049*** (0.016)	0.019 (0.021)	0.024* (0.014)	0.027* (0.015)
Mean Dep. Var. among users	0.167	0.126	0.127	0.138	0.337	0.504	0.096	0.192
Counterfactual	0.127	0.104	0.099	0.080	0.288	0.485	0.072	0.165
Observations	1804	1804	1804	1804	1804	1804	1804	1804
Panel B:	(9)	(10)	(11)	(12)	(13)	(14)	(15)	(16)
Instagram	0.046** (0.019)	0.037** (0.016)	0.043*** (0.015)	0.044*** (0.015)	0.081** (0.031)	0.067** (0.031)	0.053*** (0.014)	0.033 (0.023)
Mean Dep. Var. among users	0.182	0.145	0.130	0.130	0.395	0.565	0.116	0.219
Counterfactual	0.136	0.108	0.087	0.086	0.314	0.498	0.063	0.186
Observations	1804	1804	1804	1804	1804	1804	1804	1804
Panel C:	(17)	(18)	(19)	(20)	(21)	(22)	(23)	(24)
Facebook	0.044 (0.027)	0.030** (0.014)	0.008 (0.023)	0.032* (0.017)	−0.013 (0.023)	−0.011 (0.028)	0.039*** (0.014)	−0.038* (0.022)
Mean Dep. Var. among users	0.197	0.148	0.140	0.129	0.332	0.512	0.109	0.167
Counterfactual	0.153	0.118	0.132	0.097	0.345	0.523	0.070	0.205
Observations	1804	1804	1804	1804	1804	1804	1804	1804

The coefficients from the entropy balancing model are reported. All specifications control for the covariates used to estimate the entropy balancing weights. The counterfactual is defined as the difference between the mean dependent variable and the average treatment effect. Standard errors clustered at the prefecture level are in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.