Table 1.

Results of negative binomial models examining the association between Twitter sentiment rate (negative and positive) and the anti-Asian hate crimes in New York City from 2019 to 2022.

Modela (N=48)			Incidence rate ratio (95% CI)		P value
Crude model 1.0
	Negative sentiment (0-month lag)	1.24 (1.07-1.44)		.005
Crude model 2.0
	Positive sentiment (0-month lag)	0.88 (0.79-0.97)		.008
Adjusted model 1.1
	Negative sentiment (0-month lag)	1.16 (1.06-1.28)		.002
Adjusted model 2.1
	Positive sentiment (0-month lag)	0.93 (0.86-1.02)		.11

^aThe crude models 1.0 and 2.0 regress the number of anti-Asian hate crimes on negative or positive sentiment with a 0-month lag. The adjusted models 1.1 and 2.1 control for the monthly unemployment rate and whether it was during the COVID-19 pandemic (using a 0 value for periods before March 2020 and a value of 1 from March 2020 onward). All models use monthly data and measure sentiment in percentages.