Table 2. Panel regressions of average log nightlight intensity by district as log distance from Pyongyang increases (January 2020 to January 2022)

Variables	(1)Ordinary least squares (OLS)	(2)Two-way FE (unclustered)	(3)Two-way FE (unclustered with interaction)	(4)Two-way FE (clustering by province)	(5)Two-way FE (clustering with interaction)	(6)Random effects (with interaction)
Log distance	–0.1327***(0.008856)	–0.1327***(0.008198)	–0.1329***(0.008532)	–0.3312***(0.01224)	–0.2056***(0.01323)	–0.2027***(0.01474)
Provincial population density (people/km²)	0.0002001***(0.00001083)	0.0002001***(0.00001003)	0.0001930(0.0001002)	–0.0007460***(0.0001071)	0.00227***(0.00009951)	0.002333***(0.00014645)
Log distance×Provincial pop. density			0.00000071(0.00000988)		–0.0002233***(0.0000113)	–0.0002183***(0.00001272)
Constant	0.1587(0.1057)	–0.4472***(0.1013)	–0.4455***(0.1044)	1.9989***(0.15778)	0.4435***(0.1691)	0.74671***(0.18108)
Observations	3,534	3,534	3,534	3,534	3,534	3,534
Adjusted R2	0.2293	0.3397	0.3395	0.56	0.6041	0.2079

Standard errors in parentheses.
p<0.001.
FE, fixed effects.