Application of a Theorem in Stochastic Models of Elections

Table 8

𝛽-spatial conditional logit models for USA (2008).

	(1) Spatial	(2) Sp. & Traits	(3) Sp. & Demog	(4) Full
McCain valence 𝜆	−0.84^∗∗∗	−1.08^∗∗∗	−2.60^∗∗	−3.58^∗∗∗
	(0.11)	(0.13)	(0.93)	(1.05)
Distance 𝛽	0.85^∗∗∗	0.78^∗∗∗	0.86^∗∗∗	0.83^∗∗∗
	(0.06)	(0.07)	(0.07)	(0.08)
McCain traits		1.30^∗∗∗		1.36^∗∗∗
		(0.17)		(0.19)
Obama traits		−1.02^∗∗∗		−1.16^∗∗∗
		(0.15)		(0.18)
Age			−0.01	−0.01
			(0.01)	(0.01)
Female			0.29	0.44
			(0.23)	(0.26)
African-American			−4.16^∗∗∗	−3.79^∗∗∗
			(1.10)	(1.23)
Hispanic			−0.55	−0.23
			(0.41)	(0.45)
Education			0.15^∗	0.22^∗∗∗
			(0.06)	(0.06)
Income			0.03	0.01
			(0.02)	(0.02)
Working Class			−0.54^∗	−0.70^∗∗
			(0.24)	(0.27)
South			0.36	−0.02
			(0.24)	(0.27)
Observations	781	781	781	781
Loglikelihood	−298.63	−243.14	−250.25	−206.88
AIC	601.27	494.28	520.50	437.77
BIC	610.59	512.92	567.11	493.69

Standard errors in parentheses. ^∗prob<.05; ^∗∗prob<.01; ^∗∗∗prob<.001.
Vote for Obama is the baseline outcome.