Table 3.

Post hoc contrasts for number of faces in the person queries. Effects of gender by query and search engine and of the gender × query interaction by search engine on image representativeness. Results averaged over the levels of region, browser, wave. Main effects are adjusted for interaction influence and multiple comparison (Tukey). Gender coded: 0 = women; 1 = men. Bolded rows were found significant at the .05 level. Negative z-ratio indicates lower representation of women. For example, at odds ratio (OR) 0.23, Bing was four times as likely to show images of men than of women in the person query, while at OR of 4.5, Baidu was 4.5 times more likely to show images of women than of men in the person query.

	Search engine	Odds ratio	Standard error	z-ratio	p-value	95% CI L	95% CI H
Gender—Person	Baidu	4.551	0.822	8.388	<.001	2.861	7.239
	Bing	0.228	0.019	−17.553	<.001	0.184	0.283
	Google	1.233	0.123	2.096	.154	0.954	1.595
	Yandex	0.043	0.005	−27.178	<.001	0.032	0.058
Gender—Intelligent person	Baidu	0.721	0.146	−1.614	.37	0.428	1.213
	Bing	0.146	0.014	−19.511	<.001	0.114	0.188
	Google	0.188	0.024	−12.866	<.001	0.135	0.263
	Yandex	0.098	0.011	−20.216	<.001	0.073	0.131
Gender × Query	Baidu	0.158	0.043	−6.786	<.001	0.093	0.27
	Bing	0.641	0.083	−3.427	.001	0.498	0.827
	Google	0.152	0.025	−11.473	<.001	0.111	0.21
	Yandex	2.253	0.368	4.981	<.001	1.637	3.102