Table 4.

First-round guesses and correctness controlling for the raters’ beliefs.

Dependent variable	(1) Guess Split for all treatments	(2) Guess Split when gender revealed	(3) Correct guess when gender revealed
Label	2.28***
	(4.91)
Photo	1.77***	0.52**	0.09
	(3.99)	(3.46)	(1.34)
Video	2.41***	1.30***	0.06
	(5.32)	(8.78)	(0.84)
Belief about players	0.08***
	(13.77)
Label × belief about players	−0.04***
	(−5.47)
Photo × belief about players	−0.04***
	(−4.66)
Video × belief about players	−0.05***
	(−6.46)
Gender-specific belief		0.05***
		(30.47)
Photo × gender-specific belief		−0.01***
		(−5.04)
Video × gender-specific belief		−0.03***
		(−12.89)
Sufficiently correct stereotype			0.83***
			(14.88)
Photo × sufficiently correct stereotype			−0.28***
			(−3.78)
Video × sufficiently correct stereotype			−0.39***
			(−5.22)
Constant	−3.69***	−2.37***	−0.19***
	(−11.09)	(−21.24)	(−3.78)
Guess	39,668	29,516	29,516
Raters	422	314	314
Log-likelihood	−24,026	−18,057	−19,882
Akaike information criteria	48,070	36,128	39,779
Bayesian information criteria	48,147	36,186	39,837
Chi-squared (7/5/5 degrees of freedom)	338***	1,861***	438***

Z-Value in parentheses. *** p < 0.001, ** p < 0.01. We report the results of logit regression, where the rater is the panel and players are the trials. Regression (1) includes first-round data from all treatments. Regressions (2) and (3) include first-round data only from treatments where players’ gender is revealed: Label, Photo and Video. The variable Belief about players in regression (1) refers to the average of the rater's male and female gender-specific beliefs. Gender-specific beliefs in regression (2) refers to the applicable belief about male or female player given the player's self-described gender. ‘Sufficiently correct stereotype’ in regression (3) is a dummy variable that equals one if the rater's gender stereotype is ≥50% and players of that gender tended to be cooperative, or if the rater's gender stereotype is <50% and players of that gender tended to be non-cooperative.