“…Empirical approaches based on structural assumptions that distinguish ex ante from ex post returns to education include Carneiro et al (2003), Cunha, Heckman and Navarro (2005), Cunha and Heckman (2007), and Foley, Gallipoli and Green (2014). 1 However, there is relatively little empirical work using completion probabilities directly; notable exceptions using predicted college completion probabilities include Hussey and Swinton (2011), Fossen and Glocker (2014), and Castex (2015). In general, when studying choice under uncertainty, researchers have to make assumptions about how expectations are formed (Manski, 2004), and, most commonly, such work relies on rational expectations, e.g., that individuals' predictions are unbiased.…”