Are teachers accurate in predicting their studentsperformance on high stakesexams? The case of Russia Andrey Zakharov a , Martin Carnoy * , b a National Research University Higher School of Economics, Ul. Myasnitskaya 20, Moscow, Russia b Stanford University, School of Education, 485 Lasuen Mall, Stanford, CA 94305, USA A R T I C L E I N F O Article history: Received 21 September 2014 Received in revised form 23 March 2015 Accepted 27 April 2015 Keywords: Educational policy International education Teacher expectations High stakes exams A B S T R A C T The paper focuses on how accurate teachers may or may not be in gauging their classacademic abilities. We use a sample of classrooms in three Russian regions to identify sources of mathematics and Russian teachersinaccuracies in predicting their high school classesscores on Russian and mathematics high stakes college entrance tests (the Unied State Exam, or USE). We test the hypothesis that teachers perceptions of their relationship with their classes are good predictors of such inaccuracies. This is important because teachers often focus on their relationship with the class as an end in itself or as a means to engaging students. Good teacherstudent relations may indeed result in more students learning, but perhaps not nearly as much as teachersbelieve. We nd that both Russian and mathematics teachers make inaccurate predictions of their classhigh stakes examination results based on how they perceive their relationship with their class. Teachers who believe they have a very good relationship with the class signicantly overestimate their classperformance on the USE, and those who perceive a poor relationship, underestimate their classperformance, although this underestimate is generally not statistically signicant. ã 2015 Elsevier Ltd. All rights reserved. 1. Introduction Much of the literature on effective teaching emphasizes the ideal of teachers as reective professionals capable of individual- ized approaches to student learning (for example, Cohen, 1993; Darling Hammond, 1996). An implicit assumption underpinning this ideal is that teachers are accurate, fair judges of their students abilities, and that they can (and should) individualize broad curricular guidelines to t each students capacity and learning style. Two strands of research have questioned this assumption. One strand, going back to the 1960s, argues that teachers may not be neutral observers of studentsabilities, that teachersexpectations may vary among students, and that teachersexpectations (positive or negative) can affect studentsperformance (Rosenthal and Jacobson, 1968). More recently, this research has turned to nding students' (and teachers) characteristics that may affect teacher expectations in particular subjects, and, hence, their studentsperformance (Brophy, 1983; Rosenthal and Rubin, 1978; Rosenthal, 1994). A number of studies nd teacher gender bias teachers viewing boys as having greater math and science skills and girls as having greater literary skills (Qing, 1999; Ready and Wright, 2011; Riegle-Crumb and Humphries, 2012; Shepardson and Pizzini, 1992), but others nd no evidence of teacher gender bias (Dusek and Joseph, 1983; Madon et al., 1998). Similarly, many studies have found teacher race/ethnic bias (Ready and Wright, 2011; Rubie-Davies et al., 2006; Tenenbaum and Ruck, 2007), and social class bias (Auwarter and Aruguete, 2008; Ready and Wright, 2011). This may also relate to how teachers view students in different academic tracks (Kelly and Carbonaro, 2012; Oakes, 1985; Page, 1987; Tach and Farkas, 2006). Teacher gender and ethnicity have also been shown to play a role in affecting the performance of studentsof particular gender and ethnicity (Dee, 2005; McKown and Weinstein, 2008; Ready and Wright, 2011; Van den Bergh et al., 2010). Some of these studies estimate causal effects and show that teacherssubjective judgmentconsciously or unconsciouslycan and does affect studentsacademic outcomes. Although much less studied, the second strand of research argues that teachersexpectations for studentsperformance compared to actual results may differ not because of conscious or unconscious biases,but because of what Jussim and Harber (2005) called predictive validity without self-fullling inuence. Ferguson (2003) argued that both teacher inaccuracy and bias * Corresponding author. Tel.: +1 6508567722. E-mail addresses: ab.zakharov@gmail.com (A. Zakharov), carnoy@stanford.edu (M. Carnoy). http://dx.doi.org/10.1016/j.ijedudev.2015.04.007 0738-0593/ ã 2015 Elsevier Ltd. All rights reserved. International Journal of Educational Development 43 (2015) 111 Contents lists available at ScienceDirect International Journal of Educational Development journal homepage: www.else vie r.com/locate /ijedudev