Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Draft – Not to be cited without the consent of the authors PYGMALION EFFECTS ON FOLLOWERS’ FOLLOWERS Søren C. Winter1 Mogens Jin Pedersen1, 2 Vibeke Lehmann Nielsen2 Simon Calmar Andersen2 1 SFI – The Danish National Centre for Social Research 2 Aarhus University Corresponding Author: Søren C. Winter [email protected] January 2016 Paper prepared for presentation at the Annual Meeting of the Southern Political Science Association at San Juan, Puerto Rico, 7 - 9 January 2016. Acknowledgements We would like to thank Emil Rosenlund Andersen for assistance with statistical analyses and Ida Gran Andersen for first analyzing the role of teacher expectations. We are grateful to the teachers who participated in our survey. An earlier version of this paper was presented at the 11th Public Management Research Conference in Madison, Wisconsin, 20-22 June 2013. The research was supported by grants from Innovation Fund Denmark (Grant no. 0603-00281B), SFI – The Danish National Centre for Social Research and Aarhus University for a project on “School Management, Teaching, and Student Performance” directed by Søren C. Winter PYGMALION EFFECTS ON FOLLOWERS’ FOLLOWERS ABSTRACT Since Rosenthal’s pioneering studies of Pygmalion effects showing that teachers’ high expectation increase the achievement of their students, numerous experimental studies have showed that also managers’ expectations for followers increase followers’ performance and efforts. In this study we demonstrate that school principals’ expectations even affect followers’ followers, i.e. in increasing students’ achievement. As expected by goal setting theory this effect is strongest when the expectations are most challenging — in this case for socioeconomic disadvantaged students. We do this in a naturalistic setting using 1,130 teachers’ assessment of 440 school leaders’ expectations, archival data on test scores of 23,195 students, and detailed control for socioeconomic characteristics of parents and schools. We show that the associations between leader expectations and student performance are significant even when controlling for schools’ past performance and teachers’ own expectations. 2 INTRODUCTION Pygmalion effects are a kind of self-fulfilling prophesy in which leaders’ expectations regarding their followers increase the effort and performance of followers. In their seminal work Rosenthal and Jacobson (1968) showed that by experimentally giving teachers high expectations student achievement was increased. Later Eden (1984, 64) asked: `If raising teacher expectations enhances pupil performance, wouldn't raising manager expectations boost subordinate productivity?' A considerable number of studies have demonstrated this effect of manager expectations on subordinates. Meta-analyses of Pygmalion leadership effects show substantial and statistically significant effect sizes (McNatt 2000; Kierein and Gold 2000). In addition, more recent metaanalyses find that Pygmalion based leadership has the highest effect of three broad groups of leadership theories (Avolio et al. 2009) and that manager expectations are among the variables having most impact on leader-member exchange (Dulebohn et al. 2012). We contribute to this research by studying whether leader expectation effects extend beyond their immediate followers and affect the performance of the followers’ followers. We do so by going back to the setting of the Rosenthal studies and examine how school leaders’ expectations affect the performance of their teachers’ students. We also address some of the limitations that have been identified in existing research on Pygmalion leadership effects. Especially, there has been a call for work that does not involve deception of manager expectations. Also studies of established leader-follower relationships have been lacking (e.g., Eden et al., 2000, Kierein and Gold, 2000, and McNatt, 2000). Most Pygmalion studies involve experimental manipulation of leaders’ expectations. The strength of this method in terms of identifying causal effects is unquestionable. However, in natural settings where managers’ expectations regarding their subordinates evolve over time probably partly reflecting past performance, expectations may be less volatile and expectation effects therefore smaller (for a similar argument pertaining to teacher expectations see Jussim and Harber 2005). More generally, 3 since many studies were conducted by a small set of researchers in military settings “a greater variety of researchers, methods, settings, and so forth, would help demonstrate greater generalizability” (McNatt 2000, 320). In line with both Pygmalion theory (Bezuijen et al. 2009) and goal-setting theory (Locke and Latham 1990, 2002, 2006) there are indications that setting challenging goals is an important mediator for Pygmalion effects. We extend this line of research by studying the effect of high leader expectations when they are most difficult – namely for socioeconomically disadvantaged students that are known to have lower educational achievements, which would imply lower expectations. We find that high school leader expectations are especially beneficial to this group of followers’ followers. THEORETICAL EXPECTATIONS Leader expectations effects form a key element of several leadership theories, including the theoretical perspectives on charismatic leadership, transformational leadership, goal-setting, and more exclusively in Pygmalion leadership. According to charismatic leadership theorizing, charismatic leaders advocate an appealing vision with optimism and confidence and make emotional appeals to values. Followers tend to personally identify with the leader and imitate his behaviors (Conger and Kanungo 1987, Shamir et al. 1993). Transformational leadership theory shares some elements with charismatic leadership theory including the leader’s role in forming a vision, motivating subordinates and setting high expectations. The theory suggests that leaders can increase organizational performance by influencing values and aspirations, inducing a purpose transcending short-term goals, activating higher-order needs, and motivating people to move beyond their own self-interests (Avolio and Yammarino 2002, Bass 1985, 1990, Judge and Piccolo 2004, Yukl 2013). 4 Basically, transformational leadership is claimed to work by engaging peoples’ selfconcept by psychological mechanisms of personal identification, social identification, and value internalization. Leaders’ expression of high performance expectations is a key behavior whereby transformational leaders may engage peoples’ self-concept as such expectations contribute to enhance self-esteem, self-worth, and self-efficacy (Shamir et al. 1993). Closely related to transformational leadership theory, social cognitive theory (Bandura 1988, 1997) suggests that people may internalize a leader’s task performance expectations, i.e., integrate them into their own value hierarchies. Similarly, goal-setting research emphasizes the importance of goal content, particularly goal specificity and goal difficulty (Kanfer 1990, Locke and Latham 1990, 2006). Numerous studies show that specific and difficult goals may produce higher performance than easy or general goals: As long as the individual accepts the goal, has the ability to attain it, and does not have conflicting goals, there is a positive relationship between goal difficulty and task performance (Locke and Latham 2006). Goal-setting theory thus highlights the importance of providing individuals with a clear understanding of what is expected of them. Formulating performance expectancies is of vital importance because it facilitates the individual in focusing her efforts in a specified direction. Pygmalion Effects Studies In 1968 Rosenthal and Jacobson published their seminal work, Pygmalion in the Classroom: Teacher Expectations and Student Intellectual Development. In the experimental study, the authors administered an intelligence test to all children in Jacobson’s elementary school. However, they did not tell teachers it was an intelligence test, but explained it was a test for identifying children likely to “bloom” by showing a sudden and dramatic intellectual growth over the upcoming academic year. They then informed each of the teachers which students had been identified as “late 5 bloomers.” These students had in fact been randomly selected. One and two years later they administered the same intelligence test again. The only initial systematic difference between the controls and the “late bloomers” was in the teacher’s mind. Nevertheless, the designated “late bloomers showed IQ gains over the controls. The teachers’ false expectations had become true through a self-fulfilling prophecy. The effect was named “Pygmalion effect” after Shaw’s play, “Pygmalion,” which later gave inspiration to “My Fair Lady.” The teachers had subconsciously given the false “late bloomers” more attention by creating a friendlier and warmer climate, e.g. more smiling and by giving them more challenging assignments (White and Locke 2000). Pygmalion experimental research was extended to organizational settings focusing on manager-subordinate relationships first by King (1971, 1974) in industrial plants, and later by Eden in Israeli Defence Force training-courses (e.g. Eden 1992). Both authors found support for the Pygmalion leadership theory that inducing managers to have higher expectations for certain subordinates had self-fulfilling impacts on the performance of these subordinates. Some of the mediating variables could be identified from higher ratings of the instructors by the high-expectancy trainees in terms of the instructors’ efforts (1) to coach subordinates in effective work habits, (2) to stimulate enthusiasm for meeting a goal or achieving excellent performance, (3) to enhance others’ feelings of importance and self-worth, and (4) to encourage members to form relationships and work together as a team. Notably, the instructors were unaware of having treated the trainees differently (White and Locke 2000: 392). We note that some of these mediating variables—particularly stimulating enthusiasm and enhancing subordinates’ self-worth and self-efficacy—in combination with high leader expectations or goals form key elements of the above theories of charismatic and transformational leadership. However, later research by Eden and his colleagues (Eden et al. 2000) also demonstrated some limitations of the experimental Pygmalion approach. The induced manager expectations did not have the expected effects on workgroups that had been established for some time or on female 6 subordinates. However, in their meta-analysis Kierein and colleagues (2000) found the same impact of leader expectations for male and female subordinates (see also Natanovich and Eden, 2008). While the identified Pygmalion effects were based on deception—which is not a proper management strategy due to its ethical problems—Eden and his colleagues instead experimentally tested the role of training of managers in having higher expectations of their subordinates, however with no observable effects on the performance of subordinates. While the effect of leader expectations is well-documented and quite substantial, reviews of the research called for more studies of naturally occurring expectations in established leader-follower relationships (White and Locke 2000, McNatt 2000, Kierein and Gold 2000). More recently, studies have found Pygmalion effects in natural settings. In a study of 151 workplace leader-follower dyads in California Whiteley, Sy, and Johnson (2012) find that follower performance is higher when leader expectations are high. Previous Pygmalion and Pygmalion leadership research has shown that leaders’ or streetlevel bureaucrats´ (typically teachers’) expectations affect their immediate followers’ performance. However, the value of Pygmalion leadership theory would be much greater if leaders’ expectations affect not only the performance of their immediate followers but also the performance of more distant followers at lower levels of the organization. We have found no such research regarding Pygmalion leadership theory. When trying to understand how leaders may affect distant followers, we can draw on multilevel leadership effect theory by Waldman and Yammarino (1999) and Hannah et al. (2008). They claim that leaders can affect distant followers in two different ways. One is called the “falling dominoes” or “cascading” effect. These concepts refer to the step-by-step process when a leader’s leadership style is being reflected in a similar form at subsequent organizational levels. The leader’s leadership style is thus reflected in the leadership style of the immediate subordinate followers by some social learning process, including role-modeling, and that particular leadership style is then in turn reflected in the leadership style or performance of her 7 immediate followers at the next level (Bass et al. 1987, Waldman and Yammarino 1999, Avolio and Hannah 2008). In this situation, a leader is influencing distant followers indirectly through a multilevel hierarchical cascade of similar leadership styles. However, leaders may also affect distant followers by bypassing one or more hierarchical levels by influencing distant followers directly by skipping one or more of the hierarchical levels inbetween (Waldman and Yammarino 1999, Hannah et al. 2008). Unfortunately, only very few studies offer some empirical evidence of cascading effects/ and/or by-pass effects. These studies are focusing on charismatic leadership (Bass et al. 1987), transformational leadership (Dvir et al. 2002), and transactional leadership (Berson and Avolio 2004). Because of the very limited evidence of multilevel leadership effects, there has been a general call for more research on such relationships (e.g. Hannah et al. 2008, Yukl 2013). We would like to contribute by examining if the performance of distant followers is affected by Pygmalion leadership. Based on the substantial evidence on Pygmalion effects in street-level bureaucrat (teacher) – target group (students) relations as well as in leader-subordinate relations, we expect that the effect of leader expectations would extend beyond their immediate followers to the followers’ followers. This may happen through several channels. First, cascading, indirect effects may occur when leader expectations are raising followers’ expectations which in turn are raising the expectations or performance of followers’ followers. Second, leader expectation may also have a direct impact on distant followers’ performance by bypassing the immediate followers. Third, leader expectations may have other kinds of indirect effects on performance by affecting other aspects of followers’ behavior than the particular leadership style of communicating expectations, such as followers’ work effort including less absence (King 1971)). We know from several studies of Pygmalion leadership effects that high leader expectations increase followers’ effort (White and Locke 2000, Whiteley, Sy and Johnson 2012, McNatt 2000). And general work motivation literature suggests 8 that both how hard a person works and the direction that the person’s work efforts take have important performance consequences. Work effort is a behavioral force, which is driving the achievement of work-related goals and performance (Kanfer, Chen and Pritchard 2008; Latham 2007; Locke and Latham 2004; Pinder 1998; Porter, Bingley and Steers 2009). However, our research focus here is not the separate direct and indirect effects but the combined effects of leader expectations on the performance of followers’ followers. Accordingly, our first hypothesis is: H1: Leaders’ higher expectations of performance increase the performance of followers’ followers.A special aspect of Pygmalion theory – and as we shall see later also of Goal Setting Theory - is the role of difficult or ambitious goals in mediating or enhancing the impact of leader expectations. Based on Pygmalion studies, Rosenthal (1973: 1) identified teaching more and increasingly difficult material as one of the teacher behaviors that mediate the relationship between teachers’ expectations and students’ achievement. In their later meta-analysis Harris and Rosenthal (1985) found support for this supposition by finding positive relationships both between teachers’ or leaders’ expectations and presenting more and increasingly difficult material to the target group and between that behavior and outcomes. Extending this theorizing to Pygmalion leadership, Bezuijen et al (2009) in a recent study of 904 manager-subordinate dyads in six organizations find indications that Pygmalion effects are mediated by leaders setting more specific and difficult goals. Goal-Setting Theory The role of difficult goals in achieving better outcomes is also an important and more elaborated part of goal-setting theory. Locke’s (1968) pioneering work on goal-setting and motivation suggests that employees are motivated by clear goals and appropriate feedback. Basically, to work toward a goal is an essential source of motivation in reaching that goal—which, in turn, may improve performance. Locke and Latham (1990) reinforced the appropriateness of setting specific and difficult goals. Goal specificity may increase performance for two reasons: First, specific goals 9 serve to focus attention, i.e., they let employees know what he or she is expected to do and accomplish. Second, specific goals can focus effort, i.e., make it easier for employees to understand the relationship between effort and performance and between performance and subsequent rewards (Locke and Latham 1990; Wright 2004). The value of goal setting is widely recognized and has recently been shown to affect performance in an educational setting (Favero, Meier and O’Toole 2014). Goal difficulty is particularly important, as (a) people are often motivated by achievement, and because (b) people value a given goal based on the significance of the anticipated accomplishment. Inasmuch as goal accomplishment is realistic, the accomplishment of a challenging goal may thus entail a psychological needs satisfaction that, in itself, may boost and direct enthusiasm and work efforts toward goal accomplishment. Basically, goals set the primary standard for self-satisfaction with performance (Locke and Latham 2002, 2006). Low or easy goals may therefore be less motivating than high or challenging goals because they require a person to attain less in order to be satisfied. For these reasons, we derive our second hypothesis: H2: The positive influence of leader expectations on the performance of followers’ followers is enhanced by the difficulty in meeting the expectations for selected target groups. CONTEXT The setting for our study is the management of public primary and lower secondary schools in Denmark. These schools are governed by multi-purpose municipalities. Danish schools provide a “tough” test of the performance impact of leaders setting goals or expectations. Professionals—and not least teachers—probably tend to seek some level of autonomy in their work but this tendency may be particularly strong in Denmark. First, Denmark has a national culture characterized by a relatively small power distance (Hofstede 1980, 1983). Compared to other countries, Danes generally have less respect for authority. Therefore, teachers in Denmark may have less respect for principals than teachers in some of the Anglo-Saxon countries, where "instructional leadership" has 10 proven effective (Robinson et al. 2008, 2009). By the same token, Danish students and their parents may have less respect for their teachers and principals than would be the case in other countries. Second, Danish school principals have less decision-making autonomy than their peers in many other countries—due to a considerable degree of corporatism in policy formulation and implementation. In the area of educational policy, the teachers’ unions are thus normally involved in the formulation of most laws on schooling and very much so in the implementation of policy at the local level. For long, school principals’ authority in allocating teacher’s working hours at the school has been substantially reduced by collective agreements with the teacher unions.1 Principals’ authority is also limited by substantial influence from local branches of the teacher union or the teacher shop steward at each school (Meier et al. 2015). Thirdly—and likely as a natural extension of the small power distance and the corporatist tradition—Danish schools have had a tradition of less "strong" school leaders. Principals were often perceived as a primus inter pares, once also indicated by the term “first teacher” (head teacher). Moreover, teachers traditionally have had a high degree of autonomy. Nowadays, however, school principals can also find some support for their managerial role in the Public School Act, article 45, according to which the school manager is responsible for the administrative and educational management of the school. Variation in school managers’ involvement in setting goals and expectations presupposes that they (and their school) have some level of autonomy in decision-making. As previously mentioned, public schools are governed by multi-purpose municipalities. Within a few general constraints, the municipal council is free to decide the annual school budgets and is authorized to impose taxes on the municipal inhabitants within certain overall limits, negotiated with the national government. Local politicians can set strategic goals for the municipal schools, and some take 11 As an exception from the rule, the existing collective agreements on teacher’s working hours was cancelled by Parliament in 2013 after the Danish teachers’ unions refused to accept proposals of doing so from the central and locl government employer organizations. 11 initiatives in terms of regulating the teaching methods that are applied in the schools and personnel matters. The Danish national school legislation regulates the themes to be taught in each grade but gives municipalities and their schools substantial autonomy with respect to objectives and choice of pedagogical methods. In general, most school principals report having substantial autonomy in managing their school (Pedersen et al. 2011). Students’ parents are represented in the school board of each school but students as well as their parents also play important roles in coproducing education in primary and lower secondary education. The school managements inform students and parents in various meetings as well as through intranet communication. DATA AND METHODS Data We use two data sources. The first source comprises a survey among Danish and Math teachers in 9th grade in public schools. Their responses to a teacher survey, which was administered in 2011, form the basis for characterizing school managements. Specifically, a web-based questionnaire was sent to 1,998 teachers who taught a ninth grade class in Danish or Math in a municipal lower secondary school. A total of 1,027 teachers from 337 municipal schools responded (57 percent response rate). The distribution of the schools of the responding teachers does not differ significantly from the population of Danish schools in terms of student performance, socioeconomic student composition, and school size (analysis based on administrative register data). Information on school leader practices for quantitative analyses can be provided through either manager or teacher surveys. While our research project collected information from both managers and teachers, this article uses teacher survey information. Leader surveys may be more 12 susceptible to a social desirability bias (Jacobsen and Andersen 2015, Favero et al. 2015), and teachers are more likely to be influenced by how they perceive the behavior of their leader rather than by how the leader is presenting himself.2 Our second data source comprises archival data from administrative registers on the population of Danish lower secondary students (provided by Statistics Denmark). We merge the survey data from individual teachers with register data from their students. In this way, in our analyses we have data on 13,568 students’ test scores at the final exams at the end of ninth grade in 2011 as well as rich demographic and socio-economic information on all students and their family. Measures Dependent variables The main dependent variable in the analyses is student performance, measured by students’ test scores in summer 2011. Students in lower secondary municipal schools are required to take a final test at the end of 9th grade, which is a standardized academic subject test. All schools are required to use the same examination tests, issued by the Ministry of Education. Grades are given on a 7-point scale. Written examinations are graded by the students’ subject teacher and an external examiner appointed by the Ministry of Education (Departmental order no. 351, 19 May 2005, section 24). This procedure arguably makes the scores relatively valid and reliable. For each student, we use the mean score in the written tests in the two subjects of Danish (understanding of reading, spelling, and written presentation) and Mathematics (problem-solving and arithmetic). We use the written, open-ended (not multiple choice) math exams, because they are graded relatively objectively. We 2 In fact, we also examined the performance impact of manager reported management expectations for performance. The relationship is insignificant in similar fixed effect regression analyses at the municipal level. This manager reported expectation variable was created as the mean score of two items in which respondents were asked to rate their degree of agreement with either of two opposing statements A and B. The first set of statement was: A. “I and the other managers expect that students at our school perform better in marks at the final exams than similar students at other schools.” “B. I and the other managers have no expectations with regard to students’ performance in marks at this school compared to similar students at other schools.” The second set of statements was: A. “The school management places very high demands on teachers’ classroom teaching at our school.” B. “The school management hardly places any demands on teachers’ classroom teaching at our school. It is their own responsibility.” Scores were reversed to indicate higher scores for higher expectations. 13 use the written exam results in both Danish and math to get a broader measure encompassing both humanistic and scientific core competencies. Most school management research use test scores in Math and the national language and literature for measuring student performance (Hattie 2009; Robinson 2009). Our test score data set, comprising the mean of Danish and math test scores, provides 13,568 students’ mean test scores in Danish and math that is our unit of analysis. The 13,568 sample students are distributed around a mean score of 6.20 in Danish and 6.35 in mathematics (with a standard deviation of 2.58 in Danish and 3.24 in mathematics) on the Danish 7-point grading scale. The grading scale ranges from -3 to 12. The mean standard deviation for the two subjects is 2.91. To obtain a higher degree of normal distribution, the total distribution of the two sets of tests scores is standardized (mean = 0, standard deviation = 1). Leaders’expectations For measuring our key explanatory variable, school leaders’ expectations regaring students’ academic performance, we use two items from the teacher survey and form an index on teachers’ perceptions of the school leadership at their school. The first item is: "How would you describe the management of the school on a scale from 1 (totally agree with A) to 5 (totally agree with B)?” where (A) is “The school management expects students to do better in terms of test scores compared to similar students at other schools” and (B) is “The school management has no expectations about how students are doing in terms of test scores compared to similar students at other schools." The second item is: "The school management has high expectations regarding students' academic level" on a Likert scale from 1 ("Strongly agree") to 5 ("Strongly disagree"). Cronbach's alpha of items is .75. The index is generated using the sum of the two item scores. The index score is subtracted by one to form a scale ranging from 1 to 9, with high scores indicating that school management has high expectations for student performance. Control variables and interaction terms 14 To control for other variables potentially affecting both leader expectations and student performance, we include administrative register data on many relevant background variables. First, we include information about student gender, ethnicity, and age at the time of testing. Second, we control for the following family background variables: living with both parents, number of siblings and birth order, mother and father’s age at birth, mother and father’s length of education, mother and father’s disposable income, and whether mother and father are currently employed. These variables have a strong influence on student performance, explaining about 22 percent of the variation (Andersen and Winter 2011). In some of our analyses, we examine the conditioning effect of students’ individual socioeconomic background on the relationship between school leader or teacher expectations and student performance. We do so by employing an interaction term in the regression models. The interaction term is defined as the product of the individual student's socio-economic status and leader expectations as these are recorded by each teacher. This approach has been used in previous studies of the importance of social inheritance (Andersen 2008a; 2008b). We form an overall measure of student socioeconomic status (SES) using the following four indicators: - Mother’s length of education (years) - Father’s length of education (years) - Mother’s income (in 15 percentiles) - Fathers’ income (in 15 percentiles) SES is measured as the first extracted factor of a Principal-Component Analysis of the four variables (eigenvalue = 2.01). The first factor explains 50 percent of the overall variation, and the four variables show satisfactory factor loading scores. Cronbach’s alpha for the four items is 0.62. To ensure comparability between analyses with and without the interaction terms, we use the SES- 15 variable in all analyses, instead of the individual variables comprising the SES-measure. We use the SES-variable in addition to the other individual student and family background variables. Several studies suggest that students’ performance is affected by peer group effects, particularly class-mate peer effects (Rangvid 2008; Hanushek 2006; Somers, McEwan and Willms 2004). Moreover, research finds that school leader behaviors as well as teachers’ teaching are affected by the social background of the students (Pedersen et al. 2011). Therefore, controlling for the social background of the students in the class is important when examining the impact of leader and teacher expectations on student performance. We include the following six variables to control for student peer effects: Share of mothers and fathers in the lowest income quartile, share of mothers and fathers in the lowest length of education quartile, share of students not living with both parents, and share of descendants of parents with a non-Danish origin. Several studies show that teachers’ teaching experience affects student performance (Mikkelsen 2013; Clotfelter, Ladd and Vigdor 2007; Rivkin, Hanushek and Kain 2005). Also, and although the international evidence on the influence of teacher gender is mixed, a recent Danish study finds a relationship between teacher gender and student performance as well as between teacher gender and teaching behaviors (Mikkelsen 2013; Andersen 2013b). Therefore, we include controls for teacher experience and gender. In addition, we include measures on the number of students in the class, and school size (the number of students at the school) as these factors may affect teaching as well as student performance. To control for potential reverse causality bias, namely that leaders express high expectations because the students at the school in general are talented, we control for student past-performance— measured as the average of nine grade students’ test scores in Math and Danish in the school year 2008-2009. We choose the school year 2008-2009, instead of, say, 2009-2010, since actual leader alignment of expectations to past performance may require some time. 16 Finally, in order to make sure that principals’ expectations are not completely driven by teachers’ expectations , we control for teacher expectations. For measuring teacher expectations to student performance, we use the teacher response to the following survey item: ”The following questions concern your teaching in your subject [Danish/Math] in your ninth grade. How would you describe your teaching in your subject [Danish/Math] in your ninth grade on a scale from 1 (fully agree with A) to 5 (fully agree with B) where 3 is neutral?” A’s statement is: “I express high expectations to my students’ academic performance” while B’ statement is: “I do not express high expectations to my students’ academic performance”. We reverse the item responses to let higher scores indicate higher expectations. Due to a skew distribution (few observations expressing low expectations), the reversed values 1 and 2 have been merged to 1. Controlling for teachers’ selfreported expectations provides a stronger, more conservative test for two reasons. First, we thereby account for the indirect cascading effect on performance that leaders may have through their influence on teachers’ expectations. Second, because social desirability bias might make some respondents overstate their optimism, controlling for teachers’ self-reported expectations in their evaluation of leader expectations helps neutralize this risk of bias. Descriptive statistics for all variables in the analysis are presented in Table 1. 17 Table 1: Descriptive Statistics Variables Test score1 Leaders’ expectations to students’ performance Obs Mean Std. Dev. Min Max 13,568 .07 .98 -2.61 1.96 754 5.63 1.79 1 9 SES 13,568 3.98 1.00 .83 7.23 Male 13,568 .51 .50 0 1 Immigrant 13,568 .07 .26 0 1 Age at the time of the test 13,568 16.10 Not living in nuclear family 13,568 .18 .39 0 1 Only child 13,568 .07 .25 0 1 First born (ref.) 13,568 .36 .48 0 1 Second born 13,568 .17 .38 0 1 Youngest child 13,568 .39 .49 0 1 Mother’s age when child was born 13,568 28.42 6.58 0 46 Father’s age when child was born 13,568 30.35 8.64 0 67 Mother employed (ref.) 13,568 .86 .34 0 1 Mother unemployed 13,568 .07 .25 0 1 Mother in education 13,568 .00 .06 0 1 Mother outside workforce Father employed (ref.) Father unemployed 13,568 .04 .88 .19 .32 0 0 1 1 .06 .23 0 1 Father in education 13,568 .00 .01 0 1 Father outside workforce 13,568 .04 .20 0 1 Share of mothers in first quartile of education 13,568 .23 .13 0 .91 Share of fathers in first quartile of education 13,568 .20 .12 0 1 Share of mothers in first quartile of income 13,568 .25 .13 0 1 Share of fathers in first quartile of income 13,568 .23 .13 0 1 Peers’ proportion of non-Danes 13,568 .09 .13 0 1 Share of peers’ who are not living in nuclear family 13,568 .21 .12 0 .72 Class size 715 19.70 3.78 1 38 School size 336 507.92 163.11 140 1005 Teacher’s gender (male) 1027 .41 .49 0 1 Teacher’s teaching experience 1027 12.91 12.84 0 43 Student background: 13,568 13,568 .38 14.75 17.92 Peers: Class, School, and teacher controls: Past performance 331 6.23 .92 2.75 8.52 Teachers’ expectation to students’ performance 824 3.22 .81 1 4 1. The standardization of marks was performed for all tested students for all schools represented in the teacher survey while the listed descriptive statistics is made only for those students who are part of the analyses of the paper. 18 Statistical Modeling Our statistical model specifications must account for the hierarchical, multi-level character of our data, with students in classes in schools in municipalities. Our models are fixed effects models at the municipal level3. As leader expectations are measured by teacher perceptions of leader expectations, a school fixed effects approach would not be appropriate for examining the relationship between school leader expectations and student performance. Because teachers in the same school are observing the same leader, the teachers’ perceptions of leader expectations are likely to be characterized by limited empirical variation. Therefore, we use municipal fixed effects models for these analyses. Cluster robust standard errors are used to account for spatial autocorrelation in the hierarchical data set. In addition to modeling interaction effect of students’ SES on the relationship between expectations and performance, we use marginal effect analyses for quintiles of the SES-distribution. FINDINGS Table 2 shows the findings in relation to hypothesis H1. The results in all three models indicate that high leader expectations are positively associated with students’ test scores (even with control for past organizational performance) thus supporting H1. Moreover, Model 3 rejects that the impact of leader expectations to target group performance is fully mediated by teacher expectations as would be the case if leaders’ expectations only affected student performance through cascading Pygmalion leadership effects by the leaders’ expectations being reflected in teachers’ expectations. One interpretation is that high leader expectations have a positive direct impact on students’ test scores thus bypassing mediation through teacher expectations. However, we cannot exclude the possibility that leader’s expectations are mediated by other kinds of teacher behavior than their expectations, for example their effort. Hence, the finding supports H1: that leaders’ communication of higher task 3 In addition to the fixed effects models, as a kind of robustness check we conducted OLS modeling with robust standard errors clustered at the municipal level. These studies are not shown based on the stronger method of municipal fixed effects. The findings are similar to those presented. 19 performance expectations increases the performance of followers’ followers. Leaders signaling high expectations may influence target groups as well indirectly though cascading leadership effects reflected in teachers’ expectations and perhaps other teacher behaviors, as directly—here students and their parents. In addition, the findings indicate that the effect of teacher expectations is larger than the effect of leader expectations, even though teachers’ self-reported expectations are possibly less valid than their evaluation of their leaders’ expectation due to a possible social desirability bias when reporting their own expectations. Moreover, in line with expectations from both Pygmalion and goal-setting research, we test whether the relationship between leaders’ communication of higher task performance and the performance of followers’ followers is moderated by the difficulty in meeting leaders’ performance expectations (H2). Accordingly, we expect that the positive influence of leader expectations is enhanced by task difficulty. Our indicator for task difficulty is students’ socio-economic background, as students with low socio-economic background tend to perform poorer in school. In other words, we expect to find a negative effect of the interaction term between students’ SES and leader expectations. Table 3 shows the results of the interaction analyses. In line with our expectations, the impact of leader expectations on student performance appears to be moderated by student SES. The coefficient of the interaction variable between leaders’ expectations and SES is negative and significant. Thus, the positive impact of school leaders’ high expectations is greater among students with a lower socio-economic status. This applies in models with as well as without control for teacher expectations – and in models with as well as without control for past performance. 20 Table 2: Impact of leader expectations on student test scores. Unstandardized coefficients with standard errors in parentheses Variables Leaders’ expectations Past performance (2008-2009) Municipal fixed effects Controlling model 1 for past performance Controlling model 2 for teacher expectations Model 1 Model 2 Model 3 .016** .014* .010† (-.005) (-.006) (-.006) - .064* .061* (.025) (.024) .055** Teacher expectations (-.013) Student background: SES Male Immigrant Age at the time of the test Not living in nuclear family Only child First born (ref.) Second born Youngest child Mother’s age when child was born Father’s age when child was born Mother employed (ref.) Mother unemployed Mother in education .307** .304** .303** (-.014) (-.014) (-.014) -.150** -.151** -.152** (-.02) (.020) (.021) -.137** -.133** -.136** (-.036) (.036) (.036) -.243** -.243** -.242** (-.023) (.023) (.023) -.056* -.055* -.059* (-.024) (.024) (.024) -.029 -.030 -.029 (-.035) (.035) (.034) - - - -.190** -.189** -.186** (-.023) (.022) (.022) -.231** -.231** -.229** (-.019) (.019) (.019) .013** .013** .013** (-.002) (.002) (.002) .005* .005* .005* (-.002) (.002) (.002) - - - -0.056 -.058 -.061 (-.037) (.037) (.037) .231* .232* .227* (-.112) (.110) (.111) Table will continue on next page 21 Table 2 (continued): Impact of leader expectations on student test scores. Unstandardized coefficients with standard errors in parentheses Model 1 Controlling model 1 for past performance Model 2 Controlling model 2 for teacher expectations Model 3 .026 .022 .023 (-.049) (.049) (.050) Municipal fixed effects Variables Mother outside workforce Father employed (ref.) Father unemployed Father in education Father outside workforce - - - -0.04 -.042 -.038 (-0.04) (.040) (.039) 1.793** 1.767** 1.768** (-.156) (.142) (.115) -.162** -.163** -.161** (-.05) (.051) (.051) Peers: Share of mothers in first quartile of education Share of fathers in first quartile of education Share of mothers in first quartile of income Share of fathers in first quartile of income Peers’ proportion of non-Danes Share of peers’ who are not living in nuclear family -.212 -.196 -.159 (-.142) (.147) (.145) -.214* -.165 -.175† (-.1) (.101) (.097) .004 .020 .008 (-.118) (.113) (.110) .062 .072 .110 (-.114) (.114) (.114) .013 .101 .087 (-.138) (.131) (.132) .044 .070 .025 (-.092) (.093) (.096) 0 -.000 -.001 (-.004) (.004) (.004) .000** .000 .000 Class, school and teacher controls: Class size School size Teacher’s gender (Male) Teacher’s teaching experience Constant Observations Number of schools Adj. R2 (.0) (.000) (.000) -.009 -.008 -.001 (-.023) (.023) (.023) .001 .001 .001 (-.001) (.001) (.001) 2.305** 1.930** 1.789** (-.374) (.416) .422 13,568 330 13,568 330 13.549 324 .19 .19 .19 Cluster robust standard errors in parentheses ** p<.01, * p<.05, † p<.10 22 Table 3: Impact of leader expectations on student test scores conditioned by SES and controlling for teacher expectations and past performance. Unstandardized coefficients with standard errors in parentheses Variables Leaders’ expectations Past performance (2008-2009 Teachers' expectations Leaders' expectations*SES Student background: SES Male Immigrant Age at the time of the test Not living in nuclear family Only child First born (ref.) Second born Youngest child Mother’s age when child was born Father’s age when child was born Mother employed (ref.) Mother unemployed Mother in education Mother outside workforce Father employed (ref.) Father unemployed Father in education Father outside workforce Peers: Municipal fixed effects Model 1 .056*** (.020) Controlling Model 1 for teacher expectation & past performance -.010** (.005) Model 2 .051** (-0.02) .061** (.024) .055*** (.013) -.010** (.005) .364*** (.031) -.150*** (.021) -.138*** (.036) -.244*** (.023) -.055** (.024) -.03 (.035) -.189*** (.023) -.231*** (.019) .013*** (.002) .005*** (.002) -.055 (.037) .232** (.111) .030 (.049) -.040 (.040) 1.795*** (.158) -.162*** (.050) .361*** -.031 -.152*** (.021) -.137*** (.036) -.242*** (.023) -.058** (.024) -.031 (.034) -.185*** (.022) -.229*** (.019) .013*** (.002) .005*** (.002) -.060 (.037) .229** (.110) .027 (.049) -.038 (.039) 1.770*** (.117) -.160*** (.051) - Table will continue on next page 23 Table 3 (continued): Impact of leader expectations on student test scores conditioned by SES and controlling for teacher expectations and past performance. Unstandardized coefficients with standard errors in parentheses Variables Share of mothers in first quartile of education Share of fathers in first quartile of education Share of mothers in first quartile of income Share of fathers in first quartile of income Peers’ proportion of non-Danes Share of peers’ who are not living in nuclear family Class, school and teacher controls: Class size School size Teacher’s gender (Male) Teacher’s teaching experience Constant Observations Number of municipalities Adj.R2 Cluster robust standard errors in parentheses *** p<.01, ** p<.05, * p<.10 Municipal fixed effects Controlling Model 1 for teacher expectation & past performance Model 1 -.207 (.141) -.214** (.099) .008 (.118) .061 (.115) .011 (.138) .046 (.093) Model 2 -.154 (.144) -.175* (.097) .012 (.109) .109 (.115) .084 (.133) .027 (.097) -.000 (.004) .000** (.000) -.008 (.023) .001 (.001) 2.078*** (.401) 13,568 80 .19 -.001 (.004) .000 (.000) .000 (.022) .001 (.001) 1.560*** (.441) 13,549 80 .19 Hypothesis 2 is thus supported. In line with goal-setting theory and some Pygmalion research, this result supports the notion that low or easier-to-achieve goals are less motivating than high and challenging goals. Table 4 shows the results of the marginal effect analysis, demonstrating similar heterogeneous effects of high leader expectations to student performance as above. The positive impact of high expectations is significant for students in the three lowest SES-quintiles, but non- 24 significant for students in the two highest quintiles. Thus, high expectations appear to be particularly efficient in fostering better outcomes among more socially-disadvantaged students. Table 4: Marginal effects of leader’s expectations on student test scores conditioned by students’ socio-economic background with and without control for teacher expectations and past performance. quintile 1st quintile 2nd quintile 3rd quintile 4th quintile 5th quintile Marginal effects (municipal fixed effects) dy/dx (std. err.) With control for teacher expectation and past performance dy/dx (std. err.) .025** (.007) .019** (.006) .014* (.006) .008 (.006) -.007 (.013) ** p<.01, * p<.05, † p<.10 .019** (.007) .014* (.006) .009 (.006) .003 (.007) -.013 (.013) We diminish the risk of reverse causality bias by including a control for past performance. Moreover, the moderation test findings (i.e., the test of H2) speak against such bias. In brief, the findings appear counter-intuitive to what we would expect if the leader expectation-student performance relationship was fully driven by reverse causality (i.e. high performance leading to high expectations). We find a positive association between leader expectations and student performance, which is especially predominant among low-SES students. Because the accomplishment of high performance expectations among leaders and teachers is more difficult in the contexts of low-SES students—we expect that reverse causality bias would work by lower social status among students decreasing both performance and expectations to performance. In fact this is what typically happens as seen in table 3 and table 4 for the impact of SES on student performance, and in separate analyses (not shown here) we also find that high student social status 25 generally tends to foster higher leader expectations. Moreover, we expect that any reverse causation effects would engender a stronger expectation-performance relationship among high-SES students. To the contrary, we find this relationship for low-SES students. The implication is that in general students’ low social status tends to decrease leader expectations to their performance (not shown here) while in fact high leader expectations – when practiced - seem to increase the performance of particularly low social status students. Thus, although we cannot reject the possible issue of reverse causation, our moderation result helps to minimize this concern. CONCLUSION Since Rosenthal’s pioneering studies of Pygmalion effects showing that teachers’ high expectation increase the achievement of their students, numerous experimental studies have showed that also managers’ expectations for followers increase followers’ performance and efforts. We extend this research by showing that leaders’ expectations even have impacts across multiple hierarchical levels, by affecting the performance of followers’ followers. Thus, school principals’ communicating high expectations have impacts beyond their teachers by increasing students’ achievements. As expected by goal setting theory and some Pygmalion leadership research this effect is strongest when the expectations are most difficult to meet—in this case for the performance of socioeconomically disadvantaged students. Leaders’ expectation may affect distant followers’ performance though several channels. One is indirect by cascading leadership effects when the leaders’ expectations are reflected in the leadership style of the immediate subordinate, here in teachers setting high expectations to their students, which make the latter perform better. However, the impact of leadership expectations does not exclusively work this way as we show that leader expectations are associated with student achievement even when controlling for teacher expectations. Accordingly, some of the leader expectation impact on student achievement must work through other indirect or direct channels. One of several potential other indirect channels might be subordinate front line staffs’ efforts, which 26 in several Pygmalion leadership studies have been shown to be increased by leaders setting higher expectations. And as shown in work motivational research greater efforts among workers are driving goal-achievement and performance. However, leader expectations might also affect distant followers directly by bypassing one or more hierarchical levels. Leaders with high expectations could also influence target groups—e.g. students and their parents—directly by communicating high expectations to them. Target groups, as in this case students and their parents, have important roles in coproducing public services (Ostrom 1996, Bovaird 2007, McCulloch 2009). Leaders’ signaling high expectations for target group performance could be a crucial way of managing coproduction among target groups. With the present research design, we cannot disentangle these indirect cascading and other indirect effects and the direct effects of leader expectation on distant followers other than stating that the impact is not completely driven by cascading effects. Clearly, the self-reinforcing nature of our theorizing advises some caution in the causal interpretation of our statistical findings. High performance is likely to lead to higher expectations (Jussim and Harper 2005). While non-experimental studies are inherently challenged in terms of establishing hard evidence in this area, our data and design strategy offer several strong-points—in turn making our findings contribute to existing empirical knowledge. First, we do not use self-reported performance measures (which might have led to some common source bias), but apply students’ test scores in national tests that are graded by external examiners and thus provide more objective performance measures (Meier and O’Toole 2013, Meier et al. 2015). Second, we do not use self-reported leader expectations (which might have led to some social desirability bias), but apply teachers’ assessments of their leaders’ expectations regarding task performance. By the same token, teachers are likely to be more influenced by how they perceive their leader’s behaviors than by how leaders prefer to present themselves (Favero et al. 2015, Jacobsen and Andersen 2015). 27 Third, when examining the relation between leader expectation and target group performance, we control for teachers´ expectations to their students’ performance to make sure that leaders’ expectations are not merely reflecting those of the teachers. This control also makes it more probable that a positive relationship between leader expectations and target group performance is not merely caused by cascading effects by teachers mediating leader expectations to the performance of target groups. Leader expectations are also affecting the performance of target groups by other indirect factors or directly. Fourth, using extraordinarily detailed archival register data, we control for many aspects of students’ social and economic background, which are known to be highly correlated with student performance and may also affect leaders’ and subordinates’ expectations. Fifth, we use municipality fixed effects to control for unobserved heterogeneity at the municipality level. Sixth, we control for past performance at the school level in order to diminish the risk of adverse causality. This would be the case if the positive relationship between leader expectations and target group performance is merely reflecting that high target group performance is increasing leaders’ expectations (Jussim and Harper 2005). Finally, we find strongest associations between leader expectations and target group performance for the most disadvantaged group of students. If students’ previous high performance was driving leader expectations upward, we would expect to find the strongest associations between advantaged (hence high-performing) students and expectations among leaders and subordinates. We find the opposite, which in turn increases the confidence in our causal interpretation of the results. In sum, the findings presented in this paper support the theoretical notion that the performance of distant followers – and here organizational performance at the delivery level increases when public managers are signaling high performance expectations to followers and followers’ followers. 28 The research on Pygmalion effects and Pygmalion leadership effects has been dominated by experimental research designs in which teachers’ or leaders’ expectations are manipulated with false information on the virtues of some of their followers. The strength of this method in terms of identifying causal effects is unquestionable. However, the deception used raises some ethical concerns. It also fails to take into account that leader-follower relationships typically evolve over time with expectations affecting follower performance, and follower performance affecting leader or teacher expectations (Jussim and Harper 2005). Therefore, there has been a call for studies of established leader-follower relationships in more natural contexts (e.g., Eden et al., 2000, Kierein and Gold, 2000, and McNatt, 2000). This study makes a contribution towards meeting this call. While research on Pygmalion effects has predominantly been performed in educational settings, a few military settings have dominated Pygmalion leadership research. Thus our study of Pygmalion leadership effects is an extension of that leadership research into the field of education indicating that Pygmalion leadership effects are relevant in other settings. Public school principals in Denmark may have less decision-making authority compared to other countries—due to a corporatist tradition with strong influence from teachers unions and shop stewards (Meier et al. 2014). According to cultural research by Hofstede (1980, 1983), Denmark is also an extreme case of small power distance with little respect of authorities. In this perspective, our case choice constitutes a relatively “tough” test of the impact of leader expectations on performance, implying that our findings are likely to apply also in other settings. It is though worth considering if leader expectations might have the greatest impact on performance in public service organizations where there is a considerable goal-consensus between the organization and the target groups. However, signaling expectations also seems to have an important role in regulatory enforcement (Winter and May 2001). Our research points to some issues calling for future research. For example, what interventions may actively increase leaders’ expectations? Experimental research by Eden and 29 colleagues (2000) on the role of training as an instrument to increase leader expectations did not offer any answer to this question. Yet, as indicated by White and Locke (2000), there may be other experimental as well as observational ways to study the role of training and other instruments for increasing leader expectations. We also need studies that examine the extent to which and by what mechanisms leader expectations directly or indirectly are affecting distant followers, including target groups. There is a lack of research on multilevel leadership not only within Pygmalion leadership theory but also more generally in leadership and management research (Hannah et al. 2008, Yukl 2013). While a substantial amount of research in the last 15 years has demonstrated that management and leadership matter for organizational performance, we do not know much about the mechanisms by which they matter. Having high expectations for clients who are already performing well may not be very difficult. But to have high expectations for target group clients with the least likelihood of high performance is something different. Some leaders of public organizations seem, however, to do so pretty well. 30 REFERENCES Andersen, Ida Gran. 2013a. Undervisningsformer og –metoder. In Lærere, undervisning og elevpræstationer i folkeskolen, edited by Søren C. Winter and Vibeke Lehmann Nielsen, 10536. Copenhagen: SFI – The National Danish Centre for Social Research, 13:09. ______ . 2013b. Undervisningsstrategier. In Lærere, undervisning og elevpræstationer i folkeskolen, edited by Søren C. Winter and Vibeke Lehmann Nielsen, 137-54. Copenhagen: SFI – The National Danish Centre for Social Research, 13:09. Andersen, Simon C. 2008a. The impact of public management reforms on student performance in Danish schools. Public Administration 86: 541-58. ______ . 2008b. Private schools and the parents that choose them. Scandinavian Political Studies 31: 44-68. Andersen, Simon C. and Søren C. Winter, eds. 2011. Ledelse, læring og trivsel i folkeskolerne. Copenhagen: SFI – The National Danish Centre for Social Research 11: 47. Ashforth, Blake E., and Fred Mael. 1989. Social identity theory and the organization. Academy of Management Review 14: 20-39. Avolio, Bruce J. and Francis J. Yammarino. 2002. Transformational and charismatic leadership: The road ahead. New York: Erlbaum. Avolio, Bruce J. and Sean T. Hannah. 2008. Developmental Readiness: Accelerating Leader Development. Consulting Psychology Journal: Practice and Research 60(4): 331–47. Avolio, Bruce J., Fred O. Walumbwa, and Todd J. Weber. 2009. Leadership: Current Theories, Research, and Future Directions. Annual Review of Psychology 60: 421-449. 31 Bandura, Albert. 1988. Organizational Application of Social Cognitive Theory. Australian Journal of Management 13: 275–302. ______ . 1986. Social Foundations of Thought and Action. Englewood Cliffs, NJ: Prentice-Hall. ______ . 1997. Self-Efficacy: The Exercise of Control. New York, NY: W. H. Freeman and Company. Bass, Bernard M. 1999. Two Decades of Research and Development in Transformational Leadership. European Journal of Work and Organizational Psychology 8: 9–32. ______ . 1998. Transformational Leadership: Industry, Military, and Educational Impact. Mahwah, NJ: Erlbaum. ______ . 1995. Theory of Transformational Leadership Redux. Leadership Quarterly 6: 463–78. ______ . ed. 1990. Bass & Stogdill’s handbook of leadership. New York: Free Press. ______ . 1985. Leadership and performance beyond expectations. London: Free Press. Bass, B. M., Waldman, D. A., B.J. Avolio and M. Bebb. 1987. Transformational leaders: The falling dominoes effect. Group and Organization Studies, 12, 73−87. Berson, Y., and B. Avolio. 2004. Transformational leadership and the dissemination of organizational goals: A case study of a telecommunication firm. The Leadership Quarterly, 15, 625−46. Bezuijen, Xander M., van den Berg, Peter T., and van Dam, Karen. 2009. Pygmalion and Employee Learning: The Role of Leader Behaviors. Journal of Management 35(5): 1248-1267. Bovaird, Tony. 2007. Beyond engagement and participation: user and community coproduction of public services. Public Administration Review 67: 846-60. 32 Boyne, George A. 2003. Sources of public service improvement: a critical review and research agenda. Journal of Public Administration Research and Theory 13: 367-94. Boyne, George A., Kenneth J. Meier, Laurence J. O’Toole Jr., and Richard M. Walker, eds. 2006. Public service performance: Perspectives on measurement and management. New York: Cambridge University Press. Burns, James M. 1978. Leadership. New York: Harper & Row. Clotfelter, C.T., H.F. Ladd, and J.L. Vigdor. 2007. How and why do teacher credentials matter for student achievement? National Bureau of Economic Research Working Paper Series, 12828. Conger, J.A. and R.N. Kanungo. 1998. Charismatic leadership in organizations. Thousand Oaks, CA: Sage Publications. DeGroot, Thomas, D. Scott Kiker, and Thomas C. Cross. 2000. A meta-analysis to review organizational outcomes related to charismatic leadership. Canadian Journal of Administrative Sciences 17: 356-71. Dulebohn, James H., Bommer, William H. and Liden, Robert C. 2012. A Meta-Analysis of Antecedents and Consequences of Leader-Member Exchange: Integrating the Past With an Eye Toward the Future. Journal of Management 38(6): 1715-59. Dvir, T., D. Eden, B.J. Avolio and B. Shamir. 2002. Impact of transformational leadership training on follower development and performance: A field experiment. Academy of Management Journal 45, 735−44. Eden, D. 1984. Self-fulfilling prophecy as a management tool: Harnessing Pygmalion. Academy of Management Review, 9(1): 64-73. 33 Eden, D. 1992. Leadership and Expectations: Pygmalion Effects and other Self-Fulfilling Prophecies in Organizations. Leadership Quarterly 3(4), 271–305. Eden, D., D. Geller, A. Gewirtz, R. Gordon-Terner, I. Inbar, M. Liberman, Y. Pass, I. SalomonSegev, and M. Shalit, M. 2000. Implanting Pygmalion leadership style through workshop training: Seven field experiments. The Leadership Quarterly 11: 171–210. Favero, Nathan, Kenneth J. Meier, and Laurence J. O’Toole, Jr. 2014. Goals, trust, participation, and feedback: Linking internal management with performance outcomes. Journal of Public Administration Research and Theory, Advance Access: doi:10.1093/jopart/muu044. Favero, N., S.C. Andersen, K.J. Meier, L.J. O’Toole Jr. & S.C. Winter (2015): Is the Performance Effect of Management Underestimated? Comparing Public Managers' and Front-line Employees' Perceptions of Management. Paper for the Annual Meeting of the Southern Political Science Association in New Orleans, January 2015. Texas A&M University, Aarhus University, University of Georgia & SFI. Fuller, J. Bryan, Coleman E. P. Patterson, Kim Hester, and Donna Y. Stringer. 1996. A quantitative review of research on charismatic leadership. Psychological Reports 78: 271-87. Gellis, Zvi D. 2001. Social work perceptions of transformational and transactional leadership in health care. Social Work Research 25: 17-25. Hanushek, E.A. 2006. School resources. In Handbook of the economics of education, edited by E.A. Hanushek and F. Welch. Amsterdam, London: North-Holland. Hannah, Sean T., Bruce J. Avolio, Fred Luthans, and P.D. Harms. 2008. Leadership efficacy: Review and future directions. The Leadership Quarterly 19: 669–92 34 Harris, Monica J., and Robert Rosenthal. 1985. Mediation of interpersonal expectancy effects: 31 meta-analyses. Psychological Bulletin 97(3): 363-86. Hattie, John. 2009. Visible learning: A synthesis of over 800 meta-analyses relating to achievement. London/N.Y.: Routledge. Hofstede, G. 1980. Motivation, leadership, and organization: do American theories apply abroad? Organizational Dynamics 9: 42-62. ______ . 1983. The cultural relativity of organizational practices and theories. Journal of International Business Studies 14: 75-89. House, Robert J., and Boas Shamir. 1993. Toward the integration of transformational, charismatic, and visionary theories. In leadership theory and research: perspectives and directions, edited by Martin M. Chemers and Roya Ayman, 81-108. San Diego, CA: Academic Press. Jacobsen, C.B. & Andersen, L.B. (2015): Is Leadership in the Eye of the Beholder? A Study of Intended and Perceived Leadership Practices and Organizational Performance. Public Administration Review, 75(6): 829-41. Judge, Timothy A., and Ronald F. Piccolo. 2004. Transformational and transactional leadership: A meta-analytic test of their relative validity. Journal of Applied Psychology 89: 755-68. Jussim, Lee and Kent D. Harper. 2005. “Teacher Expectations and Self-Fulfilling Prophecies: Knowns and Unknowns, Resolved and Unresolved Controversies.” Personality and Social psychology Review 9(2): 131-55. Kanfer, Ruth. 1990. Motivation theory and industrial and organizational psychology. In Handbook of industrial and organizational psychology, edited by Marvin D. Dunnette, 75-170. Palo Alto, CA: Consulting Psychologist Press. 35 Kanfer, R., G. Chen and R.D. Pritchard. (2008). Work Motivation: Past, Present, and Future. New York, NY: Routledge, Taylor and Francis Group. Kierein, N., & Gold, M. A. (2000). Pygmalion in work organizations: A meta-analysis. Journal of Organizational Behavior, 21, 913–28. King, A.S. (1971). Self-fulfilling prophecies in training the hard-core: Supervisors’ expectations and the underprivileged workers’ performance. Social Science Quarterly, 52, 369-78. King, A.S. (1974). Expectation effects in organization change. Administrative Science Quarterly, 19, 221-30. Korman, A.K. 1971. Expectancies as determinants of performance. Journal of Applied Psychology 55 (3): 218-22. Latham, G. P. (2007). Work Motivation: History, Theory, Research and Practice. Thousand Oaks, CA: Sage. Latham, Gary P. and Edwin A. Locke. 2007. New developments in and directions for goal-setting research. European Psychologist 12 (4): 290-300. Lipsky, Michael, 1980. Street-level bureaucracy: Dilemmas of the individual in public services. New York: Russel Sage Foundation. Locke, Edwin A. 1968. Toward a theory of task motivation and incentives. Organizational Behavior & Human Performance 3: 157-89. Locke, Edwin A., and Gary P. Latham. 1990. A theory of goal setting and task performance. Upper Saddle River, NJ: Prentice Hall. 36 ______ . 2002. Building a practically useful theory of goal setting and task motivation: A 35-year odyssey. American Psychologist 57: 705–717. ______. 2004) What should we do about motivation theory? Six recommendations for the twentyfirst century. Academy of Management Review 29, 388-403.______ . 2006. New directions in goal-setting theory. Association for Psychological Science 15: 265–68. Lowe, Kevin B., K. Galen Kroeck, and Nagaraj Sivasubramaniam. 1996. Effectiveness correlates of transformation and transactional leadership: A meta-analytic review of the MLQ literature. Leadership Quarterly 7: 385-415. McCulloch, Alistair. 2009. The student as co-producer: Learning from public administration about the student–university relationship. Studies in Higher Education 34: 171–83. McNatt, D. Brian. 2000. Ancient Pygmalion joins contemporary management: A meta-analysis of the result. Journal of Applied Psychology, 85(2): 314-22. Meier, Kenneth J., and Laurence J. O’Toole, Jr. 2013. Subjective organizational performance and measurement error: Common source bias and spurious relationships. Journal of Public Administration Research and Theory 23: 429-56. ______ . 2011. Isopraxis Leadership: Self-Efficacy, Managerial Strategy, and Organizational Performance. Paper presented at the 11th Public Management Research Conference, Syracuse NY, June 2-4, 2011. Meier, Kenneth J., Simon Calmar Andersen, Laurence J. O'Toole Jr., Nathan Favero, and Søren C. Winter (2015). Taking Managerial Context Seriously: Public Management and Performance in U.S. and Denmark Schools. International Public Management Journal 18(1): 130-50). 37 Meier, Kenneth J., Søren C. Winter, Laurence J. O’Toole, Jr., Nathan Favero, Simon Calmar Andersen (20015). “The Validity of Subjective Performance Measures: School Principals in Texas and Denmark.” Public Administration. Published online. Morse, Ricardo S., Terry F. Buss, and C. Morgan Kinghorn, eds. 2007. Transforming public leadership for the 21st century. Armonk, NY: M. E. Sharpe. Natanovich, Gloria, and Dov Eden. 2008. Pygmalion Effects among Outreach Supervisors and Tutors: Extending Sex Generalizability. Journal of Applied Psychology 93(6): 1382-89. Ostrom, Ellinor. 1996. Crossing the great divide: coproduction, synergy, and development. World Development 24: 1073-87. O’Toole, Laurence J. Jr., and Kenneth J. Meier. 1999. Modeling the impact of public management: The implications of structural context. Journal of Public Administration Research and Theory 9: 505-26. Pedersen, M.J., A. Rosdahl, S.C. Winter, A.P. Langhede, and M. Lynggaard. 2011. Ledelse af folkeskolerne. Vilkår og former for skoleledelse. Copenhagen: SFI – The National Danish Centre for Social Research, 11:39. Pinder, C. C. 1998. Work Motivation in Organizational Behavior. Upper Saddle River, NJ: Prentice-Hall. Porter, L. W., Bingley, G. A, and R.M. Steers. 2009. Motivation and Work Behavior (7e). Glasgow, UK: Bell & Bain Ltd. Rainey, Hal G. 2009. Understanding and managing public organizations. 4th ed. San Francisco, CA: Jossey-Bass. 38 Rangvid, B.S. 2008. Skolegennemsnit af karakterer ved folkeskolens afgangsprøver. Korrektion for social baggrund. AKF Working paper 2008 (1). Copenhagen: AKF. Rivkin, S.G., E.A. Hanushek, and J.F. Kain. 2005. Teachers, schools and academic achievement. Econometrica 73 (2): 417-58. Robinson, V., C. Lloyd, and K. Rowe. 2008. The impact of leadership on student outcomes: An analysis of the differential effects of leadership types. Educational Administration Quarterly 44 (5): 635-74. Robinson, V., M. Hohepa and C. Lloyd. 2009. School leadership and student outcomes: Identifying what works and why. Auckland: New Zealand Ministry of Education. Rosenthal, R. and L. Jacobson. 1968. Pygmalion in the Classroom: Teacher Expectations and Student Intellectual Development. New York: Holt. Saks, Alan M., and Blake E. Ashforth. 1997. Organizational socialization: Making sense of the past and present as a prologue for the future. Journal of Vocational Behavior 51: 234-79. Shamir, Boas, Robert J. House, and Michael B. Arthur. 1993. The motivational effects of charismatic leadership: A self-concept based theory. Organization Science 4: 577-94. Somers, M.A., P.J. McEwan and J.D. Willms. 2004. How effective are private schools in Latin America? Comparative Education Review, 48: 48-69. Stryker, Sheldon. 1980. Symbolic interactionism: A social structural version. Menlo Park, CA: Benjamin Cummings. Tajfel, Henri, and John C. Turner. 1986. The social identity theory of intergroup behavior. In Psychology of intergroup relations, edited by Stephen Worschel and William G. Austin, 2nd ed., 7–24. Chicago, IL: Nelson Hall. 39 Waldman, D. A., and F.J. Yammarino. 1999. CEO charismatic leadership: Levels of management and levels of analysis. Academy of Management Review 24: 266−85. Wang, Gang, In-Sue Oh, Stephen H. Courtright, and Amy E. Colbert. 2011. Transformational leadership and performance across criteria and levels: A meta-analytic review of 25 years of research. Group & Organization Management 36: 223-70. Weber, Max. 1922. Wirtschaft und Gesellschaft. Tübingen: Mohr. White, Susan S., and Edwin A. Locke. 2000. Problems with the Pygmalion effect and some proposed solutions. Leadership Quarterly 11: 389-415. Whiteley, Paul,Thomas Sy and Stefanie K. Johnson. 2012. Leaders' conceptions of followers: Implications for naturally occurring Pygmalion effects. Leadership Quarterly 23(5): 822-834. Winter, Søren C., and Peter J. May. 2001. Motivation for compliance with environmental regulation. Journal of Policy Analysis & Management 20: 675 - 98. Winter, Søren C. and Vibeke Lehmann Nielsen, eds. 2013. Lærere, undervisning og elevpræstationer i folkeskolen. Copenhagen: SFI – The National Danish Centre for Social Research, 13:09. Wright, Bradley E. 2004. The role of work contexts in work motivation: A public sector application of goal and social cognitive theories. Journal of Public Administration Research and Theory 14: 59-78. Yukl, Gary. 1989. Managerial leadership: A review of theory and research. Journal of Management 15: 251-89. ______ . 2013. Leadership in organizations, 8th ed. Essex: Pearson. 40