Learning From Leadership: Investigating the Links to Improved Student Learning

Click here to download the full report:
 Learning From Leadership: Investigating the Links to Improved Student Learning

Key Findings

  • Collective leadership has a stronger influence on student achievement than individual leadership.
  • Almost all people associated with high-performing schools have greater influence on school decisions than is the case with people in low-performing schools.
  • Higher-performing schools award greater influence to teacher teams, parents, and students, in particular.
  • Principals and district leaders have the most influence on decisions in all schools; however, they do not lose influence as others gain influence.
  • Schools leaders have an impact on student achievement primarily through their influence on teachers‘ motivation and working conditions; their influence on teachers‘ knowledge and skills produces less impact on student achievement.


Collective leadership, as the term is used in this component of our study, refers to the extent of influence that organizational members and stakeholders exert on decisions in their schools. This relatively narrow but fundamental perspective on leadership focuses attention on the combined effects of all sources of leadership, along with possible differences in the contributions made by each of these sources (e.g., administrators, teachers, students, parents). Guided by this conception of leadership, the sub-study set out to estimate the following:

  • the relative influence on school decision making of each of the individuals or groups potentially contributing to a school‘s collective leadership;
  • the impact of collective leadership on teacher feelings and beliefs and on student learning; and
  • whether differences in the extent of influence exerted by the respective participants is related to differences in levels of student achievement.

Prior Evidence

Leadership as Influence

The conception of collective leadership used for this study overlaps with Rowan‘s conception of organic management, defined as follows:30

    a shift away from conventional, hierarchical patterns of bureaucratic control toward what has been referred to as a network pattern of control, that is, a pattern of control in which line employees are actively involved in [making] organizational decision[s,] [and] staff cooperation and collegiality supplant the hierarchy as a means of coordinating work flows and resolving technical difficulties. (Miller & Rowan, 2006, p. 219-220)

Conceptualizing collective leadership as a network of influence and control also locates our study in relation to other research about organizational control structures. A seminal paper by Tannenbaum (1961), for example, introduced the "control graph" as a means of displaying patterns of control in formal organizations. The horizontal axis of a control graph designates each of the "levels" (designated positions) in the organization, while the vertical axis represents the degree of perceived influence or control exercised at each level. Tannenbaum used the control graph to illustrate four prototypical control modes or approaches to leadership: autocratic (influence rises with the hierarchical level of the role), democratic (higher levels of influence are ascribed to those in hierarchically lower levels or roles), anarchic (relatively little influence by any level or role), and polyarchic (high levels of influence by all levels or roles). Reflecting Rowan‘s (1990) expectations for organic management under conditions of uncertainty, Tannenbaum also hypothesized that organizational effectiveness will be related to: (a) more democratic, and (b) more polyarchic forms of control.

The first of these hypotheses arises from two sets of expectations. First, more democratic forms of control will be more consistent with employees‘ beliefs and values in a democratic society and contribute to higher levels of job satisfaction and morale, whereas autocratic forms of control are expected "to reduce initiative, inhibit identification with the organization and to create conflict and hostility among members" (Tannenbaum, 1961, p. 35). Second, more control by those lower in the hierarchy will lead to greater acceptance of jointly-made decisions along with an increased sense of responsibility for and motivation to accomplish organizational goals. Such participation may also contribute to more effective coordination through mutual influence mechanisms.

The second of Tannenbaum‘s hypotheses, sometimes called the "power equalization" hypothesis, is justified, Tannenbaum claims, by certain results—by improved organizational efficiency realized when more control is exercised by those lower in the hierarchy, and by improved motivation and identification with the organization on the part of those whose power is enhanced. Reasons offered in the current literature about distributed leadership are quite similar to the justification Tannenbaum‘s offers for his two hypotheses.

Collective Leadership Effects

What evidence is there to show that democratic, supportive, and shared forms of leadership are effective? Some empirical evidence may be found in research on teacher participation with peers in planning and decision making31 and in research on transformational leadership.32 Several lines of related theory also give rise to expectations of a positive association between organizational effectiveness and the distribution of influence, including theories of organizational learning,33 distributed cognition,34 and communities of practice.35

Nonetheless, there is substantial evidence to the contrary, especially from research in which organizational effectiveness is defined as the organization‘s bottom line (some measure of productivity) and assessed using objective indicators, such as student test scores. Tannenbaum was able to provide only limited support for his hypotheses about organizational control structures. And after about 15 years of programmatic research about organic management, Miller and Rowan reported that "the main effects are weak[,] and positive effects appear to be contingent on many other conditions" (2006, p. 220). A recent, comprehensive review of research on teacher leadership found only a small handful of studies in which researchers had actually inquired about effects of teacher leadership on students, and the results were generally not supportive.36

To date, most research about school leadership has focused on the work of teachers and school administrators. It is certainly possible, however, to conceive of people acting in other roles—as parents, students, interested members of the community—to exercise influence in schools. The work of Pounder, Ogawa and Adams (1995) provides one example (there are not many) of research that examines leadership exercised by a broader array of participants. Pounder et al. test a model of the influence of principals, teachers, parents, and secretaries on a number of mediating variables, as well as a range of school outcomes, providing a useful model for our approach a decade later .

The current sub-study looks beyond the school setting in its examination of leadership. Staff members in district roles also have an obligation to influence what schools do, although most studies of collective, shared, and distributed leadership have not examined the contribution of district personnel.37 Our study concerned itself with all of these potential sources of influence.

Antecedents of Teacher Performance

Miller and Rowan (2006) sought to assess certain effects of organic management. In this effort they did not attend to variables potentially mediating the effects of leaders on student learning. This is an important limitation, given prior work (Pitner, 1988; Hallinger & Heck, 1996a) showing that the effects of leadership on students are largely indirect. Studies designed to explore direct effects of leadership rarely detect significant effects, whereas many studies of indirect effects do. Most studies since 1996 have been guided by complex causal models which include a wide array of potential mediators.38

The framework for this sub-study assumed indirect leadership effects and conceptualized as mediators a set of teacher performance antecedents including motivation, capacity, and the situations in which people work. These are variables in a general model of employee performance and how it improves. Our own modification of this framework is based on theoretical and empirical accounts of the conditions required for development of motivation and capacity on the part of school people to engage productively in improvement efforts. Our modification also incorporates accounts of organizational conditions and characteristics of the infrastructure which facilitate the successful implementation of large-scale reform, or what van den Berg, Vandenberghe, and Sleegers (1999) refer to as the organization‘s "innovative capacity."39

New Evidence


Sample. This sub-study is based on data collected in the first round of surveys for the larger study. The achieved sample included responses by 2,570 teachers (77% response rate) from a total of 90 schools in which seven or more teachers completed usable surveys and for which usable student achievement data were available.40 Table 1.1.1 below presents a summary of the characteristics of our achieved sample.


Sources of evidence. To measure student achievement across schools, we collected data from state websites. These data comprised school-wide results on statemandated tests of language and mathematics at several grade levels over three years (2003 to 2005). We represented a school‘s level of student achievement by the percentage of students meeting or exceeding the proficiency level (usually established by the state) on language and mathematics tests. We averaged these percentages across grades and subjects in order to increase the stability of scores,41 arriving finally at a single achievement score for each school for each of three years. Our analysis also included an achievement change score, calculated as the gain in percentage of students attaining or exceeding the state-established proficiency level from the first to the third year for which we had evidence.

Teacher responses to 49 items from a 104-item survey provided the remaining data for this sub study. The survey, which required about 20 minutes to complete, measured the collective leadership and teacher-performance antecedents described in our framework: 9 items measured collective leadership, 9 items measured teacher capacity, 17 items measured teacher motivation, and 14 items measured teacher work settings or conditions. Each of the nine items used to measure collective leadership pertained to a single source of influence from a set including district administrators, principals, other school administrators, some individual teachers, teachers with designated leadership roles, staff teams, some individual parents, parent advisory groups, and students. About each source of influence, we asked respondents to rate the extent of direct influence on school decisions (on a 6-point scale). We also asked respondents to rate the extent to which they agreed with statements about each of the three antecedents of teacher performance, also on a 6-point scale.

Analysis. We merged individual responses to the teacher survey, aggregated to the school level, with school-level student achievement results. We used SPSS to calculate means, standard deviations, and reliabilities (Cronbach‘s alpha) for scales measuring the four variables. We used paired-sample t-tests to compare mean ratings of various sources of leadership. We tested the factor structure of the teacher variables included in the study. We used hierarchical multiple regression to examine the moderating effects of student SES on some relationships in our framework. Finally, we used LISREL to test a model of the relationships among collective leadership, teacher motivation, capacity and setting, and student achievement. This path-analytic technique allows for testing the validity of causal inferences for pairs of variables while controlling for the effects of other variables. We analyzed data using the LISREL 8.80 analysis of covariance structure approach to path analysis and maximum likelihood estimates.42 We used four goodness-of-fit statistics to assess the fit of our path model with the data: the Root Mean Square Error of Approximation test (RMSEA), the Norm-fit index (NFI), the adjusted Goodness of Fit index (GFI) and the mean Root Mean Square Residual (RMR).


We begin with a summary of responses to the teacher survey and with information about the statistical properties of our measures, including the results of a factor analysis of the measures of teacher capacity, motivation, and setting. The remaining sections report evidence relevant to each of three questions addressed by the study: the impact of collective leadership on key teacher variables and student learning; the relative influence of different collective leadership sources; the relationship between different patterns of collective leadership and student achievement.

Table 1.1.2 reports the internal reliabilities (Cronbach‘s alpha) of the scales used to measure each of the three antecedents of teacher performance—capacity, motivation and work setting—and the measure of collective leadership. Overall mean ratings of the three antecedents are not reported because z-scores had to be calculated to accommodate the use of different response scales. We calculated variable reliabilities using z-scores. Responses to all variables ranged between slight agreement and moderate agreement, with low to moderate standard deviations. All scales achieved acceptable levels of reliability (between .72 and .96).


Note: z-scores were used to calculate the aggregate values for the capacity, motivation, and setting scales. Collective leadership was calculated from the sum of nine sources of leadership, each rated on a 6-point scale from 'no influence' to 'very great influence.'

Of the 40 items used to measure the three teacher antecedents, 9 measured capacity, 17 measured motivation, and 14 measured work setting. We analyzed the dimensionality of these 40 items using principal component factor analysis. We used the scree test and the interpretability of the factor solution to determine the number of factors to rotate. We rotated three factors using a Varimax rotation procedure. The rotated solution yielded three interpretable factors which corresponded very closely with the three variable categories: capacity, motivation, and setting. The capacity factor accounted for 14.4% of the item variance; the motivation factor accounted for 13.9% of the item variance; and the setting factor accounted for 8.6% of the item variance.

Although our initial conception of the three teacher variables suggested a number of distinct sub-dimensions, these were not supported by the factor analysis. Thus, we used aggregate scores for each of the three teacher-performance antecedents in all subsequent analyses. Also in response to the results of the factor analysis, we omitted two of the original items measuring capacity and seven of the items measuring motivation from subsequent analysis.

Collective Leadership Effects on Teachers and Students

Table 1.1.3 reports correlations among measures of all variables in the study. As these results indicate, collective leadership is significantly related to all three teacher variables. The strongest relations are with collective leadership and teachers‘ work setting (r =.58), followed by teacher motivation (r=.55). All variables but teacher capacity are significantly related to student achievement: teachers‘ work setting has the strongest relationship (r = .37), followed by teachers‘ motivation and collective leadership (r= .36 and .34). These data also indicate significant relationships among the teacher variables.


The path model described in Figure 2 (using LISREL) and Table 1.1.4 provides a further test of relationships among collective leadership, teacher capacity, motivation and work setting, and student achievement. This model is an excellent fit to the data (RMSEA = .00; RMR = .03; AGFI = .93; NFI = .99) and, as a whole, explains 20% of the variation in student achievement. Collective leadership has significant direct effects on all teacher variables. Its strongest effects are on teachers‘ work setting (r = .58), followed by teacher capacity (r = .36) and motivation (r = .25). Collective leadership accounts for only 13 % of the explained variation in teacher capacity.


Figure 2. Testing a model of collective leadership effects on student achievement43

The paths linking the three teacher variables to student achievement indicate that collective leadership influences student achievement through teacher motivation and work setting. The effect of teachers‘ work setting on achievement is significant (.25), but the effect of teacher capacity is insignificant. Total effects on student achievement are greatest for work setting, followed by teacher motivation and the indirect influence of collective leadership. The higher effect for setting is explained by its indirect effect through motivation, as indicated in the data presented in Table 1.1.4.


In order to estimate the contribution of student SES (calculated as the percentage of students in a school eligible for free or reduced lunch) to relationships described in the path model between the three teacher variables and student achievement, we computed three hierarchical regressions. In each regression equation SES was entered first, collective leadership second, and one of the teacher variables third.44 Results of these hierarchical regressions, described in Table 1.1.5, indicate that only motivation explains a unique and significant proportion of variation in student achievement after controlling for student SES. Motivation, on its own, explained 6% of the variation in achievement, whereas setting increased the variation explained by only 1% in combination with SES and leadership, and capacity decreased the explained variance by the same amount.


In sum, these results indicate the following:

  • Our model as a whole explains a significant proportion (20%) of variation in student achievement across schools.
  • Collective leadership has modest but significant indirect effects on student achievement.
  • Of the three teacher variables, the influence of collective leadership on students operates through its influence on teacher motivation and work setting.
  • While collective leadership does have a significant effect on teacher capacity, this variable is not significantly linked to student achievement.

These results confirm, in some respects, and contradict, in others, evidence from two of our earlier studies. One earlier studies incorporated approximately the same measures used in the present study of teachers‘ capacity, motivation, and work setting.45 Instead of collective leadership, however, that study used a measure of individual leaders‘ transformational practices. In that study, as in the present one, leadership was most strongly related to teachers‘ work setting and had weaker effects on teacher capacity than on teacher motivation. This earlier study also reported weaker effects of (likely individually provided) transformational leadership practices on student achievement as compared with the effects of collective leadership in the present study. This comparison of results provides encouragement, at least, for claims about benefits accruing to students when leadership is more widely distributed in schools.

Our second earlier study also differed in several important respects from the present study, but it addressed several of the same questions.46 Student engagement rather than student achievement was used as the dependent variable, and the variables mediating leaderships‘ influence on students were different from those used in the present study. The measure of collective leadership, however, was almost identical to the measure used in the present study. In contrast to the main findings of present study, this earlier study found non-significant, negative effects of collective leadership on students. This important difference in results offers at least modest support for the argument that the choice of mediating variables is a crucial matter in studies of leadership effects on students.47

The differences we have noted among our three studies might well be accounted for by non-trivial differences in their designs. To this point, consistency is greatest in respect to the effects of collective leadership on teachers‘ internal states. Specifically, collective leadership has so far not been shown to have a demonstrable impact on our measures of teacher capacity. Also, claims that collective leadership has significant impact on students have received mixed support. Evidence from other recent studies, however, seems to provide further support for this claim, although this evidence has been collected in contexts quite unlike the schools for which we have data. For example, Hiller, Day and Vance (2006) recently reported significant effects of collective leadership on supervisor-rated team performance in a road maintenance department. They also reviewed evidence from six other studies of collective leadership effects on team effectiveness, concluding that collective leadership is likely to be effective:

    when teams are engaged in complex tasks that require large amounts of interdependence, but under more routine conditions…the benefits of collective leadership have yet to be demonstrated (2006, p. 388).

The Relative Influence of Collective Leadership Sources

To address this issue, we analyzed teachers‘ ratings of the extent of influence on school decisions of the nine measured sources of collective leadership. Table 1.1.6 reports the mean response of teachers to each source. We calculated paired-samples ttests to estimate the significance of differences in these ratings. As Table 1.1.6 indicates, principals and district administrators were given the highest, almost identical ratings (M = 5.30 and 5.28, respectively). The small standard deviations of these ratings indicate considerable agreement among respondents about the perceived influence of people acting in these two roles. There is a significant drop in the rating of the next-most influential role: building-level administrators other than the principal, typically the assistant principal (M = 4.75).


Among teacher sources of influence, teachers with designated leadership roles were perceived to have the strongest influence (M = .4.43), followed by staff teams (M = 4.36) and then some individual teachers (M = 4.28); the ratings of teachers with formal leadership roles were significantly higher than the ratings of staff teams (t = 3.51, p<.01) or some individual teachers (t=5.54, p<.001), and the rating of staff teams was significantly higher than the rating of individual teachers (t=2.19, p<.05).

Ratings for parents (some individual parents, and parent advisory groups) were considerably lower than for teachers, ranging from means of 3.84 to 3.96, a statistically significant difference (t = 3.16, p<.01). Respondents perceived students to have the lowest level of direct influence on school decisions (m = 3.49). The very low standard deviation of ratings for all sources of influence, especially for principals, reduces the potential strength of relationships with any other variable in our study.

Table 1.1.7 reports the relationships between each of the individual sources of collective leadership and both teacher variables and student achievement (mean annual achievement over three years). Among the teacher variables, work setting has a significant relationship with seven of the nine sources of leadership (not principals or individual teachers). This surprising result for principals may be a reflection of the low level of variation in the ratings noted above. The strongest relationship is between motivation and staff teams (r = .71). Capacity was the only variable significantly related to principal influence (r=.22); teachers‘ work setting was the only variable related to other building administrators (r =.32) and district-level administrators (r =.41).
Teachers in formally designated roles were significantly related to all three teacher variables but not to student achievement. Staff teams, individual parents, parent advisory groups, and students all have significant relationships with student achievement. Student leadership is most strongly related to teacher motivation (r =.55). Parent advisory teams are most strongly related to motivation (r =.44) and achievement (r =.56); individual parents are most strongly related to achievement (r =.43) and weakly to setting (.34). There appears to be a differentiation between those leaders who are members of the school staff and those who are not. Staff teams have stronger relations with all three teacher variables than any of the other within-school collective leadership sources, and staff teams are the only in-school source of collective leadership related to achievement (r=.28).


We were intrigued to see that the two sources of leadership consistently showing significant relationships with all three mediating variables, and with student achievement, were collectives: staff teams and parent advisory groups had significant correlations with all our mediators and with student achievement. In schools with high levels of student achievement, and high ratings for capacity, motivation, and setting, we are more likely to see higher levels of influence from staff teams and parent advisory groups. This suggests that there may be something about the collective nature of these roles which adds to their influence in the schools.

In sum, our results indicate the following:

  • School decisions are influenced by a broad array of groups and people, reflecting a distributed conception of leadership.
  • The degree of influence exercised by these people and groups reflects a traditional, hierarchical conception of leadership in organizations. Teachers rate the influence of traditional sources of leadership much higher than they rate non-traditional sources.
  • Among teacher roles, the more formalized the leadership expectation, the greater the perceived influence.
  • Nonetheless, the influence of parents and students is significantly related to student achievement. This result may reflect the well-known effects of student SES on achievement.

If the profession has become enamored of distributed forms of leadership, as one might infer from current scholarship, the responses of teachers surveyed here suggest that few changes detectable by teachers have actually occurred in schools. The ground swell of support for distributed conceptions of leadership may well be a kind of "meta-rhetoric" denoting little reality "on the ground." This possibility is consistent with a familiar criticism of schools: that as a means of legitimizing their work, they are more concerned with the appearance than the substance of change.

Despite a decades-long effort to restructure schools—in part, at least, to give parents a greater voice in school decisions—we see little evidence that teachers perceive much influence from parents, or from students.48 This outcome probably reflects the wellknown and persistent challenges teachers and administrators face in creating authentic relationships with parents for school-improvement purposes. Our results also reinforce two other claims. First, significant change in schools requires much more than encouragement and rational argument,49 strategies which have often been relied on to promote greater parent influence. Second, as Jaques (2003) has long maintained, hierarchy is a necessary, unavoidable feature of any large organization, even when participants add structures and procedures to encourage lateral influence within the hierarchy. If Jaques is correct, current expectations about the extent to which leadership distribution is both possible and desirable in schools will need to be severely modified.

Patterns of Collective Leadership and Student Achievement

As we reported above, teachers on average perceived influence in their schools to be exercised in a distributed but still hierarchical manner. Nevertheless, prompted by widespread claims by many organizational theorists about the benefits of more distributed forms of leadership, we sought to learn whether variations in these perceptions of influence were related to levels of student achievement in schools. To address this question we returned to Tannenbaum‘s early work (reviewed above) on control graphs.50


Figure 3. Relationships between Sources of Collective Leadership Influence and Student Achievement

Schools were divided into quintiles based on the mean achievement of their students on test scores over three years. So, for example, Quintile 1 = schools with the lowest mean achievement over three years and Quintile 5 = schools with the highest mean achievement over three years.

To distinguish schools by mean levels of achievement averaged over three years, we constructed a control graph of our own. As Figure 3 indicates, we first divided the schools in our sample into quintiles on the basis of mean annual student achievement scores. Then we compared teachers‘ ratings of each source of collective leadership influence across quintiles.

Results displayed in Figure 3 indicate that teachers in the highest-achieving schools (Quintile 5) generally attributed higher levels of influence to all people and groups than did teachers in lower-achieving schools. Even though they attributed greater influence to non-traditional leadership roles in higher-achieving schools, teachers perceived that those in traditional leadership roles had the same relative amount of influence. For example, an increase in the influence of staff teams or parents does not mean less influence for principals and district administrators. Furthermore, teachers in schools whose students achieve in the highest and second-highest quintiles award significantly more relative influence to staff teams; teachers in the highest-quintile schools award significantly more relative influence, as well, to individual parents and to groups of parents.

Although we do not include a table reporting all correlations, we found SES to be significantly (and unsurprisingly) related to student achievement—a possible explanation for the high level of influence parents and students apparently exercise in schools in the higher quintiles of performance, which generally serve higher SES students. Three correlations seem especially interesting: those between SES and the influence of individual parents (r = .35), parent advisory committees (r = .53) and students (r = .36). The influence of staff teams was also related to student SES as strongly as student influence was (r = .34).51 Bidwell, Frank, & Quiroz (1997) provide evidence of the relationship between SES and parental involvement, and, more interestingly, between SES and levels of collegial control in schools. Schools in high-SES communities, Bidwell found, tend to build collegial professional practice among teachers and to have a particularly high focus on student learning.

This evidence indicates, in sum, that participants acting in traditional leadership roles remain highly influential in high-performing schools, a result not evident from the correlation analyses reported in Table 1.1.6. Reflecting a distinction by Dunlap and Goldman (1991) between power-over and power-through, our results illustrate the point that influence in schools is not a fixed sum. In the highest-performing schools, everyone seems to have more influence than participants in low-performing schools, where leadership may be "laissez-faire"—an approach to leadership almost invariably found to be ineffective.52

Overall, we also see continuing support for Jaques‘ (2003) claim about the inevitable presence of hierarchy in large organizations. Theorists who regard the attainment of "flat" organizational contours as something like a holy grail are running ahead of the evidence. Indeed, the evidence we have reviewed and the implications it suggests conform quite closely to a hypothesis prompted by Tannenbaum‘s conception of control graphs (and proposed by McMahon and Perritt). A decade after Tannenbaum‘s publication, McMahon and Perritt (1971) argued that organizational effectiveness may have less to do with "power equalization" than with perceived "concordance" or agreement across roles in control structures. Their research evaluated the degree to which people in different roles in the organization were in agreement about who was most influential. Their conclusion "emphasizes the importance of agreement on the perceptions of the control structure of various hierarchical echelons within an organization" (p. 339).

We are unable to test this claim directly with our own data, since teachers‘ perceptions are all we have; but it is a hypothesis worthy of further research, especially in light of widespread, unfounded claims about the positive consequences of distributed leadership and flat organizational structures. The pattern of leadership distribution evident among the highest-achieving schools in our study reflects none of Tannenbaum‘s prototypical models. It is, rather, a hybrid composed of "autocratic" (influence rises with hierarchical level) and "polyarchic" (high levels of influence for all) prototypes. If one were to accept the inevitability and value of hierarchy in organizing, this hybrid could serve as a best-case scenario. Let‘s call it "intelligent hierarchy" to reflect the opportunities this hybrid approach affords to ensure that organizations take advantage of the capabilities and strengths of most of their members while at the same time ensuring careful coordination of effort in a common direction.

Implications for Policy and Practice

Three implications for policy and practice emerged from this section of our study.

  1. In their efforts to improve student achievement, school- and district-level leaders should, as a matter of policy and practice, extend significant decisional influence to others in the school community. (See also Section 2.1.) Compared with lowerachieving schools, higher-achieving schools provided all stakeholders with greater influence on decisions. The higher performance of these schools might be explained as a consequence of the greater access they have to collective knowledge and wisdom embedded within their communities.
  2. Superintendents and principals working to extend influence to others should not be unduly concerned about losing their own influence. Results reported here show that higher-performing schools awarded greater influence to most stakeholders; at the same time, little changed in these schools‘ overall hierarchical structure. Our data depict the hierarchical structure of influence typically associated with roles and responsibilities in schools and districts—a structure that conforms, we believe, with Jacques‘ (2003) claim about requisite hierarchy in social organizations large enough to place significant demands on the coordination of its members‘ actions.
  3. In responding to demands that they focus sharply on improving their teachers‘ instructional capacities, school and district leaders should not overlook the influence they can have on classroom practice by continuing efforts to motivate their teachers, and to align their teachers‘ work settings with what is known about effective instructional practice.

Our results show that collective leadership is linked to student achievement indirectly, through its effects on teacher motivation and teachers‘ workplace settings. As in several of our previous studies,53 we found significant but much weaker relationships between leadership and teacher capacity. At least in part, our measure of teacher capacity may explain these results. It was primarily a measure of professional development opportunities—that is, opportunities to learn from colleagues in a variety of ways—rather than a direct measure of the knowledge and skills teachers need to foster student achievement. In effect, while principals and their co-leaders exert a significant influence on teacher access to professional learning opportunities, their power to influence the quality and impact of those activities on teacher knowledge and skills may be more limited. Thus, our finding of the absence of a strong relationship between the indirect measure of teacher capacity that we used and student achievement may simply reflect the low quality of typical professional development inputs available to teachers in schools. This qualification, however, does not diminish our finding that motivation and work settings—factors subject to leadership influence—have significant effects on student achievement. In light of this, a narrow focus on leadership efforts aimed only at building teacher capacities would be misguided.

< < Previous | Next > >


30. Miller & Rowan (2006); Rowan (1990).

31. Talbert & McLaughlin (1993).

32. Leithwood & Jantzi (2005).

33. Hutchins (1996).

34. Perkins, 1993; Tsoukas (2005).

35. Wenger, McDermott & Snyder (2002).

36. York-Barr & Duke (2004).

37. But see Firestone (1989), and Firestone & Martinez (2007).

38. For example, Leithwood & Levin (2005) and Leithwood, Louis, Anderson, & Wahlstrom (2004).

39. For a more detailed explanation of how these variables were defined and measured, see Leithwood & Jantzi (2008).

40. We were able to generate data on the SES of only 76 of these schools, so the calculations for tables drawing on SES have been adjusted to use this smaller sample.

41. Linn (2003).

42. Joreskog & Sorbom (1993).

43. While a number of iterations of our framework were run, testing relationships in a variety of ways, we present here only the results that have proved statistically significant. The LISREL model presented has Chi-square = 1.97, df = 2, p = .37.

44. Readers should note that the order in which variables are added to the model has an influence on the strength of the relationship. In our analysis, leadership adds 3.6% to the 11.3% explained variance from SES. Entering collective leadership first explains 9.2%; introducing SES at step 2 provides an additional 5.7% for the same total of 14.9%. If they are entered at the same time, SES explains 6.8%, leadership explains 4.6%, and their combined effect explains the other 3.5% to the total 14.9%.

45. Leithwood & Jantzi (2006).

46. Leithwood & Riehl (2005).

47. Hallinger & Heck (2002).

48. Beck & Murphy (1998).

49. Desimone (2006).

50. Tannenbaum (1961).

51. These correlations are all significant at the 0.01 level (2 tailed).

52. Avolio (1994).

53. Leithwood & Jantzi (2006); Leithwood et al. (2004a).