How School Leadership Influences Student Learning: A Test of “The Four Paths Model”

Classroom instruction (CI)

As conceptualized and measured in this study, CI incorporated general, rather than domain-specific, teaching practices that research from three bodies of literature indicate are effective in enhancing student learning: high-yield instructional strategies (e.g., Hattie, 2009; Marzano, Waters, & McNulty, 2005), the use of student data to inform instructional decisions (e.g., Dalton, 2009; Gates, 2008; Palucci, 2010; Rayor, 2010; J. Williams, 2011) and technology use to facilitate face-to-face instruction (Means, Toyama, Murphy, Bakia, & Jones, 2010).

Academic press (AP)

Academic press or emphasis is a key feature of high performing schools (Cannata, Smith, & Haynes, 2017). Hoy, Hannum, and Tschannen-Moran (1998) define academic emphasis as “a combination of teachers setting high, but reasonable goals, students responding positively to the challenge of these goals, and the principal supplying the resources and exerting influence to attain these goals” (p. 342). Academic press has been positively related to achievement in all types of schools including schools serving poor and minority students (Goddard, Hoy, & Hoy, 2000; Hoy, Tarter, & Hoy, 2006; Smith & Kearney, 2013), with its effect stronger in low-socioeconomic status (SES) high schools (Shouse, 1996).

Disciplinary climate (DC)

School DC includes students’ discipline concerns; class disruptions; student absenteeism and tardiness; students counseling about discipline; students’ discipline experience; the rules for behavior, race, or cultural conflicts at the school; students’ behaviors and the punishments for misbehaviors at the school; teachers’ behavior; and teacher–student relations (Ma & Willms, 2004). Disciplinary climate has a significant relationship with student learning (Hattie, 2009; Ning, Damme, Noortgate, Yang, & Gielen, 2015; Sortkaer & Reimer, 2018) and strongly influences the equitable treatment of students when it is approached through principles of “restorative justice” (Wiley et al., 2018). Its effects are larger than the effects of student SES (Ma & Crocker, 2007; Ma & Willms, 2004).

Teachers’ use of instructional time (UIT)

Evidence indicates that total instructional time matters less than how the time is spent (Bellei, 2009), the subjects on which time is spent, and the strength of the curriculum (OECD, 2013). The content of the curriculum in which students spend time studying, “opportunity to learn,” has quite strong effects on learning (Tornroos, 2005; Wang, 1998). Teachers’ UIT includes teachers’ efforts to maximize teaching and learning time, create classroom conditions that allow for an appropriate pace of instruction, and help students take charge of their own learning in age-appropriate ways.

Collective teacher efficacy (CTE)

This variable is defined as the level of confidence a group exudes in its capacity to organize and execute the tasks required to reach desired goals (Bandura, 1993; Goddard, Hoy, & Hoy, 2004). Correlations between measures of CTE and student learning range from 0.38 to 0.99, with an average r = 0.69 (e.g., Tschannen-Moran & Barr, 2004). Angelle and Teague (2014) report a strong relationship between CTE and the likelihood of them taking on leadership role in their schools. One recent study reported modest but significant relationships between CTE and the improvement of math instruction in middle schools (Berebitsky & Salloum, 2017).

Teacher trust in others (TTO)

Common to most concepts of trust is one party’s willingness to be vulnerable to another party based on the belief that the latter party is competent, reliable, open, and concerned (Handford & Leithwood, 2013). Teacher trust in others in this study included teacher trust in colleagues, students, and parents. This variable has been linked positively to school effectiveness (Goddard, Tschannen-Moran, & Hoy, 2001), school climate (Hoy, Sabo, & Barnes, 1996; Tarter, Sabo, & Hoy, 1995), and student achievement (Leithwood, Anderson, Mascall, & Strauss 2010), even when SES and other student demographics (e.g., Goddard et al., 2001).

Teacher commitment (TC)

Evidence has accumulated about at least four types of TC: commitment to teaching, students, to the school organization, and to change. This study measured only TC to the school organization. Such commitment is about an individual’s strong belief in the organization, identification and involvement in the organization, and a strong desire to remain part of the organization (Freeston, 1987; Leithwood, Jantzi, & Steinbach, 1999; Porter, Steers, Mowday, & Boulian, 1974). This type of commitment is positively associated with important organizational conditions such as teachers’ instruction (Granger, Morbey, Lotherington, Owston, & Wideman, 2002; Hendel, 1995) and student outcomes including moral growth (M. M. Williams,1993), and academic achievements (Gill & Reynolds, 1999; Harvey, Sirna, & Houlihan, 1988; Housego, 1999; Janisch & Johnson, 2003). Many factors giving rise to this form of commitment are responsive to leadership influence (Dannetta, 2002).

Safe and orderly environment (SOE)

Assuming a holistic approach to school safety and orderliness, this variable relies on the coordination of school, parents, community and community services, efficient provision of mental health services for those students who need it, threat assessment rather than violence surveys, emphasis on prevention versus suspension (on safe school vs. school violence), and increasing the use of restorative justice practices in progressive discipline (Astor, Guerra, & Acker, 2010; Borum, Cornell, Modzeleski, & Jimerson, 2010; Cornell & Mayer, 2010; Mayer & Furlong, 2010; Jimerson, Swearer, & Espelage, 2010; Swearer, Espelage, Vaillancourt, & Hymel, 2010). This variable captures the features of both an orderly, safe environment, and an inclusive environment.

Collaborative structures and culture (CSC)

This variable captures key elements of teachers’ collaborative instructional knowledge sharing, creation, and experimentation based on student progress data. This variable is especially prominent in schools making significant progress with their students’ achievement (e.g., Hill, 2010; Lomos, Hofman, & Bosker, 2011). One review of data-use research (Sun, Przybylski, & Johnson, 2016) revealed that teachers felt the opportunity to work with their colleagues, using common assessment to monitor student academic progress, engage in shared instructional decision making, and sharing best practices was an integral part of the process leading to increased academic scores. Collaborative school culture and structures are significantly correlated with teacher-perceived effectiveness in specialized programs for students with disabilities (Kristoff, 2003).

Organization of planning and instructional time (OPIT)

This variable includes two components providing time and structure for teachers’ common planning and maximizing instruction time at the school level. Common planning time is probably the support teachers need most from school administration for collaboration and professional development (Deike, 2009; Gallagher, Means, Padilla, & SRI International, 2008; Quezada, 2012). Teachers’ developing common assessment tools, sharing effective assessments and teaching strategies, identifying students’ need, and developing interventions during common planning times has been reported as one prominent feature of successful schools—a typical way to improve “social capital” in schools (DuFour & Fullan, 2013; Hargreaves & Fullan, 2012), and an effective way to move students forward (Leithwood, Aitken, & Jantzi, 2006; Sun, Johnson, & Przybylski, 2016).

Parent expectations for children’s success at school and beyond (PE)

This variable was defined as “The degree to which a student’s parents [hold] high expectations of the student’s promise of achieving at high levels” (Jeynes, 2005, p. 246). Jeynes’s (2005) meta-analysis identified “parental expectations,” among all forms of parental involvement in school, as having the greatest impact on student achievement by a large margin; a significant effect size of .58 (p. 253).

Forms of communication between parents and children (FC)

Schools typically spend considerable effort on creating meaningful ways of communicating with parents (Epstein et al., 2002) such as school newsletters, curriculum nights at school, online messaging systems and the like. However, it is the FC in the home that has by far the largest effect on student success at school. Underlying most such communication is what the literature refers to as “parenting styles” (e.g., Jeynes, 2005). Creating effective parent/child communications necessarily entails clarifying with parents the advantages of adopting a supportive yet firm approach to interacting with their children, as compared with more extreme forms of either autocratic or laissez-faire approaches. Creating effective parent/child communications about school-related matters requires school staff to focus on how productive parenting styles are applied to obviously school-relevant issues (Leithwood & Patrician, 2015).

Parents’ social and intellectual capital about schooling (PSC)

This variable includes the power and information present in parents’ social relationships that can be used to leverage additional resources helpful in furthering their children’s success at school (Leithwood & Patrician, 2015). “The more people do for themselves, the larger community social capital will become, and the greater will be the dividends upon the social investment” (Ferlazzo, 2011, p.11). Parents’ Intellectual Capital has been defined as the knowledge and capabilities of parents with the potential for collaborative action. Taken together, PSC encompass parent engagement, involvement, and assistance in student learning and school activities. Parent engagement is nurtured when parents believe they should be involved in their children’s education and schooling and have a positive sense of efficacy about the usefulness of their involvement (Hoover-Dempsey & Sandler, 1997).

School leadership (SL)

The 22-item scale measuring SL reflected each of the four domains of the integrated leadership model described earlier: Set Directions (e.g., “To what extent do your school leaders provide useful assistance to you in setting short-term goals for teaching and learning”); Develop People (e.g., “To what extend do your school leaders give you individual support to help you improve your teaching practices”); Develop the Organization (e.g., “To what extent do your schools leaders build community support for the school’s improvement efforts”); and Improve the Instructional Program (e.g., “To what extent do your school leaders encourage teachers to use data effectively to improve their instruction”). Previous uses of this measure reported relatively high reliability .94 in the prior study and .93 in Leithwood and Louis (2012).

Variable on the rational path

Four variables represented the rational path. The eight-item scale measuring CI was constructed for this study from the review of literature summarized in the framework (above). In Study 2, this scale had an alpha coefficient of .93. A sample item from the scale measuring this variable is “I provide prompt, specific scaffolding or remedial feedback to provide more precise instruction to each of the students.”

The four-item scale measuring AP was adapted from a scale used by Hoy and Tarter (1997) with an alpha coefficient of .94. In Study 2 this scale had an alpha coefficient of .84. Sample item from this scale includes “My school sets high standards for academic success” and “Students respect others who get good grades.” Disciplinary climate was measured with four items adapted from earlier research by Ma and Willms (2004), for example, “Students do not start working for a long time after my lessons begin.” The reliability of this scale used in previous research ranged from 0.45 to 0.71 (e.g., Ma & Willms, 2004). Six items developed specifically for this study were used to measure teachers’ UIT (e.g., “My classroom timetable includes large uninterrupted blocks of learning time”). Our prior study reported an alpha coefficient of .87 for this scale.

Variables on the emotions path

This path is populated by three variables. The eight-item scale for measuring CTE, with an alpha coefficient of .88 in our prior study, was originally based on a scale reported by McGuigan and Hoy (2006). A sample item is “Teachers in this school are confident they will be able to motivate their students.” The four-item scale measuring teachers’ trust in parents, students, and colleagues was adapted from a Faculty Trust scale used by Tschannen-Moran and Hoy (1998), expanded to include other stakeholders but not school administrators. This four-item scale included, for example, “Teachers can count on support from most students’ families.” The alpha coefficient for this scale in our prior study was .82. The six-item scale measuring TC was created from our own synthesis of literature on this topic and includes, for example, “I am willing to go the extra mile to help students.” The reliability of this scale in Study 2 was 0.94.

Variables on the organizational path

On this path are SOE, CSC, and OPIT all measured using items based on our own review of prior evidence. Use of the nine-item CSC scale in our Study 2 resulted in an alpha coefficient of .92. An example from that scale is “We collaborate with one another to develop common assessment tools for measuring students’ progress.” The six-item scale measuring SOE included, for example, “The learning environment in this school is safe and orderly.” Our prior study using this scale reported an alpha coefficient of .77. Finally, an example of the four-item OPIT scale is “Teachers in our school have common planning times to discuss teaching and learning.” As used in Study 2 the scale had an alpha coefficient of .80.

Family path variables

The three variables representing the family path included PE, FC, and PSC relevant to schools. The three scales measuring these variables originated in a previous study by the first author (Leithwood & Patrician, 2015). By way of example, one item from the five-item PE scale is “Most of my students’ parents or guardians make sure their children finish their homework.” One item from the five-item scale measuring FC is “Most of my students’ parents or guardians listen to their children’s ideas,” and one item from the eight-item PSC scale is “Most of my students’ parents or guardians ensure that their kids have healthy diets and enough sleeps.” The reliability of these scales (aggregated) in Study 2 was 0.83.

Student achievement

State of Texas Assessments of Academic Readiness (STAAR) program results was the source of student achievement data for the study. This program was implemented in spring 2012 and includes tests of reading and mathematics at Grades 3 and 8; writing at Grades 4 and 7; science at Grades 5 and 8; social studies at Grade 8; end-of-course assessments for English I, English II, Algebra I, biology, and U.S history. The study used results aggregated at the school level of the Texas STAAR Percentage at Phase-in Satisfactory Standard or Above, combining all subjects and all grades.

Socioeconomic status

Students’ SES, measured by EcoDis was based on the count and percentage of students eligible for free or reduced-price lunch or eligible for other public assistance (PEIMS [Public Education Information Management System], October 2013, October 2014; and TEA [Texas Education Agency] Student Assessment Division). While free or reduced-price lunch is an indirect, school-wide measure of SES, recent evidence suggests that it predicts student achievement better than family income measures alone even after controlling for such income (Domina et al., 2018).

Rational path variables

Classroom instruction was eliminated from the regression analysis because of a .02 correlation with achievement. Regression analysis indicated that, among the remaining three variables on the rational path, DC was the primary influence on student achievement (counteracting the negative impact of students’ economic status) followed by UIT. These results largely replicate results of Study 2, whereas Study 1 found that AP and DC had similar effects but UIT had no effect

Results of this study, as well as Study 2, endorse the promise of a composite or latent variable labelled academic culture (Leithwood & Sun, 2018). This composite variable, combining AP, DC, and UIT makes significantly greater contributions to student learning than any of the three composite variables alone; it is also a powerful mediator of SL effects on student learning. “Focus and opportunity” are the theoretical grounds on which attention to academic culture is justified. AP and DC focus both teachers and students on the academic goals of the curriculum and help minimize distractions from pursing those goals. Using most of the instructional time in the classroom for teaching and learning without many other distractions, provides students with opportunities to be meaningfully engaged in acquiring those academic goals.

While the .02 correlation of CI with achievement resulted in this variable being dropped from additional analyses, it does warrant further attention. The most plausible explanation for these results, in our view, is the difficulty teacher respondents often have articulating the nature of their own instruction. Louis et al. (2010), for example, reported substantial differences between teachers’ interview accounts of their CI and subsequent classroom observations of those practices. Furthermore, using school leader ratings of CI based on the same scale slightly adapted so school leaders rated their teachers’ CI, Study 2 found a significant correlation (.59) between student achievement and CI. These results are consistent with self-rater halo effects reported in much of the literature on performance appraisal (Bass & Yammarino, 1991), as well as recent evidence about the superior predictive validity of teacher versus principal ratings of principal leadership (Zheng, Li, Chen, & Loeb, 2017). These results suggest that ratings of CI included in future research should not likely depend solely on teacher respondents.

Emotions path variables

Results of regression analysis indicated that CTE was the primary influence on student learning on this Path (β = .18). This result adds to considerable amounts of evidence confirming the important contributions to student achievement of CTE. Both individual and collective efficacy make impressive contributions to performance, whether the performance of students (e.g., Pajares, 1996), teachers (Goddard et al., 2000) or school leaders (Leithwood & Jantzi, 2008). CTE is nurtured by school leaders when they help clarify school goals, provide meaningful capacity-building and collaborative opportunities for staff, such as participation in a professional learning community (Voelkel & Chrispeels, 2017), and engage staff authentically in school decisions (Brinson & Steiner, 2007).

Organization path variables

Regression analysis indicated, as in Study 2, that SOE had a larger impact on student learning (β = .18) than either CSC (β = .06) or OPIT (β = −.11), although none of these effects were significant, when student SES was entered into the equation. These results provide only weak support for OPIT and seem inconsistent with the considerable weight of evidence about the value of collaboration in schools (e.g., Hill, 2010; Lomos et al., 2011). Indeed, all three studies in our series report relatively strong effects of SL on the organization path but relatively weak effects of this path on student achievement. One plausible reason for this inconsistency may be competition from other four path variables in accounting for variation in student achievement, thus supporting the value of multivariable research rather than the typically single-variable research about influences on student achievement, in the future. All things equal, these results imply that School Leaders should not expect large payoffs in student achievement for time spent improving the status of variables on this path.

Family path variables

The regression analysis indicated that PSC had a larger impact on student learning (β = .26) than either PE (β = −.08) or FC (β = −.04), however, student learning was not significant when students’ SES was entered into the equation, possibly due to multicollinearity or suppressor variable effects. The data measuring the variables on this path reflecting parents–student relationships were provided by teachers’ estimates. This may have affected the accuracy of results. Neither Studies 1 or 2 provided evidence allowing for comparisons. The prominence of PSC in this study differs from the results of Jeynes’s meta analyses (2005) awarding largest effects to PE (an effect size of .58). But the most important practical implication of these results is to recommend Family Path variables to the attention of school leaders as part of their improvement efforts.

While viewed as lying outside the legitimate boundaries of their responsibility by many school leaders and staffs (Hornby & Lafaele, 2011), parent involvement in their children’s learning is widely acknowledged to have large positive effect on students’ academic success (Henderson & Mapp, 2002). Furthermore, while all students benefit from family involvement in education, the influence of parent engagement can mitigate differences in SES and family background (Jeynes, 2011). Evidence now indicates that those features of the home that contribute most to student success at school can be enhanced by the intentional and sustained efforts of school leaders and their staffs (Bolivar & Chrispeels, 2011; Jeynes, 2018; Leithwood & Patrician, 2015). Future research should aim to extend what is known about successful strategies school leaders and their staff can use to nurture important features of the home that contribute to student success at school.

A final regression model used two approaches. The first approach included only variables with the greatest influence on student learning as identified by the previous four regression analyses (i.e., DC, CTE, SOE, PSC). These variables were entered into the equation together with SES and SL. The regression equation yielded from this approach indicated that, including SES, the four paths accounted for 69% of the variance in student achievement, F (6, 68) = 24.91, p < .001. Of this variance, SES accounted for 62%. The only significant predictor of student learning among all path variables in this model was DC; this was the result even when DC was entered as the last path variable.

It would be difficult to predict the relative effects of the variables included on the four paths from a reading of most of the research about each of these variables. A large proportion of that research has been single-variable research as, for example, studies of academic emphasis (e.g., Hoy, Tarter, & Hoy, 2006), DC (e.g., Ma & Willms, 2004) TTO (e.g., Goddard et al., 2001), CTE (e.g., Angelle & Teague (2014), and TC (e.g., Janisch & Johnson, 2003).

In multivariable studies such as this one, each variable has to share the total explained variation with other variables and take its chances on the order in which variables are entered into regression equations. Focusing more future effort on multivariable research would help build an evidence base about explanations for variation in student achievement that reflect a much broader horizon than does most of the current evidence base.

Multivariable research of this type would have considerable practical value, as well, allowing for “optimizing” rather than “satisficing” forms of school improvement decision making. As we have argued elsewhere (e.g., Sun & Leithwood, 2016), school leaders’ improvement work entails making choices about the small number of variables in their schools on which they are able to focus their energies at any point in time. So, studies that examine the relative contribution to student success of several or more variables of interest at the same time seem to produce better estimates of the real-world effects of any one variable.

A structural equation model was tested to assess the direct and indirect effects on student achievement of the aggregate four paths variables as well as the indirect effects of SL on such achievement through each of the paths. SES, the exogenous control variable, was hypothesized to have a direct effect on student achievement and an indirect effect on student achievement through the four paths. Results (Figure 1) indicate that the rational path had a significant direct effect on student achievement (λ = .42; p < .05). The effects of the emotions, organization and family paths on student achievement were not statistically significant. While these results are consistent with Study 2 results, they are at odds with the results of Study 1 which found similar significant positive effects of the rational, emotional and family paths (b = .21 to .26; p < .05) on student achievement. Considering significant correlations between the three family path variables and SES, and the nonsignificant impact of the family path on student learning may be due to multicollinearity or suppressor variable effects between SES and the three family path variables.

OPEN IN VIEWER

Figure 1.

The impact of school leadership on student achievement through the four paths (N = 81).

A significant modification of the relations among the four path variables in the theoretical model is called for by these results. While the version of the four paths model tested in this study assumed significant direct effects on achievement of all four paths, along with unspecified interactions among paths, results indicated that only the rational path had significant direct effect on student achievement (Study 1 found such results for both the rational and family paths). The other paths influenced achievement through their influence on the rational path. These results raise questions about whether and how specific variables on the emotions, organizational, and family paths influence DC, AP, and UIT. For example, do changes in CTE have an influence on AP and if they do what are the mechanisms connecting the two? Logically, there are many such questions (27—a total of nine variables on the emotions, organizational and family paths and three variables on the rational path).

A careful analysis of existing research reveals fairly robust answers to some of those questions but certainly not the majority. Studies of teacher trust typically test the direct effect of such trust on achievement, sometimes speculating on mediating conditions without actually testing them (e.g., Goodard, Tschannen-Moran, & Hoy, 2001). So, one important direction for future theory and research would entail unpacking the complex relationships between variables on the rational path and variables on the other three paths.

SEM analysis (Figure 1) indicates significant direct effects of SL on all four paths—the rational (λ =.58), emotions (λ = .74), organizational (λ = .83), and family (λ = .34) paths. However, the indirect effects of SL on student achievement were achieved mainly through the rational path. School leadership explained 35.8% of the variance in student achievement through the rational path, controlling for SES. SES had a significant direct effect on student achievement (β = −.68; p < .05). The path model had an acceptable model fit (χ² = 0.13, df = 1; p = .72; root mean square error of approximation < .05; goodness of fit index = .99).

The size of the indirect effects of SL on student achievement is remarkably similar to estimates reported in Study 1 (β = .12; p < .05), as well as by others (e.g., Heck & Hallinger, 2014) These results suggest that whatever school leaders do to influence variables on the other paths, the goal should be to ensure improvement of variables on the rational path (AP, DC, and UIT).).