Understanding the Relationship Between Urban-Centric Locale and Teachers’ Perceptions of School Leadership as a Working Condition

Abstract

Using the 2012 North Carolina Teacher Working Conditions Survey data, school demographic information, and school’s urban-centric locale census designation, hierarchical linear modeling was conducted to examine the relationship between locale and teachers’ perceptions of school leadership as a working condition and explore any variance in the relationship between school contextual factors and teachers’ perceptions of school leadership associated with locale. After controlling for school contextual factors, the results showed urban-centric locale is not a predictor of teachers’ perceptions of school leadership, and most of the variance in teachers’ perceptions of school leadership is explained by factors within, not between, schools.

Keywords

urban schools school leadership principals working conditions

Introduction

Research has consistently identified teachers as being the most critical school-specific factor influencing student achievement (Chingos, Whitehurst, & Lindquist, 2014; Hanushek, 2010; Rivkin, Hanushek, & Kain, 2005; Wright, Horn, & Sanders, 1997). The effect of quality of instruction on student learning is even greater for low-income students than their peers from middle-class backgrounds (Gordon, Kane, & Staiger, 2006; Rivkin et al., 2005). Unfortunately, districts’ efforts to improve teacher effectiveness have been significantly hindered by high rates of teacher turnover, with the highest rates often associated with rural and urban schools who serve a higher percent of low-income students than their suburban peers (Ladd, 2011; National Commission on Teaching and America’s Future [NCTAF], 2003). Although some degree of turnover is healthy, especially if the teachers leaving are of low quality, continual high rates of turnover undermine school improvement efforts (Charlotte Advocates for Education, 2004; Johnson, 2006).

By and large, teacher retention is most influenced by a teacher’s working conditions (Borman & Dowling, 2008; Guarino, Santibañez, & Daley, 2006; Hanushek, Kain, & Rivkin, 2004; Ingersoll, 2001; Ladd, 2011; NCTAF, 2003). Furthermore, among the different types of working conditions, school leadership is consistently related to teacher retention (Certo & Fox, 2002; Ladd, 2011; Turner, 2003). For instance, national data show that for the 2007-2008 school year, 25.1% of teachers who left the profession and 34.7% of teachers who changed schools indicated dissatisfaction with school administration as a very important or extremely important factor in their decision to leave their school (U.S. Department of Education, National Center for Education Statistics [NCES], 2010). School leadership is important because principals can ease or add to the burdens teachers feel as they play an important role in shaping school leadership and management, supportive relationships, feelings of success, and opportunities to grow—all of which contribute to the level of difficulty of a teacher’s job (Reichardt, Snow, Schlang, & Hupfeld, 2008).

The purpose of this study is to improve our understanding of school leadership as a working condition. Although we consider a range of teacher and school factors in our analysis, our focus is on urban-centric locale because rural and urban schools, particularly those in low-income, high-minority areas, traditionally have greater difficulty retaining teachers as compared with suburban and town locales (NCTAF, 2003). Although some researchers have demonstrated that urban-centric locale matters to teachers’ perceptions of school leadership, the mechanisms explaining why it matters remain unclear. It could be that urban-centric locale serves as a proxy measure for teacher salary, teacher labor market conditions (such as greater competition in suburban locales), or contextual factors—such as school size, class size, home and neighborhood environments, or characteristics of the student populations (Jacob, 2007; Ladd, 2011; Monk, 2007). For example, regarding contextual factors, urban schools are more likely to have a context of higher rates of crime, unemployment, and poverty than other locales, and are less likely to have the social scaffolding within neighborhoods that provide for collaborative efforts in nurturing children (Jacob, 2007). Rural areas often have distinctions that result from a dependence on agricultural commerce. In particular, student populations in rural areas are likely to have “relatively large shares of students with special needs and with limited English skills and lower shares of students attending college” and “highly mobile children of low-income migrant farm workers” (Monk, 2007, p. 155).

To discern whether locale is simply a proxy measure or that other factors are at play, this study controls for some contextual variables. In doing so, this study provides insights into understanding the relationship between a school’s urban-centric locale and teachers’ perceptions of school leadership, ultimately expanding our understanding of working conditions and subsequently informing our efforts to reduce teacher turnover associated with school leadership as a working condition. Two research questions guide this inquiry:

Research Question 1: Controlling for teacher and school demographics, as well as other working conditions, what is the relationship between a school’s locale and teachers’ perceptions of school leadership as a working condition?

Research Question 2: To what extent does the relationship between school contextual factors and teachers’ perceptions of school leadership as a working condition vary according to a school’s locale?

To answer these research questions, we draw on data from the 2012 administration of the North Carolina Teacher Working Conditions (NC TWC) Survey and the NCES urban-centric locale designations for individual schools. We ran several hierarchical linear regression analyses with teacher- and school-level information to predict teachers’ perceptions of school leadership.

With working conditions being a strong predictor of teachers’ expressed intentions of leaving (Hirsch & Church, 2009; Johnson, 2006; Ladd, 2009), improving working conditions can be a valuable and cost-effective strategy for improving teacher retention. Many studies focus on working conditions, but they tend to highlight factors associated with the demographics of the student population. A school’s demographics can be difficult to change (as evident by the persistent inequities in resources and school resegregation). The effectiveness of a school’s leadership, however, is a working condition that can be more easily addressed (Jacob, 2007).

Some existing research addresses the relationship between working conditions and a school’s urban-centric locale. However, most of the research focusing on variances in urban and rural locales examines differences in salary and school demographics. Although salary seems to be an important factor in teachers’ decisions to work in an urban school, working conditions appear to be an even more important factor (Jacob, 2007). For example, Abel and Sewell (1999) found a “significantly higher level of stress for urban versus rural school teachers from poor working conditions . . ., and poor staff relations” (p. 920). To put it differently, teachers in rural areas are generally more likely to indicate satisfaction with their working conditions than teachers in other locales (Monk, 2007; Provasnik et al., 2007). By and large, these studies do not sufficiently account for school-level context, and as such do not allow us to understand to what extent urban-centric locale may be a proxy for other constructs, most notably student demographics. The results of this study will contribute to the field of knowledge regarding the relationship between urban-centric locale and teacher perceptions of working conditions and may provide an additional strategy for efforts to improve teacher retention. For example, knowledge of the existence, or lack, of differences in perceptions of school leadership across urban-centric locales could better inform superintendents of school districts comprised of a variety of locales to better interpret teacher feedback about school leadership.

In the next section, we provide a brief overview of the literature on school leadership as a working condition and the relationship between school locale and teachers’ perceptions of school leadership. Then, we delineate the methodology used in this study and describe the results associated with the two research questions. The last section includes a discussion of our key findings, presents theoretical and practical implications for the study’s findings, and specifies directions for future research.

School Leadership as a Working Condition

Research has frequently identified school leadership as a working condition as one of the most important factors contributing to teacher retention (Boyd et al., 2011; Certo & Fox, 2002; Ladd, 2011; Reichardt et al., 2008; Tomon, 2009; Turner, 2003; Turner, 2008). According to the Teacher Follow-Up Survey to the Schools and Staffing Survey for the base year 2008-2009, of teachers who moved to another school (“movers”), 17.4% indicated dissatisfaction with the lack of support from administration as an either very or extremely important factor in their decision to leave (U.S. Department of Education, National Center for Education Statistics [NCES], 2010). Also, data from a job satisfaction survey administered to a non-random sample of teachers in North Carolina showed that of the constructs considered—discipline, administrative duties, administrative support, parental support, salary, compensation, level taught, students’ socio-economic standing, class size, age, and teaching experience—administrative support emerged as the most salient predictor in the model (Turner, 2003).

In addition, Certo and Fox’s (2002) study of teacher attrition and retention in Virginia showed a lack of administrative support at both the district and school levels as a prominent reason for teacher attrition. Also, drawing on interview data from former and current school personnel, Reichardt et al. (2008) identified two themes regarding teacher attrition and retention. First, teacher attrition reflects “how difficult it is to work in a given place,” and, second, school leadership plays an important role in “easing or adding to the burdens that teachers feel” (p. 5). Specifically, because principals can influence (although not completely control) the factors that contribute to the difficulty of a teacher’s job—effective school leadership and management, supportive relationships, feeling successful, and opportunities to grow—they can influence teacher retention.

Ladd’s (2011) analysis of the 2006 iteration of the NC TWC Survey revealed a negative correlation between the probability a teacher will leave his or her school and the perceived quality of school leadership. One caveat to this, noted Ladd, is that at the elementary- and middle-school levels, high-quality school leadership decreased the probability a teacher will transfer to another school, but it did not protect against leaving the teaching profession. Tomon (2009) analyzed the 2006 NC TWC Survey with a specific focus on middle-school teachers. Her analysis demonstrated that middle-school teachers in North Carolina were more likely to stay at their current school if they had positive perceptions of their working conditions, particularly in the domain of school leadership. Tomon found that both beginning and career teachers valued an atmosphere of trust and mutual respect within the school, effective leadership from the school improvement team, a shared vision, adherence to high professional standards for delivering instruction, feedback that helped improve instruction, effective school leadership, and sustained efforts by school administration to address teacher concerns about facilities and resources, use of time, professional development, empowering teachers, and leadership issues.

Boyd et al.’s (2011) survey of former New York City teachers who left after the first year of teaching and second-year teachers who indicated they considered leaving their current schools the previous year found that “well over 40% of both groups identified dissatisfaction with the administration as the most important factor” influencing their decision to leave or consider leaving (p. 327). Furthermore, of the six working conditions examined in the study (teacher influence, administration, staff relations, student behavior, facilities, and safety), “the administration factor is the only one that significantly predicts teacher retention decisions after controlling for other school and teacher characteristics” (p. 323). Overall, the research indicates that school leadership is the most important predictor of teacher retention. What still remains unclear is what teacher-level and school-level factors predict teachers’ perceptions of school leadership, and to what extent school locale plays a role after controlling for other contextual variables.

School Locale and Teachers’ Perceptions of School Leadership

Hanushek and Rivkin (2007) analyzed data from the 1999-2000 Schools and Staffing Survey and surmised that teachers in urban districts across the nation were more likely to indicate poorer working conditions, including administrator support, than teachers in suburban or rural communities. They also reported teachers in urban districts were more likely to report job dissatisfaction than teachers in other districts. Similarly, a study conducted by Abel and Sewell (1999) demonstrated that, among the 98 participating secondary-school teachers in Georgia and North Carolina, there was no significant difference in stress resulting from pupil misbehavior and from time pressures between rural and urban schools; however, urban teachers self-reported greater stress resulting from poor working conditions and poor staff relations than did the rural teachers. However, due to the small sample size and limited geographic area included in their study, these results may not be generalizable.

Turner’s (2008) analysis of data from a survey regarding teachers’ perceptions of principal leadership, combined with data from the Virginia Department of Education and the NCES Common Core of Data, showed a significant difference in how both rural and urban teachers perceived four of five principal leadership domains: vision, mission, and culture; curriculum and classroom instruction; collaboration and shared leadership; and effective management. There was no difference in perceptions of family and community involvement between rural and urban teachers.

Although the research consistently shows differences in how teachers in different locales perceive school leadership as a working condition (Turner, 2008), the magnitude of this relationship, once we control for teacher-level and school-level factors, is not fully understood. Furthermore, the extent to which the relationship between school contextual factors and school leadership varies according to urban-centric locale remains unclear. This study seeks to address these gaps.

To do so, we operationalize urban education in two ways. First, in this study, urban education denotes teachers and leaders located in schools inside a census-designated urbanized area or principal city. Second, we account for working conditions (e.g., facilities and resources) and a schools’ student demographic profile (e.g., percent of student body receiving free or reduced-price lunch and percent of the student body that is minority) that are often indicative of urban education.

Method

This section explains the methodology used to answer the research questions. The data sources, variables, and statistical methods used to analyze the data are described.

Data Sources

Data for this study are drawn from the 2012 iteration of the NC TWC Survey and NCES designations of urban-centric locale based on the 2010 U.S. Census data. Designed and implemented by the New Teacher Center (NTC), the NC TWC has been administered to teachers, administrators, and all other licensed educators in North Carolina’s public schools biennially since 2002. Describing both the NC TWC and South Carolina TWC Surveys, Leithwood (2006) wrote,

The methodological rigor of these studies, impressive size of the sample, and remarkable response rate, along with the consistency of results and especially their examination of relationships among teacher working conditions, student achievement, and student retention make them exceptionally useful. (p. 49)

Given these attributes of the survey’s design and sample, it is certainly well suited to provide robust conclusions related to the research questions.

Sample

All educators employed in North Carolina public schools were eligible to participate in the 2012 survey, and 100,042 educators (1,790 principals, 1,992 assistant principals, and 87,562 teachers) chose to respond, achieving an 86.22% participation rate across the state (Hirsch & Maddock, 2012). Given that the focus of this study is on teachers’ perceptions, responses from those indicating their position as principal, assistant principal, or other educational professional (e.g., school counselor, school psychologist, social worker, etc.) were omitted. All charter schools and “special schools” (e.g., the North Carolina School for the Deaf at Morganton) were excluded from the data set because of insufficient information about school contextual factors for these schools. Thus, 2,405 schools with a total of 85,306 teacher respondents were included in this study.

Survey Instrument

The 2012 NC TWC Survey collected data related to eight working conditions: time, facilities and resources, community support and involvement, managing student conduct, teacher leadership, school leadership, professional development, and instructional practices and support (North Carolina Teacher Working Conditions Initiative, 2012). Participants also responded to three basic demographic questions (i.e., role in school, years of experience in education, and years in current school). “Analyses of the psychometric soundness of this and previous iterations of the TWC Survey indicate that it is a reliable and valid measure of the presence of teaching conditions in participating schools” (NTC, 2012, p. 1). Given that an alpha coefficient ranges from 0 to 1, with higher coefficients indicating higher levels of instrument consistency, it was deemed that “All eight constructs are reliable with alphas above 0.863” (p. 3).

Administration of survey instrument

Each school’s North Carolina Association of Educators (NCAE) representative was asked to hold a faculty meeting and disseminate a letter from Governor Beverly Perdue to each participant in the school (Strickland, 2012). Each copy of the letter explained the purpose of the survey and included a unique anonymous code, usable only once, to enter into the 2012 NC TWC Survey website. The code, although anonymous, did link each individual participant to his or her respective school because of the minimum participation requirements of the NTC research design (i.e., at least five participants and a participation rate of at least 40% per school). The NCAE representative was asked to encourage participants within the same school to trade codes before accessing the survey, if they so chose, to help ease concerns over anonymity.

Participants could complete the survey from any Internet connection and at any time during the window in which the survey was open (NC TWC Initiative, 2012). A preview of the survey was made available on the NC TWC Survey website allowing individuals to review the questions prior to participating. The first page of the study reiterated the anonymity of the given code and encouraged honest responses to increase the survey’s effectiveness in collecting useful data.

Only one answer was allowed for each question (NC TWC Initiative, 2012). A Likert-type scale (strongly disagree, disagree, agree, strongly agree, don’t know) was provided for response options for the significant majority of the survey questions. Several questions on the survey did not use this same Likert-type scale response format, using choices relevant to the question, such as no role at all, small role, moderate role, large role, and don’t know.

Variables

Outcome variable

In the 2012 NC TWC Survey, there are 20 items that measure teachers’ perceptions of the effectiveness of their respective school leader. Each item was measured on a Likert-type scale, where 1 = strongly disagree, 2 = disagree, 3 = agree, and 4 = strongly agree. Respondents could also select 5 = don’t know. For this study, “don’t know” responses were recoded as “system missing.” Confirmatory factor analysis (CFA) was conducted to determine whether there were any underlying latent factors. Conducting CFA served to reduce the number of factors considered in the research model and created a composite outcome variable comprised of different items.

Eigenvalues, scree plots, and rotated component matrices resulting from the CFA were considered in determining the number of factors to create to represent the outcome variable. In general, there is a preference for selecting factors with an eigenvalue greater than one; however, at times, this may inaccurately include or exclude factors (Rahn, n.d.). Therefore, the location of the point at which the slope of the curve created by the eigenvalues begins leveling off in the scree plot and the factor loadings shown in the rotated component matrices created are considered as well. In the rotated component matrices (using varimax rotation), if a factor did not have at least three variables with their highest loadings on that factor, it was strongly considered for exclusion. In addition, the results were reviewed by the researchers to determine whether the variables that load highly on the same factor were a logical grouping of variables (i.e., they have a common concept, such as items related to instructional time spent on assessments and items related to instructional time spent on instruction).

Then, item reliability analysis was conducted to determine whether any items needed to be excluded (Gliem & Gliem, 2003). As an indication of strong internal consistency, a Cronbach’s alpha coefficient of greater than .7 was preferred, although not strictly adhered to as there is no lower limit for this value. The results from the factor analysis and item reliability analysis informed the creation of a composite variable to represent teachers’ perceptions of school leadership as a working condition. Table 1 delineates the composition of the new composite variable, school leadership, which represents the outcome variable of interest for this study.

Table 1.

School Leadership Composite Variable.

Cronbach’s α	Variable description
.946	The faculty and staff have a shared vision.
	There is an atmosphere of trust and mutual respect in this school.
	Teachers feel comfortable raising issues and concerns that are important to them.
	The school leadership consistently supports teachers.
	The school improvement team provides effective leadership at this school.
	The faculty are recognized for accomplishments.
	Teachers are held to high professional standards for delivering instruction.
	The school leadership facilitates using data to improve student learning.
	Teacher performance is assessed objectively.
	Teachers receive feedback that can help them improve teaching.
	The procedures for teacher evaluation are consistent.

Primary predictor

The primary predictor variable of interest in this study is termed here the urban-centric locale classification of each school as designated by the NCES based on the U.S. Census Bureau’s (2010) Census. The school urban-centric locale classifications were gathered from the 2011-2012 school year data via the NCES Elementary/Secondary Information System (ElSi) table-generating tool. There are four main categories of urban-centric locale: city, suburb, town, and rural area. A city is an urban-centric locale including territories inside an urbanized area and inside a principal city. A suburb is an urban-centric locale including territories outside a principal city but inside an urbanized area. A town is an urban-centric locale including territories outside an urbanized area but inside an urban cluster. A rural area is an urban-centric locale including territories that are outside urban clusters and urbanized areas.

The single variable of school locale was reclassified into four dichotomous dummy variables (where yes = 1 and no = 0). Given that much of the research regarding teacher-working conditions discusses the conditions of urban and rural schools in comparison with suburban schools, suburb is the referent group for this study.

Control variables

Three sets of control variables were included in the model. One set of control variables includes two variables about teachers’ experience: number of years of experience in education and number of years employed in their current school. The second set of control variables represents school characteristics including school size, percent of student body receiving free or reduced-priced lunch, and percent of the student body that is minority.

The third set of control variables represents other items on the 2012 NC TWC Survey that measure teachers’ perceptions of other working conditions often associated with perceptions of school leaders’ effectiveness: time, facilities and resources, community support and involvement, managing student conduct, teacher leadership, professional development, and instructional practices and support. Steps to construct these composite variables, described in Appendix A, are identical to those used to create the outcome variable school leadership.

Data Analysis

Missing Value Analysis

Data that are not missing at random can affect data analysis and lead to misleading results. Little’s missing completely at random (MCAR) test was run using SPSS Version 21 (2012), and less than 5% of the data were missing for all variables. In fact, the significant majority of the variables were missing less than 0.5% of the data. The variables with more than 3% of missing data were years at current school (3.3% missing) and usefulness of data (3.2%). Because there was minimal missing data in this data set, multiple imputation was not necessary for this study (Fichman & Cummings, 2003).

Statistical Model: Hierarchical Linear Model

Because teachers are nested within the context of schools, school characteristics (i.e., urban-centric locale, average daily membership, percent of student body receiving free or reduced-price lunch, and percent of the student body that is minority) should not be overlooked when predicting teachers’ perceptions of school leadership. Hierarchical linear modeling (HLM) allows for simultaneous analysis of relationships within (teacher-level) and between schools (school-level) (Woltman, Feldstain, MacKay, & Rocchi, 2012).

A preliminary analysis of the data (null model) was evaluated to determine whether there was sufficient variability between schools to conduct further analysis (Wech & Heck, 2004). This model was fully unconditional and did not include any predictors (see Appendix B). The statistical significance of the chi-square tests of the null model and the intraclass correlation coefficient (ICC = σ² / [τ₀₀ + σ²]), which determines the percentage of the variance attributable to school-level characteristics, provide justification for running a two-level model. A low ICC suggests that two-level analysis would not yield results different from analysis of Level 1 data only. For this study, if ICC is too low, then between-school variance (school predictors) is not significant.

The random intercepts model (Model 2) measured within-school variance in school leadership associated with teacher-level predictors (number of years of experience the teacher has as an educator, the number of years the teacher has taught in the current school, and teacher’s perceptions of working conditions other than school leadership). There are no predictors in the Level 2 model of the random intercepts-only model (see Appendix B).

To create Model 4, interaction terms were added to Level 2 to examine whether perceptions in school leadership vary by urban-centric locale. We ran several models with different locale interaction terms (not shown here), replacing the variable suburb with city, town, and rural area, to examine interaction effects.

Multicollinearity Diagnostics

The data set was examined for multicollinearity, which occurs when an outcome variable is highly correlated with another variable (Lüchters & Chakrabarty, 2006). Collinearity diagnostics were performed using SPSS Version 21 in which tolerance was measured (1 − R²), and the variance inflation factor (VIF) was calculated (1/1 − R²) for each variable. The VIF indicates how much of the standard error (SE) of the estimation of the variable’s effect is inflated by being highly correlated with another variable. For this study, a tolerance value under 0.20 accompanied by a VIF greater than four will generally be considered an indication of collinearity (Garson, 2012). This is a conservative cutoff point for the VIF. The tolerance value was above 0.20 and the VIF below four for all variables, indicating that high multicollinearity was not a concern. In addition, because multicollinearity diagnostics can be biased downward for binary variables, bivariate statistics were run for the binary variables in the model, notably chi-square statistics using cross tabs. For nominal variables, Phi and Cramer’s V values were examined. For ordinal variables, Tau-b values were examined. Of the correlations that were statistically significant, they were slight (i.e., < |0.1|), reaffirming our conclusions based on multicollinearity diagnostics, that is, multicollinearity is not of concern for our analyses.

Results

Descriptive statistics of the teacher-level binary variables are presented in Table 2. Results for the maximum and minimum values, means, and standard deviations are provided.

Table 2.

Descriptive Statistics.

Variable description	N	Minimum	Maximum	M	SD
Years experience in education	85,216	1.00	6.00	4.23	1.46
Years at current school	82,485	1.00	6.00	3.22	1.48
School use of time	85,249	1.00	4.00	2.68	0.60
Time on planning	85,194	1.00	6.00	2.92	0.83
Time on non-instruction responsibilities	85,240	1.00	6.00	2.62	0.56
Time on assessments	85,170	1.00	6.00	2.79	0.87
Time outside workday^a: none	85,222	0.00	1.00	0.01	0.10
Time outside workday: <1 hr	85,222	0.00	1.00	0.06	0.23
Time outside workday: 1 < 3 hr	85,222	0.00	1.00	0.18	0.39
Time outside workday: 3 < 5 hr	85,222	0.00	1.00	0.23	0.42
Time outside workday: 5 < 10 hr	85222	0.00	1.00	0.27	0.44
Time outside workday: >10 hr	85,222	0.00	1.00	0.25	0.43
Resources	85,237	1.00	4.00	3.02	0.57
Facilities	85,123	1.00	4.00	3.16	0.61
Community/parent involvement	84,962	1.00	4.00	3.12	0.51
Community/parent support	84,491	1.00	4.00	2.93	0.62
Managing student conduct	83,984	1.00	4.00	2.98	0.60
Teacher leadership	85,203	1.00	4.00	3.00	0.59
Teacher role: class decisions	83,431	1.00	4.00	3.19	0.69
Teacher role: school decisions	83,904	1.00	4.00	2.58	0.72
Professional development	84,942	1.00	4.00	2.90	0.53
Usefulness of assessment data	82,564	1.00	4.00	2.75	0.60
Instructional practices	84,790	1.00	4.00	3.05	0.47
School leadership	85,142	1.00	4.00	3.05	0.59
% minority	85,298	0.00	1.00	0.48	0.26
% eligible for free/reduced lunch	85,298	0.00	1.00	0.58	0.23
Average daily membership	85,298	1.00	2,736.00	756.61	442.14
City	85,306	0.00	1.00	0.25	0.43
Suburb	85,306	0.00	1.00	0.14	0.35
Town	85,306	0.00	1.00	0.13	0.34
Rural	85,306	0.00	1.00	0.47	0.50

Note. Mean for a binary variable is the percent of respondents indicating that variable is present.

Average weekly amount of time spent on school activities.

Findings

As shown in Table 3, to account for teachers being nested within schools, we conducted HLM. HLM simultaneously accounts for variance in our dependent variable, school leadership, which is measured at the teacher level—resulting from both teacher-level factors (teacher demographics and perceptions of other working conditions) and school-level factors (school contextual variables and locale).

Table 3.

Outcome Variable: School Leadership Composite Variable.

Variable	Model 1		Model 2		Model 3		Model 4^a
Variable	β	SE	β	SE	β	SE	β	SE
Intercept	3.07***	0.01	3.07***	0.01	3.08***	0.02	3.09***	0.05
Years of experience in education			0.00**	0.00	0.00**	0.00	0.00**	0.00
Years at current school			−0.01***	0.00	−0.01***	0.00	−0.01***	0.00
School use of time			0.02***	0.00	0.02***	0.00	0.02***	0.00
Time on planning			0.00	0.00	0.00	0.00	0.00	0.00
Time on non-instruction responsibilities			−0.02***	0.00	−0.02***	0.00	−0.02***	0.00
Time on assessments			0.00	0.00	0.00	0.00	0.00	0.00
Time outside workday: none			0.00	0.01	0.00	0.01	0.00	0.01
Time outside workday: <1 hr			−0.01**	0.00	−0.01**	0.00	−0.01**	0.00
Time outside workday: 1 < 3 hr			0.00	0.00	0.00	0.00	0.00	0.00
Time outside workday: 3 < 5 hr			0.00	0.00	0.00	0.00	0.00	0.00
Time outside workday: >10 hr			0.00	0.00	0.00	0.00	0.00	0.00
Resources			0.00	0.00	0.00	0.00	0.00	0.00
Facilities			0.03***	0.00	0.03***	0.00	0.03***	0.00
Community/parent involvement			0.08***	0.00	0.08***	0.00	0.08***	0.00
Community/parent support			−0.01***	0.00	−0.01***	0.00	−0.01***	0.00
Managing student conduct			0.19***	0.00	0.19***	0.00	0.19***	0.00
Teacher leadership			0.39***	0.00	0.39***	0.00	0.39***	0.00
Teacher role: class decisions			0.00	0.00	0.00	0.00	0.00	0.00
Teacher role: school decisions			0.06***	0.00	0.06***	0.00	0.06***	0.00
Professional development			0.16***	0.00	0.16***	0.00	0.16***	0.00
Usefulness of assessment data			−0.02***	0.00	−0.02***	0.00	−0.02***	0.00
Instructional practices			0.15***	0.00	0.15***	0.00	0.15***	0.00
Percent minority					0.00***	0.00	0.00***	0.00
Percent free/reduced lunch					0.00**	0.00	0.00**	0.00
ADM					0.00***	0.00	0.00***	0.00
Locale: City (suburb is referent)					−0.02	0.02	−0.04	0.05
Locale: Town (suburb is referent)					0.01	0.02	−0.01	0.05
Locale: Rural (suburb is referent)					0.01	0.02	−0.01	0.05
% Minority × Suburb							0.00	0.00
% Free/Reduced Lunch × Suburb							0.00	0.00
ADM × Suburb							0.00	0.00

Variance (intercept)	0.07		0.08		0.07		0.07
Level 1, r	.28		.08		.08		0.08
ICC	0.20		0.50		0.47		0.47
Variance between schools	—		0.00^b		0.00^b		0.00^b
Variance within schools	—		0.72		0.72		0.72
χ²(df)***	21,654.59*** (2,402)		69,566.00***(2,402)		63,494.94*** (2,396)		63,290.35*** (2393)
Deviance	138,742.87		33,396.16		33,216.32		33,254.85

Note. ADM = Average daily membership; ICC = intraclass correlation coefficient.

Model 4 was also run with interaction terms for city and rural area locales. As was true of the interaction terms for suburb locale, these interaction terms were not statistically significant. They were not displayed in the table for simplicity of presentation.

The value is zero or less (which means r² = 0; specifically, the school-level variables we have included in the model do not explain the variance in teacher perceptions of school leadership).

p < .10. **p < .05. ***p < .01.

We first ran a fully unconditional model (null model) to compute the ICC to determine whether a multi-level model was needed—that is, if there is sufficient between-school variance to run a two-level model. As shown in Model 1, the chi-square statistic is statistically significant and the ICC (ratio of the between-group variance to total variance) was .20, indicating that 20% of the variance in teachers’ perceptions of school leadership is attributed to between-group differences (i.e., Level 2 variables). These findings suggest that we should run a two-level model.

The single-level analysis (Model 2) only included Level 1 variables. As demonstrated in Model 2, although a teacher’s years of experience in education had a statistically significant (p ≤ .01) relationship to perceptions of school leadership as a working condition, the magnitude of the relationship is negligible (β = 0.00, SE = 0.00). The number of years the teacher has been working in the current school was also a statistically significant (p ≤ .001) relationship, and the magnitude was β = −.01 (SE = 0.00). The deviance for Model 2 showed a 76% decrease from the deviance of Model 1. The variance (intercept) component for Model 2 was statistically significant (p ≤ .001), indicating that school leadership was significantly related to the predictors (Garson, 2013). Model 2 explained 72% of the variance in perceptions of school leadership as a working condition. That is, teacher demographics and the Level 1 control variables accounted for 72% of the variance within schools of teacher perceptions of school leadership as a working condition.

As for the Level 1 control variables, the predictors with a statistically significant positive relationship to school leadership were school use of time (β = .02, SE = 0.00, p ≤ .001), facilities (β = .03, SE = 0.00, p ≤ .001), community/parent involvement (β = .08, SE = 0.00, p ≤ .001), managing student conduct (β = .19, SE = 0.00, p ≤ .001), teacher leadership (β = .39, SE = 0.00, p ≤ .001), teacher role: school decisions (β = .06, SE = 0.00, p ≤ .001), professional development (β = .16, SE = 0.00, p ≤ .001), and instructional practices (β = .15, SE = 0.00, p ≤ .001).

The predictors with a statistically significant negative relationship to school leadership were time on non-instructional responsibilities (β = −.02, SE = 0.00, p ≤ .001), community/parent support (β = −.01, SE = 0.00, p ≤ .001), and usefulness of assessment data (β = −.02, SE = 0.00, p ≤ .001). The predictors that did not have a statistically significant relationship with school leadership were time on planning, time on assessments, resources, and teacher role: class decisions. Of the variables indicating the average amount of time a teacher spent each week on school-related activities outside the workday, the only variable that was statistically significant was less than 1 hr (β = −.01, SE = 0.00, p ≤ .01).

Six school contextual variables were added to Model 2 to create Model 3 (Levels 1 + 2). The variables percent minority, percent free/reduced lunch, and average daily membership were all statistically significant (p ≤ .01), but the magnitudes were negligible (β = .00, SE = 0.00). The three variables related to urban-centric locale (city, town, and rural area) were not statistically significant. Also, the direction and magnitude of the control variables remained the same after the addition of the school contextual variables. The model was statistically significant (p ≤ .001), and the deviance decreased to 33,216.32 (from 33,396.16 for Model 2). Model 3 did not explain any of the variance between schools of teachers’ perceptions of school leadership; however, it did account for 72% of the variance in teachers’ perceptions of school leadership within schools.

Considering the fact that some research has shown student demographics act as proxies for working conditions (Hanushek et al., 2004; Hanushek & Rivkin, 2007; Horng, 2009), it seemed prudent to further explore the relationships of school contextual factors and school locale with teachers’ perceptions of school leadership. To do so, the school-level variables were removed from the model, and Model 3 was analyzed again. These results (not presented) revealed a statistically significant relationship between city locale and teachers’ perceptions of school leadership, indicating that urban-centric locale seems to serve as a proxy for the school contextual factors. Last, interaction terms for urban-centric locale were added at Level 2 to Model 3 to create Model 4. None of these interaction terms was statistically significant, indicating that the relationship between school contextual factors and teachers’ perceptions of school leadership does not vary by urban-centric locale.

Discussion and Conclusion

The extant literature has shown that a school’s student demographic profile is associated with both teacher quality and teacher retention. Schools with large percentages of Black and Hispanic/Latino(a) students or students qualifying for free or reduced-price lunch are most likely to have lower teacher-retention rates. These demographic profiles reflect the majority of schools in urban settings. To improve teacher quality and retention in urban schools and subsequently promote student achievement, considerable efforts have been made to improve teacher preparation and develop culturally relevant curricula. However, research has increasingly shown that working conditions matter as well (Berry, Smylie, & Fuller, 2008). Not only do working conditions improve teacher retention, but they also function as “opportunities to learn—and as a mediator of teacher qualifications and valued student outcomes” (p. 1). Thus, it stands to reason efforts devoted to improve working conditions would subsequently improve teacher retention, teacher quality, and student achievement. Although school leaders affect working conditions, very little attention has been focused on helping leaders learn how to minimize working conditions that encumber teachers (Reichardt et al., 2008). To this end, we focused on school leadership. Other key advantages to improving our understanding of school leadership as a working condition are that it is strongly associated with teacher retention and quality, it can reasonably be directly manipulated, it can be realistically taught in a preparation program, and it shapes other working conditions.

Specifically, we sought to discern whether school leadership as a working condition varied across locales. The results indicated that teachers in urban areas do perceive school leadership as a working condition different from their peers in other urban-centric locales. However, this relationship ceases once we control for a school’s student demographic profile. This finding contradicts some research that has shown that teachers’ perceptions of school leadership as a working condition vary by urban-centric locale (e.g., Hanushek & Rivkin, 2007; Turner, 2008). This is likely due to locale serving as a proxy for school context. More important, school leadership as a working condition is negligible. So, how teachers perceive school leaders, on average, has little to do with the school’s demographic profile. Indeed, our results show that generally, what the principal can most affect—in this student, teacher leadership, managing student conduct, professional development, and instructional strategy—is related to school leadership as a working condition. Conversely, what a school leader exercises the least control over, for example, facilities, has no or only slight association with teachers’ perceptions of school leadership as a working condition. This finding suggests that when we use teachers’ perceptions of principals to measure principal leadership effectiveness in schools with disadvantaged students, they are not going to be unfairly penalized because of student characteristics or school working conditions for which they exert little control. Of course, measures of effectiveness based solely on student achievement may paint a different picture.

Similar to Berry et al. (2008), much of the variance in school leadership is explained by factors within a school rather than between schools. This begs the question of how teachers in the same school can perceive leadership differently. It may be that characteristics of the teacher—grade level, subject matter, personality style, career intention, and so forth—affect how a teacher experiences a context or interacts with a principal. This indicates, as Berry et al. (2008) conclude, one style of leadership (or set of actions) may not yield the same responses from all teachers in the same school. It is not about whether a set of specific leadership activities work, but for whom and under what circumstances.

Implications

The findings have four major implications for research and practice. First, because most of the variance in school leadership is explained by teacher-specific factors, teachers within the same school are experiencing the principal’s leadership styles and decisions differently. This suggests, for research, it is important to move beyond the assumption that schools are monolithic units. Second, our findings indicate that improvements in teachers’ perceptions of teacher leadership, managing student conduct, professional development, and instructional strategies result in improvement in teachers’ perceptions of school leadership as a working condition, which has been strongly associated with teacher retention. So, principal preparation programs need to teach preservice principals how to shape these conditions. For example, because teacher leadership is associated with school leadership effectiveness and teacher retention, principals can be taught models that promote teacher leadership, such as distributive and transformational leadership. Also, considering that teachers’ perceptions of school leadership had a negative relationship with years of experience in education and a strong, positive relationship with teacher leadership, school leaders should consider providing teacher leadership opportunities to teachers with more years of experience. This approach could serve as a means to offset the negative effects of the relationship between years of experience and teachers’ perceptions of school leadership and potentially result in improved perceptions of school leadership, and, therefore, increased retention of experienced teachers. Finally, school contextual factors, such as facilities and resources and student demographics, have little to no relationship with teachers’ perceptions of school leadership as a working condition. This finding introduces doubt about some principals’ claims that teachers’ assessment of their effectiveness is largely a result of context and not of their leadership.

Limitations

This study is limited by the fact that teacher perceptions of school leadership is a complex issue, and the only aspects of school leadership considered in this study are those identified by the NC TWC Survey. As such, other valid measures or aspects of school leadership or possible predictors are not taken into account. The design of this study, for example, does not take into account all the contextual circumstances surrounding a school’s leadership. For instance, the number of years the principal or assistant principal has been serving in the school could be an important nuance to consider, and whether the school’s leadership was in flux during the 2011-2012 school year could affect the teachers’ perceptions of school leadership. The school’s academic performance, another factor that could affect teachers’ perceptions of school leadership, is also not taken into account.

This study also encounters the potential limitations inherent in relying on survey data. Teacher responses on working-condition surveys are likely to be affected by both their perceptions of their working conditions and their overall outlook, whether positive or negative (Boyd et al., 2011). Situational factors can also affect teachers’ responses. For example, an experience of a significant student behavior matter or of a negative interaction with school administration occurring in close proximity to the completion of the survey can have a magnified effect on the teacher’s perceptions of working conditions. These limitations, however, should be mitigated by the substantially large size of the data set (Boyd et al., 2011).

Although the data set was sufficiently large, it only included North Carolina’s public schools. It is conceivable that results might vary if a comparable study was conducted in another state, perhaps in a different region of the United States. Data from the 1999-2000 Schools and Staffing Survey (SASS) do indicate the potential for a difference in perception of school leadership in the northeast as compared with other regions (Hanushek & Rivkin, 2007).

Some may speculate that salary should be considered in this model. However, Ladd (2011), who accounted for salary and cost of living, concluded that there was little variation in salary across districts in North Carolina and stated, “Given this small variation, and the fact that teachers have already made their initial job decision with full knowledge of the salary, salary differentials are likely to have little or no predictive power in this cross-sectional analysis” (p. 245). Furthermore, in North Carolina, teacher salary is a variable over which school leaders have no control.

Future Research

Considering that this study’s finding that a school’s urban-centric locale did not have a statistically significant relationship to teachers’ perceptions of school leadership as a working condition is contradictory to the conclusions of other analyses (e.g., Hanushek & Rivkin, 2007; Turner, 2008), additional research in this area could help to clarify the nature of this relationship. Are the teachers and school contextual factors in North Carolina different from those of the nation at large to a degree sufficient to cause contradictory results? Future research could include qualitative research using teacher focus groups to determine what factors account for the variance in teachers’ perceptions of school leadership within a school. Also, sufficient variance in teachers’ perceptions of school leadership between schools existed in the data set analyzed, but the school-level variables used in this study were unable to explain this variance. Quite simply, more research is needed to identify factors that explain between-school variance in perceptions of school leaders.

Footnotes

Appendix A

New Composite Variables Reflecting Teachers’ Perceptions of Working Conditions Other Than School Leadership.

New composite variable	Cronbach’s α	Variable description
School use of time	.854	Class sizes are reasonable such that teachers have the time available to meet the needs of all students.
		Teachers have time available to collaborate with colleagues.
		Teachers are allowed to focus on educating students with minimal interruptions.
		The non-instructional time provided for teachers in my school is sufficient.
		Efforts are made to minimize the amount of routine paperwork teachers are required to do.
		Teachers have sufficient instructional time to meet the needs of all students.
		Teachers are protected from duties that interfere with their essential role of educating students.
Time on planning	.474	In an AVERAGE WEEK, how much time do you devote to the following activities during the school day (i.e., time for which you are under contract to be at the school)?
		Individual planning time
		Collaborative planning time
Time on non-instructional responsibilities	.692	In an AVERAGE WEEK, how much time do you devote to the following activities during the school day (i.e., time for which you are under contract to be at the school)?
		Supervisory duties
		Required committee and/or staff meetings
		Completing required administrative paperwork
		Communicating with parents/guardians and/or the community
		Addressing student discipline issues
		Professional development
Time on assessments	.765	In an AVERAGE WEEK, how much time do you devote to the following activities during the school day (i.e., time for which you are under contract to be at the school)?
		Preparation for required federal, state, and local assessments
		Delivery of assessments
		Utilizing results of assessments
Resources	.842	Teachers have sufficient access to appropriate instructional materials.
		Teachers have sufficient access to instructional technology, including computers, printers, software, and Internet access.
		Teachers have access to reliable communication technology, including phones, faxes and e-mail.
		Teachers have sufficient access to office equipment and supplies such as copy machines, paper, pens, etc.
		Teachers have sufficient access to a broad range of professional support personnel.
		The reliability and speed of Internet connections in this school are sufficient to support instructional practices.
Facilities	.801	The school environment is clean and well maintained.
		Teachers have adequate space to work productively.
		The physical environment of classrooms in this school supports teaching and learning.
Community/parent involvement	.843	Parents/guardians are influential decision makers in this school.
		This school maintains clear, two-way communication with the community.
		This school does a good job of encouraging parent/guardian involvement.
		Teachers provide parents/guardians with useful information about student learning.
		Parents/guardians know what is going on in this school.
Community/parent support	.870	Please rate how strongly you agree or disagree with the following statements about community support and involvement in your school.
		Parents/guardians support teachers, contributing to their success with students.
		Community members support teachers, contributing to their success with students.
		The community we serve is supportive of this school.
Managing student conduct	.900	Please rate how strongly you agree or disagree with the following statements about managing student conduct in your school.
		Students at this school understand expectations for their conduct.
		Students at this school follow rules of conduct.
		Policies and procedures about student conduct are clearly understood by the faculty.
		School administrators consistently enforce rules for student conduct.
		School administrators support teachers’ efforts to maintain discipline in the classroom.
		Teachers consistently enforce rules for student conduct.
		The faculty work in a school environment that is safe.
Teacher leadership	.930	Please rate how strongly you agree or disagree with the following statements about teacher leadership in your school.
		Teachers are recognized as educational experts.
		Teachers are trusted to make sound professional decisions about instruction.
		Teachers are relied upon to make decisions about educational issues.
		Teachers are encouraged to participate in school leadership roles.
		The faculty has an effective process for making group decisions to solve problems.
		In this school we take steps to solve problems.
		Teachers are effective leaders in this school.
		Teachers have an appropriate level of influence on decision making in this school.
Teacher role: class decisions	.761	Please indicate the role teachers have at your school in each of the following areas.Selecting instructional materials and resources.
		Devising teaching techniques.
		Setting grading and student assessment practices.
Teacher role: school decisions	.828	Determining the content of in-service professional development programs.
Teacher role: school decisions		Establishing student discipline procedures.
		Providing input on how the school budget will be spent.
		The selection of teachers new to this school.
		School improvement planning.
Professional development	.949	Please rate how strongly you agree or disagree with statements about professional development in your school.
		Sufficient resources are available for professional development in my school.
		An appropriate amount of time is provided for professional development.
		Professional development offerings are data driven.
		Professional learning opportunities are aligned with the school’s improvement plan.
		Professional development is differentiated to meet the individual needs of teachers.
		Professional development deepens teachers’ content knowledge.
		Teachers have sufficient training to fully utilize instructional technology.
		Teachers are encouraged to reflect on their own practice.
		In this school, follow up is provided from professional development.
		Professional development provides ongoing opportunities for teachers to work with colleagues to refine teaching practices.
		Professional development is evaluated and results are communicated to teachers.
		Professional development enhances teachers’ ability to implement instructional strategies that meet diverse student learning needs.
		Professional development enhances teachers’ abilities to improve student learning.
Usefulness of assessment data	.803	State assessment data are available in time to impact instructional practices.
Usefulness of assessment data		Local assessment data are available in time to impact instructional practices.
		State assessments provide schools with data that can help improve teaching.
		State assessments accurately gauge students’ understanding of standards.
Instructional practices	.839	Teachers use assessment data to inform their instruction.
		The curriculum taught in this school is aligned with Common Core Standards.
		Teachers work in professional learning communities to develop and align instructional practices.
		Provided supports (i.e., instructional coaching, professional learning communities, etc.) translate to improvements in instructional practices by teachers.
		Teachers are encouraged to try new things to improve instruction.
		Teachers are assigned classes that maximize their likelihood of success with students.
		Teachers have autonomy to make decisions about instructional delivery (i.e., pacing, materials and pedagogy).

Appendix B

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Dara H. Blackwell, as founder and chief evaluation principal for Blackstone Solutions, she draws on her extensive experience as a school administrator and research skills to evaluate school and district-level programs that aim to improve school leadership and student achievement.

Tamara Young is an associate professor in Educational Evaluation and Policy Analysis in the Department of Leadership, Policy and Human Development at North Carolina State University. Her research focuses on the implementation of educational policy and programs and the politics of education.

References

Abel

Sewell

(1999). Stress and burnout in rural and urban secondary school teachers. The Journal of Educational Research, 92, 287-293. doi:10.1080/00220679909597608

Berry

Smylie

Fuller

(2008). Understanding teacher working conditions: A review and look to the future. Hillsborough, NC: Center for Teacher Quality.

Borman

Dowling

N. M.

(2008). Teacher attrition and retention: A meta-analytic and narrative review of the research. Review of Educational Research, 78, 367-409.

Boyd

Grossman

Ing

Lankford

Loeb

Wyckoff

(2011). The influence of school administrators on teacher retention decisions. American Educational Research Journal, 48, 303-333. doi:10.3102/0002831210380788

Certo

Fox

(2002). Retaining quality teachers. The High School Journal, 86(1), 57-75.

Charlotte Advocates for Education. (2004). Role of principal leadership in increasing teacher retention: Creating a supportive environment. Charlotte, NC: Author. Retrieved from http://www.educationjustice.org/assets/files/pdf/Resources/Policy/Teaching%20and%20Leadership/Role%20of%20principal%20leadership%20in%20increasing%20teacher%20retention.pdf

Chingos

Whitehurst

Lindquist

(2014). School superintendents: Vital or irrelevant? Brown Center on Education Policy at Brookings. Retrieved from http://www.brookings.edu/~/media/research/files/reports/2014/09/superintendents-whitehurst-chingos/superintendentsbrown-center9314.pdf

Fichman

Cummings

(2003). Multiple imputation for missing data: Making the most of what you know. Organizational Research Methods, 6, 282-308. Retrieved from http://repository.cmu.edu/cgi/viewcontent.cgi?article=1114&;context=tepper

Garson

(2012). Testing statistical assumptions. Statistical Associates Publishing. Retrieved from http://www.statisticalassociates.com/assumptions.pdf

10.

Garson

(2013). Introductory guide to HLM with HLM 7 software. In Garson

(Ed.), Hierarchical linear modeling: Guide and applications (pp. 55-96). Thousand Oaks, CA: Sage.

11.

Gliem

(2003). Calculating, interpreting, and reporting Cronbach’s alpha reliability coefficient for Likert-type scales. Paper presented at the Midwest Research-to-Practice Conference in Adult, Continuing, and Community Education. Retrieved from https://scholarworks.iupui.edu/bitstream/handle/1805/344/Gliem%20&%20Gliem.pdf?s

12.

Gordon

Kane

Staiger

(2006). Identifying effective teachers using performance on the job (Discussion paper, The Hamilton Project). Washington, DC: Brookings Institution.

13.

Guarino

Santibañez

Daley

(2006). Teacher recruitment and retention: A review of the recent empirical literature. Review of Educational Research, 76, 173-208.

14.

Hanushek

(2010). The economic value of higher teacher quality (Working Paper No. 56). National Center for Analysis of Longitudinal Data in Education Research. Washington, DC: The Urban Institute.

15.

Hanushek

Kain

Rivkin

(2004). Why public schools lose teachers. The Journal of Human Resources, 39, 326-354.

16.

Hanushek

Rivkin

(2007). Pay, working conditions, and teacher quality. The Future of Children, 17(1), 69-86. Retrieved from https://www-jstor-org.web.bisu.edu.cn/stable/4150020

17.

Hirsch

Church

(2009). North Carolina teacher working conditions survey brief: Working conditions influence teacher retention. Retrieved from New Teacher Center.

18.

Hirsch

Maddock

(2012). North Carolina teacher working conditions initiative: Initial results from the 2012 NC TWC survey. New Teacher Center.

19.

Horng

E. L.

(2009). Teacher tradeoffs: Disentangling teachers’ preferences for working conditions and student demographics. American Educational Research Journal, 46, 690-717. doi:10.3102/0002831208329599

20.

Ingersoll

(2001). Teacher turnover and teachers shortages: An organizational analysis. American Educational Research Journal, 38, 499-534.

21.

Jacob

B. A.

(2007). The challenges of staffing urban schools with effective teachers. The Future of Children, 17(1), 129-153. doi:10.1353/foc.2007.0005

22.

Johnson

S. M.

(2006, July). The workplace matters: Teacher quality, retention, and effectiveness (Working paper, Best Practices). National Education Association. Retrieved from http://www.nea.org/assets/docs/HE/mf_wcreport.pdf

23.

Ladd

(2009). Teachers’ perceptions of their working conditions: How predictive of policy-relevant outcomes? Retrieved from New Teacher Center.

24.

Ladd

(2011). Teachers’ perceptions of their working conditions: How predictive of planned and actual teacher movement? Educational Evaluation and Policy Analysis, 33, 235-261. doi:10.3102/0162373711398128

25.

Leithwood

(2006). Teacher working conditions that matter: Evidence for change. Toronto, Canada: Elementary Teachers’ Federation of Ontario.

26.

Lüchters

Chakrabarty

(2006). Multicollinearity detection. Bangladesh: Shahjalal University of Science and Technology.

27.

Monk

D. H.

(2007). Recruiting and retaining high-quality teachers in rural areas. The Future of Children, 17(1), 129-153. doi:10.1353/foc.2007.0009

28.

National Commission on Teaching and America’s Future. (2003). No dream denied: A pledge to America’s children. Retrieved from http://nctaf.org/wp-content/uploads/no-dream-denied_full-report.pdf

29.

New Teacher Center. (2012). Research brief: Validity and reliability of the 2012 North Carolina teacher working conditions survey. Retrieved from New Teacher Center.

30.

North Carolina Teacher Working Conditions Initiative. (2012). North Carolina teacher working conditions main survey. Retrieved from New Teacher Center.

31.

Provasnik

KewalRamani

Coleman

M. M.

Gilbertson

Herring

Xie

(2007). Status of education in rural America. Washington, DC: National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education.

32.

Rahn

(n.d.). Factor analysis: A short introduction, part 4-how many factors should I find? Retrieved from http://www.theanalysisfactor.com/factor-analysis-how-many-factors

33.

Reichardt

Snow

Schlang

Hupfeld

(2008). Overwhelmed and out: Principals, district policy, and teacher retention. Hartford: Connecticut Center for School Change.

34.

Rivkin

Hanushek

Kain

(2005). Teachers, schools, and academic achievement. Econometrica, 73, 417-458.

35.

Strickland

(2012). NC12 letter instruction. Retrieved from New Teacher Center.

36.

Tomon

(2009). The impact of principal leadership on working conditions and teacher retention in North Carolina middle schools (Doctoral dissertation). East Carolina University, Greenville, NC.

37.

Turner

(2003). Teacher attrition in North Carolina schools (Doctoral dissertation). Northcentral University, Prescott Valley, AZ.

38.

Turner

S. B.

(2008). Teacher’s perceptions of leadership characteristics of public high school principals associated with student socio-economic status, community type, race, and student achievement (Doctoral dissertation). Available from ProQuest Dissertations and Theses database. (DP19210)

39.

U.S. Census Bureau. (2010). Geographic terms and concepts.

40.

U.S. Department of Education, National Center for Education Statistics. (2010). Teacher attrition and mobility: Results from the 2008-09 teacher follow-up survey (NCES 2010-353). Retrieved from http://nces.ed.gov/pubs2010/2010353.pdf

41.

Wech

Heck

(2004). An introduction to hierarchical linear modeling for marketing researchers. Marketing Bulletin, 15(1). Retrieved from http://marketing-bulletin.massey.ac.nz/V15/MB_V15_T1_Wech.pdf

42.

Woltman

Feldstain

MacKay

J. C.

Rocchi

(2012). An introduction to hierarchical linear modeling. Tutorials in Quantitative Methods for Psychology, 8, 52-69. Retrieved from http://www.tqmp.org/Content/vol08-1/p052/p052.pdf

43.

Wright

S. P.

Horn

Sanders

(1997). Teacher and classroom context effects on student achievement: Implications for teacher evaluation. Journal of Personnel Evaluation, 11, 57-67.