Public Employee Quality in a Geographic Context

Abstract

Recruiting high quality employees is one of the key functions of public human resource managers and a critical component of effective public service delivery. This is particularly true in education but little is known about public sector or teacher hiring patterns in areas that are predominantly rural, poor, and isolated from other locales. This article begins to fill that gap. We find that rural educational agencies employ the new teachers of lowest observed aptitude, implying that organizational outcomes associated with these districts may differ in systematic ways that reinforce longstanding gaps in quality. As such, human resources strategies for increasing the attractiveness of geographically and culturally isolated regions for high quality public service are needed. These strategies are likely to require different policy prescriptions than those utilized to enhance the attractiveness to employees in urban areas.

Keywords

personnel/human resource management training and development policy analysis public management issues education policy

The sheer scale of public primary and secondary education combined with its vital importance in American society make it worthy of public administration scholars’ attention and effort. In 2009, employees of public elementary and secondary schools alone accounted for 36% of the total public workforce in the United States as well as 31% of the total payroll. In comparison, the federal civilian workforce accounted for 12% of total government employees and 17% of payroll (U.S. Census, 2012). In fiscal year 2010, state and local governments in the United States collectively spent approximately US$860 billion on education, accounting for 27.6% of total expenditures (by comparison, the next largest category—public welfare—accounted for only 14.7% (Barnett & Vidal, 2012)). The lion’s share of this funding was directed to the over 13,000 public school districts, nearly 99,000 public schools, and ultimately, the more than 3.2 million teachers employed in public schools across the 50 states (U.S. Department of Education, 2012). Despite the magnitude of public schooling, Raffel (2007) laments the lack of attention by public administration scholars to this segment of the public sector. This article addresses this void by examining a human resource issue—the hiring of new teachers—within a specific geographic region of the country.

Of all the tasks delegated to human resource managers, effective recruitment and selection of new employees is one of the most critical. As Hays and Sowa (2005) state, “Recruitment and selection are the avenues by which bureaucracy acquires its most important raw materials, human resources” (p. 121). Perhaps nowhere is this axiom stronger than in the context of public education. Teachers are arguably the ultimate example of “street-level bureaucrats”: the direct and sustained interaction that a teacher has with her clients—students—easily surpasses that of virtually all other bureaucrats employed for other functions in terms of both scale and scope. Furthermore, teachers have some degree of discretion yet are likely often confronted with a gap between ideal practice and what they can achieve in the classroom. In this sense, teachers meet Lipsky’s (2010) original criteria for the bureaucratic term (p. xvii).

Not surprisingly, given the intense nature of interaction between teacher and the student as well as the reality of teacher discretion, a robust literature has demonstrated that differences in teacher aptitude, motivation, and ability have powerful and direct effects on student achievement (Greenwald, Hedges, & Laine, 1996; Rivkin, Hanushek, & Kain, 2005; Rockoff, 2004; Harris & Sass, 2010). It is perhaps this context that explains why public education has been utilized as a setting to study the factors influencing public sector performance along a variety of dimensions (Grissom, 2011; Grissom & Loeb, 2011; Grissom, Nicholson-Crotty, & Kaiser, 2012; Meier & O’Toole, 2001, 2006; Meier, O’Toole, Boyne, & Walker, 2007; O’Toole & Meier, 2003, 2004, 2008).

Work to date suggests that the initial hire may be the most important step in matching high quality teachers to lower-performing students (Boyd, Lankford, Loeb, & Wyckoff, 2005). Scholars have long recognized the importance of extrinsic factors for public sector employment decisions such as salary, working conditions, and organizational culture, among others, even for individuals identified as having strong public service motivation (Christensen & Wright, 2011). As such, school districts in areas with better amenities, more resources, or more appealing student populations may hold inherent advantages in recruitment that lead to significant and persistent inequalities in the distribution of teacher quality (Bacolod, 2007; Boyd et al., 2005; Corcoran, Evans, & Schwab, 2004; Hanushek & Pace, 1995; Podgursky, Monroe, & Watson, 2004). These issues have led scholars to study the distribution of teacher quality across school districts of varying characteristics. To date, however, most of this literature has focused on schools in urban areas. There has been less attention to rural schools generally (Arnold, Newman, Gaddy, & Dean, 2005; Ballou & Podgursky, 1995; Ingersoll & Rossi, 1995; Sherwood, 2000) yet there is reason to believe that opportunities for transition may be less likely in rural than urban areas, making initial placement especially important for rural school success (Cowen, Butler, Fowles, Streams, & Toma, 2012).

In this article, we investigate the pattern of initial hires in rural—and particularly, Appalachian—school districts, using a unique state administrative dataset for the state of Kentucky that covers a period of nearly 20 years. We find that extrinsic characteristics matter for teacher employment decisions: teachers are most likely to obtain first employment in districts close to and with cultural similarities to the location of their college. Moreover, we find that the initial distribution of quality varies, as indicated by teacher credentials, even within rural markets. In general, the less privileged rural districts employ the new teachers of lowest observed aptitude, implying that the quality of education provided by these districts may differ in systematic ways that reinforce longstanding achievement gaps. As such, human resources strategies for increasing the attractiveness of geographically and culturally isolated regions for high quality teachers are needed—strategies that are likely to require different policy prescriptions than those utilized to enhance the attractiveness of schools in urban areas.

Geographic Context of Hiring Outcomes: Rural and Isolated Rural Conditions

Nearly 56% of public school districts are located in rural areas and over 10.3 million students, or 21% of total enrolment, are enrolled in rural schools. Almost half of these students attend schools in “remote” or “distant” rural areas (Provasnik et al., 2007). Particularly challenging from an educational policy perspective is the correlation between these rural areas and poverty. But as with urban areas, not all regions within rural areas are alike. Some rural schools are in less remote areas than others. Amenities differ greatly across rural regions, as do poverty rates. As a result, we expect the desirability of employment in these rural areas will differ, and that markets may be segmented across rural areas just as evidence shows them to be across urban areas (Boyd et al., 2005; Hanushek, Kain, & Rivkin, 2004).¹ Within these segmented markets, local differences in labor quality may persist for a long period of time and fail to change as larger, more connected markets vary.

The remainder of this article focuses on the state of Kentucky. It is a particularly appropriate state in which to examine patterns of public sector employment in the rural context: with almost 17% of its population in poverty, Kentucky currently is ranked among the four states with the highest poverty rates. Educational, health and economic outcomes have differed between the poorer, rural areas and the higher income, urban areas of the state. More than a third of Kentucky’s K-12 students are in rural areas, and over half of its school districts are rural. But of great importance for the purpose of this study, not only are there poor rural districts, but also there is significant variance by location of these rural districts. The eastern mountainous, Appalachian region of the state differs from the western portion of the state in many respects. In addition to geography, per capita annual income is almost US$6,000 less in Appalachia than in the non-Appalachian regions of the state. Educational attainment of the adult population is also less, whether measured in terms of high school or college completion. Further, there is a long-discussed attachment to place in Appalachia that has been found to work against perceived economic self-interest.²

The culture of spatial fixity, or strong attachment to place, appears especially strong for the general population in Appalachia—and qualitative evidence suggests that this phenomenon may be all the more powerful for Appalachian individuals that select into the teaching profession (Gore, Wilburn, Treadway, & Plaut, 2011). As such, we wish to look at the initial placement of teachers with special attention to the preference for location among those placing in Appalachian schools versus those initially placing in other districts within the same state. Initial placement is arguably more important for explaining teacher quality across Appalachia and the non-Appalachian region of the state than in urban areas because other work illustrates that mobility of teachers across districts in these two locales after the initial placement is lower than typically found in the literature (Cowen et al., 2012). Less than 10% of teachers who initially place in one region ever transfer to the other region in Kentucky. Although between-district mobility is uncommon in the public education sector generally—teachers tend to transfer within their districts or exit the public system entirely—the particularly low rates within Appalachia underscore the inability of these rural school systems to access more dynamic labor markets (Cowen et al., 2012). Initial placement in rural states is therefore critical—and studies of initial placement comprise a necessary first step in developing a better understanding of teacher labor markets in rural states.

Data, Method, and Variables

Data

Our data are drawn largely from a rich dataset made available by the Kentucky Educational Professional Standards Board (EPSB). This dataset comprises administrative data for all certified staff in Kentucky public school districts at the individual level. From these raw data, we draw our analytical sample of teachers, which begins in school year 1987 and continues through school year 2005.³ The data include such information as age, gender, ethnicity, and for most individuals, undergraduate institution. Additionally, for a subsample of teachers we also possess American College Testing (ACT) scores as well as scores and passage rates on the National Teacher’s Examination (NTE) core battery exam and other other certification examinations. We supplement these data with county- and district-level data from the U.S. Census Bureau and the National Center for Education Statistics’ Common Core of Data (NCES CCD). The raw dataset contains information for 21,537 unique first-time teachers for this time period (defined as appearing for the first time in the dataset with zero reported years of previous teaching experience); on average, we observe 1,267 new teachers each year, ranging from a low of 997 in 1989 to a high of 1,945 in 2005.⁴ Five hundred ninety-one teachers have either incomplete data or obtain first employment in one of the school districts with too few new hires to permit inclusion in the analysis, leaving us with 20,946 unique teachers selecting among 158 unique school districts in the most general model.⁵ Fewer teachers are included in the analyses utilizing measures of teacher quality due to the less complete nature of these variables.

Method

The outcome of interest in our analysis is the district in which a teacher is observed obtaining first employment. To capture this event, we utilize a discrete dependent variable that is set equal to one in the district in which a teacher is first employed and a zero for all other districts. A commonly utilized model for unordered, multinomial, and mutually exclusive discrete dependent variables of this type is the mixed logit model. The mixed logit model is an extension of the binary logit model. It nests the multinomial logit, which estimates the effects of the characteristics of the individual (in our case, individual teachers) on the outcome of interest, and the conditional logit, which estimates the effect of the characteristics of the alternatives (in our case, individual school districts) on the outcome of interest (Cameron & Trivedi, 2005). The mixed logit specification models the observed outcome as function of these two sets of characteristics; further, it explicitly permits the impact of the characteristics of the alternatives to vary across individuals. As such, the mixed logit is particularly appropriate for our fundamental interest in whether there are differences in the initial placement patterns across rural and nonrural districts of teachers based on observable preservice credentials—a commonly used indicator of teacher quality for new teachers. We provide details on the specific variables included in the model below.

Variables

District Characteristics

We include several relevant characteristics of the school districts.⁶ First, we include the total district student enrollment. This is included to measure district size. Since we do not have any variables that capture the demand side of teacher labor markets, this variable is intended to proxy the job opportunities available in a given district in a given year: all else equal, we would expect that larger districts would be more likely to have open positions in a given year and therefore more likely to hire new teachers. In 2005 (the most recent year in the dataset), the average school district contained 3,662 students. The largest, Jefferson County (the city of Louisville), represents over 92,000 students; conversely, 35 districts contain less than 1,000 students and 13 less than 500.

Second, we include a variable that captures the total district student population that is non-White. Previous studies have shown that, all else equal, White teachers (and especially White teachers of higher observable quality) generally prefer teaching in schools with smaller minority populations (Hanushek et al., 2004; Lankford, Loeb, & Wyckoff, 2002; Scafidi, Sjoquist, & Stinebrickner, 2007). Conversely, the literature finds that non-White teachers generally prefer to teach in schools with higher minority student populations. Broadly speaking, Kentucky is a state with a low minority population, especially in the rural, Appalachian parts of the state. This reality contrasts with the characteristics of states studied in much of the previous literature on teacher labor markets, which focuses on high minority, urban schools. For example, looking again only at 2005, on average only 7.5% of the total district student population is non-White across the state, with 51 districts having less than 2% minority student population and 24 less than 1%.⁷

Third, we include a dummy variable that is set equal to one if a particular district is located in Appalachia, and zero otherwise. Finally, we include the population density of the county in which the school district is located. This is included as a general proxy for local development and availability of amenities in the county in which a teacher employed in a given district is likely to reside. Based on the extant literature, we expect that, all else equal, teachers prefer to take jobs in more populated areas, which are assumed to offer a greater set of local amenities (Miller, 2008). We also include the square of population density in order to capture any potential negative amenities that may arise due to congestion, increased crime rates, and other problems associated with highly populated areas.

Individual Characteristics

Our data contain a variety of teacher characteristics that are fundamental to our analysis, including variables capturing a teacher’s location of training, demographics, expected compensation, and credentials. We describe these in greater detail below.

Training institution

Figure 1 graphically conveys raw counts for district of initial employment for all teachers included in the most general model, disaggregated by institution of training (Appalachian or non-Appalachian). Darker areas represent initial employment of higher numbers of teachers of a given institutional origin, scaled by the total number of teachers of that origin. The thick white line indicates the borders of Appalachia and the pluses represent the location of baccalaureate institutions granting degrees to teachers included in the analytical sample.

Figure 1.

District of teacher initial employment, disaggregated by teacher region of origin.

This figure clearly demonstrates that teachers trained in Appalachia tend to obtain first employment in that region, whereas non-Appalachian teachers seem to generally begin in non-Appalachian districts. What is particularly striking about this figure is the clear delineation of district of employment for teachers attending baccalaureate institutions near the border between regions. If distance were the only relevant factor for teachers, we would expect that non-Appalachian teachers attending college near Appalachian districts would be equally as likely to obtain employment in Appalachia and non-Appalachia, holding distance constant. This figure provides strong suggestive evidence that there is more going on: teachers from both sides of the Appalachian border appear to be less likely to cross over it for first locale of employment than one might otherwise expect, a supposition that will be tested by our empirical models.

Distance from training institution

We also include a measure of travel time by car between the street address of the teacher’s baccalaureate institution and the individual school districts.⁸ Previous research, as well as the discussion above, recognizes both the segmented nature of teacher labor markets as well as the importance of geographic distance in influencing teacher employment decisions (Boyd et al., 2005). Looking at the teachers included in our most general model specification, we find that they accept their first teaching job in a district that is on average about one hour’s travel time by car from their baccalaureate institution, with a standard deviation of approximately 50 minutes. We expect that, all else equal, as travel time between a teacher’s baccalaureate institution and a given district increases, the probability of a teacher accepting employment in that district will decrease. The square of travel time is also included in order to capture any potential nonlinearity in the effect of travel time on teacher locational choices.

Demographics

We also include variables related to teacher race, age, region of origin, and gender by interacting these variables with travel time and its quadratic. The 20,946 teachers included in our most general specification includes 633 non-White teachers, 4,581 male teachers, and 6,544 teachers of at least 30 years of age at the time of their first employment. Bacolod (2007) and others find significant differences between males and females and Whites and minorities in terms of their employment choices, both in terms of the decision to enter the teaching profession as well as choice of employment location. Broadly, we expect that, all else equal, males conduct geographically broader searches then females, while minority and older teachers conduct more narrow searches. Based on the literature, we interact teacher race with district percent minority student population, with the expectation that minority teachers in general hold stronger preferences for teaching minority students than their White counterparts and are therefore attracted to jobs in districts with higher minority student populations.⁹

We interact the Appalachian district dummy variable with a dummy variable set equal to one if a teacher obtained her bachelor’s degree at an institution located in Appalachia, zero otherwise. As suggested earlier, we expect that teachers from Appalachian institutions hold strong preferences to remain in that region to teach. Similarly, it is also consistent with both the literatures on teacher labor markets and Appalachian culture to suspect that Appalachian school administrators may hold strong preferences for hiring teachers from the Appalachian institutions of the state over teachers from outside the region. This is a hypothesis that is broadly consistent with research showing that teacher quality is one of a number of competing criteria for principals when making hiring decisions (Ballou, 1996; Harris, Rutledge, Ingle, & Thompson, 2010). We also include a variable that interacts the Appalachian district dummy variable with a dummy variable set equal to one if a teacher is non-White in order to capture demand- and supply-side preferences for employment and hiring of minorities in Appalachia.

Compensation

Kentucky, like many states, pays public school teachers according to salary schedules, which reward teachers based on rank (which corresponds to education level) and years of experience. To distinguish the effect of differing compensation patterns for teachers across school districts, we include a variable that reflects the average district salary for Rank III teachers with zero to three years of experience for the two school years prior to the year in which a teacher is first employed in order to capture differences in starting salaries for teachers across districts. Certainly, we would hypothesize that, all else equal, teachers would prefer to work in districts that pay higher salaries. Others have shown cost of living differences to matter for teacher employment decisions (Chambers, 1995; Chambers & Fowler, 1995). While we cannot directly adjust the average district salary figures to reflect these differences, we do include a variable in the model that measures the population density of the county in which the school district is located. Assuming that the less populated areas generally have lower cost of living than the more heavily populated areas, the coefficient on this variable should capture some of the effect of cost of living differences in addition to potential teacher preferences for living in more or less populated areas.¹⁰

Credentials

We employ three measures of preteaching credentials designed to distinguish among teachers of varying aptitudes. First, we include a dummy variable that is set equal to one if a teacher’s score on the ACT is 26 or higher and zero otherwise. Certainly, as Corcoran et al. (2004) note, the ACT score is at best a proxy for latent characteristics that may correlate with classroom effectiveness, and is a measure that potentially speaks only to relative quality differences between teachers rather than measuring quality in an absolute sense. Accordingly, we use this variable to disaggregate our sample to see if differences emerge between the teachers with the strongest credentials and the rest of the pool in terms of the first steps they take in beginning a teaching career. Although others have used college entrance exams as a proxy for a teacher’s quality (Corcoran et al., 2004; Ferguson & Ladd, 1996; Podgursky et al., 2004), the ACT by design measures only educational achievement in college preparatory coursework taken in high school, necessarily failing to capture the impact of any knowledge or skills developed over the course of obtaining a baccalaureate degree that may increase teacher a teacher’s abilities in the classroom. In other words, deficiencies observed in teacher capability upon matriculation to college may be overcome by an effective undergraduate curriculum.

Accordingly, we also utilize an alternative measure of teacher aptitude: a dummy variable capturing whether or not a teacher’s score on the General Knowledge portion of ETS’ National Teacher Examination (NTE) core battery test is in the top decile among all prospective teachers taking the test in Kentucky in the same year. Passing the NTE core battery test was a requirement for certification in Kentucky for the majority of years covered by our dataset. Teachers typically sat for this exam (and its more contemporary counterpart, the PRAXIS II) immediately before obtaining first employment, either in the months leading up to or immediately following graduation from college. As such, the results of this test potentially reflects the knowledge, aptitude, and ability of the teacher at the time of matriculation to college as well as any increase in these in pursuit of the baccalaureate degree. The fact that the General Knowledge exam was required for all teachers for the majority of years covered by our data means that sample sizes here are larger as compared to the ACT specification. Like the ACT specification, we are not the first to utilize certification examination scores to captures teacher quality (Clotfelter, Ladd, & Vigdor, 2010).¹¹

Finally, our most general specification of teacher aptitude is a dummy variable that indicates whether or not a given teacher is ever observed to fail a certification test of any type at any point in her teaching career. This may be a potentially noisy way to measure teacher quality: although this variable captures the results of the required, general certification exams mentioned above, it also captures the results of any additional examination taken over a teacher’s career for which EPSB has set a minimum required score. This therefore includes those examinations initially required for a teaching specialization certification (Middle School Mathematics, High School Biology), as well as those taken voluntarily by a teacher for purposes of promotion into administration and other positions (School Leadership Licensure Assessment, Kentucky Principal Test). This most general measure permits inclusion of over 13,000 of our original nearly 21,000 unique teachers.

What is particularly useful about this variable is that it presents quite a different conceptualization of quality as compared to the ACT and NTE exam specifications. Obviously, a teacher scoring in the top decile on the NTE stands well above most of her peers. Similarly, scoring a 26 or higher on the ACT represents not only high achievement among teachers, but also among all Kentucky test takers. In our raw assessment data, the average ACT composite score among all those for whom scores are known is a 22.9, whereas the average for all Kentucky test takers (from both public and private schools) has ranged from a low of 19.9 in 1990 to a high of 20.9 in 2008 (Kentucky Department of Education, 2011). Rather than identifying teachers of the highest ability, our “failing a test” specification instead identifies those at the very bottom rungs of the talent ladder—a point that is driven home particularly strongly when one takes into consideration the state’s explicit policy of utilizing licensure tests as a mechanism for ensuring a minimal quality standard rather than identifying the highest quality individuals in the candidate pool (Hibpshman, 2004).

Regional preferences

Finally, because teachers and administrators may hold geographic preferences for employment and hiring that are more finely grained than that which can be captured by the dichotomous Appalachian interaction term described above, we also create a dummy variable that is set equal to one if a particular school district is in the same Area Development District (ADD) as the institution from which a teacher obtained her bachelor’s degree, zero otherwise. The ADDs are contiguous groups of counties that were initially created in the 1960s to facilitate regional planning and foster economic development. ADDs divide Kentucky’s 120 counties into 15 regions, based on shared heritage, culture, economic conditions, and geography (Kentucky Council of Area Development Districts, 2009). Because a priori there is no reason to assume that the strength of preference is uniform across all 15 ADDs, our model interacts this dummy with a dummy for each ADD in order to permit unique intercepts to be estimated for each ADD.¹²

Results

Table 1 presents the results of our most general model. Reported standard errors are robust to heteroskedasticity and are clustered on teachers.¹³

Table 1.

General Model of Teacher Initial Employment Patterns.

Variable	Coefficient (SE)
Teacher is Appalachian X Appalachian district	1.2063***
	(0.0429)
Appalachian district	−0.5534***
	(0.0340)
Travel time (hours)	−1.6735***
	(0.0395)
Square of travel time	0.0830***
	(0.0097)
Teacher is Appalachian X travel time	0.2203***
	(0.0563)
Teacher is Appalachian X square of travel time	−0.0221
	(0.0156)
Teacher is male X travel time	0.2168***
	(0.0497)
Teacher is male X square of travel time	0.0163
	(0.0138)
Teacher is non-White X travel time	−0.1759
	(0.1249)
Teacher is non-White X square of travel time	0.0663**
	(0.0319)
Teacher is 30+ X travel time	0.0368
	(0.0575)
Teacher is 30+ X square of travel time	−0.0736***
	(0.0204)
Teacher is non-White X Appalachian district	−1.2550***
	(0.2483)
District non-White student %	−0.0125***
	(0.0011)
Teacher is non-White X district non-White student %	0.0785***
	(0.0032)
District student count (000)	0.0955***
	(0.0018)
District county’s population density	0.0190***
	(0.0009)
Square of district county’s population density	−0.0003***
	(0.0000)
Standardized median district salary	0.0418***
	(0.0114)
Square of standardized median district salary	−0.0713***
	(0.0055)
Area development district dummies	Yes
Observations	3,308,419
Unique teachers	20,946

Clustered standard errors in parentheses.

p < 0.10. **p < 0.05. ***p < 0.01.

As the table reveals, in general teachers are more likely to be initially employed in larger districts, districts located in areas with higher population densities, and districts with higher median salaries for starting teachers. Teachers are also in general much less likely to obtain employment in Appalachian districts than non-Appalachian districts, a finding that should not be surprising given the discrepencies in number of schools and number of students between Appalachian and non-Appalachian Kentucky. However, the data reveal a very different story for teachers educated at an Appalachian institution: the odds of an otherwise average teacher that attended a college or university in Appalachia obtaining first employment in an Appalachian district of average characteristics are well over 3 times that of her non-Appalachian counterpart. Combined with our earlier findings that demonstrate that midcareer switches into and out of Appalachia are rare events (Cowen et al., 2012), this analysis suggests that region of origin has a powerful influence over location for the entire arc of a teacher’s career.

We also find that minority teachers have very different first employment patterns than their White peers. We find the probability of a minority teacher obtaining first employment in a given district increases as the district minority student population increases, although the opposite is true for White teachers. Interestingly, we find that the probability of a minority teacher obtaining first employment in Appalachian school districts is dramatically lower than that of her White peers, even after controlling for the non-White district student population, which, as discussed earlier, is quite low in many rural Kentucky districts. Whether this pattern is driven by demand, supply, or some combination thereof, Appalachian districts hire few new non-White teachers.

Consistent with Boyd et al. (2005), we also find that employment patterns are strongly influenced by distance: The coefficients of travel time and its quadratic reveal that estimated effects of increased travel time between a teacher’s baccalaureate institution and a given school district on the odds of employment in that district are negative for all travel times observed in the dataset. A hypothetical teacher of average characteristics has a 40% chance of being employed in a hypothetical district of average characteristics that is a 30 minute trip by car from the teacher’s baccalaureate institution. Increasing the travel time by only 30 additional minutes reduces the probability of employment to 24%. An increase of another hour (to 2 hours total) reduces the probability of employment to 7.5%. An hour further reduces the probability to 2.3%. These are findings that are largely consistent with other analyses that find significant geographic segmentation in teacher labor markets within Kentucky and other states (Cowen et al., 2012; Streams, Butler, Cowen, Fowles, & Toma, 2011). Like others, we find differences along these lines according to teacher characteristics. Male teachers and teachers from Appalachia are less deterred by travel time than female teachers and non-Appalachian teachers, respectively. Teachers of age 30 or greater at the time of their initial employment are found to be slightly more deterred by travel time than their younger counterparts, although statistically significant differences along these lines do not emerge until travel time exceeds 2 hours. No significant differences are found for minority and White teachers.

Generally, these results indicate some important differences between Appalachian and non-Appalachian districts in terms of the new teachers that they first employ. We now refine our empirical models to introduce some evidence that speaks directly to the relationship between teacher credentials and initial patterns of teacher sorting. This allows some speculation as to the effect that such patterns may have on the overall distribution of teacher quality. Table 2 presents the results of estimating a series of models that are similar to those provided in Table 1, except that we now include a measure of teacher academic aptitude: a dummy variable indicating whether or not a teacher scored a 26 or higher on the ACT, a variable that is interacted with district Appalachian status. We present the results of three models in Table 2: the first column includes every teacher for whom we possess an ACT score, while the second and third columns disaggregate this group into teachers of non-Appalachian and Appalachian origin, respectively, which facilitates comparisons between the two groups. Note that the sample size here is substantively smaller for all specifications than the corresponding model in the more general specification presented above. This is due to the fact that we possess ACT scores for only a subsample of the full population of teachers captured in our administrative data.

Table 2.

The Effect of ACT Scores on Teacher Initial Employment Patterns.

Variable	Full sample coefficient (SE)	Non-Appalachian bachelor’s coefficient (SE)	Appalachian bachelor’s coefficient (SE)
ACT Score 26+ X Appalachian district	−0.3471	−0.0251	−0.6935**
	(0.2359)	(0.2743)	(0.3154)
Teacher is Appalachian X Appalachian district	0.9457***
Teacher is Appalachian X Appalachian district	(0.1169)
Appalachian district	−0.2554***	−0.2177**	0.4094***
Appalachian district	(0.0905)	(0.0973)	(0.0986)
Travel time (hours)	−1.4491***	−1.3903***	−1.8609***
	−0.1153)	−0.1295)	(0.1602)
Square of travel time	−0.0050	−0.0298	0.1251**
	(0.0313)	(0.0373)	(0.0488)
Teacher is Appalachian X travel time	−0.4177**
	(0.1766)
Teacher is Appalachian X square of travel time	0.1213**
Teacher is Appalachian X square of travel time	(0.0544)
Teacher is male X travel time	0.2782*	0.3720*	0.0949
	(0.1687)	(0.2145)	(0.2687)
Teacher is male X square of travel time	−0.0374	−0.0410	−0.0408
	(0.0539)	(0.0688)	(0.0867)
Teacher is non-White X travel time	−0.0060	−0.2234	2.1130
	(0.3595)	(0.3907)	(1.8117)
Teacher is non-White X square of travel time	0.0591	0.0936	−0.3750
Teacher is non-White X square of travel time	(0.0944)	(0.1077)	(0.3907)
Teacher is 30+ years old X travel time	0.5005*	0.5836*	0.7339
	(0.2662)	(0.3079)	(0.4567)
Teacher is 30+ years old X square of travel time	−0.2179**	−0.1978*	−0.4634**
Teacher is 30+ years old X square of travel time	(0.1051)	(0.1156)	(0.1997)
Teacher is non-White X Appalachian district	−0.5267	0.1497	−1.6372
Teacher is non-White X Appalachian district	(0.6687)	(0.6878)	(1.2369)
District non-White student %	−0.0101***	−0.0069**	−0.0328***
	(0.0029)	(0.0029)	(0.0096)
Teacher is non-White X District non-White student %	0.0553***	0.0561***	0.0427*
Teacher is non-White X District non-White student %	(0.0094)	(0.0103)	(0.0232)
District student count (000)	0.1168***	0.1153***	0.1330***
	(0.0056)	(0.0059)	(0.0139)
District county’s population density	0.0276***	0.0326***	0.0216***
	(0.0024)	(0.0028)	(0.0051)
Square of district county’s population density	−0.0004***	−0.0004***	−0.0004***
Square of district county’s population density	(0.0000)	(0.0000)	(0.0000)
Standardized median district salary	−0.0357	−0.0705*	0.0177
	(0.0318)	(0.0402)	(0.0531)
Square of standardized median district salary	−0.0346**	−0.0325	−0.0363
Square of standardized median district salary	(0.0176)	(0.0232)	(0.0274)
Area development district dummies	Yes	Yes	Yes
Observations	407,166	266,862	140,304
Unique teachers	2,577	1,689	888

Clustered standard errors in parentheses. ACT = American College Testing.

p < 0.10. **p < 0.05. ***p < 0.01.

These results provide some evidence confirming our expectation that Appalachian students are educated by teachers with weaker credentials than their non-Appalachian counterparts. Looking first at the pooled sample, although the sign of the interaction of high ACT and Appalachian district is negative, it is not statistically different from zero. This implies that high ACT scoring teachers are, in aggregate, no more or less likely to obtain first employment in Appalachian districts than their lower achieving peers. However, disaggregating the sample into Appalachian and non-Appalachian teachers uncovers a striking difference between these two groups. First, the coefficient on whether or not a district is located in Appalachia takes on opposite signs for the two groups, implying that while Appalachian-trained teachers are more likely to be first employed in Appalachian districts, non-Appalachian teachers are less likely, all else equal. Second, we find that the most academically accomplished teachers of Appalachian origin are statistically significantly less likely to be employed in Appalachian districts as compared to lower achieving teachers from Appalachia, all else in the model held constant. Taken together, the empirical models presented in Table 2 demonstrate that non-Appalachian teachers are unlikely to be first employed in Appalachia regardless of aptitude, while the highest achieving teachers in the Appalachian teacher pool are more likely to obtain employment in school districts elsewhere.

Table 3 utilizes the same estimation strategy utilized in Table 2, but replaced teacher ACT scores with the indicator for a top decile scores on the NTE Core Battery General Knowledge test.

Table 3.

The Effect of NTE Scores on Teacher Initial Employment Patterns.

Variable	Full sample coefficient (SE)	Non-Appalachian bachelor’s coefficient (SE)	Appalachian bachelor’s coefficient (SE)
NTE general knowledge top decile score X Appalachian district	−0.3003***	−0.1413	−0.4597***
	(0.0992)	(0.1269)	(0.1356)
Teacher is Appalachian X Appalachian district	1.2623***
	(0.0543)
Appalachian district	−0.5350***	−0.4735***	0.5039***
	(0.0444)	(0.0468)	(0.0417)
Travel time (hours)	−1.6495***	−1.5714***	−1.3362***
	(0.0518)	(0.0570)	(0.0783)
Square of travel time	0.0844***	0.0693***	0.0389
	(0.0128)	(0.0142)	(0.0245)
Teacher is Appalachian X travel time	0.3651***
	(0.0704)
Teacher is Appalachian X square of travel time	−0.0493**
	(0.0193)
Teacher is male X travel time	0.1794***	0.1450*	0.1866*
	(0.0625)	(0.0782)	(0.1119)
Teacher is male X square of travel time	0.0326*	0.0397*	0.0351
	(0.0170)	(0.0206)	(0.0334)
Teacher is non-White X travel time	−0.1412	−0.1788	−0.1536
	(0.1773)	(0.1960)	(0.4463)
Teacher is non-White X square of travel time	0.0447	0.0579	0.0402
	(0.0456)	(0.0534)	(0.1045)
Teacher is 30+ years old X travel time	0.0281	0.0605	0.0266
	(0.0852)	(0.1101)	(0.1485)
Teacher is 30+ years old X square of travel time	−0.1160***	−0.1341***	−0.1103*
	(0.0331)	(0.0439)	(0.0576)
Teacher is non-White X Appalachian district	−1.7519***	−2.0023**	−1.8420***
	(0.3637)	(0.9964)	(0.4307)
District non-White student %	−0.0181***	−0.0122***	−0.0406***
	(0.0015)	(0.0016)	(0.0042)
Teacher is non-White X district non-White student %	0.0836***	0.0842***	0.0853***
Teacher is non-White X district non-White student %	(0.0042)	(0.0048)	(0.0093)
District student count (000)	0.1065***	0.1055***	0.1216***
	(0.0025)	(0.0029)	(0.0055)
District county’s population density	0.0174***	0.0235***	0.0080***
	(0.0012)	(0.0015)	(0.0024)
Square of district county’s population density	−0.0003***	−0.0004***	−0.0003***
	(0.0000)	(0.0000)	(0.0000)
Standardized median district salary	−0.0233*	−0.1047***	0.0807***
	(0.0138)	(0.0174)	(0.0216)
Square of standardized median district salary	−0.0556***	−0.0420***	−0.0798***
	(0.0062)	(0.0070)	(0.0102)
Area development district dummies	Yes	Yes	Yes
Observations	2,049,839	1,214,547	835,292
Unique teachers	12,979	7,690	5,289

Clustered standard errors in parentheses. NTE = National Teacher’s Examination.

p < 0.10. **p < 0.05. ***p < 0.01.

The coefficients presented in this table tell a substantively identical story to those presented in Table 2. The results of the pooled model reveal that, in general, the highest scorers on the exam are less likely to be first employed in Appalachian districts. The subsamples mirror the findings for ACT scores: teachers from outside Appalachia are generally not initially employed there regardless of exam score, while the highest scoring teachers from Appalachia are statistically significantly less likely to be employed there as compared to their lower scoring peers, holding all else in the model constant.

Finally, Table 4 presents the full results of the empirical models utilizing the “failing a test” indicator variable described above. The pooled model shows that teachers that have failed a certification test are, in general, more likely to obtain employment in Appalachian school districts than their peers that have never failed a test, holding all else in the model constant. Looking only at the teachers of non-Appalachian origin, we for the first time observe an identical trend: unlike the previous models, where lower quality non-Appalachian teachers were no more or less likely to teach in Appalachia than their higher achieving peers, here we find that non-Appalachian teachers that have failed a certification test are actually more likely to be first employed in Appalachia than those non-Appalachian teachers that are never observed to fail a test. The same pattern holds for teachers from Appalachia that failed a test: they are differentially more likely to obtain first employment in Appalachian school districts than their peers who have never failed a test. In other words, this model finds that the teachers of lowest observed aptitude from outside Appalachia are more likely to cross over to Appalachia to teach, while the teachers of lowest aptitude from Appalachia are statistically significantly more likely to be first employed in their home region.

Table 4.

The Effect of Certification Test Failure on Teacher Initial Employment Patterns.

Variable	Full sample coefficient (SE)	Non-Appalachian bachelor’s coefficient (SE)	Appalachian bachelor’s coefficient (SE)
Failed a certification test X Appalachian district	0.2850***	0.4616***	0.1599**
Failed a certification test X Appalachian district	(0.0651)	(0.0982)	(0.0803)
Teacher is Appalachian X Appalachian district	1.1906***
Teacher is Appalachian X Appalachian district	(0.0546)
Appalachian district	−0.6417***	−0.6152***	0.3436***
	(0.0449)	(0.0491)	(0.0454)
Travel time (hours)	−1.6970***	−1.6622***	−1.6262***
	(0.0495)	(0.0545)	(0.0718)
Square of travel time	0.0801***	0.0691***	0.0951***
	(0.0125)	(0.0140)	(0.0212)
Teacher is Appalachian X travel time	0.0795
	(0.0691)
Teacher is Appalachian X square of travel time	0.0129
Teacher is Appalachian X square of travel time	(0.0189)
Teacher is male X travel time	0.2367***	0.2183***	0.2571**
	(0.0585)	(0.0717)	(0.1065)
Teacher is male X square of travel time	0.0055	0.0127	−0.0042
	(0.0162)	(0.0192)	(0.0317)
Teacher is non-White X travel time	−0.2518*	−0.2317	−0.6515*
	(0.1424)	(0.1561)	(0.3528)
Teacher is non-White X square of travel time	0.0837**	0.0841**	0.1433*
Teacher is non-White X square of travel time	(0.0350)	(0.0402)	(0.0733)
Teacher is 30+ years old X travel time	0.1298*	0.1718*	0.0728
	(0.0770)	(0.0916)	(0.1500)
Teacher is 30+ years old X square of travel time	−0.0955***	−0.1028***	−0.0920
Teacher is 30+ years old X square of travel time	(0.0279)	(0.0326)	(0.0575)
Teacher is non-White X Appalachian district	−1.1145***	−0.6964	−1.6658***
Teacher is non-White X Appalachian district	(0.2901)	(0.4444)	(0.3971)
District non-White student %	−0.0105***	−0.0069***	−0.0292***
	(0.0013)	(0.0013)	(0.0038)
Teacher is non-White X District non-White student %	0.0760***	0.0766***	0.0751***
Teacher is non-White X District non-White student %	(0.0037)	(0.0041)	(0.0088)
District student count (000)	0.0873***	0.0853***	0.1038***
	(0.0021)	(0.0023)	(0.0053)
District county’s population density	0.0187***	0.0222***	0.0132***
	(0.0010)	(0.0012)	(0.0020)
Square of district county’s population density	−0.0003***	−0.0003***	−0.0003***
Square of district county’s population density	(0.0000)	(0.0000)	(0.0000)
Standardized median district salary	0.0607***	0.0138	0.1221***
	(0.0136)	(0.0176)	(0.0214)
Square of standardized median district salary	−0.0660***	−0.0511***	−0.0802***
Square of standardized median district salary	(0.0074)	(0.0113)	(0.0101)
Area development district dummies	Yes	Yes	Yes
Observations	2,102,370	1,348,935	753,435
Unique teachers	13,307	8,538	4,769

Clustered standard errors in parentheses.

p < 0.10. **p < 0.05. ***p < 0.01.

Discussion

We find that teachers who received their baccalaureate degree from an Appalachian institution are much more likely to obtain first employment in that region than are those who went to a college or university outside the region, controlling for factors demonstrated in the extant literature to affect placement. Furthermore, once aptitude measures drawn from our administrative data are included in our model, we find the strongest credentialed teachers from Appalachia—those that scored a 26 or higher on the ACT or in the top decile on the NTE Core Battery General Knowledge Exam—are more likely to obtain a first placement outside of Appalachia as compared to their lower scoring Appalachian counterparts, while score is generally shown to be unrelated to Appalachian placements of teachers from outside of Appalachia. Finally, utilizing a variable which captures whether or not a teacher is ever observed to fail a certification exam, we focus on the employment patterns of the teachers of weakest credentials. We find that these teachers are differentially more likely to obtain first employment in Appalachia, regardless of whether or not they are of Appalachian origin.

Our most general model finds that teachers in Kentucky are more likely to be employed in districts close to their baccalaureate institutions and in the same region (Appalachian or non-Appalachian) as that college or university, providing support to a literature which finds that teacher labor markets are highly regionalized and segmented across states. But more importantly, our findings illustrate the potentially deleterious impact of that geographic preference through its relationship with first hire location and credentials. High achieving teachers from Appalachian institutions are more likely to be employed outside of Appalachia. Appalachian teachers that do not have these credentials are significantly less likely to leave that region to teach. Ballou (1996) and others (Baker & Cooper, 2005; Harris et al., 2010) have found that schools tend not to hire the highest quality teachers from the applicant pool, findings that remind us of the importance of the demand side of the hiring process. It may be the case that the Appalachian districts eschew hiring non-Appalachian teachers for a host of reasons, including such factors as cultural and organizational norms, or simply an administrative preference for lower achieving teachers because of their lower rates of turnoever and exit from the profession. While our analysis cannot distinguish the supply- and demand- side factors, we nonetheless can conclude that geography exerts a powerful and direct influence over labor market outcomes. Put in a larger context, the findings for teachers raises questions about the credentials of public sector employees more generally. While this article does not speak directly to the quality of other public sector employees, it certainly is suggestive. Just as with the teacher market, there is little reason to believe that local governments in remote, economically depressed areas of the country will be more successful in attracting highly credentialed, well-trained individuals from outside those regions. As a result, the quality of public service workers in these areas is likely to continue to lag those of the more privileged regions.

So if the highest quality employees tend not to locate in these regions, what type of policies can be implemented to improve the quality of human resources and, ultimately, the quality of delivery of public services? Many have suggested simply increasing salaries as way to alleviate the unattractiveness of employment in historically underserved areas. Kentucky’s education finance system, like those employed in many states, utilizes a common, statewide salary schedules for teachers and places caps on the extent to which school districts can generate funds through local taxation, placing significant restrictions on the ability of disadvantaged districts to simply outbid other areas for high quality teachers. Although general purpose governments are less restricted than school districts along these lines, the current era of declining or stagnant revenue makes the adoption of such policies unrealistic in practice, especially given the fact that many of the poorer, underserved areas were among those most strongly impacted by the current fiscal crisis.

Ultimately, in disadvantaged regions and jurisdictions, in-service training may offer the strongest option for improving public service delivery. There are strong and relevant parallels between education and the broader public sector here as well. Teachers are generally required to participate in a certain number of in-service training hours as a condition of continued licensure. While no such mandate exists in the broader public sector, the increasing popularity of certificate and other nondegree training programs and the longstanding trends of midcareer enrollment in postbaccalaureate public administration degree programs indicate the continued importance of in-service training beyond those employed in primary and secondary education. The increased availability of online and hybrid programs that eliminate the geographic and time barriers historically associated with pursuing these training opportunities makes them seemingly well positioned to play an increased role in offsetting the initial differences in workforce quality in disadvantaged areas. However, the mixed evidence regarding the success of in-service training in increasing teacher effectiveness certainly suggests that policymakers should approach such avenues cautiously.

Concluding Remarks

Within most states, whether predominantly rural or urban, there are particular geographic areas whose inhabitants are traditionally disadvantaged and, therefore, have lagged the educational attainment and economic prosperity enjoyed by others. Recruiting high quality employees for public sector positions in these areas remains a critical first step in providing pathways through which these longstanding achievement gaps can be addressed. Beyond simply increasing salaries, Hess (2010) and others note that some school districts across the United States have utilized innovative strategies to grow the teacher talent pool, including such tactics as targeted recruiting of midcareer professionals from other areas, increasing the use of emerging technology to broaden the reach of job searches, tailoring job requirements and duties to better align with the unique skills and abilities of individual teachers, and fundamentally restructuring teacher in-service training. The experiences of these innovative districts should continue to inform policy and practice.

Ultimately, research in public administration broadly supports the supposition that good government relies upon high quality employees in contexts beyond schools. For the same reasons that it is difficult to staff a school in a particular place or region with an effective teacher, it is likely to be difficult to recruit high quality employees for other critical government functions, especially for positions in local or special purpose governments. As the compostion of the public workforce at all levels continues to change, addressing these problems will become increasingly critical to ensure that publicly provided goods and services are delivered effectively.

Footnotes

Acknowledgements

The authors extend their thanks to the Kentucky Education Professional Standards Board for making administrative data available for research purposes and Terry Hibpshman for his invaluable assistance with the data. The authors also thank Heather Getha-Taylor, participants at the 2012 meeting of the Association of Education Finance and Policy and two anonymous reviewers for their very helpful comments and suggestions.

Declaration of Conflicting Interests

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

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Spencer Foundation (Grant No.201000055) provided funding for this research under the project title Teaching Careers in Rural Schools.

Notes

Author Biographies

Jacob Fowles is an assistant professor in the School of Public Affairs and Administration at University of Kansas. His research interests include education finance and policy as well as public budgeting and financial management.

J.S. Butler is a professor in the Martin School of Public Policy and Administration at the University of Kentucky. His research and teaching interests focus on applied and theoretical econometrics.

Joshua M. Cowen is an assistant professor at the Martin School of Public Policy and Administration. His research interests include school choice policy, teacher quality, and the impacts of public educational reforms.

Megan E. Streams is an assistant professor of public administration in the College of Public Service and Urban Affairs at Tennessee State University. Her research interests include public finance and budgeting, education finance, intergovernmental fiscal relations, and performance measurement.

Eugenia F. Toma is the Wendell H. Ford Professor of Public Policy in the Martin School of Public Policy & Administration at the University of Kentucky. She specializes in the economics of schooling. She currently is serving as PI of an NSF grant to evaluate the effectiveness of teacher professional development programs.

References

Arnold

M. L.

Newman

J. H.

Gaddy

B. B.

Dean

C. B.

(2005). A look at the condition of rural education research: Setting a direction for future research. Journal of Research in Rural Education, 20(6), 20-26.

Bacolod

(2007). Who teaches and where they choose to teach: College graduates of the 1990s. Educational Evaluation and Policy Analysis, 29(3), 155-168.

Baker

B. D.

Cooper

B. S.

(2005). Do principals with stronger academic backgrounds hire better teachers? Policy implications for improving high-poverty schools. Educational Administration Quarterly, 41, 449-479.

Ballou

(1996). Do public schools hire the best applicants? Quarterly Journal of Economics, 111(1), 97-133.

Ballou

Podgursky

(1995). Recruiting smarter teachers. Journal of Human Resources, 30, 326-338.

Barnett

Vidal

(2012). State and local government finances summary: 2010. Washington, DC: United States Census Bureau.

Boyd

Lankford

Loeb

Wyckoff

(2005). The draw of home: How teachers’ preferences for proximity disadvantage urban schools. Journal of Policy Analysis and Management, 24(1), 113-132.

Cameron

A. C.

Trivedi

P. K.

(2005). Microeconometrics: Methods and applications. Cambridge, NY: Cambridge University Press.

Chambers

J. G.

(1995). Public school teacher cost differences across the United States: Introduction to a teacher cost index (TCI). Paper presented at the NCES State Data Conference, Washington, DC.

10.

Chambers

J. G.

Fowler

W. J.

Jr. (1995). Public school teacher cost differences across the United States. Washington, DC: National Center for Education Statistics.

11.

Christensen

R. K.

Wright

B. E.

(2011). The effects of public service motivation on job choice decisions: Disentangling the contributions of person-organization fit and person-job fit. Journal of Public Administration Research and Theory, 21, 723-743.

12.

Clotfelter

C. T.

Ladd

H. F.

Vigdor

J. L.

(2010). Teacher credentials and student achievement in high school. Journal of Human Resources, 45, 655-681.

13.

Corcoran

S. P.

Evans

W. N.

Schwab

R. M.

(2004). Women, the labor market, and the declining relative quality of teachers. Journal of Policy Analysis and Management, 23, 449-470.

14.

Cowen

Butler

J. S.

Fowles

Streams

Toma

(2012). Teacher retention in Appalachian schools: Evidence from Kentucky. Economics of Education Review, 31, 431-441.

15.

DeYoung

(1995). Constructing and staffing the cultural bridge: The school as change agent in rural Appalachia. Anthropology & Education Quarterly, 26(2), 168-192.

16.

Ferguson

R. F.

Ladd

H. F.

(1996). How and why money matters: An analysis of Alabama schools. In Ladd

H. F.

(Ed.), Holding schools accountable: Performance-based reform in education (pp. 265-298). Washington, DC: Brookings Institution.

17.

Gore

J. S.

Wilburn

K. R.

Treadway

Plaut

(2011). Regional collectivism in Appalachia and academic attitudes. Cross-Cultural Research, 45, 376-398.

18.

Greenwald

Hedges

L. V.

Laine

R. D.

(1996). The effect of school resources on student achievement. Review of Educational Research, 66, 361-396.

19.

Gregory

R. G.

Borland

(1999). Recent developments in public sector labor markets. Handbook of labor economics, 3, 3573-3630.

20.

Grissom

J. A.

(2011). Revisiting the impact of participative decision making on public employee retention: The moderating influence of effective managers. American Review of Public Administration, 42, 400-418.

21.

Grissom

J. A.

Loeb

(2011). Triangulating principal effectiveness. American Educational Research Journal, 48, 1091-1123.

22.

Grissom

J. A.

Nicholson-Crotty

Keiser

(2012). Does my boss’s gender matter? Explaining job satisfaction and employee turnover in the public sector. Journal of Public Administration Research and Theory, 22(4), 649-673.

23.

Hanushek

E. A.

Kain

J. F.

Rivkin

S. G.

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

24.

Hanushek

E. A.

Pace

R. R.

(1995). Who chooses to teach (and why)? Economics of Education Review, 14(2), 101-117.

25.

Harris

D. N.

Rutledge

S. A.

Ingle

W .K.

Thompson

C. C.

(2010). Mix and match: What principals really look for when hiring teachers. Education Finance and Policy, 5(2), 228-246.

26.

Harris

D. N.

Sass

T. R.

(2010). Teacher training, teacher quality and student achievement. Journal of Public Economics, 95, 798-812.

27.

Hays

S. W.

Sowa

J. E.

(2005). Staffing the bureaucracy: Employee recruitment and selection. In Condrey

S. E.

(Ed.), Handbook of human resource management in government (pp. 97-124). San Francisco, CA: Jossey-Bass.

28.

Hess

(2010). Education unbound: The promise and practice of Greenfield schooling. Alexandria, VA: Association for Supervision and Curriculum Development.

29.

Hibpshman

(2004). Considerations related to setting cut scores for teacher tests. Frankfort, KY: Kentucky Educational Professional Standards Board.

30.

Ingersoll

Rossi

(1995). Which types of schools have the highest teacher turnover? Washington, DC: National Center for Education Statistics.

31.

Kentucky Council of Area Development Districts. (2009). About the area development districts. KCADD 2009 [cited June 2 2011]. Retrieved from http://www.kcadd.org/what_is_ADD.html

32.

Kentucky Department of Education. (2011). ACT and SAT scores for Kentucky Students 2010 [cited July 15 2011]. Retrieved from http://www.education.ky.gov/kde/homepagerepository/proof%20of%20progress/act%20and%20sat%20scores%20for%20kentucky%20students.htm

33.

Lankford

Loeb

Wyckoff

(2002). Teacher sorting and the plight of urban schools: A descriptive analysis. Educational Evaluation and Policy Analysis, 24(1), 37-62.

34.

Lipsky

(2010). Street-level bureaucracy: Dilemmas of the individual in public services (Updated ed.). New York: Russell Sage Foundation.

35.

Martin

(2000). Local labour markets: Their nature, performance, and regulation. In Clark

Feldman

Gertler

(Eds.), The Oxford handbook of economic geography (pp. 455-476). New York: Oxford University Press.

36.

Meier

K. J.

O’Toole

L. J.

Jr. (2001). Managerial strategies and behavior in networks: A model with evidence from US public education. Journal of Public Administration Research and Theory, 11, 271-294.

37.

Meier

K. J.

O’Toole

L. J.

Jr. (2006). Political control versus bureaucratic values: Reframing the debate. Public Administration Review, 66(2), 177-192.

38.

Meier

K. J.

O’Toole

L. J.

Jr Boyne

G. A.

Walker

R. M.

(2007). Strategic management and the performance of public organizations: Testing venerable ideas against recent theories. Journal of Public Administration Research and Theory, 17, 357-377.

39.

Miller

(2008). Valuing place: Understanding the role of community amenities in rural teacher labor markets. Unpublished doctoral dissertation, School of Education, Stanford University, Stanford, CA.

40.

O’Toole

L. J.

Jr Meier

K. J.

(2003). Plus ça Change: Public management, personnel stability and organizational performance. Journal of Public Administration, 13, 43-64.

41.

O’Toole

L. J.

Jr Meier

K. J.

(2004). Public management in intergovernmental networks: Matching structural networks and managerial networking. Journal of Public Administration Research and Theory, 14, 469-494.

42.

O’Toole

L. J.

Jr Meier

K. J.

(2008). The human side of public organizations: Contributions to organizational performance. American Review of Public Administration. doi:10.1177/0275074008323805

43.

Ozimek

Miles

(2011). Stata utilities for geocoding and generating travel time and travel distance information. Stata Journal, 11(1), 106-119.

44.

Podgursky

Monroe

Watson

(2004). The academic quality of public school teachers: An analysis of entry and exit behavior. Economics of Education Review, 23, 507-518.

45.

Provasnik

KewalRamani

Coleman

M. M. L.

Gilbertson

Herring

Xie

(2007). Status of education in rural America. Washington, DC: National Center for Education Statistics.

46.

Raffel

J. A.

(2007). Why has public administration ignored public education, and does it matter? Public Administration Review, 67(1), 135-151.

47.

Rivkin

S. G.

Hanushek

E. A.

Kain

J. F.

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

48.

Rockoff

J. E.

(2004). The impact of individual teachers on student achievement: Evidence from panel data. American Economic Review, 94, 247-252.

49.

Scafidi

Sjoquist

D. L.

Stinebrickner

T. R.

(2007). Race, poverty, and teacher mobility. Economics of Education Review, 26(2), 145-159.

50.

Sherwood

(2000). Where has all the “rural” gone? Rural education research and current federal reform. Journal of Research in Rural Education, 16(3), 159-167.

51.

Streams

Butler

J. S.

Cowen

Fowles

Toma

(2011). School finance reform: Do equalized expenditures imply equalized teacher salaries? Education Finance and Policy, 6, 508-536.

52.

U.S. Census Bureau. (2012). Statistical abstract of the United States: 2012 (131st ed.). Washington, DC: U.S. Census Bureau. Retrieved from http://www.census.gov/compendia/statab/

53.

U.S. Department of Education. (2012). Digest of education statistics, 2011. Washington, DC: National Center for Education Statistics.