The state of Florida has 67 general school districts coterminous with the state’s 67 counties, that are in a competitive relationship to attract betterqualified teachers and to accomplish high student performance in the annual statewide exam. This paper investigates the effect of teacher voluntary turnover on student achievement at the school district level, along with other factors including teacher salary, teacher’s degree level, teaching experience, local tax revenue, and the percentage of White and Asian students. Analyzing the 7year panel data of Florida’s 67 school districts, I discovered that teacher voluntary turnover, the percentage of White students, and teacher degree level were significantly related to student performance, but teacher salary level was not. The results suggest that each school district should make and implement a program to reduce teacher voluntary turnover to improve student performance, but the current performancebased pay system for teachers in Florida should be revamped.
Journal of the Korean Data Analysis Society (December 2021), 23(6), 2563-2576 https://doi.org/10.37727/jkdas.2021.23.6.2563 Impact of Teacher Turnover on Student Performance 11) Dongjin Oh Abstract The state of Florida has 67 general school districts coterminous with the state’s 67 counties, that are in a competitive relationship to attract better-qualified teachers and to accomplish high student performance in the annual statewide exam This paper investigates the effect of teacher voluntary turnover on student achievement at the school district level, along with other factors including teacher salary, teacher’s degree level, teaching experience, local tax revenue, and the percentage of White and Asian students Analyzing the 7-year panel data of Florida’s 67 school districts, I discovered that teacher voluntary turnover, the percentage of White students, and teacher degree level were significantly related to student performance, but teacher salary level was not The results suggest that each school district should make and implement a program to reduce teacher voluntary turnover to improve student performance, but the current performance-based pay system for teachers in Florida should be revamped Keywords : turnover, teacher, student performance, personnel Introduction The state of Florida has 67 general school districts coterminous with the 67 counties and other special school districts The Florida Department of Education does not have established stipulations to universally regulate teacher’s salary level across the school districts in Florida, and each school district exerts substantial discretion to set annual teacher wages, leading to substantial salary differences among school districts For example, according to the Florida Department of Education’s average teacher salary data in 2010-2011, the average teacher salary was $45,723, the highest average salary was $57,798 in the Monroe district, and the lowest average was $35,721 in the Gadsden district Since teacher capability and qualification influences student achievement (Figlio, 1997; Darling-Hammond, 2000; Lin, 2010; Chang, Ahn, 2015), and the Florida Department of Education assigns an annual grade to each school district, based on the result of annual statewide standardized exam, school districts are in competitive relationship with each other to Assistant Professor, Department of Defense Management, Korea National Defense University 1098 Hwangsanbeol-ro, Yangchon-myun, Nonsan-si 33021 E-mail : dongjinoh64@gmail.com [Received 12 October 2021; Revised November 2021; Accepted November 2021] 2564 Dongjin Oh retain highly competent and experienced teachers Various factors impact on teacher turnover: unsatisfactory salary level (Imazeki, 2005; Hendricks, 2015; Figlio, 1997; Hanushek et al., 1999; Stinebrickner, 1998), teaching environment (Darling-Hammond, 1997; Loeb, Darling-Hammond, 2005), and individual motivation factors including public service motivation (Kim, 2015; Giauque et al., 2012; Naff, Crum, 1999; Shim et al., 2015; Taylor et al., 2015) Given that teacher salary levels and teaching environments vary greatly depending on each school district’s policy, it is not surprising that there exists high variation in teacher turnover rate among school districts According to the Teacher Exit Interview Information of Florida, in fiscal year of 2012-13, the average teacher turnover rate was percent: while the highest turnover rate was 53.1% in Jefferson district, the lowest turnover rate was 0% in Liberty district High teacher turnover rate may have an adverse impact on student achievement Successful implementation of education programs has been regarded as a key factor in improving student performance (Newman et al., 2001) Since frequent teacher turnover imposes substantial challenges on the consistent and successful implementation of education programs, students who attend schools with higher teacher turnover, are less likely to show high academic performance than those studying in a school with a low teacher turnover Additionally, newly hired teachers may have little knowledge of the district’s education program and the time required for learning a district’s unique educational goals and environment can be detrimental to student achievement Despite the potential harmful effect of teacher turnover on student performance, few studies have examined the relationship between teacher turnover and student achievement at the school district level This study aims to investigate the effect of teacher turnover on student achievement along with other factors such as salary level, local tax revenue, teacher quality, and student ethnicity at the school district level To achieve this, I analyzed a panel of Florida 67 school districts for school years (from 2004-2005 through 2010-2011) King et al (1994) pointed out that a good research question should satisfy two criteria: it should be important in the real world, and it should contribute to the expansion of academic literature I anticipate this research can contribute to both the research literature and practical issues From the standpoint of the research literature, this study expands our understanding of public personnel management by examining the causal effect of teacher turnover on student achievement From the pragmatic perspective, this research provides explanations about the significant factors affecting student performance, which would be helpful for personnel managers in a school district to establish personnel management policies better suited to achieving the overall educational goal of improving student performance Explanation of Student Achievement in a School District This study investigates what factors have significant effects on student achievement in a school district Florida consists of special districts and 67 general school districts 2565 Impact of Teacher Turnover on Student Performance conterminous with the 67 counties School districts have broad range of discretion to set annual teacher salary level and have various level of local tax revenues for the school district Also, demographic attributes of students are various across the school districts In addition, different teaching environment depending on districts likely leads to high variation in teacher turnover rate among the districts Those distinctive factors in each school district can influence student performance level in each district In this section, based on previous research, I establish hypotheses of determinants of student achievement at the school district level Teacher Voluntary Turnover Regarding the effect of teacher turnover on student performance, there are two different explanations: ‘Compositional’ and ‘Disruptive’ explanations (Ronfeldt et al., 2013) Compositional effect is derived from the qualitative difference between leaving teachers and newly hired teachers If leaving teachers have a superior quality than the teachers replacing them, the turnover effect on student performance would be negative (Ronfeldt et al., 2013) While compositional explanation focuses on the effectiveness of individual teachers, the disruptive explanation emphasizes the broader impact of teacher turnover on a school district’s overall organization (Ronfeldt et al., 2013) Frequent turnover of teachers impedes the effective implementation of a district’s educational programs (Newman et al., 2001) Implicit and subtle, yet critical, knowledge in implementing educational programs can be lost with the turnover of experienced teachers (Abelson, Baysinger, 1984) Additionally, recruiting, selecting, and training novice teachers for the vacant positions requires considerable resources that could otherwise be utilized for improving the quality of education programs (Barnes et al., 2007) Voluntary turnover of teachers in search of better salary level and better working conditions mean a loss of experienced, proficient, and effective teachers, which likely has a negative impact on student achievement Moreover, high voluntary turnover of teachers suggests considerable challenges to the successful implementation of educational programs Finally, the additional resources for replacing teachers could further impede a district’s achievement of its goals Based on the previous research on the effects of teacher turnover on student achievement, I establish the following hypothesis Hypothesis 1: High teacher voluntary turnover rate has a negative effect on the student achievement in the district Salary is one of the important factors for teachers to decide whether they join the school or not High salary level may have a positive effect on student achievement by inducing highly qualified teachers in terms of degree level and competent credential Several previous studies demonstrated the positive relationship between teacher salary and teacher quality For example, Figlio (1997), using data for 3,944 teachers in 584 school districts in metropolitan area across the U.S, explained that higher salaries attract better qualified teachers, and increase the probability of recruiting teachers with competent credentials Likewise, Hendricks (2015) showed that higher teacher salaries have a positive effect on attracting and retaining qualified 2566 Dongjin Oh and experienced teachers Also, Hanushek et al (1999) and Clotfelter et al (2011) discovered that offering higher salaries has a positive impact on attracting teachers who achieved a high score in the teacher certification exam In addition, high salary level can lead to improved student performance by giving monetary incentive to current teachers Podgursky (2009) demonstrated the benefits of performance-based compensation system for teachers He argued that there are a wide range of variations in a performance-based teacher’s capability compensation system, to improve teachers student would be performance, motivated and to under raise a student achievement in short time (Podgursky, 2009) Supporters for this reformed compensation system insisted that a teacher’s effectiveness cannot be fully explained by observable attributes such as educational degree level, teaching experience, and certification, and thus, the compensation system for student performance is much more efficient Therefore, under the performance-based compensation system, high-performing teachers would be continuously motivated to be effective in teaching students for monetary incentives and their self-pride, while low-performing teacher would try to modify their teaching in ways to improve student performance or simply leave their teaching position if they find the cost of improving performance unaffordable for themselves (Podgursky, 2009) Based on the previous research on the effect of teacher salary on student achievement, I establish the following hypothesis Hypothesis 2: A district’s average teacher salary level has a positive effect on student achievement in the district Teacher Quality A growing body of research has explained that a substantial part of student achievement is attributed to teacher quality Sander, Rivers (1996) found that assigning ineffective teachers in a row to students causes significantly lower student performance than students assigned with effective teachers Darling-Hammond (2000) examined the effect of teacher quality on student achievement across states by using qualitative and quantitative analysis and found that states which adopted a policy of teacher qualification in terms of teacher preparation and certification showed high student performance Similarly, Lin (2010), analyzing data on 500 school districts in Pennsylvania for school years, reported that both teacher salaries and teacher quality had a positive effect on student achievement The author explained that higher salaries likely attract highly qualified teachers, which would help improve student performance Based on the previous research on the effect of teacher quality on student achievement, I establish the following hypothesis Hypothesis 3: A high percentage of competent teachers in a district has a positive effect on student achievement in the district Local tax revenue Each school district gets a substantial portion of revenue from local property tax as well as from federal and state grants, which makes a difference in the fiscal conditions among school districts Since revenues derived from local property tax have a direct 2567 Impact of Teacher Turnover on Student Performance impact on input resources to education, the amount of local revenue will have a significant positive effect on student achievement (Lin, Quayes, 2006) In addition, high local tax revenue can motivate a local government to provide better public services and outcomes Ostrom, Tibout, and Warren (1961) argued that poly-centricity and fragmented authority within metropolitan regions can provide high quality public service and a greater responsiveness to the preference of residents because local governments would compete with each other to induce more voters and to collect more property tax Likewise, Glaeser (1996) explained that local government officials act to maximize local revenue: they try to provide high quality of public services to raise the value of local estates, which would lead to additional local revenue Previous research demonstrated the positive effect of local property tax on student performance For example, Mensah et al (2013) investigated 217 school districts in New Jersey for the school years of 2002-2009 and found that an increase in the proportion of school district funds raised by local property tax has a positive impact on student achievement improvement Similarly, Lin, Quayes (2006), using data on 507 school districts in Pennsylvania from 1996 to 1999, discovered that an increase in local tax led to higher student performance in terms of SAT scores Also, Fiva, Rønning (2008) investigated the relationship between local property tax and school quality by measuring the external written exam scores of 10th grade students in Norway In Norway, about 90% of funding form school districts comes from the central government, and local taxation for funding school districts is not mandatory and delegated to the local government They found that local taxation is positively associated with school quality measured by student test scores in the nation-wide written exam Hypothesis 4: Local tax revenue in funding for a school district has a positive effect on the student achievement in a district The percentage of White and Asian Students Although the cause of difference of achievement in ethnic group is debated among scholars, there has been considerable agreement on ethnic difference in student achievement: White and Asian students perform better than black and Hispanic students (Steinbergetal, 1992) Steinberg et al (1992) explained the ethnic difference in student achievement with three factors: ‘parenting practices’, ‘familial values about education’, and student belief about the ‘occupational reward of academic success.’ They found that variations in those three factors led to higher performance of White and Asian students than Black and Hispanic students Similarly, Peng, Wright (1994) demonstrated that the superior performances of Asian American students derives from parents’ high educational expectation, education activities, and home environment (physical and monetary resource, and psychological support from family) Warren (1996) argued that parental socioeconomic status (SES), such as parental educational level and family income level, is the most important factor that explains the performance difference between White and Hispanic American students Kao, Thompson (2003) throughly reviewed diverse empirical research and current theories that 2568 Dongjin Oh explain the difference of student achievement in ethnic group and found that, despite student performance gaps among ethnic groups having narrowed, substantial gaps still exist between disadvantaged ethnic groups (i.e Black, Hispanic, and Native Americans) and advantaged ethnic groups (i.e White and Asian Americans) Based on the previous research on the effects of ethnic group on student achievement, I establish the following hypothesis Hypothesis 5: High proportion of White or Asian student population in a school district has a positive effect on the student achievement in the district Sample, Measurements and Method 3.1 Sample In this paper, the unit of analysis is a general school district in Florida To test the hypotheses, I analyzed a panel of 67 general school districts from the school year of 2004-2005 through 2010-2011 I used the following data sources in the Florida Department of Education: 1) Teacher Salary, Experience, and Degree level; 2) Teacher Exit Interview Information; 3) Florida School District Grades; 4) Student Membership in Florida Public Schools; 5) Financial Profile of Florida School District The data sources provided the following annual information at school the district level: average teacher salary, teacher degree level (i.e bachelor, master’s, special or doctoral degree), average teaching experience, teacher turnover rate, school district grade in the annual statewide standardized exam, student demographic information, and local tax revenue per full time equivalent (FTE) student Since the data from the Florida Department of Education was open to the public, there was no restriction on using them for academic purposes With those data, I organized a 7-year data set of the Florida School Districts The final working sample consisted of 469 observations (67districts×7years) In Table 1, I presented descriptive statistics of the variables Table Descriptive statistics of variables (N=469) Variables Description Student Achievement (%) Average student percentage point in annual standardized exam at district level Mean Std.Dev 64.98 4.71 Voluntary Turnover (%) Annual voluntary teacher turnover rate 6.69 3.85 Salary ($1,000) Average annual teacher salary in a district 42.95 4.00 Bachelor’s Degree (%) Percentage of teachers holding a bachelor’s degree 63.00 8.57 Master’s Degree (%) Percentage of teachers holding a master’s degree 33.27 7.71 Special Degree (%) Percentage of teachers holding a special degree 1.85 1.60 Doctoral Degree (%) Percentage of teachers holding a doctoral degree Experience (years) Average teaching experience Local Revenue ($1,000) Annual local property tax revenue per FTE student 4.40 2.50 White Student (%) Percentage of White student in FTE students 62.60 19.74 Asian Student (%) Percentage of Asian student in FTE students 1.38 1.06 92 62 12.57 1.80 Impact of Teacher Turnover on Student Performance 2569 3.2 Measurements Dependent variable I adopted Student Achievement as a dependent variable The Florida Department of Education annually evaluates student achievement through the annual statewide standardized exam To attain sufficient variations in student achievements across the school districts and to compare them among districts over years, this dependent variable was measured as a district’s earned percentage of total possible points in the exam Independent variables Voluntary turnover (H1) in a district was measured as percentage of teachers who left their school districts I excluded other non-voluntary turnovers such as probationary, reduction in the workforce, or retirement I gained annual teacher turnover data from the Teacher Exit Interview Information issued by the Florida Department of Education Salary (H2) was measured as the annual average salary in a school district If possible, it would be ideal to measure the average salary of specific stratum of teachers such as high school teachers with a master’s degree in computer science and ten years of teaching experience, but the Florida Department of Education did not collect such detailed information about teacher salary There was no teacher salary data specified based on the experience year of teachers, school levels, or teacher’s majors Due to the limitation of secondary data, I utilized the average salary in each school district Teacher quality (H3) was measured in terms of a teacher’s degree level and their teaching experience (Lin, 2010) Teacher degree level was categorized into four groups: bachelor’s, master’s, special, and doctoral degree I measured educational degree as the percentage of teachers holding a degree in the school district To prevent perfect collinearity problem, I included the percentage of master’s, special and doctoral degree, but excluded the percentage of bachelor’s degree for the reference group Teaching experience was measured as the average teaching experience years of teachers in each district Local tax revenue (H4) means annual local property tax revenue for funding a school district The size, number of students, number of schools, and annual amount of local property tax is diverse across the school districts Thus, to compare the local tax revenue among the school districts, this variable was measured as annual local tax revenue per Full Time Equivalent (FTE) student in the school district This data was obtained from the Financial Profile of Florida School Districts Proportion of White and Asian students (H5) was measured as percentage of White and Asian students in FTE student population in a school district This data was obtained from Student Membership in Florida Public Schools 3.3 Method To test the hypotheses, I integrated all variables into the following regression equations indicates intercept of the regression, and through means coefficients on independent variables In the regression equation, ‘i’ demotes the school district and ‘t’ denotes the year 2570 Dongjin Oh The variable means fixed effect, representing all unobserved, time constant factors that affect In addition, represents unobserved, time varying error term Exp I conducted the regression model to the investigate determinants of student achievement at school district level Before running the regression model, I conducted the regression assumption diagnostic test because 1) specification error and errors correlated with independent variables would cause biased estimators, 2) in the presence of heteroscedasticity and autocorrelation, the coefficient estimators are unbiased but not best estimators with minimum variance, and 3) high multicollinearity causes inefficiency in coefficient estimators by increasing standard errors of the estimated coefficients (Wooldridge, 2016) In addition, violation of the assumption that error term is uncorrelated with independent variables leads to biased coefficient estimators This violation is caused by the specification error that relevant independent variables are omitted from the regression model, and one or more irrelevant variables are included in the model To detect specification errors, I carried out a regression specification error test (RESET) In the RESET test, the null hypothesis is that this model has no omitted variables The p-value from RESET test was 15, suggesting no omitted variables To identify multicollinearity problems, variance inflation factors (VIFs) were examined The increase in VIF value causes a high variance of the estimator, which means an imprecise estimator Tolerance value defined as 1/VIF is widely used to test multicollinearity problems: a tolerance value smaller than implies a severe multicollinearity problem in the model (Wooldridge, 2016) The smallest tolerance value was 41 (local tax revenue) The VIF test suggested that multicollinearity is not a significant problem in this model To detect heteroscedasticity problems, I conducted the Breusch-Pagan test with null hypothesis that the variance of residuals in the model is constant The p-value from the Breusch-Pagan test was 02, which suggested heteroscedasticity problem I conducted Wooldridge test to check serial correlation in the panel data The null hypothesis in this test is that there is no first-order autocorrelation (AR(1)) The p-value of the test result was 0196, indicating an autocorrelation problem in this model Lastly, I conducted a Hausman test to check random effect or fixed effect in analyzing the panel data Since the p-value obtained from Hausman test was less than 01, I chose fixed effect option rather than random effect option To correct the heteroscedasticity and autocorrelation problems, I chose a ‘xtscc’ pooled OLS model with a fixed effect option The ‘xtscc’ pooled OLS model can produce standard errors of coefficient estimators robust to the disturbance of heteroscedasticity and autocorrelation with fixed effect (Hoechle, 2007) I used Stata Version 14 to analyze the data 2571 Impact of Teacher Turnover on Student Performance Findings I ran the student achievement model to investigate which factors have a significant effect on student achievement at the school district level The result in Table shows that this pooled regression model was statistically significant based on the F test Among independent variables, the voluntary turnover rate, percentage of White students, percentage of teachers holding master’s degree or doctoral degree were all statistically significant First, other things being equal, an increase of one in the percentage of teacher voluntary turnover rate led to, on average, a decrease of 107 point in the average score in the annual statewide exam Given that the average voluntary turnover rate of teacher in Florida is 6.7 percent and standard deviation is 3.8 percent (See Table 1), the effect on student achievement is not large, but still meaningful in that this result demonstrates a negative effect of frequent turnover of teachers on student performance, which supported Hypothesis This result is consistent with the previous research that high teacher turnover rate adversely influences student academic performance (Abelson, Baysinger, 1984; Newman et al., 2001; Barnes et al., 2007) Table Results of pooled time-series regression analysis Variables Voluntary Turnover (%) Salary ($ 1,000) Coef -.107** 062 Std.Err t-score 044 -2.45 062 1.00 -5.27 Master’s Degree (%) -.040*** 008 Special Degree (%) -.063 Doctoral Degree (%) Experience (years) Local Revenue ($1,000) White Student (%) Asian Student (%) Constant 132 -.48 692** 283 2.45 022 043 51 -.274 213 -1.29 105* -.318 58.693*** 0.055 1.90 290 -1.09 4.765 12.32 Observations = 466, F(9, 66) = 143.07, Probability > F = 0.000, R-squared: 0739 *p