Nội dung

University of Arkansas, Fayetteville ScholarWorks@UARK Economics Undergraduate Honors Theses Economics 5-2012 A Brighter Future: The Impact of Charter School Attendance on Student Achievement in Little Rock Karen Brown University of Arkansas, Fayetteville Follow this and additional works at: http://scholarworks.uark.edu/econuht Part of the Economics Commons, Education Commons, and the Education Policy Commons Recommended Citation Brown, Karen, "A Brighter Future: The Impact of Charter School Attendance on Student Achievement in Little Rock" (2012). Economics Undergraduate Honors Theses. 2. http://scholarworks.uark.edu/econuht/2 This Thesis is brought to you for free and open access by the Economics at ScholarWorks@UARK. It has been accepted for inclusion in Economics Undergraduate Honors Theses by an authorized administrator of ScholarWorks@UARK. For more information, please contact scholar@uark.edu, ccmiddle@uark.edu. A Brighter Future: The Impact of Charter School Attendance on Student Achievement in the Little Rock Area By Karen Marie Brown Advisor: Dr. Amy Farmer An Honors Thesis in partial fulfillment of the requirements for the degree Bachelor of Science in International Business with a concentration in Economics. Sam M. Walton College of Business University of Arkansas Fayetteville, AR May 11, 2012 1 Table of Contents I. Introduction……………………………………………………………………………….…....3 II. Charter Schools: An Overview………………………………………………………………...3 III: Literature Review……………………………………………………………………………..3 IV: Data and Methodology………………………………………………………………………..6 V: Results………………………………………………………………………………………...13 VI. Policy Implications…………………………………………………………………………..18 VII. Avenues for Further Research……………………………………………………...……….19 VIII. Conclusion……………………………………………………………………...………….19 IX: References…………………………………………………………………………..……….20 2 I. Introduction School choice aims to enhance educational quality and to create opportunities for students who would otherwise be deprived of a better education. Originally introduced by economist Milton Friedman, this idea creates an educational market of public and charter schools. Market forces will, theoretically, increase the quality of public schools because of competition. As school choice becomes more and more popular, pressure is being exerted on the public school system to increase quality so that the best students will not leave their schools for private or charter institutions. This paper will narrow the field of school choice and will examine the impact of charter schools on National Percentile Rankings (NPR) from standardized test scores of charter and public schools in the Little Rock, Arkansas area. The study hypothesizes that charter attendance positively impacts test score NPRs for both elementary and middle school students. It will open with a brief introduction of charter schools and the literature surrounding them. Then, the data and methodology used in this study will be discussed, followed by the results. Lastly, this paper will include suggestions on charter policy based on the outcome as well as avenues for further research. II. Charter Schools: An Overview The two main forms of school choice are voucher programs and charter schools. Vouchers award a specific amount of money to successful applicants to cover all or some of the cost of private school tuition. When the income variable is less of a consideration for parents, economics tells us that they will choose to send their student to the best school, usually a private school. Vouchers are considered to be the more controversial of the two forms because public funds are used to pay student’s tuition to private, often religious schools. Private schools, in turn, also fear that this paves the way for the government to control their curriculum. Charter schools, or “charters”, are “publically funded, privately operated schools that families can select outside of their zoned schools. They promise greater school-level autonomy in exchange for greater accountability” (Loeb, Valant, Kasman, 2011). Charters are less controversial than voucher programs because operate under a management contract in which the authorization agency may revoke the charter and close the school if at any time it doesn’t meet its requirements and obligations (Scholmer, Shober, Weimer, Witte, 2007). There are two types of charter schools: conversion schools and startup schools. Conversions initially start out as public schools and usually retain existing faculty and students. The motivation to convert is explained by either a need to escape bureaucracy from the public school districts or because the school does not like its mandated curriculum. Conversely, startup charter schools are entirely new schools that “acquire facilities, faculty, and students at their inception.” The motivation for a startup usually derives from the need to create a new “holistic approach to schools”. Because startups tend to be more radical than their conversion counterparts, there is a greater expected difference between startup charters and public schools than conversions and public schools (Buddin, Zimmer, 2005). III. Review of Literature There is an abundance of literature on the impact of charter schools not only on student test scores, but also the test scores of public schools, minority student, and student behavior. Since test score gains are the most direct indicator of educational improvements, the majority of 3 research has been conducted using samples of public and charter schools. A number of studies conclude that charter school attendance leads to some degree of positive test score gains. Studies in Arizona (Nelson, Hollenbeck, 2001) and Boston (Abdulkadiroglu, Angrist, Dynarski, Kane, Pathak, 2011) school districts have determined that charter attendance is positively correlated with an increase in test scores. Another study conducted by renowned school choice researcher, Dr. John Witte, and his colleagues which looks at longitudinal data from schools in the Milwaukee area draws the same conclusion (Witte, Weimer, Shober, Schlomer, 2007). Grosskopf, Hayes, and Taylor (2009) found that Texas schools have positive gains in Math and Reading scores, in which they measured the “value added” to standardized test scores. MacIver and Farley-Ripple declare strong support for the charter school system in Baltimore and say that the city’s Knowledge is Power Program (KIPP) charter schools have shown high achievement levels that have greatly surpassed their Baltimore City Public School System (BCPSS) counterparts. Lastly, Curto and Fryer (2011) found that attendance of SEED schools (a combination of a charter school and a five-day-a-week boarding school) increase achievement by 0.189 standard deviations in Reading test scores and 0.230 in Math test scores per attendance with over an 18% return on investment. It is also important to note that SEED schools have a lottery-based admissions system and are therefore less susceptible to selection bias. Despite the plethora of studies which conclude that charter attendance leads to positive test scores gains, there have also been a significant amount of studies which have concluded just the opposite. Two separate analyses of Michigan charter schools found that students do not reach the same level of achievement as their public school counterparts by 2-9% in Reading, Writing, Math, and Science standardized test scores. In their models, the researchers controlled for student, building, and district characteristics. However, they note that they did not account for selection bias in their study (Eberts, Hollenbeck, 2001). Bettinger (2005) also uses schoollevel data from Michigan to conclude that test scores are negatively affected. In a paper titled “Explaining Charter School Effectiveness”, the authors go as far as to generalize that all nonurban charters are ultimately ineffective because of school-level homogeneity (Angrist, Pathak, Walters, 2011). Given that economists have drawn conclusions on both sides of the spectrum, declaring that charters lead to positive and negative test scores effects, it would be logical to assume that there are a number of “mixed effects” conclusions, which several do. In the paper “Student Achievement in Charter Schools in San Diego”, Tang (2007) finds that charter attendance results in the same gains as public schools overall with the exception of elementary charter Math and Reading score, which drop significantly. Another group of researchers believe that test score gains are possible, but only over a certain period of time. Studies conducted in Wisconsin, New Jersey, and Florida have all suggested that although charter scores may start off lower than or equal to public scores, “performance improves as the charter schools gain experience” (Barr, 2007). When analyzing Florida schools, Sass (2006) supports this claim and found that achievement for charters improves after five years and proposes that market forces due to competition may lead to these gains. If charter schools do in fact have a positive impact on test scores, it seems to be most observable in an urban setting. Several studies suggest that urban areas are the only place which charters can make a significant positive impact. The paper “Explaining Charter School Effectiveness” states “estimates using admissions lotteries suggest that urban charter school boost student achievement, while charters in other setting do not.” Angrist, Pathak and Walters reach also this conclusion after studying student and school-level data from schools throughout 4 Michigan. Zimmer and Buddin propose that this might be the case simply because of demographics. Urban charters tend to serve the most “disadvantaged students” and therefore are more effective because of their impact on below-average achievers. The objective of charter schools in not only to provide an alternative means of a quality education, but to service those who have less access to it. In “Are High Quality Schools Enough to Increase Achievement Among the Poor?” Dobbie and Fryer use data from Harlem Children’s Zone, an experimental program which combines community programs and charter schools. They find that achievement effects are large enough to close the racial gap in elementary, middle, and high schools and believe that “high quality schools are enough to significantly increase academic achievement among the poor”. In another study by Fryer, he urges policy-makers to “take these examples to scale” so that they may have a significant positive impact on the disadvantaged communities throughout the country. Just as with overall charter achievement, there are skeptics who believe charters actually increase the racial gap between whites and minorities. In a scathing paper titled “No Excuses: A Critique of the Knowledge is Power Program (KIPP) within Charter Schools in the USA”, the author Brian Lack argues that KIPP fosters capitalistic and militaristic ideals that preserve the “status quo” and “institutionalized racism” (Lack). North Carolina charter schools are shown to further segregate white and black students. Bifulco and Ladd used time-series data to track the test scores of individual students and find that charter schools “increase racial isolation for both black and whites…and [widen] the achievement gap”. They believe this may be because of asymmetric preferences of each race to attend the charter school where they are the majority. This may explain why there are so few racially balanced charter schools (Bifulco, Ladd). Enrollment of minorities in charters is also the main subject of many other research papers. Along the same lines as the study of North Carolina charter schools, data from 1,006 charter schools households in Texas find that race is a good predictor of parents choose to send their students to a charter school or not. Tedin and Weiher support this argument and say that “Whites, African-Americans, and Latinos transfer into charters schools where there is a 11-14% more of that ethnic group in the student body”. One paper pushes the segregation issue even further and proposes that black enrollment in charters is a function of public school district segregation and state policy which determines school choice legislation. In “Choice, Charter Schools, and Household Preferences”, Kleitz and Bretten point out that although there are differences in school choice among races and socio-economic strata, they do not show a difference in the concern for academic excellence. While most researchers of charter schools focus on more debated topics such as achievement gains, others concentrate on the externalities of these schools. Impacts on the surrounding public schools and student behavior are the most discussed externalities. Renowned economist Milton Friedman believed that the introduction of school choice will create a market for education and competitive pressures will force public schools to increase their quality. Numerous studies have shown that charter schools have a positive impact on the test scores of public schools in surrounding areas (Booker, Gilpatric, Gronberg, Jansen). North Carolina public school test scores are shown to have increased by 1% after the introduction of charters (Holmes, DeSimone, Rupp). Evidence from Michigan and Arizon has also found that charters may lead to the same effect. Nevertheless, other researchers have concluded that charters may cause public school test scores to decline because they drain resources. In Arizona, the studentteacher ratio increased by 6% after charters enticed teachers to work in the more flexible charter environment (Dee, Fu). One paper proposes that public schools become less efficient as resources and taken away (Ni). 5 IV. Data and Methodology To determine if charter attendance has a significant impact of test score NPRs for both elementary and middle school students, I employ the Ordinary Least Squares (OLS) estimation procedure. The intercept parameter “β1” denotes that the dependent variable “Test Score” will not take a value of zero if all other independent variables are controlled for. The charter dummy variable is used as an intercept dummy variable where: Regression 1: E(Test Score NPR)i = (β1 + δ) + β2(%FLP)i + β3(%White)i + β4(%Black)i + β5(%Other Minority)i + β6(School English Language Learner) + β7(School Poverty Index) + εi, when C=1 and, E(Test Score NPR)i = β1 + β2(%FLP)i + β3(%White)i + β4(%Black)i + β5(%Other Minority)i + β6(School English Language Learner)i + β7(School Poverty Index) + εi, when C=0. When C=0, it will denote that a particular school is a public school (or a “non-charter”) and will be the base group for the models, while a C=1 will denote that a school is a “charter” school. Therefore if δ is significant, it will offer evidence that charter schools to have an impact on the test score NPR of a given subject. It is important to note that the Least Squares Estimator’s properties are not affected by the intercept dummy variable. Because “School Percent White”, “School Percent Black” and “School Percent Hispanic” and “Percent Other Minority” would all equal to one and “Percent Overall School Minority” would be equal to 1“School Percent White”, I omitted the “School Percent Hispanic” and “School Percent Overall Minority” variables in each of the equations to mitigate multi-collinearity. Collinearity is where economic variables move together in systematic ways. To compensate for this, any significance in “School Percent Hispanic” will be present in the β1 intercept variable. Next, I use a more refined regression to determine if poverty significantly impacts test score NPRs in all subjects. These two models throw out all race independent variables as well as “School English Language Learner”, only using “School Poverty Index” and “Percent Free Lunch Program”. The two did not show signs of collinearity, so they are both used in the model. Regression 2: E(Test Score NPR)i = (β1 + δ) + β2(%FLP)i + β3(School Poverty Index) + εi, when C=1 and, E(Test Score NPR)i = β1 + β2(%FLP)i + β3(School Poverty Index) + εi, when C=0. Thirdly, this study uses two other models to determine if being a minority significantly impacts test score NPRs in all subjects. The use of “Overall Minority” as a collective group can point towards selection bias in charter schools, which will be discussed in more detail later in this section. 6 Regression 3: E(Test Score NPR)i = (β1 + δ) + β2(%Overall Minority)i + εi, when C=1 and, E(Test Score NPR)i = β1 + β2(%Overall Minority)i + εi, when C=0. The final two models employed in this study take into account both poverty and overall minority variables. Unlike the previous regressions, the charter data for both of these models are separated into “poor-performing charters” (Regression 4) and “well-performing charters” (Regression 5). If all other independent variables are controlled for, these models determine how the charter variable impacts these charter categories. Regressions 4 and 5: E(Test Score NPR)i = (β1 + δ) + β2(%FLP)i + β3(%Overall Minority)i + β4(School Poverty Index) + εi, when C=1 and, E(Test Score NPR)i = β1 + β2(%FLP)i + β3(%Overall Minority)i + β4(School Poverty Index) + εi, when C=0. The school-level data used for this analysis is provided by the University of Arkansas Office for Educational Policy. The data set includes all public and charter schools in the Little Rock, North Little Rock, and Pulaski school districts for the 2010-2011 academic year. The “test score NPR” data used in the study is taken from the Iowa Test of Basic Skills (ITBS) exam as a Norm-Reference Test for all of the schools used in the data set. The ITBS is administered in conjunction with the Arkansas Criterion-Referenced Exam (CRT) to form the augmented benchmark examination. The ITBS contains subtests in Reading, Mathematics, Language, and Science. Table 1 shows all of the variables used in this paper. Table 1: Definitions of all Variables Variable School Name District Name Charter Reading NPR Math NPR Description School name School district name A value of “1” denotes that a school is a Charter and a value of “0” denotes that a school is a public school. School National Percentile Rank (NPR) on the reading subject area of the Iowa Test of Basic Skills (ITBS). School National Percentile Rank (NPR) on the math subject area of the Iowa Test of Basic Skills (ITBS). 7 Language NPR Science NPR Overall NPR % FRL School Poverty Index % White % Hispanic % Black % Other Races % Overall Minority % English Language Learner School National Percentile Rank (NPR) on the language subject area of the Iowa Test of Basic Skills (ITBS). School National Percentile Rank (NPR) on the science subject area of the Iowa Test of Basic Skills (ITBS). Overall School National Percentile Rank (NPR) is the average of the Normal Curve Equivalent for each ITBS Subtest (Reading, Math, Language, and Science). The actual percentage of students in each school who qualify for the Free and Reduced School Lunch Program. The Poverty Index Range is a poverty indicator which gives a greater weight to students with greater need. Percent of students who identify as White. Percent of students who identify as Hispanic. Percent of student who identify as Black. Percentage of students who identify by another race that is not stated above. Percent of overall minority (non-white) students. Percent of students who identify as English Language Learner. Charter students are not a random sample of public school students. They usually enroll as disproportionate amount of either low-achieving and at-risk student or more astute students who seek the freedom or rigorous environment of charter schools (Buddin, Zimmer, 2005). Therefore a difference in test score NPRs may largely be attributed to selection bias within charters. This model will control for “% FLP”, “School Poverty Index”, “% White”, “Percent Black”, “% Other Minority, “% English Language Learner” in order to determine if the “Charter” variable significantly impacts test scores. This paper hypothesizes that attendance of a charter significantly impacts student test score NPRs due to selection bias. An initial comparison shows that charter schools have significantly different demographics, which suggests that selection bias is occurring. Tables 2, 3, and 4 below are the charter’s standard deviations for each racial variable: Table 2: Elementary Charter School Standard Deviation for Race Variables Elementary School Name Arkansas Virtual Academy Standard Deviation of “% White” 2.07 Standard Deviation of “% Black” -1.92 8 Standard Deviation of “% Hispanic” -0.69 Standard Deviation of “% Overall Minority” -2.07 Dreamland Academy eStem Elementary Charter Lisa Academy -1.14 0.41 1.33 -0.45 -0.08 -0.32 1.14 -0.41 0.86 -0.93 -0.32 -0.86 Table 3: Middle Charter School Standard Deviation for Race Variables Middle School Name Arkansas Virtual Academy Cloverdale Aerospace Covenant Keepers Charter eStem Middle Charter Lisa Academy Ridgeroad Charter Standard Deviation of “% White” 2.74 Standard Deviation of “% Black” -2.46 Standard Deviation of “% Hispanic” -0.63 Standard Deviation of “% Overall Minority -2.74 -1.08 -0.75 1.78 1.08 -1.16 0.32 3.70 1.16 0.45 -0.45 -0.31 -0.45 -0.02 -0.78 -1.44 0.79 0.17 0.49 0.02 0.78 Table 4: Elementary and Middle Public School Standard Deviation for Race Variables Public Schools Standard Deviation of “% White” 0.26 Elementary Middle 0.20 Standard Deviation of “% Black” 0.25 0.20 Standard Deviation of “% Hispanic” 0.08 Standard Deviation of “% Overall Minority” 0.26 0.03 0.20 The standard deviations for each race variable disproportionately high for both elementary charters and middle school charters. The highest standard deviation for public elementary and middle schools are only 0.26 and 0.20, respectively. This is in complete contrast to charter schools, which have standard deviations up to 2.74. Although almost all of the charters have high standard deviations in all race variables, eStem Elementary and Middle schools have consistently low deviations. These values, however, are not as low as the highest public school standard deviation. Lisa Academy also has particularly low standard deviations for “% White”, ‘% Hispanic”, and “% Overall Minority”. We can conclude then that public schools have consistent demographics and charter schools tend to have skewed demographics. In an initial comparison of charter and non-charter mean Subtest NPRs, elementary and middle school charter students consistently surpass their non-charter counterparts (Figures 1 and 2). The exception to this trend is the mean NPR of the Language Subtest in which the non- 9