Identification of Comparable Elementary Schools: Fact or Fiction?

Rita O'Sullivan
ritao@unc.edu
University of North Carolina - Chapel Hill
Brent Cooper, University of North Carolina - Chapel Hill, EvAP School of Education, CB 3500, Chapel Hill, NC  27599
"cooperjb@email.unc.edu" <cooperjb@email.unc.edu>
Emily Cook, EvAP School of Education, CB 3500, Chapel Hill, NC  2759
emily cook <emilyemilyc@yahoo.com>
Mary Julia Moore, EvAP School of Education, CB 3500, Chapel Hill, NC  2759
"mjmoore@erskine.edu" <mjmoore@erskine.edu>
Beth Bader, EvAP School of Education, CB 3500, Chapel Hill, NC  2759
"bbader@email.unc.edu" <bbader@email.unc.edu>

Currently much emphasis is placed on the use of experimental or quasi-experimental designs is education and social science research, which require the use of equivalent control or comparison groups. When the school is the unit of analysis, typically researchers “match” schools based on a variety of criteria and then randomly assign one of the matched pairs into the treatment and the other becomes part of the control groups. The effectiveness of this procedure has been questioned in the past (O’Sullivan et al, 2003; O’Sullivan et al 2005) and shown to be somewhat suspect with one large data set from an urban school district in the United States. The purpose of this study is to further expand the discussion around the usefulness of different matching procedures to form equivalent groups.   Using data from the North Carolina Schools Report Card reports from 2001-06, this study will test the following research questions for the four largest school districts in North Carolina:

o       Which predicators are the best for determining equivalent school achievement?
o       How well do the best predictors of achievement hold up across multiple years to support school equivalency?

Correlational analyses and group mean differences will be used to answer the research questions, holding school districts constant and controlling for school size (large, medium and small). Preliminary results suggest that identifying equivalent schools, even in the same school district, can be problematic.


Objectives
Currently much emphasis is placed on the use of experimental or quasi-experimental designs is education and social science research, which require the use of equivalent control or comparison groups. When the school is the unit of analysis, typically researchers “match” schools based on a variety of criteria and then randomly assign one of the matched pairs into the treatment and the other becomes part of the control groups. The effectiveness of this procedure has been questioned in the past (O’Sullivan et al, 2003; O’Sullivan et al 2005) and shown to be somewhat suspect with one large data set from an urban school district in the United States.

The purpose of this study is to further expand the discussion around the usefulness of different matching procedures to form equivalent groups.   Using data from the North Carolina Schools Report Card reports from 2001-06, this study will test the following research questions for the four largest school districts in North Carolina:

o       Which predicators are the best for determining equivalent school achievement?
o       How well do the best predictors of achievement hold up across multiple years to support school equivalency?

Perspective or Theoretical Framework
        According to past research, equivalence is often assumed on the basis of only one or two school characteristics.  For example, Josendal, Aaro, Torsheim, and Rashbash (2004 divided schools into groups to test the effectiveness of a school-based smoking program.  The only factor they took into consideration when determining equivalence was school size.  They assumed that as long as schools were within plus or minus 10% of the same size, they were equivalent.
        Another common practice used when establishing equivalent schools is matching schools based on demographic characteristics.  Lynch, Taymans, Watson, Ochsendorf, Pyke, and Szesze (2007) matched five pairs of diverse middle schools on several demographic characteristics including: gender, ethnicity, free and reduced lunch status, and English as a Second Language (ESOL) status.  Riordan and Noyce (2001) used a similar set of criteria to put schools into comparison groups based on the following demographics: (1) Eligibility for Free/Reduced Lunch, (2) Limited English Proficiency, (3) Special Education, or (4) Ethnicity (Asian, Black, Hispanic, White) as well as prior school performance.  Using a combination of demographic variables and prior school performance, as indicated by achievement scores appears to be one of the most common methods of establishing equivalency in school-based research.  Both McDougall, Saunders, and Goldenberg (2007) and Zhang, Fashola, Shkolnik, and Boyle (2006) also used a combination of prior years achievement scores and demographic variables to establish equivalent schools.
O'Sullivan, et al (2003) used data from a large urban school district to examine equivalence of 122 elementary schools.  Using school size, as a matching variable they examined student's standardized achievement test scores in mathematics and reading. Their findings showed less than 10% of school were equivalent on both measures.  These percentages declined as more matching criteria were added. A second study, using the same data set revealed that pre-achievement was the best predictor of current achievement but that consistency from Year 1 to Year 2 was very unreliable.


Methods or Techniques

The sample for this research study began with all the elementary schools within the four largest school districts in the state of North Carolina for the 2001-06 school years. Elementary schools were chosen because end-of-grade test score data exist for all elementary schools for grades three through five. The four largest school districts were chosen to control for school district effects as well as because there were large numbers of schools to use for comparisons.

Within the school districts several factors were considered consideration as potentially important predictor variables: average student expenditure, student needs, percentage of students eligible for free and reduced lunch, percentage of students identified as Limited English Proficient (LEP), and grade range of students within a school.

Pairs of elementary schools in each of the four school districts were chosen that had similar school characteristics according to the previous factors identified as important in the literature.  Specific school characteristics this research group selected as important when choosing pairs of schools in each district included: (1) K-5 grade configuration, (2) school size; small, medium, or large, and (3) previous achievement of students at these schools as determined by reading end-of-grade test scores.  Data for the 2002-03, 2003-04, and 2004-05 school years were used to test for school equivalence amongst pairs of schools chosen for each school district.  Median splits were used to distinguish small, medium, and large schools.  Pairs of schools were selected as potentially equivalent based on the percentage of students scoring three or above on the reading end-of-grade test in 2001 and 2002.  Effect sizes were calculated for each school once standard deviations and means were calculated.  Pairs of schools were defined as equivalent if difference in effect size was less than or equal to .5.          

Data Source & Preliminary Results

Data from the 2001-02, 2002-03, 2003-04, and 2004-05 school years from the North Carolina ReportCard data were used to test for school equivalence amongst pairs of schools chosen for each school district. Correlational analyses and group mean differences will be used to answer the research questions, holding school districts constant and controlling for school size (large, medium and small). Preliminary results suggests that identifying equivalent schools, even in the same school district, can be problematic.

References

Josendal, O., Aaro, L., Torsheim, T., & Rashbash, J. (2004). Evaluation of the school-based smoking prevention program “BE smokeFREE”. Scandinavian Journal of Psychology, 46, 189-199.
Lynch, S., Taymans, J., Watson, W. A.,&  Ochsendorf, R. J., Pyke, C., Szesze, M. J. (2007). Effectiveness of a highly rated science curriculum unit for students with disabilities in general education classrooms. Council for Exceptional Children, 73, 202-223.
McDougall, D., Saunders, W. M., & Goldenberg, C. (2007). Inside the black box of school reform: Explaining the how and why of school change at getting results schools. International Journal of Disability, Development, and Education, 54, 51-89.
O'Sullivan, R. G., Broadway, E.,, Bundy M., Gould, T., Hedgpeth, M. W., Petterson, M. & Poole, L. (2005). Exploration of Comparison of School Equivalence.  Paper presented at the annual meeting of the North Carolina Association for Research in Education.
O'Sullivan, R. G., Fedora, P., Levine, S., MacKInnon-tucker, D. McCullough, A. K. & Shaw, T. (2003). Comparison School Equivalence: Reality or Myth.  Paper presented at the annual meeting of the North Carolina Association for Research in Education, Holly Springs, NC.
Riordan, J. E., & Noyce, P. E. (2001). The impact of two standards-based mathematics curricula on student achievement in Massachusetts. Journal for Research in Mathematics Education, 32, 368-98.
Zhang, Y., Fashola, O., Shkolnik, J., & Boyle, A. (2006). Implementation of comprehensive school reform and its impact on increases in student achievement. Journal of Education for Students Placed at Risk, 11, 309-329.