Factors Affecting High School Graduation Rates in the State of Maryland
Advisor: Linden McBride
Economic Statistics - Spring 2019
Abstract
One of the leading questions in public education is what factors contribute a significant amount to the graduation rate of high school students. That is, if there were to be a focus on altering factors that positively or negatively affect graduation, what factors would those be and what impact may occur when they are changed? In this paper, multiple factors and their effects on high school graduation rates in the state of Maryland are analyzed using multiple regression. The factors found to have a significant correlation with graduation rates were the Attendance Rate, Percentage Qualifying for Free/Reduced Meals, and AP Scores.
Introduction
The crisis of the American education system has long been a problem that is acknowledged by the federal and state governments as well as the school systems themselves. This crisis is shaped by America falling behind in educational standards and rankings compared to those of other industrial and technologically advanced nations in the world. Out of 71 countries, the United States has recently been ranked 38th in math and 24th in science, at the high school level [1] (DeSilver, 2017). A multitude of programs have been created and implemented without much success, such as “Race to the Top,” “No Child Left Behind,” and “Common Core”. This paper focuses on factors affecting graduation rates in the state of Maryland, across all counties, so as to identify the most critical factors in keeping students on track to graduate and have an effective education, where they have the necessary skills to enter into the workforce or pursue higher education. This paper identifies the factors that have a significant impact on graduation rates, including attendance rate, expenditure per student, student-teacher ratio, the percentage of total students qualifying for free/reduced meals, teacher qualification level, student AP score, student SAT scores, the median income of the township in which the school is located, and the poverty rate of the same township. Within the admittedly limited academic literature on the subject (not including all the state and official reports), this paper includes both metropolitan and rural school districts similar to previous articles and also includes more suburban areas as well. The previous literature on the subject follows students who graduated prior to the year 2008 and thus does not include the class of students who have participated in the more recent attempts at addressing America’s education problems (Race to the Top and Common Core notably).
[1] This ranking is generated by the three-year Programme for International Student Assessment (PISA) measure of key educational variables.
Graduation Rates by County
Literature Review
Education is an important component of a country’s infrastructure, as a higher standard of education leads to a better work force and increases human capital. Due to this importance, understanding the key factors which contribute to high school graduation, and what factors do not, would create better tools for generating policy that assists students in graduating. Policies focusing on unimportant factors is inefficient and in the long run will not help to solve the problem, and may in fact produce more. The following are studies conducted in various regions and states throughout the country, both urban and rural, which investigate through regression analysis, the effect of multiple factors have on graduation rates.
Lyttle-Burns (2011) had four main research questions about graduation and its relationship with dropping out, social variables (ethnicity, gender, mobility, family status, and grade retention), and perceptions by teachers/school leaders about student success through a study of a class of students (over a period of time) in rural Appalachia.
The data for the research was based upon the structure and methodology of the research, as the research was divided into two phases: gathering quantitative data to answer the first two questions and qualitative data through interviews to address the last two questions. For Phase I, the sample was pulled from nine different schools (n = 391 children) and focused on the graduating class of 2008, with the data gathered on the students starting from the time they entered kindergarten (1995/96) to gauge the true retention and graduation rates (taking into account students dropping out, moving, or withdrawing for other reasons) outside of the normally studied middle and high school years. The study began by examining the cumulative student records on enrollment, educational outcomes (e.g. homeschooled, moved, passed away, still enrolled, dropout or graduate), gender, ethnicity, family status, and the number of grade retentions. The focus on this school district (and the nine schools) was due to the fact of it having one of the highest dropout rates in the state (and being one of the poorest counties in the country) and most of the population in the county had not obtained a high school diploma. Phase II developed questions based on the results of Phase I for interviews with school leaders (principals) and school teachers, as well as graduating students.
The study found that in regards to both graduation and dropping out, gender was not statistically significant, though more females graduated than males and more males dropped out than females. Grade retention was significant for graduation and dropping out: “a total of 83 percent of the students who graduated had never been retained” (Lyttle-Burns, 2011, 102) and 76.7 percent of those students who were retained (at least once) dropped out. Mobility was also found to be statistically significant to graduation and dropping out, as 68.9 percent of students who graduated never moved (Lyttle-Burns, 103) and there exists a positive correlation between the number of moves and the percentage of students dropping out: 31.1 percent dropped out with no moves, 52 percent with one move, and 55 percent with two moves (Lyttle-Burns, 2011, 106).
Ogbuagu (2011) focused on the relationship between high school graduation and the following variables: students’ gender, classroom size, teacher experience, teacher qualification, school attendance, students’ ethnicity, socioeconomic status of students, Limited English Proficiency (LEP), disabilities, school location, subject areas percent passed, principal’s leadership style, and quality of instruction. Ogbuagu also studied what subjects had the highest and lowest pass rates and whether or not that also affected graduation. The research was conducted on 30 schools that were stratified-randomly chosen from 6 various metropolitan school districts in the state of Georgia, where 5 schools were then randomly selected from within each district. The data was collected on-line from each school to be analyzed and compared, and a survey was conducted within one district to determine if the quality of instruction and leadership also had any effect on high school graduation. The use of moderating factors (language arts, mathematics, science, social studies, and writing) were also used as controls in the investigation into some of the variables such as: ethnicity, disability, and gender. Limitations of the research mentioned were that the socioeconomic statuses of students was not taken into consideration and that there was no investigation into why previous strategies to improve graduation rates were not effective.
The results of the quantitative analysis found that the factors that were not statistically significant were classroom size, school attendance, socioeconomic status, LEP, and school location. Other factors were statistically significant in some way to graduation rate. In regards to gender, females had a higher pass rate than males in mathematics, science, English and writing (though males had a higher pass rate in social studies, the margin of which was almost negligible). Ethnicity had varying significance: there was no significant correlation between Asian, Hispanic, white, and others in subject area pass rate (Ogbuagu, 2011, 87); however, Asian, whites, and others did have a higher pass rate than black and Hispanic students. Disability and graduation rates were found to have a significant relationship with graduation rates, varying subject to subject. Teacher experience and teacher qualifications were also found to have a significant relationship with graduation rates. If a school had a higher percentage of teachers with eleven or more years of experience, there was a lower graduation rate and if there was a higher percentage of teachers with ten or less years of experience meant that that there was a higher corresponding graduation rate (Ogbuagu, 2011, 72). A similar relationship was found with teacher qualifications: a teacher with a higher degree of qualification (e.g. M.A., Ed.S., Ed.D) had lower graduation rates than schools with the opposite, teachers with B.A. qualifications.
Ritter (2015) conducted an investigative report into what the most important factors that influence on-time high school graduation are. The data was collected from where available at a state and nationwide level and a summarized various studies that identified graduates and drop outs and the factors that characterized student success and failure. Data categories included economic, demographic, and student factors; where each factor was not considered singularly, but how they were correlated with each other depending upon whether a student was a graduate or a drop out.
The findings of the study showed that of the economic indicators that had a significant impact on whether or not a student graduated were family characteristics such as: socioeconomic status, family stress (death in the family, family move, etc.), family structure, and the mother’s age. It was also determined that students who “fall off track” (Ritter, 2015, 5) in their freshman year have an extremely low chance of graduating. Attendance and student engagement (rigor of the curriculum and challenge posed by teachers) also had a significant impact on student graduation, “attendance even in the first few weeks or month of the freshman year is related to whether students will eventually graduate” (Ritter, 2015, 6). Course failure (or GPA) is another strong indicator: students with a GPA 2.0 or less freshman year, on average, had a significantly lower graduation rate than those with a 2.5 or greater. Demographic results showed that males dropped out more than females; Native American students had the highest drop-out rate of all the ethnicities (followed by Pacific Islander, Hispanic, and Black students); and other demographics with high drop-out rates included: children in foster care, limited English, special education, and homeless students.
Lyttle-Burns (2011) demonstrated an in-depth discussion into factors affecting graduation rates in rural Appalachia. A strong point of the paper was the amount of data retrieved and reviewed, as it included non-publicly available records at the individual level instead of data that has already been categorized and/or grouped in some way. However, the limitations of this paper was that the research was conducted in the poorest (and lowest-ranked amongst graduation) in the area. While this is effective for determining the factors that have a strong correlation to drop-outs and graduation rates, this does not necessarily represent the entire area, or take into account other factors that may be successful in other districts in helping to retain students for graduation. In contrast, Ogbuagu’s (2011) research illustrated a more diverse sample size that would be more representative of the population as a whole, as the data was used from various districts and [multiple] randomly selected schools from within those districts, therefore having a larger sample size. Ritter (2015) was not as in-depth as the previous two papers, as it was more of a report than a true research paper, though it focused more on factors that affect graduation rates starting around freshman year of high school. A limitation of this paper is that it was not as in-depth or as statistically driven. The three papers, Lyttle-Burns (2011), Ogbuagu (2011), and Ritter (2015) all found similar conclusions with the variables that they had in common, with a few exceptions. Both Lyttle-Burns (2011) and Ogbuagu (2011) though they had different data collection methods, used similar analytical methods and formulas.
In relation to the discussed papers, my research is most similar to that of Lyttle-Burns (2011) and Ogbuagu (2011), addressing [approximately] the same questions (i.e. the variables and their relationship to graduation rates) as the two, with a data collection method most similar to that of Ogbuagu (2011). How my paper relates to current and past research is that I will be studying the factors that affect Maryland graduation rates overall, which contrasts to the previous studies mentioned which focused specifically on either rural or metropolitan areas, whereas Maryland school districts have both.
Table 1: Variable Descriptions
Note: ' * ' designates a variable not used in the regression model.
Data and Methods
Data Description and Summary Statistics
The data analyzed in this paper was collected (and summarized) by the Maryland Department of Education as reported from the schools and counties, U.S. News: World Report (compilation of public U.S. government data), and Data USA (compilation of public U.S. government data). The analysis is cross-sectional, focusing on the 2016-2017 school year.
Distribution of Graduation Rates
There are a total of 194 high schools in Maryland. As some counties had over 20 schools and some counties had only 1 school, 3 schools were drawn at random from larger counties and from counties that only had 1 or 2 schools, those schools were kept. Therefore, with 24 counties, there is a total of 63 chosen high schools (. There are 14 variables (Table 1 provides descriptions of the variables in the dataset), though only 10 (11, including the dependent variable, graduation rate) variables were used in the analysis, as some variables were included to calculate the values of the used variables
In Figure 1, the geographic distribution of the average graduation rate by county is displayed. Of the average graduation rates, the outliers include Baltimore City and Prince George’s counties, with the former having an average graduation rate of 56.4 percent and the latter having an average graduation rate of 69.5 percent. The frequency distribution of graduation rates is provided in Figure 2 and illustrates that the majority (40 schools out of 63) had a high graduation rate, that is exceeding 90 percent while 17 schools had a graduation rate of 80 percent. There were 6 schools that had a graduation rate of 70 percent or less.
From the box plots provided in Figures 3-12 we can observe the general distribution of each of the variables as well as any possible outliers. Attendance Rate (Figure 3) is generally high (above 90 percent), with three outliers, two belonging to Baltimore City schools and one belonging to a school from Prince George’s county; Baltimore City having the lowest attendance rates of all the schools: 56.4 percent and 60.5 percent. Expenditure per Student (Figure 4) is on average between $13,027 and $14,903 with no outliers. The Student-Teacher Ratio (Figure 5) has one outlier, with a high school in St. Mary’s County having a high student-teacher ratio of 22 students to one teacher. The Percentage Total Qualifying for Free/Reduced Meals (Figure 6) is generally low, between 17.8 percent and 45.6 percent, with two outliers, both from Baltimore City which are 99.3 percent and 100 percent of students qualifying for free or reduced-price meals. The Average AP Scores (Figure 7) are between 2.2 and 3.1 with no outliers (though the lowest two average scores of 1.2 and 1.3 are both from the Baltimore City schools. The Mean Total SAT Score (Figure 8) is between 1003 and 1114 with two outliers: 784 and 795, both from the Baltimore City schools. The percentage of teachers with a Standard Professional Certificate (Figure 9) is on average between 14.3 percent and 26.3 percent with no outliers. The percentage of teachers with a Advanced Professional Certificate (Figure 10) is on average between 56.3 percent and 77.6 percent, with one outlier of 23.1 percent, belonging to a Baltimore City school. The average Median Income (Figure 11) is between $46,125 and $87,533 with two outliers: $150,000 and $153,000 in Montgomery and Calvert counties respectively. The Poverty Rate (Figure 12) is generally between 7.0 percent and 21.9 percent with no outliers.
Table 2: Pearson Correlation Coefficients
Table 2 and the corresponding scatter plot matrices (Figure 13) illustrate the correlation between the independent variables and the dependent variable (Graduation Rate). The variables that have a strong positive relationship, that is as the value of the independent variable increases, the Graduation Rate increases, are: Attendance Rate, AP Score, Mean Total SAT Score, and Median Income. As the Attendance Rate increases and the Median Income, AP and Mean Total SAT Scores rise, so too does the Graduation Rate. Both Standard and Advanced Professional Certificates have weak positive relationships. The variable that has a strong negative relationship is the Percentage Total Qualifying for Free/Reduced Meals, which means that as the number of students that qualify increases, the Graduation Rate decreases. All other variables (Expenditure per Student, Student-Teacher Ratio, and the Poverty Rate all have weak negative relationships.
Methods and Hypothesis
The planned method of analysis is a multiple OLS regression analysis that will determine if there exists a significant relationship exists between the independent variables and the dependent variable, Graduation Rate.
The regression model is:
Findings and Discussion
Estimated Model
Table 3: Parameter Estimates
Analysis
The regression (Table 3 and) shows that Attendance Rate, Percentage Total Qualifying for Free/Reduced Meals, and AP Scores are statistically significant at the 95% confidence level. The parameter estimate for the Attendance Rate is , which is not large (although it is large as compared to other parameter estimates), though it is positive, indicating that for every percentage point increase graduation rate increases. The Pearson correlation coefficient for Attendance Rate is which also indicates a strong positive relationship between Attendance Rate and Graduation Rate.
For the percentage of students that qualified for the free or reduced meals, the parameter estimate is which like Attendance Rate is not large in magnitude (though, like the Attendance Rate it is large as compared to the other variables’ parameters) and is negative indicating that for every percentage point increase in the percent of students that qualify, there is a decrease in the Graduation Rate. The Pearson correlation coefficient for students qualifying for Free/Reduced meals is -0.7992, which is indicative of a strong negative relationship with the Graduation Rate.
The final statistically significant variable is AP Scores, the parameter estimate is the largest of all the parameter estimates (excluding the intercept) and negative in nature. This means that (contrary to the Pearson correlation coefficient of 0.503), there exists a negative relationship between the two variables.
Implications and Limitations
The implications of the analysis include that attendance is, in fact, important for and indicative of graduating from high school, which is supported by Ritter (2015). This is contrast to the findings of Ogbuagu (2011), in which Attendance Rate was found to not be statistically significant. This difference could be due to the limitations of the data of this paper, as Ogbuagu had more data (at an individual level) with which to conduct analysis. Also contrary to Ogbuagu’s study, the qualifications of the teacher were not found to be statistically significant (at an α of 5%). This difference could also be due to the difference in data; Ogbuagu had individual qualifications at the school(s) of interest and could compare those qualifications to the individual graduation rates, whereas the data in this paper was based on the percentage of teachers within the school which had the SPC or APC qualifications (which does not indicate total years of experience or the level of degree that the teachers had attained).
Assuming that the Percentage Total Qualifying for Free/Reduced Meals is indicative of the socioeconomic status of the students at the school, which was also found to be true in Ritter (2015). Ogbuagu (2011) and Lyttle-Burns (2011) however, did not consider the socioeconomic status of students; though Ogbuagu (2011) did mention this consideration was omitted from the study.
Overall, the findings of this paper do fit with the literature. Any discrepancies or differences may be attributed to the difference in data level (the county and school level vs. the individual level of the mentioned literature).
Assumptions made in regard to the data include that the majority of students taking AP and SAT tests are seniors. While mostly true for the latter (though some juniors also participate), AP classes and tests are available to students at all grade levels; though there also is the assumption that the students who take the tests will graduate (as most take the classes for college credit). Another limitation would be that the data is given at the school level and not at grade level (or ideally, the individual level).
While the graduation rate does take into account students leaving and students coming it, the impact of moving schools is not taken into account for factors that may or may not effect graduation rates. It was also assumed that the attendance rate (which was also school-wide) was reflective of the attendance rate of the class of 2017, the limitation of having data only at the school level as opposed to at the individual level.
Conclusion
While there are many factors that affect graduation, the only significant factors found in this paper are Attendance Rate, the Percentage Total Qualifying for Free/Reduced Meals, and AP Scores. In summary, the major limitations of this paper is that the data used is not limited to that of the senior class or graduating class and is inclusive of the other classes and therefore leads to unaccounted for bias in the regression. The findings of this paper show that as found in Ritter (2015), attendance rate and the socioeconomic status of students are linked to graduation rates. The next step in analyzing factors that affect graduation rates in Maryland would be to conduct more focused research at the individual level of a specific class, from when they begin school to when they end school; this would mean a more accurate approximation of the significance of the factors studied in this paper.
Works Cited
DeSilver, D. (2017, February 15). U.S. students' academic achievement still lags hat of their peers in many other countries. Pew Research Center.
Lyttle-Burns, A. (2011). Factors That Contribute to Student Graduation and Dropout Rates: An In-Depth Study of a Rural Appalachian School District. Eastern Kentucky University.
Ogbuagu, A. B. (2011). Factors affecting high school graduation rates in metropolitan Atlanta public schools. ETD Collection for AUC Robert W. Woodruff Library(Paper 214).
Ritter, B. (2015). Factors Influencing High School Graduation. Washington Student Achievement Council.
