Student achievement or academic performance in public schools in the United States is a topic that has become a hot topic socially, politically and also economically because it is one of the issues that affects not only our young people today, but also our society as a whole for future generations. This issue has become so important in fact, that in 2001, President George W. Bush enacted the policy of â€œNo Child Left Behind.â€ This policy requires all states to implement their own system of measuring student achievement within their schools and based on that system of measurement, reward or punish schools that donâ€™t meet required standards (Toutkoushian and Curtis). Schools that are not â€œmaking the gradeâ€ so to speak would need to be reformed or undergo changes to make sure that they meet standards in the future. The problem has been placed on the individual states of how to measure student achievement, therefore a whole new set of questions have arisen about other factors that can affect achievement or student performance.
The first problem is how do the schools measure student performance and academic achievement? Do they use metrics that are already in place such as graduation rates, dropout rates, SAT scores, etc; do they introduce a set of standardized tests; or use a combination of both? The answer to this first question has been mainly left up to the individual states. Indiana, for example, uses the state standardized ISTEP test, which was already in place before the national policy. On the other hand, New Hampshire is one of the few states that have yet to come up with any measurement tool (Toutkoushian and Curtis).
The second problem has been the evaluation of the actual metrics or the measurement tool, which is where I will focus my attention and where much of the literature I have reviewed focuses. When analyzing the data, the policymakers must decide which factors have the greatest impact on the performance of the students. If one looks at an underperforming school and compares it to a high performing school, are those differences attributed merely to differences within the schools themselves, or are there external factors at play? Does the data merely say that the high performing school have better teachers, facilities, technology, and administrators than the underperforming school or could there be something outside the school affecting the performance within the school?
This paper will attempt to examine some of those external factors and determine quantitatively, the impact that they have on the performance or achievement levels that the students can achieve within the schools. I will examine some of the social and economic factors that are hindering our schools and preventing our young people from reaching their fullest potential.
The disparities between students that perform well in primary and secondary schools and those that do not have been attributed to a variety of external factors including race, gender, family income level, family background, education levels, and even the type of neighborhood in which the school is located. The two most commonly considered factors in the National Assessment of Educational Progress results however are race and poverty (Achievement Gap). Results have shown that a disparity still exists between white students and their black and Hispanic peers despite efforts within the schools to address these population groups, especially in poverty stricken neighborhoods (Achievement Gap). While there have been some efforts to address this problem, such as reducing class and school size, improving the quality of teachers in these neighborhoods, and expanding programs in these schools, progress has been slow (Achievement Gap). Research shows that poverty is a major factor influencing the performance of young children especially since families in less affluent families lack the necessary resources to provide books for students, trips to the local library or even available time to read to younger children (Lara-Cinisomo, Pebley and Vaiana). The RAND Corporation also points out that the prior education experience of the parents not only influences the educational experience of the children, but also exacerbates the poverty problem. Their study indicates that these children do not get exposed to early reading related opportunities and do not get the assistance with Math and Literature that children with more highly educated parents would get simply because their parents are ill-equipped to do so. Furthermore, long-term research supports the belief that as educational gains of parents are made, the effects of poverty will be lessened which in turn, increases the scholastic achievement levels of the students (Lara-Cinisomo, Pebley and Vaiana). However, the RAND study found that these gains are not occurring at the same rate among different people groups, which has resulted in an ever-widening gap in the achievement rates among the children of these groups.
According to research, the early years of development (birth to 5 years), is a critical time of development and children that live in lower middle class families or those below the poverty line begin kindergarten approximately 18 months behind children from more affluent families (The Individualist). Research also indicates that these under privileged children begin school with a vocabulary of approximately 4,000 words as opposed to 12,000 of their more affluent counterparts. These statistics can be attributed to poverty and lack of education of the parents as a factor insomuch as â€œdevelopment without meaningful content has only limited usefulnessâ€ (The Individualist).
Based on the research of Jennifer Barry in her bachelorâ€™s thesis, students are greatly influenced by the conditions they face in the home as this is the primary social network during early development. Barryâ€™s regression analyses determined that family factors such as financial resources, family discussions, and parental involvement in daily student lives all had positive influences in student test scores. Similarly, New Jersey has created what is called â€œDistrict Factor Groupsâ€ as a method of studying these types of socio-economic factors that affect student achievement. In the DFG study, New Jersey uses census data and test results to analyze the relationship between certain socio-economic indicators and student academic performance. The socio-economic indicators that are used for this study include: percentage of adults with no high school diploma, percentage of adults with some college education, occupational status, unemployment rate, percentage of individuals in poverty and median family income (NJ Department of Education). In an effort to place all school corporations on equal footing, the state of New Jersey, only reported schools with a certain number of respondents (over 70) and eliminated communities with a disproportionally high number of students enrolled in non-public schools. This second omission is a result of other studies that have been performed showing that academic performance of private schoolchildren tend to be significantly higher than those of public schoolchildren (Braun, Jenkins and Grigg). By eliminating communities with low numbers of respondents or high percentages of private school enrollees, New Jersey increased the probability of a more homogeneous set of data. The 2000 New Jersey study found that the factors with a positive influence on student academic performance were the occupational status of the parents, parents with at least some college, and median family income while the factors that contributed negatively to academic performance were parents without a high school diploma, families living below the poverty rate and the unemployment rate (NJ Department of Education). This would seem to support other research available and anecdotal information collected.
Given all of this information about the impact of socio-economic factors on student performance, the question still remains about how schools can overcome these factors to improve academic performance within the schools. Robert Toutkoushian and Taylor Curtis addressed issue in their paper entitled â€œEffects of Socioeconomic Factors on Public High School Outcomes and Rankings. They concluded through regression analysis that the above factors indeed did impact academic performance of students in grades 3, 6, and 10 (Toutkoushian and Curtis). Unfortunately, because these factors are external to the school and beyond the control of the school, Toutkoushian and Curtis concluded that it made it difficult to provide clearly defined recommendations to implement at the school level. However, they did learn through their study that the per-student expenditure at the school level has had some success in overcoming these external factors (Toutkoushian and Curtis).
Data Model and Description
The data was collected from the Indiana State Department of Education where census data and ISTEP scores are made available for over 294 school corporations from the last U.S. national census. In this case, the data used included the school year 2000-2001 and only includes 294 of the Indiana school corporations as there were approximately 30 school corporations for which census and/or ISTEP data were not available or incomplete for the desired date range.
The dependent variable during the analysis phase of this investigation and in all of the models was the percentage of students that has passed the ISTEP during the given school year (2000-2001). This variable offered three variations that will be shown later in the discussion of this analysis: students that had passed both portions of the ISTEP (English and Mathematics), students that had passed only the English portion of the ISTEP, and students that has passed only the Mathematics portion of the ISTEP. Since all of this information was made available, the decision was made to make use of it and compare the results through regression analysis.
Based on my findings from literature and data available in the census reports, I started my research with seven independent variables including several relating to adult education status, poverty level, and parents marital status. Specifically, my seven initial independent variables were: 1) percentage of persons greater than age 25 in the corporation without a high school education, 2) percentage of families below the poverty line, 3) percentage of persons greater than age 25 in the corporation with a high school education, 4) percentage of persons greater than age 25 with at least some college (with or without at least some degree), 5) percentage of families that have not moved in the last 5 years, 6) percentage of single parent families (father as caregiver) and 7) percentage of single parent families (mother as caregiver). The literature and research available shows that each of these variables has, at least in other locales, had an impact on student academic performance. So the goal of my research to determine the impact, if any, of these variables on the ISTEP scores of Indiana students.
In others studies that I have studied the first two variables: a parent (or an adult) population without a high school education and a high poverty rate were found to be the highest contributors to low student performance. It was my hypothesis that my next two variables would have an impact because in areas where there were higher rates of adults had successfully completed high school and/or gone on to attend college, there would be a greater incentive for those adults/parents to help their students achieve. The fifth variable relating to the families that had moved recently was of my own choosing and since it was available, I decided to test the impact of this variable on student performance. Finally, the last two were offered by the census statistics either separately and together so I chose to use them separately. By doing this, I decided to show the difference in homes where the father was a single parent caregiver and those where the mother was the single parent caregiver. We live in a society where we have both types of single parent homes, so I decided to study both in this analysis.
Empirical Analysis and Discussion
The analysis of the ISTEP scores and student performance of Indiana students during the 2000-2001 school year began with running linear regressions using the independent variables above against each dependent variable chosen. These initial regression statistics would give me a baseline for further analysis and also help me decide which dependent variable was the best indicator of student performance.
The issue with the dependent variables was determining whether or not they could themselves be used as a good indicator for good or bad student performance because the ISTEP test itself has always had its opponents that say that the test is flawed as an indicator of student cognitive development or memory retention. Also, given that the test is broken into Mathematics and English portions, it might be expected for the numbers of students passing the Mathematics portion to be lower (and consequently the number of students passing BOTH portions to be affected). So, I decided to also test my dependent variables while testing the independent variables for these reasons.
Using all of the above independent variables my baseline linear regressions resulted in an adjusted R2 value for students passing both portions of the ISTEP of 24.4%, a value of 38.9% for students passing just the English portion and a value of 31.0% for students passing just the Math portion. While none of these regressions are presenting ideal numbers (as the adjusted R2 gives us the % of variation due to the given independent variables), the English portion of the ISTEP is by far the best indicator available in this study for student performance. (Detailed Regression statistics can be found in Appendix A)
Moving forward as the percentage of students passing the English portion of the ISTEP as my dependent variable given the adjusted R2 results above, I found the coefficients of regression as shown in Table 1 below:
Shown in the table above, the variables that all have a negative impact on student performance are NOSCHED, PVRTY, HSED, SINGLEDAD, and SINGLEMOM (Which denote: % of adults without a HS education, % of families living in poverty, % of adults with a high school with a HS education, % of families with single father caregiver, and % of families with a single mother caregiver respectively). Those variables that positively impact student performance include COLLEGE and NOMOVE (which denote % of adults with college experience and % of families that have not moved in the last 5 years). To say that each of these variables either negatively or positively impacts a studentâ€™s performance is to say that a 1% increase in one of these variables results in a 1% increase (positive influence) or a 1% decrease (negative influence) in the overall student populationâ€™s performance levels. However, while each of these variables contributes to an increase or decrease in achievement, I also had to determine whether that impact was significant. In the case of NOHSED, PVRTY, NOMOVE and SINGLEDAD, the impact was not significant because and were eliminated from subsequent regressions. (Details of the above tests can be found in Appendix A)
I ran several regressions on the data with multiple combinations of the above variables and each time, the above variables always had the same significance test results, so my final regression variables were HSED, COLLEGE, and SINGLEMOM with the coefficients shown in Table 2.
The results in Table 2 above demonstrate that HSED and SINGLEMOM both have a negative impact on student academic performance and COLLEGE has a positive impact. Of the three remaining variables previously mentioned, SINGLEMOM has the greatest impact of all of the variables with a -0.73124 impact on student performance. That is, for every 1% increase in single family households within a school corporation with the primary caregiver being the mother, there will be a resulting drop of 0.73124% in scores on the English portion of the ISTEP test.
Several of the problems with regression models are multicollinearity (the independent variables are linearly related), heteroscedasticity (the variance of the random error is not a constant for all i), and serial correlation (the random errors are correlated). I ran tests for all of these potential regression problems on the three remaining independent variables to eliminate or confirm any problems with the data. Tests do not reveal any indication of a multicollinearity problem among the remaining independent variables (HSED, COLLEGE, and SINGLEMOM) as shown in the results in Appendix C. While I cannot rule out a multicollinearity problem, I also do not suspect one due to the tests performed on the data.
Unfortunately, tests for heteroscedasticity uncovered such errors as shown in Appendix C. These errors are not fixable with the tools available to me, but indicate that the estimators for the coefficients of my variables are not the most efficient available. Finally, to test for serial correlation, I ran a regression to calculate the Durbin-Watson Statistic (also shown in Appendix C). In this analysis, I have concluded that there is no serial correlation present.
In the beginning of this research project, it was my goal to show that there are factors beyond the control of our school administrators, faculty and teachers that ultimately are affecting the educational performance of our students. Much pressure has been placed on them by the state and local governments since the inception of the No Child Left Behind policy in 2001 to monitor, report, and improve the performance and achievement levels within our schools, but it is my belief that there is only so much that they can doâ€”especially with external influences factoring into the equation. There exists much evidence and literature on the topic of academic performance but gathering the data necessary to fully understand the problem so that we can develop comprehensive problems proves difficult.
My search found what initially looked to be reliable data from the Indiana State Department of Education, but upon further review, many schools did not have complete data for the ISTEP schools which slightly compressed my sample size. The census data also proved challenging as I had to reformat and reorganize data into a usable format before running my regression sets. Furthermore, while I found little supporting evidence of multicollinearity or serial correlation problems, there was a problem with heteroscedasticity which could have been caused by the collection of data, formatting of the data for use in the analysis or the software used to run the regressions.
In conclusion though, the data collected and the methods used show that, at least in Indiana, communities with a higher percentage of adults without a high school diploma and a higher percentage of single mothers show lower test scores on the English portion of the ISTEP test and communities with higher percentages adults with at least some college show higher scores on the English portion of the ISTEP test.
Using these results, educators can at the very least begin to target these social issues within their communities at the school level. For example, they can begin to stress the importance of strong marriages in family relationships and home economics classes. Another idea may be to offer programs for adults in the community for GEDs, etc.
By better understanding why our children are not succeeding, we will be able to reach out to them and create programs to help them succeed. As shown by this and other studies, sometimes, that means taking the indirect route. Our schools have tried nearly everything within the schools, perhaps there is enough research to show that the community needs some attention as well.
“Achievement Gap.” Education Week (2004).
Barry, Jennifer. The Effect of Socio-Economic Status on Academic Achievement. Masters Thesis. Wichita State University. Wichita: Department of Sociology, 2006.
Braun, Henry, Frank Jenkins and Wendy Grigg. Comparing Private Schools and Public Schools Using Hierarchical Linear Modeling. NCES(2006-461). National Center for Educational Statistics. Washington, DC: U.S. Government Printing Office, 2006.
Indiana Department of Education. Extract Indiana Education Data. 24 11 2011 <http://mustang.doe.state.in.us/SAS/sas1.cfm>.
Lara-Cinisomo, Sandraluz, et al. “A Matter of Class.” 01 October 2006. Rand Corporation. 19 January 2011 .
NJ Department of Education. “District Factor Groups for School Districts.” DFG Report. 2010.
The Individualist. “How Do Socio-economic Factors Affect Early Literacy?” Hubpages. 20 11 2011 .
Toutkoushian, Robert K and Taylor Curtis. “Effects of Socioeconomic Factors on Public High School Outcomes and Rankings.” Journal of Educational Research 98.5 (2005): 259-271.