Defining the Causes of Educational Achievement:
|
Variables |
Observations |
Mean |
Std. Deviation |
Min |
Max |
Dependent Variables |
|
|
|
|
|
Pursuing Post-Secondary Education |
12003 |
0.522 |
0.5 |
0 |
1 |
Diploma Status Two Years after Class Graduation |
12047 |
0.865 |
0.341 |
0 |
1 |
Independent Variables |
|
|
|
|
|
Parental Social Capital Index |
2134 |
0 |
1.677 |
-18.607 |
1.34 |
Family attends church |
10329 |
0.243 |
0.429 |
0 |
1 |
Parent volunteers at school |
10393 |
0.389 |
0.488 |
0 |
1 |
Parent attends school activities |
9473 |
0.655 |
0.476 |
0 |
1 |
Parent involved in student's academics |
3117 |
0.768 |
0.422 |
0 |
1 |
Parent/Student discuss activities |
10464 |
0.943 |
0.25 |
0 |
1 |
Parent/Student discuss college |
10451 |
0.93 |
0.256 |
0 |
1 |
Parent/Student discuss studies |
10435 |
0.955 |
0.227 |
0 |
1 |
Parent/Student discuss grades |
10435 |
0.974 |
0.16 |
0 |
1 |
Parent/Student discuss Troubles |
10463 |
0.967 |
0.18 |
0 |
1 |
Parents are married |
10585 |
0.793 |
0.405 |
0 |
1 |
Family spends time talking together |
10410 |
0.944 |
0.229 |
0 |
1 |
Parent helps with homework |
9484 |
0.634 |
0.482 |
0 |
1 |
Family attends social functions together |
10392 |
0.857 |
0.351 |
0 |
1 |
Parent/Student visited colleges |
7312 |
0.697 |
0.459 |
0 |
1 |
Family rule about homework |
9313 |
0.79 |
0.407 |
0 |
1 |
Family rule about attendance |
9321 |
0.906 |
0.292 |
0 |
1 |
Parent involved in the neighborhood |
10490 |
0.778 |
0.417 |
0 |
1 |
Parent learned about financial aid |
9526 |
0.713 |
0.452 |
0 |
1 |
Community Social Capital Index |
6742 |
0 |
1.309 |
-6.13 |
2.087 |
Many parents attend P/T conferences |
8637 |
0.746 |
0.435 |
0 |
1 |
Many parents volunteer at school |
8613 |
0.23 |
0.421 |
0 |
1 |
Many parents participate in the PTA |
8320 |
0.436 |
0.496 |
0 |
1 |
Many parents are involved in an organization |
8744 |
0.541 |
0.498 |
0 |
1 |
The school is safe |
10439 |
0.798 |
0.402 |
0 |
1 |
The town is safe |
10510 |
0.942 |
0.235 |
0 |
1 |
Teachers are interested in Students |
10340 |
0.777 |
0.417 |
0 |
1 |
School assists with applications |
8990 |
0.998 |
0.408 |
0 |
1 |
School assists with financial aid |
8992 |
0.996 |
0.065 |
0 |
1 |
School has a truancy/dropout program |
8723 |
0.553 |
0.497 |
0 |
1 |
School provides a values education |
8817 |
0.898 |
0.302 |
0 |
1 |
School promotes citizenship |
8858 |
0.95 |
0.2218 |
0 |
1 |
Alcohol is not a problem |
8863 |
0.932 |
0.252 |
0 |
1 |
Drugs are not a problem |
8765 |
0.804 |
0.397 |
0 |
1 |
Teenage Pregnancy is not a problem |
8850 |
0.649 |
0.477 |
0 |
1 |
Friend Social Capital Index |
7866 |
0 |
1.3 |
-9.74 |
0.45 |
Drugs are not a problem among friends |
10074 |
0.934 |
0.248 |
0 |
1 |
Grades are important to my friends |
10092 |
0.93 |
0.255 |
0 |
1 |
High School is important to my friends |
10091 |
0.973 |
0.162 |
0 |
1 |
I am often with friends |
9995 |
0.889 |
0.315 |
0 |
1 |
I have few friends who dropped out |
9937 |
0.921 |
0.27 |
0 |
1 |
Friends are important |
10082 |
0.97 |
0.172 |
0 |
1 |
Control Variables |
|
|
|
|
|
Asian/Pacific Islander |
10544 |
0.056 |
0.23 |
0 |
1 |
Hispanic |
10544 |
0.108 |
0.31 |
0 |
1 |
Black, not Hispanic |
10544 |
0.098 |
0.298 |
0 |
1 |
White, not Hispanic |
10544 |
0.728 |
0.445 |
0 |
1 |
American Indian/Alaskan Native |
10544 |
0.01 |
0.1 |
0 |
1 |
Parent's work status |
10330 |
0.767 |
0.423 |
0 |
1 |
Family income up to $25K |
10194 |
0.328 |
0.45 |
0 |
1 |
Family income between $25-$50K |
10194 |
0.346 |
0.476 |
0 |
1 |
Family income between $50-$100K |
10194 |
0.25 |
0.433 |
0 |
1 |
Family income exceeds $100 |
10194 |
0.078 |
0.265 |
0 |
1 |
Parent did not complete high school |
11090 |
0.107 |
0.309 |
0 |
1 |
Parent completed high school but no PSE |
11090 |
0.601 |
0.49 |
0 |
1 |
Parent is a college graduate |
11090 |
0.292 |
0.455 |
0 |
1 |
Student attends school frequently |
6188 |
0.964 |
0.187 |
0 |
1 |
Most variables used in my research were coded as binary, "dummy" variables. The two possible values for each variable were zero and one, whereas one represents the affirmative response and zero the negative response. For example, take the variable "Parent is a college graduate": one is equivalent to yes and zero is equivalent to no. The indexes measuring social capital were each formed using a principal component analysis.
Each variable comprising the parental social capital index examined decisions made by parents that could plausibly be said to affect the family and the student. For this reason, church attendance, family participation in activities, parental involvement at the school, parental involvement in the community, and parental knowledge of financial aid were selected. The other variables that comprised the parental social capital index included direct instances of parent-student interaction. For this reason, parental involvement in the student's academic life, parent-student discussions, parent-student homework collaboration, college visitation, and family rules were chosen. In selecting these variables, it is possible to determine how social capital among the parents can directly and indirectly affect the educational achievement of the student.
The variables that comprise the community social capital index included questions that were asked of the teachers and administrators of the student's school. The few exceptions to this were the variables about community safety and teacher interest in the performance of the students; those questions were asked of the parent. The variables that measured parent-teacher conference attendance, parent volunteerism at the school, parent participation in the PTA, and parent participation in other organizations demonstrate citizen involvement in the community. The questions related to whether or not the parents felt the school and town were safe provided an idea about the relative safety of the community. The variables that indicated teacher interest, college assistance for students available in the schools, the existence of a truancy program, and whether or not the school provided educational value demonstrated resources available to the school and community. The variables that ask about alcohol, drugs, and teen pregnancy accounted for the problems afflicting the community.
The variables that compose the friend social capital index accounted for the behavior of friends and the student's feelings toward friends. Drug use among friends, importance of school among friends, and whether or not the student had many friends who dropped out indicate behavior. Whether or not the student found friends to be important and whether or not the student spent a lot of time with friends indicated the student's feelings about friends. This index received less emphasis partly because the high means suggested that there was a tendency to answer in the affirmative. The nature of these questions invited biased responses. For instance, the mean answer for "drugs are not a problem among my friends" was .934 on a scale of 0 to 1.
The NELS offered individual level time series, cross sectional data on over 10,000 students from 1988 to 2000. Each "follow-up" survey in the study offered a cross section of students at different points in their lives. In tandem, the follow-up surveys allowed for time series or cross sectional analysis. Data measuring social capital, and all control variables, were pulled from the second follow-up of the NELS. The second follow-up was in 1992 and examined students in their senior year. This follow-up was chosen to provide the independent variables, in part, because it also offered data from a survey asked of teachers and school administrators of each student. To measure educational achievement, the third follow up was used. It took place in 1994, two years after the anticipated graduation of students.
Logistic and OLS regressions were computed to demonstrate the effect of the social capital indexes and the control variables on educational achievement. The logistic regression is ultimately more appropriate because of the binary nature of all of the variables' outcomes. Below is the logistic formula used in the research.
Yi = SCPIiSCCIi +SCFIiXi4 +i
The subscript i represents an individual student. Y represents educational achievement. Following that are the various indexes, control variables, and error terms. In addition to treating family income as a control variable, I ran regressions to assess whether or not social capital had a differing effect on educational achievement for students below and above the median income level. The beta coefficients of the OLS regressions helped to determine the relative strength of each variable on educational achievement.
Below are two regressions of the main social capital indexes on the two dependent educational achievement variables. The logistic regression, being more appropriate for binary dependent variables, provides the most insight, while the OLS regression is used throughout the paper to confirm the findings of the logistic regression. This regression below demonstrates that, without control variables, the effects of parental social capital and community social capital are significant at the 5 percent and 1 percent levels.
Logit Regression | OLS Regression | |
(1) | (2) | |
Dependent Variable: Pursuing Post-Secondary Education | ||
Parental Social Capital Index | 0.0840*** (0.0300) |
0.0165*** (0.00816) |
Community Social Capital Index | 0.183*** (0.0445) |
0.0325*** (0.00816) |
_cons | 1.249*** (0.0586) |
0.773*** (0.0103) |
N | 1739 | |
Standard errors in parentheses ="* p<0.1 ** p<0.05 *** p<0.01 | ||
Logit Regression | OLS Regression | |
(1) | (2) | |
Dependent Variable: Diploma Status Two Years after Class Graduation | ||
Parental Social Capital Index | 0.1071** (0.0464) |
0.0014 (0.0087) |
Community Social Capital Index | 0.622*** (0.1472) |
0.006* (0.0016) |
_cons | 5.222*** (0.3373) |
0.9916*** (0.0024) |
N | 1742 | |
Standard errors in parentheses ="* p<0.1 ** p<0.05 *** p<0.01 |
The following table now introduces other control variables. This dependent variable asked whether or not the student was currently enrolled in a post-secondary institution and pursuing a degree. The logistic and OLS regressions confirmed that parental social capital, community social capital, income, parent education level, and student high school attendance have a statistically significant effect on post-secondary educational achievement.
Logit Regression OLS Regression (1) (2) Dependent Variable: Pursuing Post-Secondary Education |
Each variable was estimated relative to an omitted category. For example, income above $100K was omitted; the three other income categories had negative coefficients relative to families with income above $100K in predicting educational achievement. This was also the case for the education level of a student's parents. Parents with a college education were omitted. Parents that did not complete high school had a negative coefficient relative to parents with a college education in predicting the educational achievement of the child.
The variables that did not have a statistically significant effect on educational achievement are as interesting to discuss as those that did. Other works of literature frequently cite the disadvantages of race on educational achievement due to discrimination. These regressions seemed to suggest that, if there are disparities in educational achievement among races, they must manifest themselves in socio-economic differences that are captured by the income variables. That is to say, all other circumstances notwithstanding, outcomes in educational achievement were not affected by race.
This is not to say that racial discrimination has no negative effects for minorities in the United States. Racial discrimination can still contribute to higher levels of unemployment among Black and Hispanic citizens. Racial discrimination can also contribute to socioeconomic disparities among different races, disparities in school quality, and disparities in other determinants of educational prosperity.
Whether or not students' parents work also proved to not have a significant effect on educational achievement after high school. This seems somewhat surprising given that other literature suggests that children are impressionable by the work habits of their parents—that is to say, if a student's parents doesn't work, he or she is less likely to work and vice versa. Subsequently, one would postulate that if a child has parents that are not working, he or she would be less likely to achieve in other realms such as in education. This seemed not to be the case for post-secondary education.
Given the nature of the questions that comprise the friend social capital index, it is not terribly surprising that this index did not prove to have a significant effect. The student responses were highly skewed towards the affirmative; students may have been reluctant to be honest about the behavior of their friends. It could also be the case that friends just don't influence educational achievement after high school or that the influence of students' parents trumps that of their friends.
It is plausible that the effect of parental social capital on the educational achievement of a child may change from one level of family income to another. To better understand the true effect of social capital, it was therefore appropriate to test whether or not social capital had a stronger effect on educational achievement for lower income people relative to higher income people. To do this, the family income variable was recoded to reflect people below the median income level in one category and people above the median income level in another. People below the median income level were coded as 1 and those above the median income level were coded as zero.
The mean of the responses for this binary variable was .674, suggesting a reasonably balanced set of observations with slightly more observations below the median income level. The following table demonstrates that social capital exhibited in a student's parents has a stronger effect on students below the median income level than students above the median income level.
Logit Regression (1) |
|
Dependent Variable: Pursuing Post-Secondary Education | |
Parental Social Capital Index | -0.142 (0.0996) |
Community Social Capital Index | 0.113** (0.0493) |
Below median income* Parent SCI | 0.219*** (0.105) |
Control Variables | |
Below Median Income | -0.388** (0.149) |
Hispanic | -0.399 (0.399) |
Black, not Hispanic | -0.548 (0.388) |
White, not Hispanic | -0.474 (0.334) |
Native America/Alaskan | -0.512 (0.744) |
Parent's work status | -0.236 (0.164) |
Parent did not complete high school | -1.018*** (0.326) |
Parent completed high school | -0.301** (0.143) |
Student attends school frequently | 0.860** (0.336) |
_cons | 1.538*** (0.487) |
N | 1641 |
Standard errors in parentheses ="* p<0.1 ** p<0.05 *** p<0.01 |
OLS Regression With Beta Coefficients (1) |
|
Dependent Variable: Pursuing Post-Secondary Education | |
Parental Social Capital Index | 0.0667** (0.0074) |
Community Social Capital Index | 0.0641** (0.0101) |
Friend Social Capital Index | 0.0352 (0.105) |
Control Variables | |
Hispanic | -0.435 (0.0646) |
Black, not Hispanic | -0.016 (0.0602) |
White, not Hispanic | -0.0594 (0.0431) |
Native America/Alaskan | 0.0056 (0.1219) |
Family Income Level | 0.0655** (0.0185) |
Parent's work status | -0.0282 (0.0265) |
Parent Education Level | 0.1214*** (0.001) |
Student attends school frequently | 0.060* (0.086) |
N | 1451 |
Standard errors in parentheses ="* p<0.1 ** p<0.05 *** p<0.01 |
The beta coefficients tell the following story. Parent education level had the largest effect on educational achievement compared to all other variables. For every standard deviation increase in parent education level, there was a .1214 standard deviation increase in a student's educational achievement. Parent social capital proved to be the second largest predictor of educational achievement—each standard deviation increase in a parent's social capital yielded a .0667 standard deviation increase in the educational achievement of the child. The third, fourth, and fifth strongest determinants of educational achievement were family income level, community social capital, and student school attendance respectively.
The effect of parent education level is about double that of all other determinants. Whereas the effects of parent social capital, community social capital, income, and attendance are relatively close. There are several plausible explanations for those results. Parents with higher education levels likely placed a high value on post-secondary education and understood the process. This seemed consistent with the data recorded in the logistic regression. It stated that relative to parents who graduated college, parents who did not graduate from college have children that are significantly less likely to pursue post-secondary education. Parents without a high school degree, who may or may not understand the value of a post-secondary education, have no experience with the rigorous visitation, application, and financial aid processes.
Parents with high levels of social capital were likely to encourage a number of positive characteristics in their children but had, perhaps, a less direct impact on determining educational achievement. Parents with high levels of social capital were likely to pass on an affinity for volunteerism, a high regard for the family, and the importance of meeting obligations. A high level of social capital encouraged educational achievement indirectly, but did not necessarily demand it.
The explanation for the effect of family income on educational achievement was parallel to that of social capital. Families with higher income levels did have greater access to resources that encourage education: tutors, college savings, and extracurricular programs are among them. Families with higher income levels also tended to live around other families with high income. But access to resources was neither necessary nor sufficient in explaining educational outcomes. Ultimately, having a high level of income provided access to a good education but did not guarantee educational achievement.
To conclude, having parents with a high level of education had a larger effect on a student's educational achievement, perhaps because income and social capital were likely to predict other attributes and encouraged educational achievement indirectly.
Disparities in educational achievement are directly related to growing levels of income inequality in our country. The National Education Longitudinal Study made possible the analysis of behavioral factors and educational outcomes recorded of thousands of subject students and their families. Meaningful relationships between causal factors and educational achievement were subsequently discovered.
The effect of parent education level per standard deviation almost doubled that of any other variable. The reason for this may relate to the direct influence a parent's level of education may have on their child whereas social capital and income have a more indirect effect. However, in terms of creating constructive policy, the fact that a parent's education level has the largest effect on educational achievement means little. No policy can directly generate more educated parents. An increase in the number of educated parents is likely to increase the number of educated students but to have more educated parents, more educated students will have to become parents. Whereas parent education level proved to have the largest effect on determining educational outcomes, meaningful policies will likely need to be directed towards other predictors of achievement.
Logistic regressions were ultimately more appropriate for the analysis given the binary nature of each dependent outcome. OLS regressions confirmed the findings of the logistic regressions and helped in assessing the relative strength of each variable that proved to be a significant determinant of educational achievement. Ultimately, in order of relative strength, parent education level, parent social capital level, family income, community social capital level, and student attendance were determined to have a significant effect on educational achievement at the post-secondary level.
The regressions also confirmed an important non-determinant of educational achievement. Race was not found to be a significant predictor of educational achievement. Other works of literature cite race and racial discrimination in explaining disparities in achievement. If this is the case, racial disparities in educational achievement must manifest themselves in variables besides race itself, such as socioeconomic status. The main take away from this is that lower income people are less likely to achieve post-secondary education, irrespective of race, and higher income people are more likely to achieve post-secondary education, irrespective of race.
The determinants of high school graduation were less concrete and completely different from the determinants of post-secondary education. This outcome could be attributable to a number of factors. The variable for high school diploma had a lop-sided outcome in the affirmative: over 85 percent of subjects obtained a high school degree and another 4.5 percent obtained a GED by 1994, two years after their anticipated graduation. This lopsided outcome made finding common causes of achievement or lack of achievement more difficult. The results and an in-depth analysis of the determinants of high school graduation are located in the Appendix.
Friend social capital and parent work status were found to be significant at the 10 percent level in determining high school graduation. The friend social capital index is comprised of variables that measure the utility placed by a student's friends on education and whether or not those friends completed their education. It is therefore plausible that the behavior of a student's parents and friends can influence his or her high school achievement. Having parents who work and friends that place importance on education influence the student's understanding of obligation. Nonworking parents and underachieving friends negatively affect a student's understanding of obligation and subsequently whether or not they go on to graduate from high school
This research helps to contribute to the debate about educational achievement and its implications on inequality in this country. It also confirms that there is no one clear cause in predicting educational achievement. Educational achievement is instead attributable to a number of causes: income related but also behaviorally related. Parent and community behaviors, in the form of social capital, contribute significantly to the educational achievement of the students influenced.
Any politically constructive solution must therefore incorporate solutions that address behavioral and socioeconomic causes. Community capital for instance, is derived from community and school resources available to the student. Large disparities in educational resources exist between schools. This disparity is perpetuated by a system of school funding that relies on property taxes to finance a school district's operations. Lower income communities produce less revenue, and subsequently less resources are available to the school. To create more egalitarian levels of educational achievement, government investment is needed to improve the resources available to schools, especially those in lower income communities.
How to encourage parental social capital seems less clear. Charles Murray measured growing cultural divergences between socioeconomic classes, especially with respect to marriage. Constructive policy must foster strong families but much of that responsibility rests in the hands of the citizenry. Government can, however, incentivize marriage financially, hold absent parents accountable, increase access to knowledge about financial aid, and otherwise advertise positive behavior.
Income redistribution will not solve income inequality; only opportunity redistribution can do so. In addition to a more progressive tax system to make critical investments, the United States needs to equip students with the skill sets demanded of the 21st century, both in skills training and post-secondary education.
The data found in the National Education Longitudinal study is between 13 and 25 years old at this point. Since then, there have been even more pronounced cultural changes, socioeconomic changes, and technological changes that have altered the landscape of educational achievement. One possible avenue of research that I did not account for is the assessment of the effect of access to technology on achievement.
Additionally, the leisure activities that children focus their attention upon are changing. It would be interesting to research what impact the prevalence of video games and social media have on educational achievement because both now occupy a significant amount time among young people. Alternative solutions to educational achievement have also become more pronounced. Public charter schools now compete with standard public schools, and ambitious undergraduates are being sent to the lowest performing schools through Teach for America. Universities that exist solely online have emerged as an alternative to a standard college education.
In tandem, these other factors, which have been unaccounted for in the NELS, are changing the landscape of what determines educational achievement. Each represents a possible point of departure for research in this critically important problem of educational disparity in our society.
1. Larry M. Bartels. Unequal Democracy: The Political Economy of the New Gilded Age. (Princeton, NY: Princeton University Press, 2008) pp. 15
2. Charles Murray. Coming Apart: The State of White America 1960-2010. (New York, NY: Crown Publishing, 2012) pp. 286
3. Lawrence M. Mead, ed. The New Paternalism: Supervisory Approaches to Poverty.(Washington, DC: Brookings, 1997), pp. 33
4. Lawrence M Mead, The New Politics of Poverty: The Nonworking Poor in America. (New York, NY: Basic Books, 1992), pp. 54
5. Murray, Figure 8.3
6. Larry M. Bartels. Unequal Democracy: The Political Economy of the New Gilded Age. (Princeton, NJ: Princeton University Press, 2008), Figure 1.2
7. Julia B. Isaacs, Isabel V. Sawhill, and Ron Haskins. Getting Ahead or Losing Ground: Economic Mobility in America (Washington, DC: Brookings, 2008), Figure 4
8. Claude S. Fischer and Michael Hout. Century of Difference: How America Changed in the Last 100 Years. (New York: Russell Sage, 2006)
9. Ron Haskings. Education & Economic Mobility. (Washington, DC: Brookings, 2006) Figure 7
The dependent variable diploma status gave puzzling results. Here, parent's social capital index and the community social capital index were insignificant. The significance of each control variable was also dropped, including income and parent education level. Interpreting the results of the effect of independent variables on high school diploma achievement was difficult, given that the vast majority of subjects in the data set received their high school diploma. For example, the mean response for diploma status was 0.865 on a 0 to 1 scale. The numbers of students that did receive a high school diploma were so large that it was difficult to find common causes for failure to earn a diploma.
Roughly half the students interviewed in the study went on to pursue a college degree. It may be less difficult to assess common causes of pursuing a college degree because the balance of those observed to pursue a college degree and those observed not to allowed us to assess common predictors that influence each group's educational outcome.
The following tables demonstrate that, given all controls, only parent work status and the friends' social capital were significant variables that explained high school graduation, at least at the 10 percent level. This perplexing finding may be attributable to a number of factors.
Over 85 percent of the subjects interviewed in the NELS study received a high school diploma. Another 4.5 percent of subjects received a GED by the time they were two years out of high school. Perhaps there is no one explanation as to why over 85 percent of the subjects graduated high school, especially considering receiving a high school degree or equivalent is so common.
On the other hand, perhaps getting a high school diploma is not very difficult in most cases. Maybe social capital and family income help a little bit here and there. High schools are designed to help students graduate, so most students go on to do just that. The friends' social capital does have a significant effect on high school graduation. It is reasonable to expect that if your friends care about grades, reject drug use, and go on to graduate, it is likely to influence you to do the same. Conversely, if your friends are doing drugs, not trying in school, and dropping out, you are susceptible to mimicking those same behaviors.
The same mimicking behavior can occur with parents that work or don't work. If a student sees their parent or parents live up to the daily obligation of work, it is plausible that they will construct their own lives around the daily obligation of school. This is also plausible for non-work. If a parent does not meet their daily work obligations and proceeds to get by, why should the student?
|