Rethinking High School Graduation Rates and Trends
Susan Notes: The authors take on a formidable task--trying to cut through the political bias of calculating high school graduation rates. And the reviewer gives us a picture of whether they succeed or not.
Mishel, Lawrence and Roy, Joydeep. (2006). Rethinking High School Graduation Rates and Trends. Washington DC: Economic Policy Institute.
Reviewed by James Hall
Arizona State University
The calculation of high school graduation rates doesn't seem, at first glance, to be a topic for an entire book. But the rate of high school completion is not merely an education statistic – it is a central issue in the ideological battle over the efficacy of public education, brought to the forefront by requirements in No Child Left Behind and the proponents of school choice. Lawrence Mishel and Joydeep Roy, economists for the Economic Policy Institute, provide a reasoned analysis of the two major methods of estimating high school graduation rates without proselytizing, a daunting task in our current political climate.
The 2005 National Governor's Association Task Force reported that "about three-fourth of white students graduate from high school, but only half of African American and Hispanic students do." (p. 2) These figures have become the "new conventional wisdom" and reflect calculations offered by Jay Greene and Marcus Winters (2005) of the Manhattan Institute that are in conflict with the substantially higher estimates of the Department of Education's Condition of Education, which relies on survey information from the Current Population Survey (CPS). Mishel and Roy analyze the basis of the new figures and present a case in support of the accuracy of the CPS estimates of high school completion in Rethinking Graduation Rates and Trends. They suggest that the most reliable estimates of high school completion can only be obtained through longitudinal data tracking students over time. Furthermore, existing longitudinal studies at the national, state, and city level do not support the fifty percent dropout rate for minorities suggested by Greene and Winters. The authors examine reported flaws in data collection and methodology of both the Greene and Winters model and in the use of CPS data. Mishel and Roy conclude that traditional survey estimates, such as the CPS, more closely correspond to the estimates provided by reliable longitudinal data and are better indicators of school completion rates.
Calculating graduation rates seems simple enough-divide the number of ninth grade students into the number of diplomas granted four years later. But since many urban high schools turnover 30-40% of the student body annually, it is difficult to know what happens to students who leave school. Most states do not track individual students and there is no system to follow students who move to different states. Immigration and retention in grade further complicate the issue, as these factors vary tremendously from district to district and state to state.
Traditionally, the US Department of Education has reported national drop out rates in the Condition of Education report using the Census Bureau's Current Population Study. The CPS is a representative household survey that is used to track poverty, income, and factors relating to income-levels including educational achievement. The Condition of Education report for 2004 indicated that there was an 89.7% high school completion rate for all people ages 16-24 years, an 88.2 % rate for African Americans, and a 76.2% rate for Hispanics. Critics, like Greene, have contended that these estimates are flawed and that high school completion rates are substantially lower. Greene and Winters (2005), along with Swanson (2004) and Warren (2005) have proposed procedures for estimating high school completion rates using administrative data sets available from the National Center for Education Statistics. This information, called the Common Core of Data, provides nation-wide enrollment figures and numbers of diplomas granted. Each of the researchers' models compares ninth grade enrollment with the number of diplomas granted four years later. Greene and Winters (2005) offer the more sophisticated calculation that attempts to adjust for both population growth over the four years and to compensate for inflated ninth grade enrollment numbers caused by high retention rates for high school freshmen. Greene and Winters suggest that high school completion rates are 69% overall with rates of 55% and 53% for African Americans and Hispanics, respectively. As previously mentioned, CPS estimates for high school completion were 90% overall, 88% for African Americans and 76% for Hispanics. Which estimate is more accurate? This question has huge policy implications, since the acceptance of Greene's calculations by many education reformers has bolstered demands for high school reform and increased high-stakes testing. The stage has been set for Mishel and Roy's analysis.
Since the ultimate, but quixotic, solution to this debate lies in the construction of a nation-wide, longitudinal database to track all students in the US, Mishel and Roy look to existing longitudinal data that scholars could agree provide a reasonably accurate measure of high school completion rates. They select two: the National Longitudinal Surveys (NYSL) conducted by the Bureau of Labor Statistics and the National Educational Longitudinal Study of 1988 (NELS:88). The NELS:88 is singled out as the gold standard of school completion data because it tracks a large, representative sample of students who were in the eighth grade in 1988, surveys them every two years, and checks school transcripts to verify school completion. The NELS:88 found that 83% of all students, 74% of African American students, and 74% of Hispanic students achieved a standard diploma by age 26. There were also a significant number of students completing GEDs, especially African Americans (13.6%) and Hispanics (9.4%). Data from the NYSL found similar results. These numbers, while lower than traditional CPS estimates, are significantly higher than the 50% completion rates for minorities suggested by Greene and Winters (2005), leading Mishel and Roy to question the validity computing graduation rates using enrollment and diploma data.
Mishel and Roy then examine the methods used by the various models that use administrative data that might cause a mis-calculation of high school completion rates. They find the following:
* None of the models is able to effectively address the "ninth grade bulge." Large numbers of ninth grade students are retained, inflating the ninth grade enrollment numbers. The authors estimate that 14% overall and 26% or more of blacks and Hispanics are not promoted to tenth grade each year. Greene's model, for example, tries to compensate by averaging 8th, 9th, and 10th grade enrollments to "smooth" the bulge. Michel and Roy contend that the averaging of enrollments does not address the variation between the retention rates of whites and minorities.
2. Immigration of Hispanic students during high school years is not accounted for. Greene attempts to account for immigration by adding the average of population increases for high schools students into the total estimate of ninth grade students, but he assumes that all high-school aged immigrants enroll in school and are prepared to graduate on time. This is a huge supposition that probably belies many of the stronger motives of Hispanics immigrating to the US.
3. Administrative data are problematic and not reported consistently from state to state. For example, there are many categories of diploma, with some states not reporting vocational diplomas or certificates of completion.
The authors conclude their analysis by applying the Greene model to longitudinal administrative data from the state of Florida, New York City, and Chicago, finding the actual graduation rates in Florida and New York to be much higher than the estimates provided Greene. Once again, Mishel and Roy consider this to be evidence of the invalidity of the Greene model.
Critics of CPS, on the other hand, believe that it overstates high school completion and is biased in the following ways:
* The lowest income people, including the homeless, are more likely to be high school drop outs, but are historically under-represented in Census surveys.
* Institutionalized people are not included in the surveys, including prisoners, many of whom are dropouts.
* Data are self-reported and the possibility exists that the household member completing the survey could be in error or lying.
* The CPS does not distinguish between on-time graduation, graduation at a later date, and receiving a GED.
Mishel and Roy challenge these assertions by first noting that there has been no systematic assessment of these biases that establishes their quantitative magnitude (p. 29). They then address each individual concern:
* Improvements in sampling measures and procedures over the last ten years have made under-representation of minorities less of a problem. They admit that it is a concern, but it does not account for a significant difference in high school completion rates reported by CPS.
* Using micro-data from the 2000 decennial Census, the authors conclude that the non-reporting of prisoners is offset by the similar non-reporting of those in the military in CPS surveys. Many prisoners are dropouts but most military personnel have completed high school. The authors believe, however, that the high percentage of incarcerated young black men causes an overestimate of the black graduation rate reported by the CPS of 1.7%
* There is no evidence of self-reporting bias. Surveys, like the CPS are routinely used for economic research with little concern for significant self-reporting bias. The NELS:88 transcript verification found less than 5% reporting error, with errors in both under-reporting and over-reporting.
* The authors believe it is a mistake to discount completely the GED, especially considering that the cognitive ability needed to complete the GED is higher than the average ability of high school graduates. Attaining a GED leads to the availability of higher education and higher lifetime earnings compared to a high school dropout. They point out that GED rates can be determined and then deducted from survey results–13.6% for blacks and 9.4% for Hispanics–to estimate diploma students.
It is important to note that Mishel and Roy argue each point of contention by examining what they believe to be the best data available and offering specific remedies to overcome the possible weakness of CPS estimates. This is in direct contrast to researchers relying on administrative data who devise broad formulas to account for specific shortcomings in the data.
Mishel and Roy complete the analysis of administrative data models by examining historical trends derived from Census data going back to 1960. They report that graduation rates have steadily climbed until about 1995, after which they have remained relatively flat. They note, in particular, the rise in African American graduation rates from 41.6% in 1960 to 88% in 2004. These increases in African American school completion follow similar patterns of increased income and college completion rates for this group over the last forty years. Sherman Dorn (2003) found similar historical trends reporting that the gap between white and black graduation rates shrunk from 43% in 1940 to 13% in 1980. Mishel and Roy find these results in sharp contrast to the 50% failure rate for African Americans and the 25% gap between white and black graduation rates reported by Greene and others.
In many ways, this debate is complicated by two very different viewpoints about graduation rates. Jay Greene, for example, is a critic of public education and a supporter of school choice. His interest is in determining the effectiveness of public schools by their ability to graduate students on time. Mishel and Roy are economists who use high school completion as an indicator of lifetime earning potential. They see high school completion as a stage in the process of education, not as its culmination. The criticism could be made that Mishel and Roy spend little time addressing the issue of on-time graduation and how it could be better estimated at the state or district level. They make the case, rightly so, that district and state graduation rates cannot be accurately estimated from administrative data due to a lack of reliable information on the key issues of grade retention and immigration. Nevertheless, there is a need to calculate local and state graduation rates, and for that purpose, models like Greene's remain the only game in town.
To their credit, Mishel and Roy analyze many types of data to determine a reasonable standard by which a computation of graduation rates could be measured. Their argument that the NELS: 88 longitudinal study represents the gold standard of measurement of graduation rates is convincing and casts serious doubt on the accuracy of administrative data estimations.
Jaye Greene, however, does not see it this way. In a debate over this issue with Lawrence Mishel at the Center for Educational Policy, April 26, 2006, Greene dismissed the validity of estimates of high school graduation from measures like NELS: 88, stating that relying on survey data is much like believing the exit polls showing John Kerry winning the 2004 Presidential election over official polling results. Lawrence Mishel described this position as "hand waving" – disregarding evidence with the wave of the hand rather than presenting a reasoned argument. Michel and Roy can be commended for making a compelling case that high school graduation rates are substantially higher than the 50% success rate for minorities claimed by Greene. Unfortunately, the Greene numbers are already well established in the discourse of those who wish to promote the myth of public school failure. Time will tell if Mishel and Roy's attempt to insert rationality into the debate over high school graduation rates will have an impact on education policy.
Dorn, Sherman. (2003). High-stakes testing and the history of graduation. Education Policy Analysis Archives, 11(1). Retrieved November 10, 2006 from http://epaa.asu.edu/epaa/v11n1/.
Greene, Jay P. and Lawrence Mishel. (2006, April 26). Debate on High School Completion. Center on Educational Policy. Retrieved November 10, 2006 from http://archive.epinet.org/real_media/060427/
Greene, Jay P. and Marcus Winters. (2005). Public School Graduation and College Readiness Rates: 1991-2002. New York, N.Y.: Manhattan Institute for Policy Research. Retrieved November 10, 2006 from http://www.manhatthan-institute.org/html/ewp-08.htm
National Governor's Association. (2005). Graduation Counts: A Report of the National Governor's Association Task Force on State High School Graduation Data.
Swanson, Christopher. (2004). Who Graduates? Who Doesn't? A Statistical Portrait of Public High School Graduation, Class of 2001. Washington, D.C.: The Urban Institute. Retrieved November 10, 2006 from http.urban.org/UploadedPDF410934-WhoGraduates.pdf
Warren, John R. (2005). State-level high school completion rates: Concepts, measures, and trends. Education Policy Analysis Archives, 13(51). Retrieved November 1, 2006 from http://epaa.asu.edu/epaa/v13n51.
About the reviewer
James Hall is a PhD student in the Educational Policy and Leadership program at Arizona State University. He taught fifth and eighth grades for seven years and was a principal and district administrator for eleven years in Phoenix. He also spent six years as superintendent/principal on the White Mountain Apache Reservation in Arizona. His research interests range from the hidden curriculum of media reality to the effects of NCLB on teachers, the curriculum, and dropout rates.
Education Review: A Journal of Book Reviews
November 11, 2006
INDEX OF RESEARCH THAT COUNTS