The new conventional wisdom—that the dropout rate is rising—can’t be squared with the simple economic reality.
There’s a consensus that our high schools are in crisis. The conventional view is that only about two-thirds of students graduate, and that only about half of black youths do. These data, generated mostly by Manhattan Institute senior fellow Jay P. Greene and former Urban Institute research associate Christopher B. Swanson (now the director of the Editorial Projects in Education Research Center), have had enormous influence. Foundations have directed major resources to stem the crisis, while the Bush administration and the governors have shifted their attention to high school reform.
I’m new to education issues, with a career as a labor-market economist, concentrating not on test scores but on wages, employment, and household incomes. The more I’ve read about the dropout crisis, the more skeptical I’ve become, because the story is inconsistent with the economic data I know.
Consider this: Knowledge is becoming more important in the economy, and “returns to skill”—higher wages for workers with more education—should be growing. Yet the ratio of high school graduates’ wages to dropouts’ wages has not changed for 30 years. The most plausible explanation, accepted by most economists, is that the share of graduates in the workforce has grown, while the share of dropouts has fallen. If the demand for graduates relative to dropouts has gone up, the wage ratio can remain stagnant only if the supply of graduates relative to dropouts has also gone up.
The new conventional wisdom—that the dropout rate is rising—can’t be squared with this simple economic reality.
Economists rely on U.S. Census or U.S. Bureau of Labor Statistics surveys to estimate the educational attainment of the workforce. From these, we understand that from 80 percent to 90 percent of all Americans have a regular high school diploma, and from 70 percent to 80 percent of all African-Americans have one.
Greene, Swanson, and some other education researchers, in contrast, use enrollment and diploma data that the U.S. Department of Education collects from state governments and that states, in turn, collect from districts. The researchers use these administrative data to calculate their extraordinarily low graduation rates. To oversimplify a bit, these researchers divide the number of diplomas districts say they awarded in any year by the number of students who were 9th graders three years earlier, or by the average of the number of 8th, 9th, and 10th graders. The education researchers dismiss surveys that economists use because, the researchers say, people may lie to census-takers, claiming to have diplomas when they don’t.
Of course, some people may lie, but if this were a major issue, we’d have more serious problems than mistaken graduation rates, because our national economic and social policies rely on Census and Bureau of Labor Statistics data that the education researchers are quick to dismiss. Policies about unemployment, income inequality, marriage rates, poverty, health insurance, and so on all rely on these surveys. Try telling the Federal Reserve Board or the Congressional Budget Office that the data can’t be trusted.
It’s not as though government statisticians have never thought about the reliability of their surveys, whose strengths and limitations are amply documented. Substantial resources are spent to monitor accuracy. In contrast, education researchers have offered no evidence to support claims that the economic surveys are mistaken.
Nor have the education researchers undertaken serious studies to establish the reliability of the administrative data, which, in contrast to the Census, have little documentation. I have not been able to learn whether school districts are consistent in their reports to states, how data are checked, or even whether all diplomas are reported. I am skeptical that administrative data should be taken on faith while throwing overboard the entire statistical basis of federal social and economic policy.
This skepticism provoked me and my colleague, Joydeep Roy (also an economist), to venture into these education data ourselves. We found that the administrative data are contradicted even by other respected education sources. Consider, for example, the Education Department’s National Education Longitudinal Study, which surveyed a national random sample of students who were in 8th grade in 1988. Called NELS:88, it continued to interview these students in later years. To ensure truthful answers, the department checked them against actual transcripts in school district files, and tracked students down at home if they were no longer registered in school.
Education researchers have offered no evidence to support claims that the economic surveys are mistaken.
Longitudinal studies like these can be flawed by attrition, when students in the original sample can’t be found later, but the extraordinary efforts the department has made to control attrition and check transcripts has made NELS:88 the “gold standard” of education research.
NELS:88 shows that in 1992, the “on time” graduation year for this cohort, 78 percent of all students, and 63 percent of blacks, received a regular diploma, not including the General Educational Development, or GED, certificate. By 1994, 83 percent of all students, and 74 percent of blacks, had received one.
These last numbers provide an “apples to apples” comparison with administrative data, because diplomas reported by districts and counted by education researchers also include those for students who took more than four years to graduate.
The new conventional wisdom seems to have exaggerated the African-American dropout rate by a remarkable amount—doubling it from about 25 percent to 50 percent.
One problem with these education researchers’ approach is obvious. Their graduation-rate denominator, the number of 9th graders three years earlier, is artificially inflated by students who are held back in 9th grade, a growing phenomenon. In 2000, there were 26 percent more blacks in 9th grade than were in 8th grade in 1999. Swanson ignores this problem. Greene only partially addresses it by averaging 8th, 9th, and 10th graders.
There have also been several local longitudinal studies using the same school district records upon which the administrative data rely. Yet these studies fail to confirm the education researchers’ picture. A Florida study following individual students found graduation rates of 73 percent overall and 60 percent for blacks, from 10 to 15 points higher than we can get by applying the Greene or Swanson methodologies. A New York City longitudinal study found graduation rates from 10 to 27 points higher.
What objections might education researchers raise to our findings? They might say that NELS:88 was only one cohort, and schools have since deteriorated. But the importance of NELS:88 is that its data are so consistent with those obtained from the Census’ Current Population Survey, or CPS, and from microdata of the Census itself. Thus, NELS:88 gives us confidence in the ongoing reliability of the premier national economic and demographic surveys.
The importance of NELS:88 is that its data are so consistent with those obtained from the Census’ Current Population Survey, or CPS, and from microdata of the Census itself.
Researchers might object that the Current Population Survey, by leaving out those in prison, undercounts dropouts. But the CPS also leaves out military personnel, whose graduation rates are higher. For all students, these two omissions offset each other, although for African-Americans, graduation rates in the CPS are artificially high because more are in prison than in the military. However, by using Census microdata (which includes prisoners), we were able to adjust for this distortion. It lowers African-American graduation rates in the Current Population Survey by about 3 percent, leaving them still far higher than administrative data suggest.
Some object that surveys like the CPS combine GED recipients with high school graduates, inflating the true graduation rate. So we obtained annual GED-certificate data and subtracted these from the appropriate age groups of high school graduates in the CPS. That’s how we concluded that 73 percent of blacks get regular high school diplomas (compared with 86 percent, if GEDs are included).
My report with Joydeep Roy documenting these claims will soon be available. We’ll send one to anyone who writes to us at email@example.com.
Don’t get me wrong. I am not satisfied with a national black dropout rate of 25 percent, or a much higher urban rate. We must fix this glaring social problem.
But to solve it, we need to get our facts right, including the truth that high school graduation rates have been improving. In 1979, 77 percent of black young adults were high school completers. By 2000, it was 88 percent. For whites, the growth was from 89 percent to 95 percent. Even with the most extreme assumptions about increased numbers of GEDs and incarcerations, there would still remain a real growth in regular diplomas for blacks, and a narrowing of the black-white graduation gap. Something right has happened to propel this long-term improvement, and we should figure out what it is, so we can do more of it.