Panel Paper: Is the Increase in High School Graduation Rates Real?

Friday, November 4, 2016 : 10:55 AM
Columbia 6 (Washington Hilton)

*Names in bold indicate Presenter

Nathan Barrett, Douglas N. Harris and Lihan Liu, Tulane University


Research shows that educational attainment is a strong predictor of long-term life outcomes and, for this reason, policymakers have been searching for ways to increase high school graduation. In particular, the federal government requires states to hold high schools accountable for high school graduation. While the reasons are unclear, graduation rates have increased steadily for decades, especially for disadvantaged students. Are these increases real or might they reflect changes in the way students are classified, perhaps driven by accountability?

We study this issue using detailed exit codes from the State of Louisiana, distinguishing between exit codes that are easy for the state to validate versus those that are hard to validate and therefore more subject to manipulation. If the graduation rates are valid, then we would expect no changes in the relative frequency of different types of exit. 

After Hurricane Katrina, the stakes attached to high school graduation in New Orleans were especially intense. Since the theory is that the validity of graduation data may be influenced by such pressures, we use a difference-in-differences framework to examine whether New Orleans experienced especially large increases in certain exit codes compared with other districts that are under less pressure. 

The stakes for high school graduation also increased sharply throughout the state of Louisiana in 2012. Did this alter the exit code patterns? Also, did schools that experienced unexpected shocks in their school grades see changes in their exit code patterns in ways that suggest strategic behavior? Again, we use difference-in-differences estimation to examine these questions. 

All the analyses are carried out for different definitions of graduation rates, including delayed graduation and different credential levels (e.g., GEDs). We also examine the rate at which students exit the data but have no verifiable exit code.