While many policies appear effective in a small-scale setting, often the large-scale implementation of such policies gives rise to indirect effects that work in offsetting or reinforcing ways.  For instance, consider a policy that is often at the forefront of discussion in terms of improving public education: class size reduction.  Here, a landmark randomized experiment found that randomly assigning students to smaller classes in Tennessee in the 1980s significantly improved student learning.  When class size policies were implemented at-scale in California and Florida, however, researchers struggled to find these large positive impacts.  

One such offsetting indirect effect is that reducing class size also requires the hiring of a substantial number of new teachers.  To investigate this, I use data from New York City and isolate the effect of class size reductions by looking at classes that moved just above or below the 32-student cap required for elementary classes. Classes that moved just above 32 students would have to reduce class size by hiring a new teacher, while classes that moved from 33 to 32 students would change their class sizes without adding a new teacher.  I find large achievement gains for students in small classes among schools whose class size changed without requiring a new teacher, while finding no achievement increase among students whose class size was reduced by adding a teacher. This indicates that reducing class sizes can significantly increase student learning, but those gains are often canceled out in the short run by the lower-quality teachers who wind up staffing them and by disruptions linked to their quick hiring.

Conversely, these indirect effects may reinforce policy goals. My research with Robert McMillan and Hugh Macartney investigates California’s massive class size reduction program of the late 1990s. We look at whether the class size reform made public schools more attractive to families sending their kids to private schools: We find about a 10 percent decline in private school enrollment in grades that saw their class size drop with corresponding increases in enrollment in public schools with a nearby private school.  Drawing these high-performing students into the public sector appeared to boost scores of the incumbent public school students.  Thus the class size reduction policy appeared to reinforce policy goals of decreasing test score gaps by drawing high-performing private school students into the public sector.

Finally, education is an inherently cumulative process and so reforms are likely to have persistent effects as student skills are developed, with benefits for learning later on.  Thus, the benefits of a given reform may reinforce over time as student learning in the future depends on the skills they have developed in the past.  Economists, led by James Heckman and coauthors, call such interactions over time ‘dynamic complementarities’ and have important implications for when to invest in students and thus the design of education policy.

I reveal that these interactions over time exist by showing that students who are given extra attention by educators for two consecutive years perform better than the sum of the effect of getting extra attention in one year or the other.  Given this positive interaction, education policies should focus on investing early and continuously in students: while single year policy interventions can produce substantial positive impacts initially, they are unlikely to meaningfully increase student achievement and reduce test score gaps unless they are followed up with recurring investment. 

---

List of Papers

A Method to Disentangle Multiple Treatments from a Regression Discontinuity Design – Working paper, 2018.

Education Reform in General Equilibrium: Evidence from California’s Class Size Reduction  – NBER Working paper No. 24191, 2018, with Hugh Macartney and Robert McMillan.

School Accountability and the Dynamics of Human Capital Formation – Working paper, 2018.