Current Research

Last updated at 2025-03-31

Working Papers

• A Quantitative Theory of Teacher Quality: Evidence from a Dynamic Structural Model Estimated using an Incentive Pay Experiment. Funded by SSHRC IDG Grant 430-2020-01098.
  Abstract

  Teacher quality is an important determinant of the distribution of student achievement. Researchers have documented considerable cross-sectional variation in the quality of even novice teachers, but this variation is not well-explained by teachers’ observable characteristics. Although the determinants of teacher quality are not yet well understood, researchers have documented the effects of output-based incentives on teacher quality and substantial growth in teacher quality over teachers’ careers. I combine classic human capital and asymmetric information frameworks to develop an estimable dynamic model of teacher quality. Teacher human capital may be generated by on-the-job investments and/or by learning-by-doing. The model allows for variation in teachers’ initial human capital (hidden types) as well as for teachers to make unobserved effort inputs (hidden actions). I then estimate the model using experimental variation from Muralidharan and Sundararaman (2011), a teacher incentive pay experiment in Andhra Pradesh. Preliminary results indicate that hidden human capital types explain the lion’s share of quality variation, and that on-the-job investments are responsible for the majority of teacher quality growth. Under the incentive scheme in Muralidharan and Sundararaman (2011), the increase in teacher quality that took place during the intervention is attenuated by the reduction in future quality.

• Multidimensional Health Capital and the Production of Health (with Tian Liu and Seth Richards-Shubik). Funded by the UWO Social Science Research Fund.

Needing Resurrection

These are zombie papers, but instead of regular zombies (which need only brains), these need time from me and/or someone else (you?!)…

• Supply, Demand, and Social Learning: An Application to Learning About Charter School Quality

• The Impact of Primary Health Care Reform on Population Health: Theory and Evidence from Ontario

Number of Papers at Each Stage

• This chart summarizes the number of papers at different stages (e.g., “under review” or “accepted”). (Note: the number of working papers (“dev. wp”) isn’t very accurate for the earlier periods because I hadn’t been tracking that. Also, there are some invisible zombies.)

Research Tools

• Programming, listed in order of decreasing start-up time (for me):

  • R: Free, reasonably easy to use, and reasonably fast to get started with. Can be slow if you’re not super clever.

  • Julia: Free, rapidly changing (in my view), and can be pretty speedy if you’re clever enough to use it correctly.

  • FORTRAN: Free and non-free, and for when you don’t feel like feeling clever, but also feel like it’s crunch time!

  I’ve compiled a table summarizing which languages I’ve used in some projects:

  Project Name	Fortran 95	R	Julia
  Competition in Public School Districts	[x]	[x]
  Ability Tracking, School and Parental Effort, and Student Achievement	[x]	[x]
  Generalizing Findings from Regression-Discontinuity Designs		[x]
  Measuring Quality for Use in Incentive Schemes		[x]
  Optimal Contracting with Altruistic Agents		[x]
  Social Interactions, Mechanisms, and Equilibrium		[x]
  Teacher Quality: Partial Identification of Human Capital Technology		[x]
  Teacher Quality: Dynamic Structural Model		[x]	[x]
  Multidimensional Health Production		[x]

• etc.

  • DNS: To remotely access my office machine (to check programs I’m running), I use Free DNS, which makes it very easy to ssh into my office machine via a memorable name (which then points to the machine’s IP address).

  • It can be hard to keep track of all the tasks involved with revising a paper (or even undertaking a big project). I sometimes use taskwarrior to track my tasks and task2dot, which takes task output from taskwarrior and generates a graph using graphviz, to help me figure out how to prioritize my next steps. For example, if many tasks depend on task A it might make the most sense to complete A first; less obvious dependency structures might be easier to understand with the visualization provided by task2dot.