The Partial Monotonicity Parameter: A Generalization of Regression Monotonicity

• Wed 7, 2020 @ 14:00. Webinar!!! Moved to 14:00!!
  

  • 
    
  • Speaker: Scott Kostyshak (U Florida)
  • Invited by: M. Sartarelli  (ICAE-UCM)

The Partial Monotonicity Parameter: A Generalization of Regression Monotonicity 

By Scott Kostyshak 

Abstract: Monotonicity is perhaps the most common predicted relationship between variables in economics; and yet, it is almost never tested in practice because tests of monotonicity are structured to reject if even a small violation is detected. Our paper fills this gap between theory and practice by proposing a continuous measure of monotonicity, with a corresponding test statistic that is directly interpretable.   

We define the Partial Monotonicity Parameter (PMP) as the proportion of the population for which a small increase in the independent variable is associated with an increase in the dependent variable. The framework applies to parametric and non-parametric estimation, and is the first to provide inference on the policy-relevant question of whether a small increase in a variable would have a positive average effect and would benefit the majority of the population. We generalize the notion of partial monotonicity to partial derivatives of multivariate functions.

→  See our calendar and last recorded sessions