Teaching

Demand estimation

Nested logit

Here I describe an exercise developed to help students learn how to estimate nested logit demand for differentiated products. In this dataset, the variable g indicates nest ID, the variable t is the market ID, the variable j indicates product ID, the variable w is a cost shifter, the variable p is the product's price, the variable x is an exogenous product characteristic, and the variable q is the quantity sold. The market size is 100. The indirect utility specification is here. The recommended instrumental variables for the analysis described below are (i) the cost shifter, (ii) the sum of x among products in the same nest/market as (j, t), and (iii) the maximum of x among products in the same nest/market at (j, t).

1. Estimate a multinomial logit demand system using OLS and then using instruments. How does the estimated price coefficient differ between the estimators?

2. Estimate a nested logit demand system using OLS and then using instruments. Based on your estimate of the nesting parameter, comment on whether the multinomial logit model is appropriate in analyzing substitution patterns among the products in the data.

Upon completing the exercise, you can check whether your estimated parameters are near the true nested logit parameters by clicking here.

Random coefficients logit

I additionally provide an exercise in estimating a simple random coefficients logit demand system. In this dataset, the variable t is the market ID, the variable j is the product/market ID,  the variable p provides products' prices, the variable w is a cost shifter, the variables x1 and x2 are observable product characteristics, and the variable q is the quantity sold. The market size is 100. The indirect utility specification is here (note that only the second product characteristic x2 has a random coefficient). The goal of this exercise is to estimate the unknown parameters appearing in the indirect utility specification using the approach of BLP 1995. The recommended instrumental variables for the analysis described below are (i) the cost shifter w, (ii) the sum of x2 among rival products in the same market as (j, t), and (iii) the maximum of x2 among rival products in the same market at (j, t). There is no requirement to use the optimal weighting matrix. If you would like to check whether your parameter estimates are close to the true parameter values, please contact me.

Teaching at UWO

I have had the pleasure of teaching ECON2210, Mathematical Economics I (Honours), at the University of Western Ontario. Here is a short video describing the course.