Unlock document.
This document is partially blurred.
Unlock all pages and 1 million more documents.
Get Access
42 Chapter Highlights
Chapter 17
Measurement
The purpose of this chapter is to “bridge the gap” between the undergraduate
theory and econometric classes. It can be covered in a lecture or two and can fit
in almost anywhere in the course once you have covered the basics.
I describe the 5 basic things you do with data: summarize, estimate, test,
forecast, and predict. Most students will already be familiar with tables and bar
charts. Conditional expectation is likely a novel term, but the concept is pretty
clear if you use concrete examples.
Simpson’s paradox is great fun. Wikipedia has other examples that can be
used.
I give a brief introduction to the logic of hypothesis testing, but it is necessarily
quite incomplete.
The real meat comes in demand estimation. Here I try to relate the theory of
demand estimation to what is done in practice, which is to run an experiment. If
we have to use observational data, then we would normally have to worry about
identification, which I discuss extensively. I raise the important issue of causal
inference, but that is mostly just to make the students aware of the concept,
since I can’t do a lot with it in this short chapter.
Measurement
A. What do you do with statistics?
B. Example: Coffee consumption
1. Summarize: how many cups of coffee consumed per person per day?
2. Estimate: what is the elasticity of demand for coffee?
3. Test: do men and women drink the same amount of coffee on average?
4. Forecast: What will the price of coffee be next year?
5. Predict: What will happen to consumption of coffee if a tax is imposed?
C. Summarizing data with tables and graphs
1. Summary statistics (mean, median, model)
2. Conditional mean: an average over those observations that satisfy some
other condition
D. Simpson’s paradox
1. Coffee example, admissions example
Chapter 17 43
E. Hypothesis testing
1. Do men and women drink the same amount of coffee on average?
2. Sampling variation
F. Estimating demand using experimental data
1. Importance of randomized treatment
G. Estimating demand using observational data
H. How does demand for coffee in the U.S. change as price changes?
1. Regression analysis: conditional expectation again
I. Specifying demand function
1. p1= price of coffee, p2= price of everything else, m= income.
2. Demand for single consumer = D(p1,p
2,m).
3. Can divide by p2to get demand for = D(p1/p2,m/p
2)
J. Specifying statistical model
1. Error term: cumulative effect of other factors
2. Critical assumption: price and expenditure are not correlated with error
3. What happens if only the price changes, holding everything else constant?
K. Identification problem
1. Is demand shifting or supply shifting (or both?)
2. What can go wrong?
a) When income is high, consumption is high, price sensitivity is low and
firms raise prices
L. Policy evaluation
1. Natural experiments
