Chapter 2 Social Policy Analysis I

Problem: Will Subsidies for Electric Cars Help Rural Towns?

Background information:

  • Wellington County is made up of seven lower-tier municipalities, predominantly rural in nature (wiki).
  • The County has been experienced a decrease in heavy industry jobs for the last few decades.
  • In order to attract more businesses, one of the municipality council decided to stimulate the purchase of electric cars for the residents in 2013.

The city of Fergus/Elora asks you to determine whether the city subsidy had a causal effect on the car purchase behavior of the town’s residents.

2.1 Data source

The city of Fergus/Elora collected a lot of relevant information about their experiment.

  • First, they shared a series of annual surveys they collect about town-member’s demographic’s and car-purchases.
  • Fergus/Elora suggests that you compare the rate of electric car purchases in 2015 for citizens who were below the necessary income threshold for attaining a subsidy to citizens who were above the necessary income threshold for attaining a subsidy (i.e. those who did not qualify).
  • They expect that individuals right below the income threshold would be more likely to purchase electric cars, controlling for other demographic characteristics.

Variables:

  1. An indicator whether a resident is below an income threshold or not.
  2. An indicator whether a resident purchased an electric car or not.
  3. Demographic variables: College education, gender, and miles driven in that year.

2.2 Methodological workflow

Aim: Estimate the association between qualifying a car purchase and purchasing an electric car controlling for demographic variables.

Data: Cross-sectional data, i.e., data is available for each resident for only 2015.

Method: Lgistic regression or OLS regression.

Aim: Estimating fixed-effects model. The reason is that there might be other variables (confounders) that effect the independent variables as well as dependent variable.

Moreover, to more directly test whether falling below or above the subsidy threshold over time motivates households to purchase electric cars.

Data: Panel data with long format, i.e., each resident has two records: 2013 and 2015.

Method: Fixed effects estimation method to see whether people who qualified for subsidies in 2015 but not 2013 (or vice-versa) were more likely to purchase electric cars, controlling for miles driven.

Aim:

  • Fixed-effects model may not perform properly with less data points.

  • Moreover, the causal relationship between miles driven and buying an electric car is not clear in this model; driving more miles may incentivize electric car ownership, but buying an electric car may incentivize driving more miles. This unclear causal relationship may have confounded the relationship between qualifying for subsidies and purchasing an electric car.

Data: Wide format. Each resident has a column of threshold indicator in 2013 and purchase an electric car indicator in 2015.

Method:

  • OLS regression with lagged dependent variable.
  • To test the effect of qualifying for a car purchase in 2013 on purchasing an electric car in 2015 controlling for other 2013 variables, including college education in 2013, gender in 2013, and miles driven that year 2013.

2.3 Comment on findings

The city council then asks: “The subsidies seem to motivate people near the subsidy threshold to purchase more electric cars, but does this really matter in the grand scheme of things? Did the policy really improve rates of electric car ownership in Fergus/Elora?”

Solution 1:

Generate a graph that displays the share of electric cars within all cars in Fergus/Elora and neighboring municipality on a year-by-year basis, including when Fergus/Elora created a subsidy for electric cars in 2013.

If the rates of electric car ownership have grown a great deal over the past few years, does that mean that the subsidy must have been effective.

Solution 2:

Let’s think of a way to use the Difference-in-Differences approach to find out if this is a causal effect. In the Difference-in-Differences approach, we should try to compare rates of electric car ownership among otherwise very similar cities. The mayors of Fergus/Elora suggests we compare Fergus/Elora to the nearby city, which did not subsidize electric car purchases over the same years. If the cities are very similar across key demographics, then we can feel confident that we can compare them with a difference-in-differences approach.

First, compare statistics of city demographics in both years.

Then, let’s assume that the two cities are similar enough to warrant comparison. By illustrating a comparison of the share of electric cars owned in Fergus/Elora and nearby city and assuming the differences are statistically significant, try to answer with relative confidence whether the subsidy worked in Fergus/Elora.