5.3 Application: Economic Effects of Basque Terrorism
Research Question: What is the economic impact of terrorism?
- Factual (\(Y(1)\)): Economy given Basque region hit with terrorism in early 1970s
- From 1973 to late 1990s, ETA killed almost 800 people
- Activity localized to Basque area
- Counterfactual (\(Y(0)\)): How would Basque economy have fared in the absence of the terrorism?
- Basque was the 3rd richest region in Spain at onset
- Dropped to the 6th position by late 1990s
- Would this fall have happened in the absence of terrorism?
Problem: We can’t observe the counterfactual. We can’t go back in time to manipulate the experience of terrorism.
5.3.1 Applying 3 Identification Strategies
- Compare Basque to others after 1973 (Cross-section comparison)
- Compare Basque before and after 1973 (Before-and-after)
- Compare others before and after 1973 and subtract the difference from Basque’s difference (Difference-in-differences)
For a video explainer of the code for this application, see below. (Via youtube, you can speed up the playback to 1.5 or 2x speed.)
<- read.csv("basque.csv") basque
head(basque)
## region year gdpcap
## 1 Andalucia 1955 1.688732
## 2 Andalucia 1956 1.758498
## 3 Andalucia 1957 1.827621
## 4 Andalucia 1958 1.852756
## 5 Andalucia 1959 1.878035
## 6 Andalucia 1960 2.010140
Variables
region
: 17 regions including Basqueyear
: 1955 – 1997gdpcap
: real GDP per capita (in 1986 USD, thousands)
Subset Basque Data into Four Groups
## Basque before terrorism
<- subset(basque, (year < 1973) &
basqueBefore == "Basque Country"))
(region ## Basque after terrorism
<- subset(basque, (year >= 1973) &
basqueAfter == "Basque Country"))
(region ## others before terrorism
<- subset(basque, (year < 1973) &
othersBefore != "Basque Country"))
(region ## others after terrorism
<- subset(basque, (year >= 1973) &
othersAfter != "Basque Country")) (region
What is the economic impact of terrorism?
Cross-section comparison
mean(basqueAfter$gdpcap) - mean(othersAfter$gdpcap)
## [1] 1.132917
Before-and-after design
mean(basqueAfter$gdpcap) - mean(basqueBefore$gdpcap)
## [1] 2.678146
Difference-in-Differences design
<- mean(basqueAfter$gdpcap) -
treatDiff mean(basqueBefore$gdpcap)
<- mean(othersAfter$gdpcap) -
controlDiff mean(othersBefore$gdpcap)
- controlDiff treatDiff
## [1] -0.48316
Here is a way to visualize this difference-in-differences. Our estimated causal effect is the difference between the observed growth in the Basque region and what we assume the growth would have been in the absence of terrorism (the treatment).
What should we conclude from each approach?
- Each approach resulted in a different estimate of the impact of terrorism on the economy. We should choose the approach for which we think the underlying assumptions are most plausible.