colombia {sensemakr}

Data from the 2016 referendum for peace with the FARC in Colombia.

Description

Data on support for the 2016 referendum for peace with the FARC in Colombia, as discussed in Hazlett and Parente (2020). The main "treatment" variables are santos2014, which indicates the share of town population voting in support of Santos in the 2014 Presidential election, and fat_2011to2015_gtd, which indicates the number of fatalities due to FARC violence between 2011 and 2015, again at the town level. The main outcome of interest is yes_vote, the proportion (0-100) at the town-level voting in support of the peace referendum. The question of interest in Hazlett and Parente (2020) is what can be said about the causal effect of either violence (fatalities) or of political affiliation with Santos, recognizing that analyses of either cannot likely rule out all confounding.

Usage

colombia

Format

A data frame with 1123 rows and 16 columns.

department

Name for the provincial level unit, called departments or departamentos, of which there are 32 in the country.

dept_code

Short code for the department

town

Name for the town, which is the smallest electoral unit available and is the unit of analysis.

town_code

Code for the town.

total_eligible

Total eligible voters in the town

yes_vote

Proportion (out of 100) voting in favor of the peace deal.

santos10

Proportion (out of 100) voting for Santos in 2010 presidential election.

santos14

Proportion (out of 100) voting for Santos in the 2014 presidential election.

gdppc

The town-level GDP per capita.

pop13

Town-level population in 2013.

elev

Town's mean elevation.

fat_all

Sum of all known fatalities due to FARC violence in the town (from Global Terrorism Database, GTD).

fat_2001to2005_gtd

Sum of fatalities due to FARC in the town in 2001 to 2005 (from GTD).

fat_2006to2010_gtd

Sum of fatalities due to FARC in the town in 2006 to 2010 (from GTD).

fat_2011to2015_gtd

Sum of fatalities due to FARC in the town in 2011 to 2015 (from GTD).

fat_2010to2013

Sum of fatalities due to FARC in the town in 2010 to 2013 (from GTD).

References

Hazlett, C., and Parente, F. (2020). "Who supports peace with the FARC? A sensitivity-based approach under imperfect identification"

Examples

# loads data
data(colombia)

#-----------------------------------------------------
# Violence Models
#-----------------------------------------------------

### Model 1 (bivariate)
model1 <- lm(yes_vote ~ fat_2001to2005_gtd, data = colombia)

### Model 2 (more controls, and lagged violence.)
model2 <- lm(yes_vote ~ fat_2001to2005_gtd + fat_2006to2010_gtd +
               fat_2011to2015_gtd + total_eligible + santos10 + gdppc ,
             data = colombia)

### Sensitivity analysis - Model 2, for effect of most recent violence.
sense.model2 <- sensemakr(model2,
                          treatment = "fat_2011to2015_gtd",
                          benchmark = "santos10",
                          kd = 1)

### contour plot point estimate
plot(sense.model2)

### contour plot t-value
plot(sense.model2, sensitivity.of = "t-value")

#---------------------------------------------
# Political Affiliation Model
#---------------------------------------------

### Model 3: santos2014 as measure of political support for Santos, with control variables.
model3  <- lm(yes_vote ~ santos14 + fat_2010to2013 + elev + gdppc + pop13,
              data = colombia)

### Sensitivity analysis - Model 3
sense.model3 <- sensemakr(model3, treatment = "santos14",
                          benchmark = c("gdppc","elev"),
                          kd = 3)
summary(sense.model3)

### contour plot point estimate
plot(sense.model3, lim = .9)

### contour plot t-value
plot(sense.model3, sensitivity.of = "t-value", lim = 0.9)