+ - 0:00:00
Notes for current slide
Notes for next slide

diff-in-diff vis w ggplot

Evangeline Reynolds

5/18/2018

1 / 5
knitr::opts_chunk$set(warning = F, message = F)
library(flipbookr)

Using a difference-in-difference framework to look at the effect of policy interventions is a popular research design. A binary exlanatory variable that turns on and stays on lends itself to ease of interpretation!

In the context of such a design, what is the essential visual inspections? Below I suggest four plots that researchers and their audiences may find useful in visually inspecting the timing of interventions and the relationships between a policy intervention and outcomes.

I use ggplot2 to implement the visualization. The ability to overwrite global aesthetics, using the aes() function, means that we move from one plot to another with little additional code. A little over a year ago I learned about declaring the aes() on it's own line and maybe novelty bias is at work here, but I find the capability to be a lot of fun to play with!

2 / 5

The five plots are as follows:

  • two plots showing the timing of intervention and the cross-sectional cases (one aligning the moments of interventions)
  • two plots showing how the intervention relates to the response variable of interest (one aligning the moments of interventions)
  • a plot that breaks up the data into individual time-series for the response variable for each cross-sectional unit

Another option for visualizing such data is using the new package panelView, which certainly gave me additional inspiration for this exercise and looks useful! 

library(tidyverse)
library(gapminder)

Simulating interventions

We'll just use

3 / 5
min_year <- min(gapminder$year)
max_year <- max(gapminder$year)
span <- max_year - min_year
gapminder %>% 
  select(country) %>% 
  distinct() %>% 
  sample_frac(.5) %>% 
  mutate(intervention_year = runif(n = n(), 
                 min = min_year, 
                 max = max_year + span)) %>% 
  mutate(intervention_year = ifelse(max_year < intervention_year, 
                                    NA, 
                                    intervention_year)) %>% 
  mutate(intervention_year = 
           round(intervention_year)) %>% 
  sample_n(16) ->
synthetic_interventions
4 / 5
synthetic_interventions %>% 
  inner_join(gapminder) %>% 
  mutate(treatment = 
           case_when(
             year >= intervention_year ~ "treated",
             year < intervention_year ~ "not treated",
             is.na(intervention_year) ~ "not treated")) %>%
  group_by(country) %>% 
  mutate(mean_treated = mean(treatment == "treated")) %>% 
  arrange(mean_treated) %>% 
  ungroup() %>% 
  mutate(country = forcats::fct_inorder(as.character(country))) -> 
panel_prepped
5 / 5
ggplot(panel_prepped)

5 / 5
ggplot(panel_prepped) +  
  aes(x = year)

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country)

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country)

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds")

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5)

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5)

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5) +  
  aes(color = treatment)

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5) +  
  aes(color = treatment) +  
  theme_minimal()

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5) +  
  aes(color = treatment) +  
  theme_minimal() +  
  scale_color_manual(values = c("lightgrey", "steelblue4"))

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5) +  
  aes(color = treatment) +  
  theme_minimal() +  
  scale_color_manual(values = c("lightgrey", "steelblue4")) +  
  labs(color = NULL)

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5) +  
  aes(color = treatment) +  
  theme_minimal() +  
  scale_color_manual(values = c("lightgrey", "steelblue4")) +  
  labs(color = NULL) +  
  labs(x = NULL)

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5) +  
  aes(color = treatment) +  
  theme_minimal() +  
  scale_color_manual(values = c("lightgrey", "steelblue4")) +  
  labs(color = NULL) +  
  labs(x = NULL) +  
  labs(y = "country")

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5) +  
  aes(color = treatment) +  
  theme_minimal() +  
  scale_color_manual(values = c("lightgrey", "steelblue4")) +  
  labs(color = NULL) +  
  labs(x = NULL) +  
  labs(y = "country") +  
  labs(title = "Plot 1: Treatment-control summary")

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5) +  
  aes(color = treatment) +  
  theme_minimal() +  
  scale_color_manual(values = c("lightgrey", "steelblue4")) +  
  labs(color = NULL) +  
  labs(x = NULL) +  
  labs(y = "country") +  
  labs(title = "Plot 1: Treatment-control summary") +  
  # overwriting y position mapping
  aes(y = gdpPercap) +
  scale_y_log10() +
  labs(y = "GDP per cap") +
  labs(title = "Plot 2: Treatment-control response overlaid summary")

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5) +  
  aes(color = treatment) +  
  theme_minimal() +  
  scale_color_manual(values = c("lightgrey", "steelblue4")) +  
  labs(color = NULL) +  
  labs(x = NULL) +  
  labs(y = "country") +  
  labs(title = "Plot 1: Treatment-control summary") +  
  # overwriting y position mapping
  aes(y = gdpPercap) +  
  scale_y_log10() +  
  labs(y = "GDP per cap") +  
  labs(title = "Plot 2: Treatment-control response overlaid summary") +  
  # overwriting x position mapping
  aes(x = year - intervention_year) +
  labs(x = "year from intervention year") +
  labs(title = "Plot 3: Treatment-control response overlaid summary, aligned intervention")

5 / 5
ggplot(panel_prepped) +  
  aes(x = year) +  
  aes(y = country) +  
  aes(group = country) +  
  labs(caption = "Four essential plots for diff-in-diff | Gina Reynolds") +  
  geom_line(alpha = .5) +  
  geom_point(size = 1.5) +  
  aes(color = treatment) +  
  theme_minimal() +  
  scale_color_manual(values = c("lightgrey", "steelblue4")) +  
  labs(color = NULL) +  
  labs(x = NULL) +  
  labs(y = "country") +  
  labs(title = "Plot 1: Treatment-control summary") +  
  # overwriting y position mapping
  aes(y = gdpPercap) +  
  scale_y_log10() +  
  labs(y = "GDP per cap") +  
  labs(title = "Plot 2: Treatment-control response overlaid summary") +  
  # overwriting x position mapping
  aes(x = year - intervention_year) +  
  labs(x = "year from intervention year") +  
  labs(title = "Plot 3: Treatment-control response overlaid summary, aligned intervention") +  
  # overwriting x position mapping again
  aes(x = year) +
  labs(x = NULL) +
  scale_y_continuous() +
  labs(title = "Plot 4: Treatment-Response faceted") +
  facet_wrap(~ country)

5 / 5
knitr::opts_chunk$set(warning = F, message = F)
library(flipbookr)

Using a difference-in-difference framework to look at the effect of policy interventions is a popular research design. A binary exlanatory variable that turns on and stays on lends itself to ease of interpretation!

In the context of such a design, what is the essential visual inspections? Below I suggest four plots that researchers and their audiences may find useful in visually inspecting the timing of interventions and the relationships between a policy intervention and outcomes.

I use ggplot2 to implement the visualization. The ability to overwrite global aesthetics, using the aes() function, means that we move from one plot to another with little additional code. A little over a year ago I learned about declaring the aes() on it's own line and maybe novelty bias is at work here, but I find the capability to be a lot of fun to play with!

2 / 5
Paused

Help

Keyboard shortcuts

↑, ←, Pg Up, k Go to previous slide
↓, →, Pg Dn, Space, j Go to next slide
Home Go to first slide
End Go to last slide
Number + Return Go to specific slide
b / m / f Toggle blackout / mirrored / fullscreen mode
c Clone slideshow
p Toggle presenter mode
t Restart the presentation timer
?, h Toggle this help
Esc Back to slideshow