Research Interests

I am interested in policy made by unelected institutions within democracies, especially how activists and social movements influence bureaucratic policymaking in the United States.

Methodologically, this means that I am trying to figure out how to measure policy change and identify its causes.

Questions I hope to answer

How do we distinguish grassroots and astroturf activism? Does either affect policy? When and why do Members of Congress lobby federal agencies? How do we measure policy change and attribute influence?

Methods

Text analysis (reuse, topic models and other Bayesian classifiers, content analysis), various regression models including duration models, field experiments, and process tracing.

Relevent data

Example data viz:

How much ideological diversity is there in U.S. state governments? Which state governments are all conservative? Which are all liberal. Which are mixed?

Using DIME ideology scores (based on campaign donations):

## load data from a web address
load(url("https://github.com/judgelord/state-level-DIME/raw/master/data/BonicaStateLevel.Rdata"))
d %<>%
  ## naming things! 
  rename(Governor = gov_cf) %>% 
  rename(Lower = l_cf_median) %>% 
  rename(Upper = u_cf_median) %>%
  mutate(state_ideology = ifelse(dist_1_x_zero == 0 & Governor < 0, "All Liberal", "Mixed")) %>% 
  mutate(state_ideology = ifelse(dist_1_x_zero == 0 & Governor > 0, "All Conservative", state_ideology)) %>%
  mutate(state_ideology = ifelse(is.na(dist_1_x_zero) & Governor > 0, "All Conservative", state_ideology)) %>%
  ## identify min and max DIME scores for each state (each row is a state)
  mutate(min = pmin(Governor, Lower, Upper, na.rm = T), 
         max = pmax(Governor, Lower, Upper, na.rm = T) )

And a function for plotting them:

## Define a function for plotting state-level DIME data with ggplot
DIMEplot <- function(d){ 
  d %>%
    # define y as state for all layers
  ggplot(aes(y = reorder(state, distance_1_gov_leg_med))) +
    # x = lines for the distance between max and min actors
  geom_segment(aes(x = min, 
                   xend = max, 
                   yend = reorder(state, distance_1_gov_leg_med)), 
               color = "grey", 
               size = 2.1, 
               lineend = "round", 
               alpha = .3) + 
    # x = points for each actor
  geom_point(aes(x = Governor, color = Governor), shape = "G") + 
  geom_point(aes(x = Lower, color = Lower), shape = "L") + 
  geom_point(aes(x = Upper, color = Upper), shape = "U") + 
    # labels 
  labs(y = "States by Preference Divergence", 
       x = "Mean Ideology Score", 
       title = paste("Preference Divergence of Political Principals,", unique(d$year) ), color = "") +
    # scales
  scale_x_continuous(breaks = c(-1, 0, 1)) + 
  scale_colour_gradient2(low = "blue", mid = "black", high = "red") +
    # facet by liberal, conservative, or mixed
  facet_grid(state_ideology ~ ., scales = "free_y", space = "free_y") + 
    # adjust text size and drop legend
  theme(axis.text.y = element_text(size = 5), 
        title = element_text(size = 7), 
        legend.position = "none") 
}

We can compare alignment among governors (G) and upper (U) and lower (L) chambers of state legislatures between 2004 and 2008:

d %>% 
  filter(year == 2004) %>% 
  DIMEplot()
d %>% 
  filter(year == 2008) %>%
  DIMEplot()

Step by step

1. Load trimmed-down DIME data from URL:

## load data from a web address
load(url("https://github.com/judgelord/state-level-DIME/raw/master/data/BonicaStateLevel.Rdata"))

2. Give variables better names

## with `rename()`
d %<>%
  ## naming things! 
  rename(Governor = gov_cf) %>% 
  rename(Lower = l_cf_median) %>% 
  rename(Upper = u_cf_median)

3. Make new variables

## with `mutate()`
d %<>%
  ## identify min and max DIME scores for each state (each row is a state)
  mutate(min = pmin(Governor, Lower, Upper, na.rm = T), 
         max = pmax(Governor, Lower, Upper, na.rm = T),
         distance = pmax(
           abs(Governor-Lower),
           abs(Governor-Upper),
           abs(Upper-Lower)
           ) )

4. Make a new variable indicating “All Liberal,” “All Conservative,” or “Mixed”

## with `mutate()` and `ifelse()`
d %<>%
  mutate(state_ideology = ifelse(min < 0 & max < 0,
                                 "All Liberal",
                                 ifelse(min > 0 & max > 0,
                                        "All Conservative",
                                        "Mixed") 
                                 )) %>%
  mutate(state_ideology = ifelse(
    (min > 0 & max < 0) | (min > 0 & max < 0), 
    "Mixed", state_ideology))

5. Plot

A geom_point() layer (x = Governor, y = State):

d %>%
  ggplot() +
  geom_point() + 
  aes(y = state, x = Governor)

aes() = “aesthetic” mapping — i.e. how to show some kind of variation - position (e.g. along the x or y-axis) - color - shape - size - opacity

Add the positions of legislatures:

## with more layers of `geom_point`
d %>%
  ggplot() +
  # define y = state for all layers
  aes(y = state) +
  # x = points for each actor
  geom_point( aes(x = Governor) ) + 
  geom_point( aes(x = Lower) ) + 
  geom_point( aes(x = Upper) )

Add axis labels:

## with  `labs()`
d %>%
  ggplot() +
  # define y = state for all layers
  aes(y = state) +
  # x = points for each actor
  geom_point( aes(x = Governor) ) + 
  geom_point( aes(x = Lower) ) + 
  geom_point( aes(x = Upper) ) + 
  # axis labels! 
  labs(y = "States", 
       x = "Mean Ideology Score")

Adjust shape of points:

## with `shape =`
d %>%
  ggplot() +
  # define y = state for all layers
  aes(y = state) +
  # x = points for each actor
  geom_point(aes(x = Governor), shape = "G") + 
  geom_point(aes(x = Lower), shape = "L") + 
  geom_point(aes(x = Upper), shape = "U") + 
  # axis labels! 
  labs(y = "States", 
       x = "Mean Ideology Score")

Add a distance line:

## with `geom_segment`
d %>%
  ggplot() +
  # define y = state for all layers
  aes(y = state) +
  # x = points for each actor
  geom_point(aes(x = Governor), shape = "G") + 
  geom_point(aes(x = Lower), shape = "L") + 
  geom_point(aes(x = Upper), shape = "U") + 
  # axis labels! 
  labs(y = "States", 
       x = "Mean Ideology Score",
       color = "") + 
  # x = lines for the distance between max and min actors' DIME scores
  geom_segment(aes(x = min, 
                   xend = max, 
                   y = state,
                   yend = state))

Adjust the color, size, line end shape, and opacity of the distance line:

## `color`, `size`, line end shape (`lineend`), and opacity (`alpha`)
  d %>%
  ggplot() +
  # define y = state for all layers
  aes(y = state) +
    # x = points for each actor
  geom_point(aes(x = Governor), shape = "G") + 
  geom_point(aes(x = Lower), shape = "L") + 
  geom_point(aes(x = Upper), shape = "U") +
  # labels 
  labs(y = "States", 
       x = "Mean Ideology Score") +
  # x = lines for the distance between max and min actors
  geom_segment(aes(x = min, 
                   xend = max, 
                   yend = state), 
               color = "grey", 
               size = 2.1, 
               lineend = "round", 
               alpha = .3)

Break out (“facet”) by groups of interest:

###  with `facet_grid()` or `facet_wrap()`
  d %>%
  ggplot() +
  # define y = state for all layers
  aes(y = state) +
  # x = points for each actor
  geom_point(aes(x = Governor), shape = "G") + 
  geom_point(aes(x = Lower), shape = "L") + 
  geom_point(aes(x = Upper), shape = "U") +
  # labels 
  labs(y = "States", 
       x = "Mean Ideology Score") + 
  # x = lines for the distance between max and min actors
  geom_segment(aes(x = min, 
                   xend = max, 
                   yend = state), 
               color = "grey", 
               size = 2.1, 
               lineend = "round", 
               alpha = .3) +
  # facet by liberal, conservative, or mixed
  facet_grid(state_ideology ~ ., scales = "free_y", space = "free_y")

Adjust x axis and color scales:

### with `scale_x` and `scale_color`
  d %>%
  ggplot() +
  # define y = state for all layers
  aes(y = state) +
  # x = points for each actor
  geom_point(aes(x = Governor, color = Governor), shape = "G") + 
  geom_point(aes(x = Lower, color = Lower), shape = "L") + 
  geom_point(aes(x = Upper, color = Upper), shape = "U") +
  # labels 
  labs(y = "States", 
       x = "Mean Ideology Score",
       color = "") + 
  # x = lines for the distance between max and min actors
  geom_segment(aes(x = min, 
                   xend = max, 
                   yend = state), 
               color = "grey", 
               size = 2.1, 
               lineend = "round", 
               alpha = .3) +
    # facet by liberal, conservative, or mixed
  facet_grid(state_ideology ~ ., scales = "free_y", space = "free_y") +
  # scales
  scale_x_continuous(breaks = c(-1, 0, 1)) + 
  scale_color_gradient2(low = "blue", mid = "black", high = "red")

Adjust axis text size and legend:

## with `theme()`
  d %>%
  ggplot() +
  # define y = state for all layers
  aes(y = state) +
  # x = points for each actor
  geom_point(aes(x = Governor, color = Governor), shape = "G") + 
  geom_point(aes(x = Lower, color = Lower), shape = "L") + 
  geom_point(aes(x = Upper, color = Upper), shape = "U") +
  # labels 
  labs(y = "States", 
       x = "Mean Ideology Score",
       color = "") + 
  # x = lines for the distance between max and min actors
  geom_segment(aes(x = min, 
                   xend = max, 
                   yend = state), 
               color = "grey", 
               size = 2.1, 
               lineend = "round", 
               alpha = .3) +
  # scales
  scale_x_continuous(breaks = c(-1, 0, 1)) + 
  scale_color_gradient2(low = "blue", mid = "black", high = "red") + 
  # facet by liberal, conservative, or mixed
  facet_grid(state_ideology ~ ., scales = "free_y", space = "free_y") + 
  # adjust text size and drop legend
  theme(axis.text.y = element_text(size = 5), 
        legend.position = "none")

Make all of that a function:

## Define a function for plotting state-level DIME data with ggplot
DIMEplot <- function(d){ 
  d %>%
    # define y as state for all layers
  ggplot(aes(y = reorder(state, distance))) +
    # x = points for each actor
  geom_point(aes(x = Governor, color = Governor), shape = "G") + 
  geom_point(aes(x = Lower, color = Lower), shape = "L") + 
  geom_point(aes(x = Upper, color = Upper), shape = "U") + 
  # x = lines for the distance between max and min actors
  geom_segment(aes(x = min, 
                   xend = max, 
                   yend = reorder(state, distance)), 
               color = "grey", 
               size = 2.1, 
               lineend = "round", 
               alpha = .3) + 
    # labels 
  labs(y = "States by Preference Divergence", 
       x = "Mean Ideology Score",
       title = paste("Preference Divergence of Political Principals,", unique(d$year) ), 
       color = "") + 
  # scales
  scale_x_continuous(breaks = c(-1, 0, 1)) + 
  scale_colour_gradient2(low = "blue", mid = "black", high = "red") +
    # facet by liberal, conservative, or mixed
  facet_grid(state_ideology ~ ., scales = "free_y", space = "free_y") + 
    # adjust text size and drop legend
  theme(axis.text.y = element_text(size = 5), 
        title = element_text(size = 7), 
        legend.position = "none") 
}

So we can easily can compare alignment among governors (G) and upper (U) and lower (L) chambers of state legislatures between 2004 and 2008:

d %>% 
  filter(year == 2004) %>% 
  DIMEplot()
d %>% 
  filter(year == 2008) %>%
  DIMEplot()