Analysis stuff here.

Table 1: Summary of hypotheses, treatments, and variables
Hypothesis Treatments Outcomes
H1 If transit agencies provide larger incentives, riders will reload passes with higher values and with greater frequency in the future. Four levels of incentives: (1) $10, (2) >$10, (3) shorter pass, (4) subsidized annual pass Three measures of future use after six months: (1) binary pass reloading, (2) total value of reloadings, (3) total count of reloadings
H2 If transit agencies provide larger incentives, lower-income riders will re-enroll in specially designed programs like ORCA LIFT. Four levels of incentives: (1) $10, (2) >$10, (3) shorter pass, (4) subsidized annual pass Binary reenrollment in ORCA LIFT
H3 If transit agencies provide subsidized passes, riders will reload passes with higher values and with greater frequency in the future. Subsidized annual pass Three measures of future use after six months: (1) binary pass reloading, (2) total value of reloadings, (3) total count of reloadings
H4 If transit agencies provide subsidized passes, lower-income riders will re-enroll in specially designed programs like ORCA LIFT. Subsidized annual pass Binary reenrollment in ORCA LIFT
Category Variable Source Level
Demographics Age King County Individual
Demographics Race King County Individual
Demographics Language King County Individual
Demographics Population ACS Block group
Demographics % female ACS Block group
Demographics % nonwhite ACS Block group
Demographics % working age ACS Block group
Social % married ACS Block group
Social % with kids ACS Block group
Social % with high school education ACS Block group
Social % enrolled in college ACS Block group
Social % veterans ACS Block group
Social % with access to the internet ACS Block group
Housing % with high rent ACS Block group
Housing % with no car ACS Tract
Economic % employed ACS Block group
Economic % uninsured ACS Block group
Economic % per capita ACS Block group
Economic % commuting using public transportation ACS Block group
Economic % working from home ACS Block group
Economic % working in an industry heavily affected by COVID ACS Block group
Economic % below 200% of the poverty line ACS Block group
Economic % below 100% of the poverty line ACS Block group
Economic % using SNAP ACS Tract

DAG

Code 
library(tidyverse)
library(ggdag)
library(dagitty)

clrs <- PNWColors::pnw_palette("Sunset2", n = 6)
Code 
simplified_dag <- dagitty('
dag {
  Demographics [adjusted,pos="0.400,0.600"]
  Economics [adjusted,pos="0.500,0.400"]
  Housing [adjusted,pos="0.400,0.400"]
  Outcome [outcome,pos="0.600,0.500"]
  Social [adjusted,pos="0.500,0.600"]
  Treatment [exposure,pos="0.300,0.500"]
  Demographics -> Outcome
  Demographics -> Treatment
  Economics -> Outcome
  Economics -> Treatment
  Housing -> Outcome
  Housing -> Treatment
  Social -> Outcome
  Social -> Treatment
  Treatment -> Outcome
}
') %>% 
  tidy_dagitty() %>% 
  node_status() %>% 
  dag_adjustment_sets() %>% 
  mutate(highlight = coalesce(status, adjusted))

simplified_dag %>% 
  ggplot(aes(x = x, y = y, xend = xend, yend = yend)) +
  geom_dag_edges(edge_color = "grey70", edge_width = 1) +
  geom_dag_point(aes(color = highlight)) +
  geom_dag_label_repel(
    aes(label = name), force = 0,
    family = "Mulish", size = 4, color = "black", 
    fill = scales::alpha("white", 0.75), label.size = 0
  ) +
  scale_color_manual(values = clrs[c(4, 6, 1)], guide = "none") +
  theme_dag()