Figures and tables in paper
Andrew Heiss
Last run: July 6, 2018
knitr::opts_chunk$set(cache = FALSE, fig.retina = 2,
tidy.opts = list(width.cutoff = 120), # For code
width = 120) # For outputlibrary(tidyverse)
library(pander)
library(countrycode)
library(ggrepel)
library(ggstance)
library(gridExtra)
library(here)
# Load pre-cleaned data
edb_clean <- read_rds(file.path(here(), "output", "data", "edb_clean.rds"))
edb_reforms <- read_rds(file.path(here(), "output", "data", "edb_reforms.rds"))
# Load helpful functions
source(file.path(here(), "lib", "model_stuff.R"))
source(file.path(here(), "lib", "graphics_stuff.R"))Tables
Table 2: Overview of De Jure Reform Measures
Not generated with this script.
Table 3: Summary of Coefficients for 8 OLS Models
Important: All models only use countries included in the original 2001 report.
These difference models use the lag of the outcome variable and a year indicator variable to determine the change in outcome variable.
Formally, this is defined as:
\[ y_t = \beta_0 + \beta_1 y_{t-1} + \beta_2 X + \epsilon \]
- t = Year
- X = {1 if t = cutpoint year, 0 otherwise}
- β0 = Constant
- β1 = Change in outcome from the previous year
- β2 = Effect of the event
Or an example R formula:
sb_proced ~ sb_proced_lag + ranked_2005# Define all the models that need to be run
# For the paper we just do 2005 and 2006 for all countries. In the appendix we
# run models from 2003-2013 for all countries, countries with EDB committees,
# and countries without EDB committees
models_to_run <- expand.grid(year = 2005:2006,
outcome = c("sb_proced", "sb_days_ln",
"sb_cost_ln", "sb_capital_ln"),
grouping = "All countries",
stringsAsFactors = FALSE)
# This function generates an R model formula based on the name of the dependent
# variable and the year provided. i.e., given "sb_proced" and "2005", it will
# create the formula "sb_proced ~ sb_proced_lag + ranked_2005" and run the model
run_lagged_ols_model <- function(outcome, year, df) {
outcome_lag <- paste0(outcome, "_lag")
year_variable <- paste0("ranked_", year)
form <- as.formula(paste0(outcome, " ~ ", outcome_lag, " + ", year_variable))
lm(form, data = df)
}
# Run models only on countries that were included in the 2001 report
models_to_run_full <- edb_clean %>%
filter(in_2001 == 1) %>%
# Add grouping column so this can be joined with models_to_run
mutate(grouping = "All countries") %>%
group_by(grouping) %>%
nest() %>%
right_join(models_to_run, by = "grouping")
# Run all the models within the data frame
ols_models_lagged <- models_to_run_full %>%
mutate(model = pmap(.l = list(outcome, year, data), run_lagged_ols_model),
# Add robust clustered SEs
robust_se = pmap(.l = list(model, data, "ccode"), robust_clusterify),
# Add model summary statistics
glance = model %>% map(broom::glance),
# Add a data frame of model parameters with correct SEs
tidy_robust = robust_se %>% map(~ broom::tidy(.$coef)))
# Extract the ranking coefficients from all models
ols_coefs <- ols_models_lagged %>%
# Spread out the model results
unnest(tidy_robust) %>%
# Only look at the ranked* coefficients
filter(str_detect(term, "ranked")) %>%
# Clean up the estimates, labels, and add stars
mutate(value = paste0(sprintf("%.3f", round(estimate, 3)), p_stars(p.value)),
term = str_replace(term, "\\.\\d+TRUE", "")) %>%
# Get rid of extra columns
select(-c(estimate, std.error, statistic, p.value)) %>%
spread(outcome, value) %>%
# Give table clean column names
select(Subset = grouping, Year = year,
Procedures = sb_proced, `Cost (log)` = sb_cost_ln,
`Days (log)` = sb_days_ln, `Capital (log)` = sb_capital_ln)caption <- 'Summary of β~2~ coefficients (i.e. "ranked_200x") for difference models'
tbl_ols <- pandoc.table.return(filter(ols_coefs, Year %in% c(2005, 2006)),
caption = caption)
cat(tbl_ols)| Subset | Year | Procedures | Cost (log) | Days (log) | Capital (log) |
|---|---|---|---|---|---|
| All countries | 2005 | -0.176** | -0.081*** | -0.040* | -0.030 |
| All countries | 2006 | -0.240*** | -0.071*** | -0.049** | -0.097* |
Table 4: Summary of Coefficients for 8 ITS Models
These interrupted times series models use an interaction term of the years since an intervention and the intervention itself to determine the change in slope after the intervention.
\[ y_t = \beta_0 + \beta_1 T + \beta_2 X_t + \beta_3 (T \times X_t) + \epsilon \]
- t = Year
- T = Years before/after cutpoint
- X = {1 if t = cutpoint year, 0 otherwise}
- β0 = Constant: pre-period intercept - baseline pre intervention
- β1 = pre-period slope - baseline time trend - level of increase prior to intervention
- β2 = immediate effect of the event - change in intercept at point of experiment
- β3 = change in slope after the experiment - what happens after
Or an example R formula:
sb_proced ~ year_centered_2005 + ranked_2005 + year_centered_2005 * ranked_2005# As with run_lagged_ols_model(), this function generates an R model formula and
# runs it. i.e., given "sb_proced" and "2005", it will create the formula
# "sb_proced ~ year_centered_2005 + ranked_2005 + year_centered_2005 * ranked_2005"
# and run the model
run_its_model <- function(outcome, year, df) {
year_variable <- paste0("ranked_", year)
year_centered <- paste0("year_centered_", year)
form <- as.formula(paste0(outcome, " ~ ", year_centered, " + ", year_variable,
" + ", year_centered, " * ", year_variable))
lm(form, data = df)
}
# Run all the ITS models within the data frame
its_models <- models_to_run_full %>%
mutate(model = pmap(.l = list(outcome, year, data), run_its_model),
robust_se = pmap(.l = list(model, data, "ccode"), robust_clusterify),
glance = model %>% map(broom::glance),
tidy_robust = robust_se %>% map(~ broom::tidy(.$coef)))
# Extract the ranking coefficients from ITS models
its_coefs <- its_models %>%
# Spread out the model results
unnest(tidy_robust) %>%
# Only look at the coefficients from interaction terms (they have ":" in their names)
filter(str_detect(term, ":")) %>%
# Clean up the estimates, labels, and add stars
mutate(value = paste0(sprintf("%.3f", round(estimate, 3)), p_stars(p.value)),
term = str_replace(term, "\\.\\d+TRUE", "")) %>%
# Get rid of extra columns
select(-c(estimate, std.error, statistic, p.value)) %>%
spread(outcome, value) %>%
# Give table clean column names
select(Subset = grouping, Year = year,
Procedures = sb_proced, `Cost (log)` = sb_cost_ln,
`Days (log)` = sb_days_ln, `Capital (log)` = sb_capital_ln)caption <- 'Summary of β~3~ coefficients (i.e. "year.centered.200x × ranked.200x")'
tbl_its <- pandoc.table.return(filter(its_coefs, Year %in% c(2005, 2006)),
caption = caption)
cat(tbl_its)| Subset | Year | Procedures | Cost (log) | Days (log) | Capital (log) |
|---|---|---|---|---|---|
| All countries | 2005 | -0.088 | -0.059** | 0.009 | 0.054 |
| All countries | 2006 | -0.105 | -0.041** | 0.008 | -0.019 |
Table 5: Countries with Reform Committees Directly Using the EDB Data
Not generated with this script.
Table 6: Experimental Results Status Comparisons on Importance of EBD and Business Climate Improvements
Not generated with this script.
Table 7: Experimental Results of Ranking Differences on Investment Likelihood
Not generated with this script.
Figures
Figure 1: Doing Business Website Visits, Annually (2003-2016)
Estimated data based on a screenshot of internal EDB web traffic reports.
edb_web <- read_csv(file.path(here(), "data_raw", "edb-traffic_2013-2016.csv"),
col_names = c("Year", "Visits"))
fig_edb_web <- ggplot(edb_web, aes(x = Year, y = Visits)) +
geom_line() +
labs(x = NULL, y = "Annual visits") +
scale_y_continuous(labels = scales::comma) +
scale_x_continuous(breaks = seq(2002, 2016, 2)) +
theme_edb() +
theme(panel.grid.minor = element_blank())
fig_edb_web
ggsave(fig_edb_web,
filename = file.path(here(), "output", "figures", "edb_website_traffic.pdf"),
width = 6, height = 2.5, units = "in", device = cairo_pdf)
ggsave(fig_edb_web,
filename = file.path(here(), "output", "figures", "edb_website_traffic.png"),
width = 6, height = 2.5, units = "in", type = "cairo", dpi = 300)Figure 2: Average change in select subindicators, before and after public ranking (2006)
df_plot_edb_2001 <- edb_clean %>%
# Only look at countries in the 2001 report
filter(in_2001 == 1) %>%
select(year, sb_days, sb_proced, sb_cost, sb_capital, con_days, con_proced) %>%
# Calculate average of each variable by year
gather(variable, value, -year) %>%
group_by(year, variable) %>%
summarise(avg = mean(value, na.rm = TRUE)) %>%
filter(!is.nan(avg)) %>%
# Only use 2004 and beyond
filter(year >= 2003) %>%
mutate(variable = fct_recode(variable,
`Contract—Days` = "con_days",
`Contract—Procedures` = "con_proced",
`Starting a business—Capital` = "sb_capital",
`Starting a business—Cost` = "sb_cost",
`Starting a business—Days` = "sb_days",
`Starting a business—Procedures` = "sb_proced"))
# Mark the intervention years
plot_interventions <- data_frame(year = 2005:2006,
intervention = c("2005", "2006"))
# Technically, facets in gpplot are for different aspects of the same value, not
# for separate variables. This function creates a miniature ggplot object which
# then gets combined below with gridExtra::arrangeGrob()
make_faux_facet <- function(x, ymin, ymax, ylab) {
plot_data <- df_plot_edb_2001 %>%
filter(variable == x)
faux_facet <- ggplot() +
geom_vline(data = plot_interventions, aes(xintercept = year,
colour = intervention),
linetype = "dashed", size = 0.5) +
geom_line(data = plot_data, aes(x = year, y = avg)) +
scale_color_manual(values = c(its_1, its_2), name = NULL) +
scale_x_continuous(limits = c(2000, 2015), breaks = seq(2000, 2015, 5)) +
coord_cartesian(xlim = c(2000, 2015), ylim = c(ymin, ymax)) +
guides(color = FALSE) +
labs(x = NULL, y = ylab, title = x) +
theme_edb() +
theme(plot.title = element_text(size = rel(1), hjust = 0.5))
return(faux_facet)
}
# Create a data frame with plotting parameters and then create ggplot objects in a new column
plots_2001 <- tribble(
~ymin, ~ymax, ~ylab, ~variable,
0, 610, "Days", "Contract—Days",
0, 45, "Procedures", "Contract—Procedures",
0, 200, "Capital", "Starting a business—Capital",
0, 100, "Cost", "Starting a business—Cost",
0, 50, "Days", "Starting a business—Days",
0, 10, "Procedures", "Starting a business—Procedures"
) %>%
mutate(facet = pmap(.l = list(variable, ymin, ymax, ylab), make_faux_facet))
# Plot all those plots in one big mega plot
plot_edb_2001 <- arrangeGrob(grobs = plots_2001$facet, nrow = 2)
grid::grid.newpage()
grid::grid.draw(plot_edb_2001)
ggsave(plot_edb_2001,
filename = file.path(here(), "output", "figures", "edb_its_2001.pdf"),
width = 9, height = 4.5, units = "in", device = cairo_pdf)
ggsave(plot_edb_2001,
filename = file.path(here(), "output", "figures", "edb_its_2001.png"),
width = 9, height = 4.5, units = "in", type = "cairo", dpi = 300)Figure 3: EDB Rankings, 2005 versus 2014, by Committee Status Based on 2015 report
# Select only 2005 and 2014 rankings
edb_rankings <- edb_clean %>%
select(ccode, year, p_edb_rank, has_bureau) %>%
filter(year %in% c(2005, 2014)) %>%
spread(year, p_edb_rank) %>%
mutate(countryname = countrycode(ccode, "cown", "country.name"),
countryname = ifelse(countryname == "Lao People's Democratic Republic",
"Laos", countryname),
add_label = countryname %in% c("Georgia", "Rwanda",
"Bangladesh", "Nigeria", "Pakistan",
"Norway", "New Zealand",
"Laos", "Niger"),
label = ifelse(add_label, countryname, NA))
# Extra stuff that will go on the plot
annotations <- data_frame(x = c(max(edb_rankings$`2005`, na.rm = TRUE), 0),
y = c(0, max(edb_rankings$`2014`, na.rm = TRUE)),
text = c("Outliers\nimproving", "Outliers\nworsening"),
hjust = c("right", "left"),
vjust = c("bottom", "top"))
reference_line <- data_frame(x = c(0, max(edb_rankings$`2005`, na.rm = TRUE)),
y = c(0, max(edb_rankings$`2014`, na.rm = TRUE)))
# Set the seed every time geom_label_repel is called so the labels are repositioned the same way
set.seed(1234)
fig_rankings_corr <- ggplot(edb_rankings, aes(x = `2005`, y = `2014`,
colour = has_bureau, label = label)) +
# geom_smooth(aes(fill = has_bureau), method = "lm", size = 0.5, alpha = 0.1, fullrange = TRUE) +
geom_line(data = reference_line, aes(x = x, y = y, label = NULL), colour = "grey70", size = 0.5) +
geom_point(size = 1) +
geom_label_repel(aes(fill = has_bureau), colour = "white",
size = 2, alpha = 0.9, segment.color = "grey40") +
geom_text(data = annotations, aes(x = x, y = y, label = text, hjust = hjust, vjust = vjust),
colour = "grey40", size = 2, lineheight = 1) +
scale_color_manual(values = c(color_no_committee, color_committee), name = NULL) +
scale_fill_manual(values = c(color_no_committee, color_committee), name = NULL, guide = FALSE) +
coord_cartesian(xlim = c(0, 160), ylim = c(0, 190)) +
labs(x = "Rank in 2005", y = "Rank in 2014") +
theme_edb()
set.seed(1234)
fig_rankings_corr
set.seed(1234)
ggsave(fig_rankings_corr,
filename = file.path(here(), "output", "figures", "bureau_rankings.pdf"),
width = 6, height = 4, units = "in", device = cairo_pdf)
set.seed(1234)
ggsave(fig_rankings_corr,
filename = file.path(here(), "output", "figures", "bureau_rankings.png"),
width = 6, height = 4, units = "in", type = "cairo", dpi = 300)Figure 4: Ranking Gains by Number of Reforms, by Committee Status (95% CI)
Every reform leads to a negative change in rankings (which is good) in the following year for countries with special EDB reform committees. For instance, doing an enforcing contracts reform will move a country down (i.e. to a better position) nearly 6 positions. This is also the case when normalizing rankings to a 0-100 scale to account for the changing number of rank positions over time.
# Add new columns for leaded rankings and changes in rankings
change_in_rankings_norm <- edb_clean %>%
select(ccode, country_name, year, p_edb_rank, p_edb_rank_normalized, has_bureau) %>%
mutate(rank0 = p_edb_rank,
rank1 = lead(rank0, 1),
rank2 = lead(rank0, 2),
change1 = rank1 - rank0,
change2 = rank2 - rank0,
rank0_norm = p_edb_rank_normalized,
rank1_norm = lead(rank0_norm, 1),
rank2_norm = lead(rank0_norm, 2),
change1_norm = rank1_norm - rank0_norm,
change2_norm = rank2_norm - rank0_norm) %>%
filter(year >= 2005)
# Calculate the changes--or bumps--in rankings per reform undertaken
edb_reforms_rankings_bumps <- edb_reforms %>%
filter(!str_detect(reform_type, "lag")) %>%
group_by(ccode, year, reform_type_clean) %>%
summarise(num_reforms = sum(reform_num_no_na)) %>%
left_join(change_in_rankings_norm, by = c("ccode", "year")) %>%
filter(num_reforms != 0) %>%
mutate(change_per_reform1 = change1 / num_reforms,
change_per_reform2 = change2 / num_reforms,
change_per_reform1_norm = change1_norm / num_reforms,
change_per_reform2_norm = change2_norm / num_reforms)
# Calculate summary statistics of bumps in rankings
edb_reforms_rankings_bumps_summarized <- edb_reforms_rankings_bumps %>%
group_by(reform_type_clean, has_bureau) %>%
summarise(avg_bump_norm = mean(change_per_reform1_norm, na.rm = TRUE),
stderr_bump_norm = sd(change_per_reform1_norm, na.rm = TRUE) /
sqrt(length(change_per_reform1_norm)),
lower_norm = avg_bump_norm + (qnorm(0.025) * stderr_bump_norm),
upper_norm = avg_bump_norm + (qnorm(0.975) * stderr_bump_norm),
avg_bump = mean(change_per_reform1, na.rm = TRUE),
stderr_bump = sd(change_per_reform1, na.rm = TRUE) /
sqrt(length(change_per_reform1)),
lower = avg_bump + (qnorm(0.025) * stderr_bump),
upper = avg_bump + (qnorm(0.975) * stderr_bump)) %>%
ungroup() %>%
arrange(desc(has_bureau), desc(avg_bump)) %>%
filter(!is.na(reform_type_clean)) %>%
mutate(reform_type_clean = factor(reform_type_clean,
levels = unique(reform_type_clean),
ordered = TRUE))plot_ranking_bumps_norm <- ggplot(edb_reforms_rankings_bumps_summarized,
aes(x = avg_bump_norm, y = reform_type_clean, colour = has_bureau)) +
geom_vline(xintercept = 0, colour = "black", size = 0.5) +
geom_pointrangeh(aes(xmin = lower_norm, xmax = upper_norm), size = 0.5,
position = position_dodgev(0.5)) +
scale_color_manual(values = c(color_no_committee, color_committee), name = NULL) +
labs(x = "Average change in EDB rankings one year after reform (normalized rankings)",
y = "Type of reform") +
theme_edb()
plot_ranking_bumps_norm
ggsave(plot_ranking_bumps_norm,
filename = file.path(here(), "output", "figures", "bureau_rankings_bumps.pdf"),
width = 6, height = 4, units = "in", device = cairo_pdf)
ggsave(plot_ranking_bumps_norm,
filename = file.path(here(), "output", "figures", "bureau_rankings_bumps.png"),
width = 6, height = 4, units = "in", type = "cairo", dpi = 300)t-tests for changes in rankings
edb_reforms_rankings_bumps_tests <- edb_reforms_rankings_bumps %>%
ungroup() %>%
select(has_bureau, reform_type_clean, change_per_reform1_norm) %>%
gather(variable, value, -has_bureau, -reform_type_clean) %>%
group_by(reform_type_clean, has_bureau) %>%
summarise(data = list(value)) %>%
spread(has_bureau, data) %>%
mutate(t_test = list(t.test(unlist(`No EDB reform committee`),
unlist(`Special EDB reform committee`))),
t_value = t_test %>% map_dbl("statistic"),
p_value = t_test %>% map_dbl("p.value"),
estimate = t_test %>% map("estimate"),
mean_committee = estimate %>% map_dbl(~ .x[2]),
mean_no_committee = estimate %>% map_dbl(~ .x[1]),
diff_in_means = estimate %>% map_dbl(~ abs(.x[2] - .x[1]))) %>%
mutate(t = paste0(sprintf("%.2f", round(t_value, 2)), p_stars(p_value))) %>%
arrange(mean_committee) %>%
select(Reform = reform_type_clean,
`Average change in rankings, committee` = mean_committee,
`Average change in rankings, no committee` = mean_no_committee,
`Difference in means` = diff_in_means, `*t*` = t)
caption <- 'Differences in average change in EDB rankings by reform, by committee status'
tbl_edb_reforms_diffs <- pandoc.table.return(edb_reforms_rankings_bumps_tests,
caption = caption, split.tables = Inf,
digits = 2, justify = "lcccc")
cat(tbl_edb_reforms_diffs)| Reform | Average change in rankings, committee | Average change in rankings, no committee | Difference in means | t |
|---|---|---|---|---|
| Enforcing contracts | -3.2 | -0.38 | 2.8 | 2.01* |
| Getting electricity | -2.9 | -1.2 | 1.7 | 1.47 |
| Construction permits | -1.9 | -0.43 | 1.4 | 2.10** |
| Paying taxes | -1.4 | -0.3 | 1.1 | 1.88* |
| Resolving insolvency | -1.3 | 0.14 | 1.4 | 2.15** |
| Protecting minority investors | -1.2 | -1.8 | 0.55 | -0.58 |
| Registering property | -1.2 | -0.38 | 0.84 | 1.03 |
| Starting a business | -1.1 | -0.072 | 1.1 | 1.86* |
| Trading across borders | -1.1 | -0.32 | 0.74 | 1.11 |
| Getting credit | -1 | -0.11 | 0.92 | 1.36 |
cat(tbl_edb_reforms_diffs,
file = file.path(here(), "output", "tables", "table_edb_reforms_diffs.md"))*p<0.1; **p<0.05; ***p<0.01
Figure 5: Public Assessments of Importance of Improving India’s Business Climate and EDB Rankings, by Exposure to EDB Information
chn_ind_assessments <- read_csv(file.path(here(), "data_raw", "chn-ind-assessments.csv")) %>%
rename(Frame = X1) %>%
gather(key, value, -Frame)
# Split into two data frames for faceting, then recombine
chn_ind_both <- chn_ind_assessments %>%
filter(key %in% c("India Higher", "No Rank", "China Higher")) %>%
mutate(comparison = "Both")
ind_only <- chn_ind_assessments %>%
filter(key %in% c("No Rank", "India's Rank Only")) %>%
mutate(comparison = "India only")
blank_country <- data_frame(Frame = unique(chn_ind_assessments$Frame),
key = " ", value = c(0, 0),
comparison = "India only")
chn_ind_plot_data <- bind_rows(chn_ind_both, ind_only, blank_country) %>%
mutate(key = factor(key, levels = c("India Higher", "India's Rank Only",
"No Rank", "China Higher", " "),
ordered = TRUE),
Frame = fct_inorder(Frame, ordered = TRUE))
chn_ind_annotations <- data_frame(x = c(4, 5), y = c(0.6, 0.6),
text = c("Lower importance", "Higher importance"),
hjust = c("left", "right"))
# Plot
plot_chn_ind <- ggplot(chn_ind_plot_data, aes(x = value, y = fct_rev(key),
color = fct_rev(Frame))) +
geom_pointrangeh(aes(xmin = 0, xmax = value), size = 0.5,
position = position_dodgev(height = 0.5)) +
geom_text(data = chn_ind_annotations,
aes(x = x, y = y, label = text, hjust = hjust),
vjust = "center", color = "grey40", size = 2, lineheight = 1) +
geom_text(aes(label = value), position = position_dodgev(height = 0.5),
hjust = -0.4, size = 2.5) +
scale_color_manual(values = c(edb_importance, climate_importance), name = NULL,
guide = guide_legend(reverse = TRUE)) +
scale_x_continuous(breaks = seq(0, 5, 0.5)) +
coord_cartesian(xlim = c(4, 5)) +
labs(x = NULL, y = NULL) +
theme_edb() +
theme(strip.text = element_blank(),
panel.grid.minor = element_blank(),
panel.grid.major.y = element_blank()) +
facet_wrap(~ comparison, scales = "free_y")
plot_chn_ind
ggsave(plot_chn_ind,
filename = file.path(here(), "output", "figures", "chn_ind_exposure.pdf"),
width = 6, height = 3, units = "in", device = cairo_pdf)
ggsave(plot_chn_ind,
filename = file.path(here(), "output", "figures", "chn_ind_exposure.png"),
width = 6, height = 3, units = "in", type = "cairo", dpi = 300)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