Load libraries and data
Variables
- Dependent variables
- Explanatory variables
Models
Bayesian tinkering

Load libraries and data

library(magrittr)
library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

library(purrr)

## 
## Attaching package: 'purrr'

## The following object is masked from 'package:dplyr':
## 
##     order_by

## The following object is masked from 'package:magrittr':
## 
##     set_names

library(broom)
library(ggplot2)
library(Cairo)
library(stargazer)

## 
## Please cite as:

##  Hlavac, Marek (2015). stargazer: Well-Formatted Regression and Summary Statistics Tables.

##  R package version 5.2. http://CRAN.R-project.org/package=stargazer

full.data <- readRDS(file.path(PROJHOME, "Data", "data_processed",
                               "full_data.rds"))

my.seed <- 1234
set.seed(my.seed)

Variables

Dependent variables

Civil society regulatory environment:

v2csreprss: V-Dem CSO repression (“Does the government attempt to repress civil society organizations?). Higher numbers = less repression.
v2csreprss_ord: Ordinal version of v2csreprss (labels = “Severely”, “Substantially”, “Moderately”, “Weakly”, “No”)
v2cseeorgs: V-Dem CSO entry and exit (“To what extent does the government achieve control over entry and exit by civil society organizations (CSOs) into public life?”). Higher numbers = less control.
cs_env_sum: Sum of v2csreprss and v2cseeorgs. V-Dem includes a “Core civil society index” that uses Bayesian factor analysis to combine repression, entry/exit, and participatory environment. I’m less interested in the participatory environment. The paper describing the original factor analysis model, however, does not explain how it was built, so here I make a simple additive index of the two variables.

Explanatory variables

Internal stability:

icrg.stability:
icrg.pol.risk.internal.scaled: Political risk rating—external measures removed and index rescaled to 0-100 scale
yrsoffc:
years.since.comp:
opp1vote:

External stability:

neighbor.pol.risk.XXX

International reputation:

Variable from ICEWS / shaming data from Murdie

Controls:

e_polity2:
physint:
gdpcap.log:
population.log:
oda.log:
countngo:
globalization:

Models

Internal factors

Just internal political risk

model.int.simple <- full.data %>%
  split(.$polity_ord2) %>%
  map(~ lm(cs_env_sum.lead ~ icrg.pol.risk.internal.scaled,
           data=.))

stargazer(model.int.simple, type="html", 
          dep.var.caption="CSRE",
          dep.var.labels.include=FALSE, no.space=TRUE,
          column.labels=names(model.int.simple),
          omit="factor\\(year", 
          add.lines=list(c("Year fixed effects",
                           rep("No", 2))))


	CSRE

	Autocracy	Democracy
	(1)	(2)

icrg.pol.risk.internal.scaled	-0.020^***	0.072^***
	(0.006)	(0.003)
Constant	0.848^**	-1.007^***
	(0.332)	(0.168)

Year fixed effects	No	No
Observations	784	1,776
R²	0.014	0.318
Adjusted R²	0.013	0.318
Residual Std. Error	2.270 (df = 782)	1.382 (df = 1774)
F Statistic	10.987^*** (df = 1; 782)	828.026^*** (df = 1; 1774)

Note:	p<0.1; p<0.05; p<0.01

All internal variables

model.int.full <- full.data %>%
  split(.$polity_ord2) %>%
  map(~ lm(cs_env_sum.lead ~ icrg.pol.risk.internal.scaled +
             yrsoffc + years.since.comp + opp1vote + as.factor(year.num),
           data=.))

stargazer(model.int.full, type="html", 
          dep.var.caption="CSRE",
          dep.var.labels.include=FALSE, no.space=TRUE,
          column.labels=names(model.int.full),
          omit="factor\\(year",
          add.lines=list(c("Year fixed effects",
                           rep("Yes", 2))))


	CSRE

	Autocracy	Democracy
	(1)	(2)

icrg.pol.risk.internal.scaled	-0.009	0.079^***
	(0.007)	(0.003)
yrsoffc	-0.060^***	-0.036^***
	(0.011)	(0.011)
years.since.comp	-0.038^***	0.019
	(0.008)	(0.017)
opp1vote	0.038^***	0.007^**
	(0.009)	(0.003)
Constant	1.363^***	-0.876^***
	(0.480)	(0.275)

Year fixed effects	Yes	Yes
Observations	409	1,377
R²	0.302	0.370
Adjusted R²	0.258	0.359
Residual Std. Error	1.940 (df = 384)	1.348 (df = 1352)
F Statistic	6.922^*** (df = 24; 384)	33.094^*** (df = 24; 1352)

Note:	p<0.1; p<0.05; p<0.01

All internal variables + controls

model.int.all <- full.data %>%
  split(.$polity_ord2) %>%
  map(~ lm(cs_env_sum.lead ~ icrg.pol.risk.internal.scaled +
             yrsoffc + years.since.comp + opp1vote +
             physint + gdpcap.log + population.log +
             oda.log + countngo + globalization + as.factor(year.num),
           data=.))

stargazer(model.int.all, type="html", 
          dep.var.caption="CSRE",
          dep.var.labels.include=FALSE, no.space=TRUE,
          column.labels=names(model.int.all),
          omit="factor\\(year", 
          add.lines=list(c("Year fixed effects",
                           rep("Yes", 2))))


	CSRE

	Autocracy	Democracy
	(1)	(2)

icrg.pol.risk.internal.scaled	0.016	0.055^***
	(0.011)	(0.008)
yrsoffc	-0.056^***	-0.038^***
	(0.011)	(0.015)
years.since.comp	-0.013	0.021
	(0.009)	(0.036)
opp1vote	0.015^*	0.006
	(0.009)	(0.004)
physint	0.165^**	0.286^***
	(0.069)	(0.042)
gdpcap.log	-0.942^***	0.215^**
	(0.162)	(0.096)
population.log	0.301^**	0.145^***
	(0.129)	(0.055)
oda.log	0.419^***	-0.066
	(0.126)	(0.058)
countngo	-0.008^*	0.004^*
	(0.004)	(0.002)
globalization	-0.008	-0.022^**
	(0.015)	(0.009)
Constant	-6.931^***	-2.472
	(2.307)	(1.574)

Year fixed effects	Yes	Yes
Observations	299	770
R²	0.604	0.285
Adjusted R²	0.562	0.258
Residual Std. Error	1.472 (df = 270)	1.473 (df = 741)
F Statistic	14.680^*** (df = 28; 270)	10.567^*** (df = 28; 741)

Note:	p<0.1; p<0.05; p<0.01

All models (to LaTeX)

all.models <- c(model.int.simple, model.int.full, model.int.all)

coef.labs <- c("Internal political risk (ICRG)", "Years executive in office", 
               "Years since competitive election", "Opposition vote share",
               "Physical integrity rights", "GDP per capita (log)", 
               "Population (log)", "Foreign aid (log)", 
               "Number of INGO members", "Globalization")
extra.lines <- list(c("Year fixed effects",
                      c(rep("No", 2), rep("Yes", 4))))

capture.output({
  stargazer(all.models, type="latex", 
            out=file.path(PROJHOME, "Output", "tables", "1-internal-models-all.tex"),
            covariate.labels=coef.labs,
            dep.var.caption="Civil society regulatory environment (CSRE) in following year",
            dep.var.labels.include=FALSE,
            column.labels=names(all.models),
            omit="factor\\(year", no.space=TRUE,
            add.lines=extra.lines)
}, file="/dev/null")

External factors

Just neighbors

model.ext.neighbors <- full.data %>%
  split(.$polity_ord2) %>%
  map(~ lm(cs_env_sum.lead ~ neighbor.pol.risk.min + 
             neighbor.coups.activity.bin,
           data=.))

stargazer(model.ext.neighbors, type="html", 
          dep.var.caption="CSRE",
          dep.var.labels.include=FALSE, no.space=TRUE,
          column.labels=names(model.int.simple),
          omit="factor\\(year", 
          add.lines=list(c("Year fixed effects",
                           rep("No", 2))))


	CSRE

	Autocracy	Democracy
	(1)	(2)

neighbor.pol.risk.min	-0.018^***	0.049^***
	(0.006)	(0.003)
neighbor.coups.activity.bin	0.197	-0.025
	(0.260)	(0.175)
Constant	0.639^**	0.968^***
	(0.270)	(0.161)

Year fixed effects	No	No
Observations	788	1,647
R²	0.015	0.165
Adjusted R²	0.012	0.164
Residual Std. Error	2.231 (df = 785)	1.550 (df = 1644)
F Statistic	5.776^*** (df = 2; 785)	161.882^*** (df = 2; 1644)

Note:	p<0.1; p<0.05; p<0.01

Neighbors and subregion

model.ext.neighbors.subregion <- full.data %>%
  split(.$polity_ord2) %>%
  map(~ lm(cs_env_sum.lead ~ neighbor.pol.risk.min + 
             icrg.pol.risk.subregional +
             neighbor.coups.activity.bin + 
             coups.activity.subregional,
           data=.))

stargazer(model.ext.neighbors.subregion, type="html", 
          dep.var.caption="CSRE",
          dep.var.labels.include=FALSE, no.space=TRUE,
          column.labels=names(model.int.simple),
          omit="factor\\(year", 
          add.lines=list(c("Year fixed effects",
                           rep("No", 2))))


	CSRE

	Autocracy	Democracy
	(1)	(2)

neighbor.pol.risk.min	0.017^***	0.018^***
	(0.006)	(0.003)
icrg.pol.risk.subregional	-0.116^***	0.069^***
	(0.011)	(0.005)
neighbor.coups.activity.bin	-0.423^*	-0.024
	(0.249)	(0.189)
coups.activity.subregional	0.483^***	0.173^**
	(0.105)	(0.075)
Constant	5.692^***	-2.055^***
	(0.613)	(0.270)

Year fixed effects	No	No
Observations	779	1,646
R²	0.201	0.257
Adjusted R²	0.196	0.255
Residual Std. Error	2.009 (df = 774)	1.463 (df = 1641)
F Statistic	48.547^*** (df = 4; 774)	142.020^*** (df = 4; 1641)

Note:	p<0.1; p<0.05; p<0.01

Neighbors and region

model.ext.neighbors.region <- full.data %>%
  split(.$polity_ord2) %>%
  map(~ lm(cs_env_sum.lead ~ neighbor.pol.risk.min + 
             icrg.pol.risk.regional +
             neighbor.coups.activity.bin + 
             coups.activity.regional,
           data=.))

stargazer(model.ext.neighbors.region, type="html", 
          dep.var.caption="CSRE",
          dep.var.labels.include=FALSE, no.space=TRUE,
          column.labels=names(model.int.simple),
          omit="factor\\(year", 
          add.lines=list(c("Year fixed effects",
                           rep("No", 2))))


	CSRE

	Autocracy	Democracy
	(1)	(2)

neighbor.pol.risk.min	0.001	0.036^***
	(0.006)	(0.003)
icrg.pol.risk.regional	-0.058^***	0.045^***
	(0.018)	(0.007)
neighbor.coups.activity.bin	-0.011	0.017
	(0.258)	(0.179)
coups.activity.regional	0.116	0.100^**
	(0.073)	(0.045)
Constant	3.074^***	-1.467^***
	(1.126)	(0.434)

Year fixed effects	No	No
Observations	787	1,646
R²	0.058	0.187
Adjusted R²	0.053	0.185
Residual Std. Error	2.185 (df = 782)	1.530 (df = 1641)
F Statistic	12.000^*** (df = 4; 782)	94.659^*** (df = 4; 1641)

Note:	p<0.1; p<0.05; p<0.01

Neighbors and subregion + controls

model.ext.neighbors.subregion.ctrl <- full.data %>%
  split(.$polity_ord2) %>%
  map(~ lm(cs_env_sum.lead ~ neighbor.pol.risk.min + 
             icrg.pol.risk.subregional +
             neighbor.coups.activity.bin + 
             coups.activity.subregional + 
             gdpcap.log + population.log + oda.log + 
             countngo + globalization + as.factor(year.num),
           data=.))

stargazer(model.ext.neighbors.subregion.ctrl, type="html", 
          dep.var.caption="CSRE",
          dep.var.labels.include=FALSE, no.space=TRUE,
          column.labels=names(model.int.simple),
          omit="factor\\(year", 
          add.lines=list(c("Year fixed effects",
                           rep("Yes", 2))))


	CSRE

	Autocracy	Democracy
	(1)	(2)

neighbor.pol.risk.min	0.007	0.005
	(0.007)	(0.005)
icrg.pol.risk.subregional	-0.085^***	0.026^***
	(0.014)	(0.010)
neighbor.coups.activity.bin	-0.154	-0.101
	(0.250)	(0.236)
coups.activity.subregional	0.446^***	0.132
	(0.119)	(0.095)
gdpcap.log	0.028	0.371^***
	(0.117)	(0.090)
population.log	-0.343^***	-0.147^***
	(0.078)	(0.043)
oda.log	0.556^***	0.058
	(0.080)	(0.057)
countngo	-0.005^*	-0.005^*
	(0.003)	(0.002)
globalization	0.012	-0.004
	(0.011)	(0.008)
Constant	-2.211	0.111
	(1.905)	(1.412)

Year fixed effects	Yes	Yes
Observations	584	998
R²	0.287	0.133
Adjusted R²	0.252	0.109
Residual Std. Error	1.771 (df = 556)	1.572 (df = 970)
F Statistic	8.272^*** (df = 27; 556)	5.527^*** (df = 27; 970)

Note:	p<0.1; p<0.05; p<0.01

All models (to LaTeX)

all.models <- c(model.ext.neighbors, model.ext.neighbors.region, 
                model.ext.neighbors.subregion,
                model.ext.neighbors.subregion.ctrl)

coef.labs <- c("Lowest political risk in all neighboring countries", 
               "Average political risk in subregion", 
               "Average political risk in region",
               "Coup activity in neighboring countries (binary)",
               "Coup activity in subregion (binary)",
               "Coup activity in region (binary)",
               "GDP per capita (log)", 
               "Population (log)", "Foreign aid (log)", 
               "Number of INGO members", "Globalization")
extra.lines <- list(c("Year fixed effects",
                      c(rep("No", 6), rep("Yes", 2))))

capture.output({
  stargazer(all.models, type="latex", 
            out=file.path(PROJHOME, "Output", "tables", "1-external-models-all.tex"),
            covariate.labels=coef.labs,
            dep.var.caption="Civil society regulatory environment (CSRE) in following year",
            dep.var.labels.include=FALSE,
            column.labels=names(all.models),
            omit="factor\\(year", no.space=TRUE,
            add.lines=extra.lines)
}, file="/dev/null")

International reputation

All three at once?

model.everything <- lm(cs_env_sum.lead ~ icrg.pol.risk.internal.scaled +
                         yrsoffc + years.since.comp + opp1vote +
                         neighbor.pol.risk.min +
                         e_polity2 +
                         physint + gdpcap.log + population.log +
                         oda.log + countngo + globalization,
                       data=full.data)
summary(model.everything)

## 
## Call:
## lm(formula = cs_env_sum.lead ~ icrg.pol.risk.internal.scaled + 
##     yrsoffc + years.since.comp + opp1vote + neighbor.pol.risk.min + 
##     e_polity2 + physint + gdpcap.log + population.log + oda.log + 
##     countngo + globalization, data = full.data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -3.8000 -1.0443  0.0321  1.0299  5.5844 
## 
## Coefficients:
##                                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                   -2.1801746  1.2598195  -1.731  0.08385 .  
## icrg.pol.risk.internal.scaled  0.0258183  0.0063801   4.047 5.60e-05 ***
## yrsoffc                       -0.0457290  0.0095454  -4.791 1.92e-06 ***
## years.since.comp              -0.0269695  0.0084376  -3.196  0.00144 ** 
## opp1vote                      -0.0014489  0.0041757  -0.347  0.72868    
## neighbor.pol.risk.min          0.0003976  0.0048706   0.082  0.93496    
## e_polity2                      0.2278247  0.0137361  16.586  < 2e-16 ***
## physint                        0.2236433  0.0361012   6.195 8.58e-10 ***
## gdpcap.log                    -0.0694932  0.0751787  -0.924  0.35552    
## population.log                 0.1064540  0.0501490   2.123  0.03402 *  
## oda.log                        0.0751036  0.0526832   1.426  0.15431    
## countngo                       0.0019172  0.0022356   0.858  0.39133    
## globalization                 -0.0213493  0.0067471  -3.164  0.00160 ** 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.495 on 979 degrees of freedom
##   (2440 observations deleted due to missingness)
## Multiple R-squared:  0.5824, Adjusted R-squared:  0.5772 
## F-statistic: 113.8 on 12 and 979 DF,  p-value: < 2.2e-16

Bayesian tinkering

See:

# library(rstanarm)
# options(mc.cores = parallel::detectCores())  # Use all possible cores
# 
# model.simple.b <- stan_glm(cs_env_sum.lead ~ icrg.stability + e_polity2,
#                           data=full.data, family=gaussian(),
#                           prior=cauchy(), prior_intercept=cauchy(),
#                           seed=my.seed)
# print(model.simple.b, digits=2)
# plot(model.simple.b, pars=c("icrg.stability", "e_polity2"))
# pp_check(model.simple.b, check="distributions", overlay=FALSE, nreps=5)

▲ Back to top ▲

Determinants of international civil society restrictions

Andrew Heiss

March 3, 2016

Load libraries and data

Variables

Dependent variables

Explanatory variables

Models

Internal factors

Just internal political risk

All internal variables

All internal variables + controls

All models (to LaTeX)

External factors

Just neighbors

Neighbors and subregion

Neighbors and region

Neighbors and subregion + controls

All models (to LaTeX)

International reputation

All three at once?

Bayesian tinkering