Descriptive Statistics
Paula
5/9/2021
Abstract
This study will seek to explain variation in US Public Diplomacy spending in 189 countries over seven years for the period of 2013 and 2019. Employing a cross-sectional, time-series data set, the study will examine whether US public diplomacy determinants are the same as those for US ODA and military spending. Additionally, the study will examine whether US PD spending goes primarily to countries already economically / politically / culturally similar to the US in order to maintain positive public images. Or whether US PD spending goes primarily to countries dissimilar or hostile to the US. Furthermore, the study will examine the extent to which US PD spending is consistent across the years and across countries. Or alternatively, whether spending is reactionary to international events. Finally, the paper will examine whether PD spending is correlated with or independent of concurrent Chinese or Russian foreign policy or PD actions.
Hypotheses
Hypothesis 1
Null 1.1:
US PD spending follows same contours as US ODA (humanitarian) spending
Null 1.2:
US PD spending follows same contours as US military aid spending
Alt 1.1:
US PD spending is distinct from ODA and military spending
IV: ODA spending, military spending
Hypothesis 2
Null 2.1:
US PD spending goes primarily to countries already similar to the US (i.e. to maintain existing positive image in (economically/politically/culturally) similar populations)
Alt 2.1:
US PD spending goes primarily to countries dissimilar to the US (i.e. to create positive image in dissimilar populations)
Alt 2.2:
US PD spending goes primarily to countries hostile to the US (i.e. to salvage positive image in hostile population)
IV: similar regime type, similar religion, similar civil liberties, similar voting pattern in UN, PEW attitudes to US, US alliance or not
Hypothesis 3:
Null 3.1:
US PD spending is consistent across the years and across countries
Alt 3.1:
US PD spending is reactionary (i.e. used to punish or reward countries for their behaviour vis-a-vis US foreign policy)
Alt 3.2:
US PD spending is reactionary (i.e. used to combat PR crisis or anti-US sentiment that occur stochastically)
IV: Anti-US sanctions, anti-US protests, US invasions, US army presence, US realpolitik strategies
Hypothesis 4:
Null 4.1:
US PD spending is not affected by concurrent Chinese or Russian foreign policy or PD actions
Alt 4.1:
US PD spending is affected (i.e. correlation with PD spending with Chinese / Russian spending)
IV: Chinese ODA, Russian ODA, Chinese-led international organisations, Russian-led international organisation membership.
Methodology
Panel data allow the analysis of dynamic effects (effects among variables over time; Bergh, 1993; Greve & Goldeng, 2004; Pitariu & Ployhart, 2010); provide a way to control for unmeasured stable variables (the so-called unobserved heterogeneity in the econometric literature); and can address some sources of endogeneity of regressors (Hamilton & Nickerson, 2003) that prevent estimates from being interpreted as causal effects (Antonakis, Bendahan, Jacquart, & Lalive, 2010)
‘System GMM’
It is an estimator designed for situations with:
- ‘small T, large N’ panels: few time periods and many individual units a linear functional relationship
- one left-hand variable that is dynamic, depending on its own past realisations
- right-hand variables that are not strictly exogenous: correlated with past and possibly current realisations of the error
- fixed individual effects, implying unobserved heterogeneity
- heteroskedasticity and autocorrelation within individual units’ errors, but not across them
The Arellano–Bond estimator sets up a generalized method of moments (GMM) problem in which the model is specified as a system of equations, one per time period, where the instruments applicable to each equation differ (for instance, in later time periods, additional lagged values of the instruments are available).
Time-fixed effects are a way of accounting for the influence of global business cycles, as was done in Burnside and Dollar (2000), and for potential changes in the multilateral donors‟ inclination to provide aid.
Country-fixed effects can account for individual characteristics of the countries in the sample that could affect their receipts of multilateral aid, regardless of the values of the explanatory variables. These effects can be introduced into the error term eit through a country-specific intercept
In paper on determinants of multilateral aid by Hlvac
Although the two time-fixed panel regressions yield statistically significant coefficients, their adjusted R2 is very low – only around 0.19. Such low values indicate that the panel regression overall does not fit the sample data well, and that other factors – perhaps country-specific qualities that remain constant across time – might be important.Comparing PD Budgets non-logged versus logged
We see that the logged PD Budget variable approaches normal distribution so it is more suitable for regression analysis.
Across all seven years, the logged PD budgets approach normal distribution.
If we look at the relationship between GDP (total) an PD budget (total) per country, there is a positive linear trend. However, we see evidence that there could be heteroskedasticity issues as we go higher with both variables
Will need to investigate and remedy non-random error. Also is it slightly polynomic?
If we look at the relationship between a country’s trade openness index and PD budget (total) per country, there is no strong trend… it is slightly negative
If we look at the relationship between a country’s trade imports from US and PD budget (total) per country, there is no strong trend… it is slightly negative
The US State Department Diplomatic Bureaus
#The US State Department has six geographical regions for its bureaus
- African Affairs (AF),
- East Asian and Pacific Affairs (EAP),
- European and Eurasian Affairs (EUR),
- Near Eastern Affairs (NEA),
- South and Central Asian Affairs (SCA),
- Western Hemisphere Affairs (WHA).
We can quickly look at whether there are any clear trends related to which party was in power in the US Senate, US House and US Presidency.
We can look at the distribution of budgets across continents
#Smaller budgets went to Pacific and African countries, largest budgets went to Asian countries
If we look at budgets per capita, this is reversed. More money goes per capita to Pacific countries and Europe.
Asiam countries receive vey little PD budget per capita
We see African countries recieves very little per capita and in absolute terms.
Examine the relationship between UN voting similarities and PD budgets
Is there a very steep positive correlation between votes and PD
The more far away a country is from the US liberal order vote value, the more PD spending they receive?
However, regression analysis with the panel data shows that if we lag the UN vote value, the countries that voted in a way dissimilar to the US liberal order voting record, they were “punished” less PD spending?Also note worthy is the distinct clustering among “Free countries” … if we look down later, this can be somewhat explained by whether the country is Western / NATO member or not
If we only look at 2015
pd <- read.csv("C:/Users/Paula/Desktop/PD_original_datasets/vg.csv")
library(stargazer)##
## Please cite as:## Hlavac, Marek (2018). stargazer: Well-Formatted Regression and Summary Statistics Tables.## R package version 5.2.2. https://CRAN.R-project.org/package=stargazerpd %<>%
dplyr::mutate(over_100_army = ifelse(army_personnel > 99, 1, 0)) %>%
dplyr::mutate(exports_percent_gdp = exports / gdp_current) %>%
dplyr::mutate(imports_percent_gdp = imports / gdp_current) %>%
dplyr::mutate(trade_openess_percent_gdp = exports + imports/ gdp_current)
pd$ln_mil <- log(pd$sum_military_aid + 0.000001)
pd$ln_econ <- log(pd$sum_economic_aid + 0.000001)## Warning in log(pd$sum_economic_aid + 1e-06): NaNs producedrandom_lagged <- plm(log(budget) ~
lag(log(budget)) +
as.factor(isWest) +
civil_lib +
percent_christian +
as.factor(freedom_house) +
lag(IdealPointDistance) +
as.factor(us_alliance) +
as.factor(over_100_army) +
lag(as.factor(house_maj)) +
# lag(as.factor(senate_maj)) +
lag(as.factor(president)) +
lag(log(trade_openess_percent_gdp)),
index = c("country.name.x", "year"),
model = "random",
random.method = "swar",
data = pd)
lag_econ <- plm(ln_econ ~
lag(ln_econ) +
as.factor(isWest) +
civil_lib +
percent_christian +
as.factor(freedom_house) +
lag(IdealPointDistance) +
as.factor(us_alliance) +
as.factor(over_100_army) +
lag(as.factor(house_maj)) +
# lag(as.factor(senate_maj)) +
lag(as.factor(president)) +
lag(log(trade_openess_percent_gdp)),
index = c("country.name.x", "year"),
model = "random",
random.method = "swar",
data = pd)
lag_mil <- plm(ln_mil ~
lag(ln_mil) +
as.factor(isWest) +
civil_lib +
percent_christian +
as.factor(freedom_house) +
lag(IdealPointDistance) +
as.factor(us_alliance) +
as.factor(over_100_army) +
lag(as.factor(house_maj)) +
# lag(as.factor(senate_maj)) +
lag(as.factor(president)) +
lag(log(trade_openess_percent_gdp)),
index = c("country.name.x", "year"),
model = "random",
random.method = "swar",
data = pd)
stargazer( random_lagged, lag_mil, lag_econ,
type = "text")##
## ======================================================================
## Dependent variable:
## ----------------------------------
## log(budget) ln_mil ln_econ
## (1) (2) (3)
## ----------------------------------------------------------------------
## lag(log(budget)) 0.787***
## (0.018)
##
## lag(ln_mil) 0.642***
## (0.027)
##
## lag(ln_econ) 0.498***
## (0.033)
##
## as.factor(isWest)1 -0.151*** -5.917*** -4.954***
## (0.058) (1.068) (0.682)
##
## civil_lib -0.019 2.380 -0.693
## (0.150) (2.316) (1.509)
##
## percent_christian -0.001* -0.005 -0.003
## (0.0005) (0.007) (0.005)
##
## as.factor(freedom_house)2 0.137*** 0.128 1.447***
## (0.050) (0.757) (0.503)
##
## as.factor(freedom_house)3 0.153* -0.419 0.604
## (0.086) (1.319) (0.860)
##
## lag(IdealPointDistance) -0.077*** -0.798* 0.299
## (0.027) (0.428) (0.275)
##
## as.factor(us_alliance)1 0.117*** -0.385 -0.495
## (0.039) (0.601) (0.392)
##
## as.factor(over_100_army)1 0.068 0.705 -0.655
## (0.043) (0.688) (0.450)
##
## lag(as.factor(house_maj))R -0.002 -2.082*** 0.699
## (0.047) (0.738) (0.483)
##
## lag(as.factor(president))Trump 0.207*** -0.281 0.478
## (0.035) (0.549) (0.358)
##
## lag(log(trade_openess_percent_gdp)) 0.044*** -0.142 -0.073
## (0.009) (0.126) (0.082)
##
## Constant 2.605*** 8.709*** 7.750***
## (0.309) (2.933) (1.948)
##
## ----------------------------------------------------------------------
## Observations 870 828 824
## R2 0.802 0.521 0.512
## Adjusted R2 0.800 0.514 0.504
## F Statistic 3,478.116*** 888.181*** 849.816***
## ======================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01pd$ln_budget_pc <- log(pd$pd_budget_per_capita)
random_lag_pc <- plm(ln_budget_pc ~
lag(ln_budget_pc) +
as.factor(isWest) +
civil_lib +
percent_christian +
as.factor(freedom_house) +
lag(IdealPointDistance) +
as.factor(us_alliance) +
as.factor(over_100_army) +
lag(as.factor(house_maj)) +
# lag(as.factor(senate_maj)) +
lag(as.factor(president)) +
lag(log(trade_openess_percent_gdp)),
index = c("country.name.x", "year"),
model = "random",
random.method = "swar",
data = pd)
pd$ln_econ_pc <- log(pd$economic_aid_per_capita + 0.000001)## Warning in log(pd$economic_aid_per_capita + 1e-06): NaNs producedlag_econ_pc <- plm(ln_econ_pc ~
lag(ln_econ_pc) +
as.factor(isWest) +
civil_lib +
percent_christian +
as.factor(freedom_house) +
lag(IdealPointDistance) +
as.factor(us_alliance) +
as.factor(over_100_army) +
lag(as.factor(house_maj)) +
# lag(as.factor(senate_maj)) +
lag(as.factor(president)) +
lag(log(trade_openess_percent_gdp)),
index = c("country.name.x", "year"),
model = "random",
random.method = "swar",
data = pd)
pd$ln_mil_pc <- log(pd$military_aid_per_capita + 0.000001)
lag_mil_pc <- plm(ln_mil_pc ~
lag(ln_mil_pc) +
as.factor(isWest) +
civil_lib +
percent_christian +
as.factor(freedom_house) +
lag(IdealPointDistance) +
as.factor(us_alliance) +
as.factor(over_100_army) +
lag(as.factor(house_maj)) +
# lag(as.factor(senate_maj)) +
lag(as.factor(president)) +
lag(log(trade_openess_percent_gdp)),
index = c("country.name.x", "year"),
model = "random",
random.method = "swar",
data = pd)
stargazer( random_lag_pc, lag_mil_pc, lag_econ_pc,
type = "text")##
## ==========================================================================
## Dependent variable:
## --------------------------------------
## ln_budget_pc ln_mil_pc ln_econ_pc
## (1) (2) (3)
## --------------------------------------------------------------------------
## lag(ln_budget_pc) 0.855***
## (0.016)
##
## lag(ln_mil_pc) 0.782***
## (0.022)
##
## lag(ln_econ_pc) 0.672***
## (0.029)
##
## as.factor(isWest)1 -0.084 -2.292*** -1.956***
## (0.059) (0.429) (0.310)
##
## civil_lib 0.397** 1.665* 0.380
## (0.158) (0.898) (0.677)
##
## percent_christian -0.0002 -0.003 -0.0005
## (0.0005) (0.003) (0.002)
##
## as.factor(freedom_house)2 0.0003 0.013 0.689***
## (0.051) (0.292) (0.228)
##
## as.factor(freedom_house)3 0.152* 0.126 0.362
## (0.088) (0.510) (0.386)
##
## lag(IdealPointDistance) -0.051* -0.276* 0.191
## (0.028) (0.167) (0.124)
##
## as.factor(us_alliance)1 0.166*** 0.010 -0.275
## (0.041) (0.232) (0.176)
##
## as.factor(over_100_army)1 0.112** 0.376 -0.138
## (0.045) (0.266) (0.203)
##
## lag(as.factor(house_maj))R -0.015 -0.895*** 0.309
## (0.048) (0.285) (0.217)
##
## lag(as.factor(president))Trump 0.231*** -0.106 0.295*
## (0.036) (0.212) (0.161)
##
## lag(log(trade_openess_percent_gdp)) -0.034*** -0.085* -0.153***
## (0.009) (0.050) (0.038)
##
## Constant -0.620*** 0.750 -0.048
## (0.194) (1.113) (0.839)
##
## --------------------------------------------------------------------------
## Observations 870 828 821
## R2 0.826 0.719 0.705
## Adjusted R2 0.824 0.715 0.701
## F Statistic 4,072.837*** 2,087.067*** 1,931.059***
## ==========================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01pd$budget_per_gdp <- pd$budget / pd$gdp_current
pd$ln_budget_per_gdp <- log(pd$budget_per_gdp)
random_lag_gdp <- plm(ln_budget_per_gdp ~
lag(ln_budget_per_gdp) +
as.factor(isWest) +
civil_lib +
us_distance_ln +
percent_christian +
as.factor(freedom_house) +
lag(IdealPointDistance) +
as.factor(us_alliance) +
as.factor(over_100_army) +
lag(as.factor(house_maj)) +
# lag(as.factor(senate_maj)) +
lag(as.factor(president)) +
lag(log(trade_openess_percent_gdp)),
index = c("country.name.x", "year"),
model = "random",
random.method = "swar",
data = pd)
pd$econ_aid_per_gdp <- pd$sum_economic_aid / pd$gdp_current
pd$ln_econ_aid_per_gdp <- log(pd$econ_aid_per_gdp + 0.000001)## Warning in log(pd$econ_aid_per_gdp + 1e-06): NaNs producedlag_econ_gdp <- plm(ln_econ_aid_per_gdp ~
lag(ln_econ_aid_per_gdp) +
as.factor(isWest) +
civil_lib +
us_distance_ln +
percent_christian +
as.factor(freedom_house) +
lag(IdealPointDistance) +
as.factor(us_alliance) +
as.factor(over_100_army) +
lag(as.factor(house_maj)) +
# lag(as.factor(senate_maj)) +
lag(as.factor(president)) +
lag(log(trade_openess_percent_gdp)),
index = c("country.name.x", "year"),
model = "random",
random.method = "swar",
data = pd)
pd$mil_aid_per_gdp <- pd$sum_military_aid / pd$gdp_current
pd$ln_mil_aid_per_gdp <- log(pd$mil_aid_per_gdp + 0.000001)
lag_mil_gdp <- plm(ln_mil_aid_per_gdp ~
lag(ln_mil_aid_per_gdp) +
as.factor(isWest) +
us_distance_ln +
civil_lib +
percent_christian +
as.factor(freedom_house) +
lag(IdealPointDistance) +
as.factor(us_alliance) +
as.factor(over_100_army) +
lag(as.factor(house_maj)) +
# lag(as.factor(senate_maj)) +
lag(as.factor(president)) +
lag(log(trade_openess_percent_gdp)),
index = c("country.name.x", "year"),
model = "random",
random.method = "swar",
data = pd)
stargazer(random_lag_gdp, lag_mil_gdp, lag_econ_gdp,
type = "text")##
## ============================================================================================
## Dependent variable:
## --------------------------------------------------------
## ln_budget_per_gdp ln_mil_aid_per_gdp ln_econ_aid_per_gdp
## (1) (2) (3)
## --------------------------------------------------------------------------------------------
## lag(ln_budget_per_gdp) 0.771***
## (0.018)
##
## lag(ln_mil_aid_per_gdp) 0.870***
## (0.018)
##
## lag(ln_econ_aid_per_gdp) 0.922***
## (0.013)
##
## as.factor(isWest)1 -0.286*** -0.596*** -0.183
## (0.062) (0.189) (0.115)
##
## civil_lib 0.304* 0.904** -0.165
## (0.157) (0.408) (0.265)
##
## us_distance_ln -0.116*** -0.042 -0.171***
## (0.038) (0.101) (0.066)
##
## percent_christian -0.0002 -0.003** -0.0004
## (0.0005) (0.001) (0.001)
##
## as.factor(freedom_house)2 0.149*** 0.193 0.261***
## (0.051) (0.133) (0.092)
##
## as.factor(freedom_house)3 0.226** 0.314 0.070
## (0.088) (0.230) (0.150)
##
## lag(IdealPointDistance) 0.004 -0.087 0.067
## (0.029) (0.075) (0.048)
##
## as.factor(us_alliance)1 0.111*** -0.033 -0.201***
## (0.041) (0.105) (0.070)
##
## as.factor(over_100_army)1 0.084* 0.016 -0.003
## (0.045) (0.120) (0.079)
##
## lag(as.factor(house_maj))R -0.060 -0.406*** -0.003
## (0.048) (0.129) (0.084)
##
## lag(as.factor(president))Trump 0.113*** -0.123 0.069
## (0.036) (0.095) (0.062)
##
## lag(log(trade_openess_percent_gdp)) -0.105*** -0.023 -0.047***
## (0.012) (0.024) (0.017)
##
## Constant -1.282*** -0.648 1.090
## (0.427) (1.094) (0.713)
##
## --------------------------------------------------------------------------------------------
## Observations 867 825 823
## R2 0.902 0.828 0.949
## Adjusted R2 0.900 0.825 0.948
## F Statistic 7,833.929*** 3,908.586*** 15,050.840***
## ============================================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01library(stargazer)
library(lmtest)## Loading required package: zoo##
## Attaching package: 'zoo'## The following objects are masked from 'package:base':
##
## as.Date, as.Date.numericlibrary(sandwich)
# pd$ideal_distance_groups <-
# cut(pd$ideal_point_distance, # Numeric input vector
# 3, # Number or vector of breaks
# labels = c("very similar", "sometimes similar", "not similar"), # Labels for each group
# include.lowest = FALSE, # Whether to include the lowest 'break' or not
# right = TRUE, # Whether the right interval is closed (and the left open) or vice versa
# dig.lab = 3, # Number of digits of the groups if labels = NULL
# ordered_result = FALSE # Whether to order the factor result or not
# ) # Additional arguments
#
# summary(pd$ideal_distance_groups)
#
#
# stargazer(plm(log(pc_budgetcap) ~
# log(gdp_pc) +
# as.factor(temporary_unsc) +
# as.factor(muslim_majority) +
# as.factor(isWest) +
# log(exports) +
# as.factor(us_bases) +
# as.factor(us_invaded) +
# # log(tradeflow_comtrade_d) +
# democracy_vdem.x +
# as.factor(president_party) ,
# data = pd,
# index= c("country_name.x", "year"),
# method = "random",
# random.method = "swar"), type = "text")# pd$exports_pc <- pd$exports / pd$population
#
# pd$imports_pc <- pd$imports / pd$population
# stargazer(plm(log(pc_budgetcap) ~
# lag(log(imports_pc +0.000001)),
# data = pd,
# index= c("country_name.x", "year"),
# method = "random",
# random.method = "swar"),
# plm(log(pc_budgetcap) ~
# lag(log(exports_pc +0.000001)),
# data = pd,
# index= c("country_name.x", "year"),
# method = "random",
# random.method = "swar"),
# plm(log(pc_budgetcap) ~
# lag(log(exports_pc)),
# data = pd,
# index= c("country_name.x", "year"),
# method = "random",
# random.method = "swar"),
# type = "text")