Executive Summary
This study explores how Western countries respond when aid recipient countries choose alternatives to Western foreign aid. A relatively new alternative is South-South development finance, where funding is provided by emerging market and Global South countries. This alternative differs from Western funding in several ways.
Firstly, China presents its development funding as a "no strings attached" alternative to Western funds. This means that China commits to not interfering in the domestic politics of the countries it supports. In contrast, Western donors typically exert pressure on aid recipients to meet certain criteria, such as promoting democracy, protecting human rights, and combating corruption.
Secondly, the majority of Chinese development funding comes in the form of loans with interest, whereas Western funding is often provided as grants (where repayment is not expected) or concessional loans (with very low or zero interest rates).
To summarize, Chinese loans have higher economic cost, while Western aid carries a higher political cost.
My argument is that Chinese development finance, particularly through the Belt and Road Initiative (BRI), allows recipient leaders to bypass the economic pressures that Western donors use to promote their political aims abroad. When Western donors lose their economic leverage, they intensify their focus on funding projects that promote political reforms, even if the effectiveness of these efforts is questionable. BRI agreements cause Western donors to shift from using aid and the threat of aid withdrawal to influence leaders to using aid to promote political liberalization through developmental means.
To investigate this, I apply a recently-developed method called doubly-robust difference-in-difference modeling, which accounts for multiple treatment periods. The aim is to analyze how the composition of Western foreign aid to different sectors changes after a recipient country's leader signs a Belt and Road Initiative Memorandum of Understanding with China. The study uses data from 2013 to 2018 and examines the impact of 80 countries signing their first BRI agreements on Western foreign aid.
The findings reveal that Western donors do not increase their economic sector aid, which can be used flexibly by recipient country leaders, to compete with China. Nor do they decrease this type of aid to punish leaders. Instead, Western donors increase their support for the governance sector, promoting projects related to democracy, human rights, and anti-corruption. This effect is more prominent in countries with high levels of severe human rights violations. When geopolitical competition makes it impossible for Western donors to use aid as a source of economic coercion, they increasingly use political liberalization as a strategy to reduce state violence.
Western donors maintain their support for human rights and democracy even when faced with great power competition from China. However, a qualitative case study focusing on the actions of the United States and China in Burma (Myanmar) suggests that Western donors continue to use aid to promote democracy, even in situations where it may further destabilize countries.
Firstly, China presents its development funding as a "no strings attached" alternative to Western funds. This means that China commits to not interfering in the domestic politics of the countries it supports. In contrast, Western donors typically exert pressure on aid recipients to meet certain criteria, such as promoting democracy, protecting human rights, and combating corruption.
Secondly, the majority of Chinese development funding comes in the form of loans with interest, whereas Western funding is often provided as grants (where repayment is not expected) or concessional loans (with very low or zero interest rates).
To summarize, Chinese loans have higher economic cost, while Western aid carries a higher political cost.
My argument is that Chinese development finance, particularly through the Belt and Road Initiative (BRI), allows recipient leaders to bypass the economic pressures that Western donors use to promote their political aims abroad. When Western donors lose their economic leverage, they intensify their focus on funding projects that promote political reforms, even if the effectiveness of these efforts is questionable. BRI agreements cause Western donors to shift from using aid and the threat of aid withdrawal to influence leaders to using aid to promote political liberalization through developmental means.
To investigate this, I apply a recently-developed method called doubly-robust difference-in-difference modeling, which accounts for multiple treatment periods. The aim is to analyze how the composition of Western foreign aid to different sectors changes after a recipient country's leader signs a Belt and Road Initiative Memorandum of Understanding with China. The study uses data from 2013 to 2018 and examines the impact of 80 countries signing their first BRI agreements on Western foreign aid.
The findings reveal that Western donors do not increase their economic sector aid, which can be used flexibly by recipient country leaders, to compete with China. Nor do they decrease this type of aid to punish leaders. Instead, Western donors increase their support for the governance sector, promoting projects related to democracy, human rights, and anti-corruption. This effect is more prominent in countries with high levels of severe human rights violations. When geopolitical competition makes it impossible for Western donors to use aid as a source of economic coercion, they increasingly use political liberalization as a strategy to reduce state violence.
Western donors maintain their support for human rights and democracy even when faced with great power competition from China. However, a qualitative case study focusing on the actions of the United States and China in Burma (Myanmar) suggests that Western donors continue to use aid to promote democracy, even in situations where it may further destabilize countries.
Demonstrating why the BRI matters
I include a visualization below of the count of project commitments (bar chart, left axis) and the dollar value of commitments (line chart, right axis) that China has made toward economic development projects in foreign countries over time. The scale of China's involvement in foreign development finance has grown significantly over this time and now either rivals or dwarfs Western donors' foreign assistance.
Preparing the data
This study uses most of the same data as the "State Violence and Foreign Aid" project, but limits the time period to 2013-2018, when the BRI became active (initially as the "One Belt One Road" and similar agreements, which I include in "BRI agreements"). See the "State Violence and Foreign Aid" project for more information on the data selection, scope, limitations, and modifications.
For this project, I also collected information about the initial BRI agreements signed between China and developing country leaders. Countries that had never signed a BRI agreement are coded as 0. The first year a country signs a BRI agreement is coded as a 1, which is carried forward so that if a country had ever signed a BRI agreement, it would always be coded as a 1.
Identifying observable implications of the theory
Theory in brief
Theoretically, I am interested in understanding how Western donor countries are strategically responding to growing geopolitical competition. Furthermore, I am interested in whether this geopolitical competition undermines donors' ability to respond to severe human rights abuses using foreign aid, as it did during the Cold War.
Coercive strategy is more likely to be effective if donors can coordinate their strategies. If all donor countries are able to coordinate their coercive strategies, then the threat of decreased aid for non-compliance with human rights norms, and the promise of increased aid for improvements, may be substantial. If only one donor out of many uses coercive strategy, then the pain inflicted by aid cuts on a violent recipient state will be minimal (assuming that the donor does not provide a large share of the recipient's aid). The greatest challenge to the ability of coercive threats to influence recipients is when one donor is willing to increase its economic sector aid in response to another donor decreasing its aid. By offsetting another donor's punishment strategy, this action renders coercion powerless.
In contrast, catalytic strategy relies less on coordination and may benefit from a larger set of donors with more diverse interests, so long as the set of donors is broadly interested in promoting human rights. Catalytic strategy benefits from a broader pool of democratic donors if those donors pool their resources or pursue governance improvements in a specialized manner.
This study investigates the rise of Chinese foreign assistance to answer several questions about Western donors' strategies for promoting human rights: Have OECD donors continued to use foreign aid to promote human rights, since violent states can simply turn to China for their development finance needs? Is there any indication that donors increase their economic sector aid to counter Chinese influence? To what extent does catalytic strategy substitute for coercive strategy, and to what extent does catalytic strategy rely on underlying, unobserved coercive threats to secure approval from recipient countries?
Aligning the analysis to the theory
Want to know how foreign aid changes when recipient country leaders can credibly demonstrate that they are willing and able to circumvent Western economic pressure.
- Interested in the effect of signing a BRI agreement on western donors' foreign aid strategy.
- Therefore, the goal is to establish causality. Strong preference for causal inference methods over using descriptive or inferential statistics to identify correlations.
- The theory is more concerned with how Western donors respond when recipient leaders decide to pursue outside options for development finance, allowing them to offset Western pressures that use aid as leverage to gain policy concessions.
- Therefore, we are not interested in finding the average treatment effect (ATE)
- Instead, we are interested in finding the average treatment effect on the treated (ATT).
This presents several challenges:
There's no way to turn this into an experiment. Relying on observational data.
- In an ideal experiment, we would randomly assign aid recipient countries to receive Chinese development finance (or not), ensure that these recipient countries are balanced on important covariates, and compare the aid commitments that Western donor countries make between the treatment and control groups.
- Lacking this level of control over world events, the difference-in-difference approach is attractive.
Recipient leaders signed BRI agreements in different years, and over time nearly all countries that are eligible to receive foreign aid have signed these agreements.
And leaves open a few options for estimators:
- Ordinary least squares (logistic regression)
- Two-way fixed effects (TWFE) panel models
- The doubly-robust difference-in-difference model with multiple treatment periods (Callaway and Sant'Anna 2021)
Visualizing variation in year of BRI agreement
After collecting data on the year that countries signed their first BRI agreements, I used Tableau to visualize the variation in BRI agreement timing. This is sufficient variation to warrant using a difference-in-difference with multiple time periods design, and suggests that the counterfactual should be based on not-yet-treated observations. By 2018, there were not enough "never treated" countries to provide a meaningful counterfactual.
Conducting event study using R
# Set working directory
setwd("D:/replication/Chapter 4/")
library(readstata13)
library(ggplot2)
# Brantly Callaway and Pedro Sant’Anna “Difference-in-differences with multiple
# time periods”. Journal of Econometrics, Forthcoming, December 2020.
library(did)
# Load the data
data <- read.dta("analysis_all_donors_1973-2020_with_bri.dta")
# Filter the data
data <- data %>%
filter(year > 2011,
year < 2019,
dac_donor == 1)
data$bri_group <- ifelse(data$bri_group >= 2019, 0, data$bri_group)
# Generate high_sv variable
data$high_sv <- 0
data$high_sv[data$recipientname %in% c("Afghanistan", "Angola", "Azerbaijan", "Bangladesh", "Brazil",
"Burundi", "Cambodia", "Cameroon", "Central African Republic", "Chad",
"China (People's Republic of)", "Colombia", "Congo", "Cuba", "Côte d'Ivoire",
"Democratic People's Republic of Korea", "Democratic Republic of the Congo",
"Dominican Republic", "Egypt", "Eritrea", "Ethiopia", "Gambia", "Honduras",
"India", "Indonesia", "Iran", "Iraq", "Jamaica", "Kazakhstan", "Kenya",
"Kyrgyzstan", "Libya", "Madagascar", "Mali", "Mexico", "Myanmar", "Nigeria",
"Pakistan", "Philippines", "Rwanda", "Somalia", "South Africa", "South Sudan",
"Sri Lanka", "Sudan", "Syrian Arab Republic", "Tajikistan", "Thailand",
"Turkey", "Turkmenistan", "Uganda", "Ukraine", "Uzbekistan", "Venezuela",
"Viet Nam", "Yemen", "Zimbabwe")] <- 1
data <- data %>%
filter(year > 2006)
# DiD estimation using Callaway and Sant'Anna method (inverse probability weighting)
model <- did(data = data, formula = csdid ~ lneconaidpc + repression_L1 + lnexport_L1 + xdem_L1 + lnpop_L1 +
lngdp_L1 + statecap_L1 + highdissent_L1 + highterror_L1, treatment = "bri_group",
group = "dyad_id", time = "year", unit.variable = "dyad_id", time.variable = "year")
summary(model)
did_estimate <- as.data.frame(effect(model))
ggdid(did_estimate$time, did_estimate$eff.est, type = "b", main = "Effect of Signing BRI MoU on OECD Economic Aid Commitments")
# DiD estimation using Callaway and Sant'Anna method (inverse probability weighting)
model <- did(data = data, formula = csdid ~ lneconaidpc + repression_L1 + lnexport_L1 + xdem_L1 + lnpop_L1 +
lngdp_L1 + statecap_L1 + highdissent_L1 + highterror_L1, treatment = "bri_group",
group = "dyad_id", time = "year", unit.variable = "dyad_id", time.variable = "year")
summary(model)
did_estimate <- as.data.frame(effect(model))
ggdid(did_estimate$time, did_estimate$eff.est, type = "b", main = "Effect of Signing BRI MoU on OECD Economic Aid Commitments")
# Filter the data to only include high state violence recipients
data <- data %>%
filter(year > 2011,
year < 2019,
high_sv == 1)
# DiD estimation using Callaway and Sant'Anna method (inverse probability weighting)
model <- did(data = data, formula = csdid ~ lneconaidpc + repression_L1 + lnexport_L1 + xdem_L1 + lnpop_L1 +
lngdp_L1 + statecap_L1 + highdissent_L1 + highterror_L1, treatment = "bri_group",
group = "dyad_id", time = "year", unit.variable = "dyad_id", time.variable = "year")
summary(model)
did_estimate <- as.data.frame(effect(model))
ggdid(did_estimate$time, did_estimate$eff.est, type = "b", main = "Effect of Signing BRI MoU on OECD Economic Aid Commitments for recipients with high state violence")
# DiD estimation using Callaway and Sant'Anna method (inverse probability weighting)
model <- did(data = data, formula = csdid ~ lngovaidpc + repression_L1 + lnexport_L1 + xdem_L1 + lnpop_L1 +
lngdp_L1 + statecap_L1 + highdissent_L1 + highterror_L1, treatment = "bri_group",
group = "dyad_id", time = "year", unit.variable = "dyad_id", time.variable = "year")
summary(model)
did_estimate <- as.data.frame(effect(model))
ggdid(did_estimate$time, did_estimate$eff.est, type = "b", main = "Effect of Signing BRI MoU on OECD Governance Aid Commitments for recipients with high state violence")Conducting event study using Stata
* Created using Stata MP/14.1
clear all
set more off
cd "D:\replication\Chapter 4\"
use "analysis_all_donors_1973-2020_with_bri.dta", clear
keep if year > 2011 & year < 2019 & dac_donor == 1
replace bri_group = 0 if bri_group >= 2019
gen high_sv = 0
local countries "Afghanistan Angola Azerbaijan Bangladesh Brazil Burundi Cambodia Cameroon Central African Republic Chad China (People's Republic of) Colombia Congo Cuba Côte d'Ivoire Democratic People's Republic of Korea Democratic Republic of the Congo Dominican Republic Egypt Eritrea Ethiopia Gambia Honduras India Indonesia Iran Iraq Jamaica Kazakhstan Kenya Kyrgyzstan Libya Madagascar Mali Mexico Myanmar Nigeria Pakistan Philippines Rwanda Somalia South Africa South Sudan Sri Lanka Sudan Syrian Arab Republic Tajikistan Thailand Turkey Turkmenistan Uganda Ukraine Uzbekistan Venezuela Viet Nam Yemen Zimbabwe"
foreach country in `countries' {
replace high_sv = 1 if recipientname == "`country'"
}
keep if year > 2006
xtset dyad_id year
csdid lneconaidpc repression_L1 lnexport_L1 xdem_L1 lnpop_L1 lngdp_L1 statecap_L1 highdissent_L1 highterror_L1, ivar(dyad_id) time(year) gvar(bri_group) notyet method(stdipw)
estat all
estat event
csdid_plot, title("Effect of Signing BRI MoU on OECD Economic Aid Commitments")
xtset dyad_id year
csdid lngovaidpc lnexport_L1 xdem_L1 lnpop_L1 lngdp_L1 statecap_L1 highdissent_L1 highterror_L1, ivar(dyad_id) time(year) gvar(bri_group) notyet method(dripw)
estat all
estat event
csdid_plot, title("Effect of Signing BRI MoU on OECD Governance Aid Commitments")
keep if high_sv == 1
xtset dyad_id year
csdid lneconaidpc repression_L1 lnexport_L1 xdem_L1 lnpop_L1 lngdp_L1 statecap_L1 highdissent_L1 highterror_L1, ivar(dyad_id) time(year) gvar(bri_group) notyet method(stdipw)
estat all
estat event
csdid_plot, title("Effect of Signing BRI MoU on OECD Economic Aid Commitments (high state violence recipients)")
xtset dyad_id year
csdid lngovaidpc lnexport_L1 xdem_L1 lnpop_L1 lngdp_L1 statecap_L1 highdissent_L1 highterror_L1, ivar(dyad_id) time(year) gvar(bri_group) notyet method(dripw)
estat all
estat event
csdid_plot, title("Effect of Signing BRI MoU on OECD Governance Aid Commitments (high state violence recipients)")
Event study graphs
Average treatment effect on the treated (ATT) with Governance Aid dependent variable
The first event study graph shows the ATT for all recipients. The second limits the analysis to recipients with high levels of state violence. The subgroup analysis demonstrates that the ATT is being driven by recipients with high levels of state violence, demonstrating that donors are doubling down on their governance sector projects when BRI agreements are signed by more violent recipient countries.
Average treatment effect on the treated (ATT) with Economic Aid dependent variable
The event study graph shows the ATT for all recipients. There is no evidence that OECD donors respond to BRI agreements by trying to outbid or compete with China. There is simply no indication that donors' economic sector aid changes significantly after a recipient country signs a BRI agreement.
Robustness checks for the influence of outliers
It is possible that the relationship is driven by outliers. This section replicates the above high state violence analysis, but drops the United States as a donor, and Venezuela, Afghanistan, and Iraq as recipients. This check determines whether the relationship holds when these outliers are omitted. Eliminating these outliers slightly decreases the effect size for the governance sector dependent variable, but does not undermine the statistical or substantive results.
