Deepening Our Understanding of the Effects of US Foreign Assistance on Democracy Building, 1990-2004...

Deepening Our Understanding of the Effects of US Foreign

Assistance on Democracy Building, 1990-2004

Steven E. Finkel, University of Pittsburgh (USA) and

Hertie School of Governance (Berlin)

Aníbal Pérez-Liñán, University of Pittsburgh (USA)

Mitchell A. Seligson, Vanderbilt University (USA)

C. Neal Tate, Vanderbilt University (USA)

http://www.pitt.edu/~politics/democracy/democracy.html

Goals of the Project• Phase I (2004-2005):

– Does USAID Democracy Assistance (DG) Matter for Recipient Countries’ Democratic Trajectories Over Time?

– What Are the Impacts of Specific Sub-Sectoral (e.g. Civil Society) and Sub-Sub-Sectoral (e.g. Human Rights) Assistance?

• Phase II (2006-2007):

– Confirm Phase I Findings with 1990-2004 data– Explore Conditions Under Which USAID DG Assistance Matters More

or Less• Country Structural and Political Characteristics, Political Culture• US-Recipient Relations and Other “External” Factors• USAID Funding Strategies/Patterns

– Explore Cumulative or Longer-Range Impacts of USAID DG Assistance– Conduct More Intensive Analyses to Rule out “Endogeneity,” i.e., that

USAID Funds “Democratic Winners”– Explore Negative Finding Concerning Human Rights Assistance

Independent Variables: USAID Democracy Assistance

Based on World-Wide USAID Activities Data Base, over 40,000 Activities Aggregated into:

• Democracy and Governance (DG) Assistance (millions 2000 $, 2-year “Actual Appropriations”)

• Sub-Sector and Sub-Sub-Sector Assistance:– Elections and Political Processes (elections, parties)– Rule of Law (human rights, legal and judicial development)

• Human Rights Programs

– Civil Society (media, civic education, labor unions)• Free Media Programs

– Governance Programs (transparency, decentralization)

Dependent Variables: Democratic Development

• Freedom House (1-13) and Polity IV (-10, +10)

• Five composite indices:– Free Elections (electoral competition, competitive legislative

elections, women’s rights, competitiveness of participation, and democratic accountability)

– Human Rights (political killings, disappearances, torture, political imprisonment, political terror)

– Civil Society (organization of minorities, freedom of assembly, religious freedom, workers’ rights, freedom of movement, and women’s economic rights)

– Free Media (freedom of the press, freedom of speech, and freedom of expression for minorities)

– Governance (perceptions of corruption, conditions for investment, administrative corruption, and bureaucratic quality)

0

200

400

600

800

1000

1200

Mill

ion

s o

f 2

00

0 d

olla

rs

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

Elections Rule of Law Civil Society Governance

Evolution of USAID DG Funding (by Sub-Sectors)

$128 M

G: 11%

CS: 50%

RL: 33%

E: 7%

$1134 M

G: 42%

CS: 25%

RL: 13%

E: 20%

Examples of Democratic Growth Trajectories

0.40

3.70

7.00

10.30

13.60

Fre

ed

om

Ho

us

e In

de

x

1989 1993 1997 2000 2004

YEAR

0.40

3.70

7.00

10.30

13.60

Fre

ed

om

Ho

us

e In

de

x

1989 1993 1997 2000 2004

YEAR

BelarusMali

Hierarchical or Multilevel Growth Models• Two Level (“Hierarchical”) Model

– Level 1: Each Country’s 1990-2004 Trajectory of Democratic Growth, Including Starting Point (“Intercept”) and Rate of Change (“Slope”)

(1) yti= π0i + π1iati + πkivkti + εti

Intercept + Slope*Time + Slope*Covariate + Error

– Level 2: Country Factors that Influence the Magnitude of the Level 1 Intercept and Slope

(2) π0i = Β00 + Β0mXmi + r0i

“Fixed “Population Value + Slope*Country Factor + Error

(3) π1i = Β10 + Β1mXmi + r1i

(4) πki= Βk00

Hierarchical or Multilevel Growth Models• Two Level (“Hierarchical”) Model

– Level 1: Each Country’s 1990-2004 Trajectory of Democratic Growth, Including Starting Point (“Intercept”) and Rate of Change (“Slope”)

(1) yti= π0i + π1iati + πkivkti + εti

Intercept + Slope*Time + Slope*Covariate + Error

– Level 2: Country Factors that Influence the Magnitude of the Level 1 Intercept and Slope

(2) π0i = Β00 + Β0mXmi + r0i

“Fixed” Population Value + Slope*Country Factor + Error

(3) π1i = Β10 + Β1mXmi + r1i

(4) πki= Βk0

The “Baseline Model”• Level 1 Growth Trajectory: Freedom House Index, Polity IV Index

– Country-Specific Intercept (π0i) + Linear Slope (π1i) for Time

• Covariates: Democracy And Other Foreign Assistance– USAID DG – USAID Non-DG– National Endowment for Democracy (NED)– US Assistance Other Than USAID and NED– Regional/Sub-Regional USAID DG & Non-DG Assistance – Other Donor DG and Other Donor Non-DG Assistance

• Covariates: Economic and Political Factors– Annual GDP Growth– Democratic Diffusion (average of democracy in other countries lagged by

one year, weighted by distance to recipient nation)– Political and Social Conflict– State Failure– US Military Assistance (percent of US security assistance)

• Level 2 Control Variables:

– Years Rated “Free,” 1972-1989 (Freedom House)– Pre-1990 USAID Presence– UNDP Human Development– Income Per Capita (thousands $, PPP, CIA World Factbook)– Ethno-Linguistic Fractionalization (0-1, Annett and Fearon Data)– Income Inequality (percent share of income, Top 20%, WB)– Years of State Failure, 1960-1989 (Polity Data)– Population– Land Area of Country

Growth Model for Freedom House and Polity ScoresLevel 1 Freedom House Polity IV

Democracy and Other Assistance

USAID DG (in millions $) .029** .044**

USAID Non-DG (in millions $) .000 .000

National Endowment for Democracy .071 -.279

US Assistance Other Than USAID or NED -.000 .001

Regional-Subregional DG -.018 -.051

Regional-Subregional Non-DG -.010 .017

Other Donor Assistance DG 4.5E-005 .001

Other Donor Assistance Non-DG -6E-007 -.000

Economic and Political Factors

Annual Growth in GDP Per Capita .009** -.004

US Military Assistance Priority -.029 -.025

Democratic Diffusion .181** 1.156**

Social and Political Conflict -.001** -.001

State Failure Indicator -.740** -1.994**

Iraq 2004 -4.509** -10.309**

Note: ** coefficients (in bold red) are significant at p<.05 (two-tailed)

Summary of Baseline Model Results

• USAID-DG Appropriations Affect Overall Level of Democracy, Controlling for Country-Specific Growth Trajectories and Level 1/ Level 2 Control Variables

• Effects Nearly Identical for Freedom House and Polity IV Dependent Variables

• Effects Modest in Absolute and Relative Magnitude

• No Other Aid Variable is Statistically Significant (though measurement deficiencies and possible indirect effects)

• Confirms Phase I (1990-2003) Findings, but Iraq 2004 Exerts Extreme “Leverage,” Necessitating A Separate Control For This Case

Additional Analyses I: Cumulative Effects

10.0

10.5

11.0

11.5

12.0

12.5

1990 1992 1994 1996 1998 2000 2002 2004

Year

Pre

dic

ted

FH

Sco

re

Permanent intervention ($120M, 1993-2004)

No intervention (latent trend)

Cumulative Effect= .020(1+.65+.652+.653…..+.65t)= .020/(1-.65)=.056

Additional Analysis II: The “Endogeneity” Challenge

• Endogeneity: A Country’s Level of Democracy May Cause USAID Appropriations and Not the Reverse.– Countries (e.g., North Korea) May Receive no Funding Because

They Score at the Lowest Level of Democracy.

– Other Countries May Receive Funding Because They are Already Making Democratic Progress

– If True, This Would Produce the Illusion of a Positive Effect from USAID DG to Democracy.

Additional Analysis II: The “Endogeneity” Challenge• Endogeneity: A Country’s Level of Democracy May Cause

USAID Appropriations and Not the Reverse.– Countries (e.g., North Korea) May Receive no Funding Because

They Score at the Lowest Level of Democracy.– Other Countries May Receive Funding Because They are Already

Making Democratic Progress – If True, This Would Produce the Illusion of a Positive Effect from

USAID DG to Democracy.

• Yet– USAID Often Provides Assistance to “Tough Cases” (e.g., Haiti)– At a Certain Level of Democracy, Countries Often “Graduate” (e.g.,

Costa Rica, Botswana, Poland)– In Such Cases, There is a Negative Relationship Between USAID

DG and Democracy

Empirical Refutation of the Endogeneity Challenge

• Statistical Models Controlling for Endogenous DG Assistance Confirm the Significant Effects Found in the Baseline Model

– “Instrumental Variables” Models

• “State Department Priority” (mentions by Secretary or Assistant Secretary of Country in NYT) as Primary Instrument

• Deeper Lags of DG Appropriations used as Additional Instruments in Dynamic Econometric (“Arellano-Bond”) Models

• All Models Pass Statistical Tests Regarding Assumptions of Exogeneity and Relevance of Instruments, and Assumptions About Appropriate Error Term Correlations

– “Fixed Effects” and “First Differences” Models Control for Unobserved Stable Variables, Confirm USAID DG Effects

Under What Conditions Does DG Assistance Matter Most?

• Region

• Socioeconomic Conditions – Human Development– Ethnic and Linguistic Fractionalization– GDP Growth– Income Inequality– Land Area

• Domestic Political Conditions

– Social and Political Conflict– State Failure– Prior Democracy

(Note: variables in red statistically significant )

Under What Conditions Does DG Assistance Matter Most?

• U.S. Relations and International Factors– U.S. Military Assistance Priority– Democratic Diffusion

• USAID Funding Strategies/Patterns– Trends in Investment– Volatility in Investment– Sub-Sectoral Emphases

(Note: variables in red statistically significant)

Statistical Model for Conditional Effects

Level 1: Each Country’s 1990-2004 Democratic Trajectory

yti= π0i + π1iati + πki USAID-DGkti + εti

Intercept + Slope*Time + Slope*USAID-DG +Error

Level 2: Country Factors that Influence the Magnitude of the Level 1 USAID-DG Slope

πki= Βk0 + Βk1X1i + Βk2X2i

where X1, X2, etc. are Socioeconomic, Political, International, and USAID-related Variables from Previous Slides

Implications:– Effects of USAID-DG on Democracy Differ at Each Level of X– Standard Errors and Confidence Intervals Calculated at Each

Level of X– Substantive Interest in the Level of X Where USAID DG is No

Longer Statistically Significant, or Where it Becomes Significant

USAID-DG Effects for Different RegionsCoefficient Std. Error

Democracy Assistance

Baseline Coefficient for USAID DG 0.029** 0.012

Africa 0.052** 0.023

Asia 0.009 0.016

Latin America -0.030 0.020

Eurasia -0.018 0.021

Middle East -0.009 0.014

Latent Democratization Trend

Baseline Slope for Growth Curve 0.073* 0.038

Africa 0.084 0.058

Asia -0.027 0.051


Eurasia -0.254** 0.064

Middle East -0.074 0.050

Initial Level of Democracy (Intercept)

Baseline Intercept 8.640** 0.589

Africa -2.212** 0.962

Asia -2.941** 0.833


Eurasia -1.768* 1.036

Middle East -4.388** 0.834

Note: Baseline coefficients correspond to the average country in Europe and Oceania. Coefficients for regions indicate the distance between the coefficient for the typical country in the region and the average coefficient.

** Significant at p<.05; * Significant at p<.10

USAID DG Effect at Different Levels of Human Development

-.010

.000

.010

.020

.030

.040

.050

.060

.070

0.00 0.20 0.40 0.60 0.80 1.00

Effe

ct o

f US

AID

DG

on

Fre

edom

Hou

se In

dex

Human Development Index

USAID DG Effect at Different Levels of Ethnic and Linguistic Fractionalization

-.010

.000

.010

.020

.030

.040

.050

.060

.070

0.00 0.20 0.40 0.60 0.80 1.00

Effe

ct o

f US

AID

DG

on

Fre

edom

Hou

se In

dex

Ethnic and Linguistic Fractionalization

.018

.042

-.010

.000

.010

.020

.030

.040

.050

.060

.070

0 1

USAID DG Effect at Different Levels of State FailureE

ffect

of U

SA

ID D

G o

n F

reed

om H

ouse

Inde

x

No State Failure

State Failure

USAID DG Effect at Different Levels of US Military Assistance Priority

-.010

.000

.010

.020

.030

.040

.050

.060

.070

0.0 1.0 2.0 3.0 4.0 5.0

Effe

ct o

f US

AID

DG

on

Fre

edom

Hou

se In

dex

% of Annual US Military Assistance

USAID DG Effect at Different Levels of Investment Volatility

-0.08

-0.06

-0.04

-0.02

0.00

0.02

0.04

0.06

0.08

0.00 0.20 0.40 0.60 0.80 1.00

Effe

ct o

f US

AID

DG

on

Fre

edom

Hou

se In

dex

Investment Volatility

Summary: Conditional Effects of USAID Democracy Assistance

• USAID DG Effects are Greater When Countries:– Have Lower Levels of Human Development– Are More Ethnically Divided– Face Major Instances of Political Instability– Are Not Recipients of Large Amounts of US Military Assistance

• US Military Assistance “Explains” the Iraq 2004 Effect (i.e., Iraq 2004 No Longer Significant Once US Military Assistance Priority is Taken into Account)

• USAID DG Effects are Greater for Countries with Less Volatility in DG Investment

Summary of Effects from Sub-Sector and Sub-Sub-Sector Analyses

Sub-Sector and Sub-Sub-Sector Democracy Assistance

Elections HumanRights

CivilSociety

Free Media Governance

Elections 0.092* 0.152* n.s. 0.185** n.s.

Rule of Law (all sub-sub-sectors) n.s. -- n.s. n.s. n.s.

– Human Rights -- -0.664** -- -- --

– Non Human Rights -- n.s. -- -- --

Civil Society (all sub-sub-sectors) n.s. n.s. 0.292** -- n.s.

– Media -- --- -- 0.573** ---

– Non-Media -- --- -- 0.151** ---

Governance 0.060** n.s. 0.098* n.s. 0.070**

Understanding the Impact of Human Rights (HR) Assistance

• First phase of project found human rights assistance – human rights abuse conundrum

• Negative impact of human rights obligations on sub-sectoral human rights outcomes

• Clearly counterintuitive and disturbing• Especially given other positive findings for effect of

DG assistance on democracy growth• Why does it occur?

Can We Explain the HR Assistance – HR Abuse Conundrum?

• Our plan: investigate alternative explanations to demonstrate spuriousness of relationship

• Analytical strategy?– Reexamine "reverse causality" explanation– Reexamine "measurement/reporting error" explanation– Investigate the effects of potential omitted variables – Explore theory/analysis to explain a "genuine relationship"

Testing Reverse Causality Explanation

• Aid goes to countries with records of abuse– tested reciprocal causation models – models gave no support for reverse causality

Testing Measurement/Reporting Error Explanation

• HR aid yields higher reporting of abuse, not higher actual abuse

• Proxies for increased reporting of HR abuses:– Data on press freedom– Data on within country presence IGOs/INGOs (weak data)

• Findings– Press freedom increases respect for HR (encouraging

finding) –does not support reporting error hypothesis. – IGO/NGO associates with increased abuse (a discouraging

finding?) –does support reporting error hypothesis.

Testing Omitted Variables Explanation

• Relationship between HR aid & abuse due to omission of key variables

• Worthy omitted variables to consider: – Formal constitutional structures to protect/promote HR

– Formal and actual judicial independence

• Findings– No formal constitutional provisions predicted respect for HR– Actual judicial independence strongly associated with

respect for HR • a most encouraging finding, given AID rule of law concerns

Testing Genuine Relationship Explanation

• Leaders under pressure to improve their HR performance respond by becoming more repressive

• They feel their grip on power is threatened• Measure “threat” with indicators of events perceived as threatening by

potentially repressive leaders– Organized Nonviolent Protest– Organized Nonviolent Rebellion– Organized Violent Rebellion, and – Civil War

• Findings: Increased threat strongly associated with HR abuse– supports proposition that at least some human rights abuse may be the

result of perceptions of threat by political leaders

Understanding the Impact of Human Rights Assistance Redux: Conclusions

• Despite important findings such as– Rejection of reverse causality hypothesis– Positive effect of Press Freedom on HR respect– (Tentative) negative impact of IGO/NGO growth on HR

respect (support for reporting error hypothesis)– Strong positive effect of actual judicial independence on HR– Strong negative effects of “threat” on HR

• Negative relationship between DG Rule of Law HR assistance and HR respect persists

Political Culture: Does it play a facilitative role?

• First phase of project: made no attempt to investigate if political culture mediates the impact of DG assistance

• Second phase: extensive effort to identify useful cross-national opinion data on culture variables

• Survey data in usable form exists on 60-80 countries for a list of eight cultural variables

• Data drawn from multiple survey sources (see text) including the surveys that USAID helps fund (the LAPOP AmericasBarometer, AfroBarometer).

Working hypothesis:

• Countries with cultures that promote trust and social engagement are ones in which democracy assistance will have a stronger impact.

• Conversely, in countries with less trusting and engaged political cultures, the impact of DG assistance will be attenuated.

Political Culture Variables Chosen

• Interpersonal trust• Support for democracy as form of government• Institutional trust (government, parliament, justice

system)• Satisfaction with democracy• Happiness• Life satisfaction• Interest in politics• Nationalism

Analyses and Results

• Separate analyses of role of each dimension of culture in mediating the impact of U.S. democracy assistance on Freedom House democracy scores over time

• All three showed that culture has positive facilitative effect on DG assistance:– The more democratic a country’s political culture, the stronger

the effects of democracy assistance on democracy scores– DG assistance is most effective when citizens trust one

another, are engaged with politics, and less strongly nationalistic

Conclusions

• USAID’s bold risk in commissioning study: could have found no impact or negative impact of Democracy assistance

• First report: U.S. democracy assistance (1990-2003) had a positive impact on national levels of democracy

• This study: added one (important) year of data and many new indicators, enriching analysis

• New effort found results similar to those in first study, controlling for impact of 2004 spending in Iraq, i.e.,

• When USAID expends funds to promote democracy, the effort has same degree of impact as in first study

Conclusions (continued)

• New or enhanced findings:

• No evidence of “endogeneity” (reverse causation)

• No regional effects except for Africa (larger impact)

• Countries with greatest need benefit most from given amount of USAID DG assistance

• Democracy assistance is less effective when the U.S. provides larger amounts of military assistance– Warrants further investigation– Appears to explain the “Iraq 2004 Effect”

• Limited evidence suggests that USAID-DG is less effective if investment is more volatile

Conclusions (continued - 2)

• New or enhanced findings (continued)

– The more democratic a country’s political culture, the stronger the effect of U.S. DG aid on democracy scores

– New study unable to “wash out” the only important negative effect of aid on democracy, the impact in human rights area

• This counter-intuitive finding remains a puzzle

• Positive impact of USAID on democracy clear:

– The 15 years of data we have analyzed here provide a robust basis for drawing the conclusion that DG assistance in the post-Cold War period has worked.

– http://www.pitt.edu/~politics/democracy/democracy.html

Deepening Our Understanding of the Effects of US Foreign Assistance on Democracy Building, 1990-2004...

Documents

Transcript of Deepening Our Understanding of the Effects of US Foreign Assistance on Democracy Building, 1990-2004...