Government Distortion in Independently Owned Media: Evidence from U.S. Cold War News Coverage of Human Rights

NBER WORKING PAPER SERIES

GOVERNMENT DISTORTION IN INDEPENDENTLY OWNED MEDIA:

EVIDENCE FROM U.S. COLD WAR NEWS COVERAGE OF HUMAN RIGHTS

Nancy Qian

David Yanagizawa-Drott

Working Paper 15738

http://www.nber.org/papers/w15738

NATIONAL BUREAU OF ECONOMIC RESEARCH

1050 Massachusetts Avenue

Cambridge, MA 02138

February 2010

This paper supersedes the previous version entitled, “Watchdog and Lapdog...”. We are indebted to

Matthew Gentzkow, Mikhail Golosov and David Stromberg for their many thoughtful comments;

Abhijit Banerjee, Stefano DellaVigna, Raquel Fernandez, Dean Karlan, Brian Knight, Michael Kremer,

Justin Lahart, Suresh Naidu, Nathan Nunn, Torsten Persson, Jesse Shapiro, Jakob Svensson and Chris

Udry for their insights; and the seminar participants at Stanford University, Yale University, New

York University, Boston University, the University College of London, Stockholm University IIES,Warwick

University, Universitat de Pompeu Fabra, Paris School of Economics, Universitet du Toulouse, McGill

University, NBER Summer Institute Political Economy, BEROC, BREAD CIPREE and NEUDC

for useful comments; and Carl Brinton and Aletheia Donald for invaluable research assistance. All

mistakes are our own. Comments or suggestions are very welcome. The views expressed herein are

those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.

NBER working papers are circulated for discussion and comment purposes. They have not been peer-

reviewed or been subject to the review by the NBER Board of Directors that accompanies official

NBER publications.

to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including

Government Distortion in Independently Owned Media: Evidence from U.S. Cold War News

Coverage of Human Rights

Nancy Qian and David Yanagizawa-Drott

NBER Working Paper No. 15738

February 2010, Revised September 2013

JEL No. L82,P16

ABSTRACT

This paper investigates the extent to which strategic objectives of the U.S. government influenced

news coverage during the Cold War. We establish two relationships: 1) strategic objectives of the

U.S. government cause the State Department to under-report human rights violations of strategic allies;

and 2) these objectives reduce news coverage of human rights abuses for strategic allies in six U.S.

national newspapers. To establish causality, we exploit plausibly exogenous variation in a country's

strategic value to the U.S. from the interaction of its political alliance to the U.S. and membership

on the United Nations Security Council. In addition to the main results, we are able to provide qualitative

evidence and indirect quantitative evidence to shed light on the mechanisms underlying the reduced

form effects.

Nancy Qian

Department of Economics

Yale University

27 Hillhouse Avenue

New Haven, CT 06520-8269

and NBER

[email protected]

David Yanagizawa-Drott

79 JFK Street

Cambridge, 02138 MA

USA

[email protected]

1 Introduction

Mass media plays a powerful role in society. It reaches an immense audience, and its content

can aﬀect a wide range of outcomes, including political behavior such as voting.

In democratic

regimes such as the United States, the importance of the media is reﬂected in its being called the

fourth estate, which is supposed to report on the activities of the government in the interest of the

public and act as a “watchdog” of democracy. However, the ability of the media to perform its

prescribed role has come under question as observers point to an increasing number of instances

when media content is distorted by the government. Numerous books written by political scientists

and former journalists voice this concern. Prominent examples include Bennett, Lawrence, and

Livingston (2008), Cook (1998) and Thomas (2006).

The examples of government distortion come

from many diﬀerent time periods, including the war with Vietnam during the 1960s, interventions in

Central America during the 1980s, as well as the recent war with Iraq that began in 2003 (Bennett,

Lawrence, and Livingston, 2008). Bennett, Lawrence, and Livingston (2008, p. 8) summarizes

the motivation behind these concerns: “The democratic role of the press is deﬁned.. by those

moments when government deception or incompetence compels journalists to ﬁnd and bring credible

challenges to public attention and hold rulers accountable... This accountability function of the

U.S. press has been weakened in the contemporary era, and its standing is sorely in need of greater

examination”. Similarly, in her textbook, Mass Media and American Politics, Graber (2006) urges

a closer examination of the political economy of media with a quote from Joseph Pulitzer, “An able,

disinterested, public-spirited press, with trained intelligence to know the right and courage to do

it, can preserve that public virtue without which popular government is a sham and a mockery...

The power to mould the future of the Republic will be in the hands of the journalists of future

generations”, to which Graber follows with, “How well is the U.S. press meeting [this challenge]?”

Graber (2006, p.20).

This paper attempts to make progress on this important question by assessing the extent of

Recent studies have shown that media can aﬀect voting behavior (e.g., Prat and Stromberg, 2005; Gentzkow,

2006; DellaVigna and Kaplan, 2007; Chiang and Knight, 2011; and Enikolopov and Zhuravskaya, 2011), other

political behavior (Paluck, 2008; Gerber, Karlan, and Bergan, 2009; Olken, 2009), and social outcomes such as

literacy (Gentzkow and Shapiro, 2008a) female empowerment (Jensen and Oster, 2009) and fertility (La Ferrara,

Chong, and Duryea, 2008).

See the works referenced in Bennett, Lawrence, and Livingston (2008) and Cook (1998) for the large body of

work about the media and the U.S. government from media and political science scholars.

The quote from Pulitzer was originally printed in the North American Review (1904).

government distortion of the news in the United States. Our study aims to determine whether the

anecdotal and case evidence on government manipulation reﬂects isolated incidents or whether they

reﬂect systematic distortion that could be a symptom of deeper and more fundamental concerns.

In other words, we ask whether a democratic government can systematically distort news coverage

from independently owned outlets. With the exception of Besley and Prat (2006), this question has

not yet been rigorously studied.

For practical reasons that we discuss later, we focus on human

rights news coverage during the latter period of the Cold War.

Our study proceeds in three steps. First, to motivate our investigation, we document that the

U.S. government often attempted to manipulate news coverage of human rights practices of their

political allies during the Cold War. We rely on qualitative evidence from political scientists as well

as internal government memos that explicitly state government objectives and tactics for a large

number of cases. These memos, which are typically not available to the public, were declassiﬁed

as part of the Iran-Contra investigation. The availability of rich case evidence is the one of the

advantages of focusing on the Cold War era.

Second, we combine the recent theories of endogenous news coverage developed by Prat and

Stromberg (2005, 2011) and Stromberg (1999, 2004a, 2004b) with a theory of media manipulation

by Besley and Prat (2006) to develop a framework for understanding government manipulation in

our context. In our model, domestic voters care in part about the foreign policy pursued by the

U.S. government. Voters cannot directly and fully evaluate the foreign policy (preferences) of the

incumbent and partly base their inferences on the behavior of allied foreign countries that vote

with the United States in the United Nations (UN). News reports about human rights violations of

allies serve as indicators of U.S. foreign policy and aﬀect the probability that the incumbent U.S.

government will be voted out of oﬃce. We assume that “worse” countries are more likely to commit

human rights violations and are more likely to vote with the United States if the U.S. government’s

foreign policy is “bad”. U.S. voters observe voting behavior and read about human rights violations

to make their inferences about the U.S. government’s type. There are two groups of voters. The ﬁrst

group has broad interests, reads the news about all foreign countries and make inferences based on

the behavior of all countries. The second group is only interested in the countries that are currently

A recent working paper by Gentzkow, Petek, Shapiro, and Sinkinson (2012) examines the historical U.S. context.

We discuss this study more later in the introduction.

on the UNSC. That voters infer the quality of government policies from news reports and the insight

that newspaper coverage aﬀects the posterior beliefs of the voters about the quality of those policies

is very similar to the framework in Durante and Knight (2012).

It follows that obtaining a seat on the Security Council generates two opposing eﬀects on U.S.

news coverage of the human rights practices of foreign countries: i) the demand eﬀect, an increase

in coverage when countries become Council members because more readers are interested; and ii)

the distortion eﬀect, a reduction in coverage due to the incentives of the government to manipulate

the media. Since there are many fewer countries on the Council than not on the Council (e.g., in

the UN General Assembly), it is cheaper for the U.S. government to manipulate public opinion by

suppressing news about Security Council countries than non-Security Council countries. Moreover,

we show that the closer the foreign country is aligned with the United States, the more severe the

distortion on news coverage. Our model implies that if there is government distortion, then we

should observe that the increase in news coverage that occurs with Council membership is declining

in magnitude with the level of alliance. For strongly allied countries, Council membership can

actually reduce the amount of news coverage.

The ﬁnal and most important step in our paper is to estimate the diﬀerential relationship between

Council membership and U.S. news coverage for foreign countries of diﬀerent levels of alliance with

the United States during the Cold War. We follow an earlier paper, Qian and Yanagizawa-Drott

(2009) and proxy for alliance with the degree to which a country votes with the United States in the

UN General Assembly on issues for which the United States votes in opposition to the Soviet Union

during the end of the Cold War.

Data on news coverage is collected from the text-analysis of ﬁve

large U.S. newspapers (New York Times, Wall Street Journal, Washington Post, Chicago Tribune

and Los Angeles Times).

We examine the interaction eﬀect of a time-invariant measure of Cold War alliance with the

U.S. and a time-varying measure of Council membership, while controlling for year and country

ﬁxed eﬀects. Country ﬁxed eﬀects control for all time-invariant characteristics across countries

such as cultural aﬃnity with the United States, which can aﬀect the degree of alliance, Council

membership and/or news coverage. Year ﬁxed eﬀects control for all changes over time that inﬂuence

Qian and Yanagizawa-Drott (2009) studies the relationship between political alliance with the United States and

human rights abuses reported by the U.S. State Department and Amnesty International. This earlier work does not

examine media news coverage or the United Nations Security Council.

all countries equally. The coeﬃcient of the interaction term reveals the degree to which Council

membership diﬀerentially inﬂuences the news coverage of countries with higher levels of alliance and

the coeﬃcient of the uninteracted Council membership term reveals the eﬀect of Council membership

for the least allied countries with the United States in our sample. In the context of our framework,

ﬁnding a negative interaction would be consistent with the presence of government distortion, and a

positive coeﬃcient for the uninteracted Council membership variable would be consistent with the

presence of the demand eﬀect. These two forces are not mutually exclusive.

The results indicate that news coverage of human rights abuses of foreign countries during

the Cold War was inﬂuenced by both government distortion and reader demand. Taken literally,

our estimates imply that Council membership for a country that is on the 75th percentile of the

distribution of U.S. alliance (e.g., Ecuador, Egypt) will result in approximately 40% fewer news

articles about human rights abuses than Council membership for a country that is on the 25th

percentile of the distribution of alliance (e.g., Mali, the Republic of Congo).

Our results are consistent with U.S. news coverage being partly determined by government

distortions and partly driven by reader demand. This interpretation relies on the assumption

that there were no other forces that were simultaneously correlated with news coverage, Council

membership and political alliance with the United States. We do not take this as given and provide

a large body of evidence in support of our interpretation. For example, we show that the timing of

our reduced form eﬀects corresponds with entry onto and exit from the Council. We also show that

our estimates are robust to controlling for a large number of potentially confounding factors such

as the interactions of the full vector of year ﬁxed eﬀects with variables such as Council membership

and U.S. alliance. See the section 6 for a detailed discussion.

In addition, we provide several pieces of evidence to support our interpretation of the main

results. First, we show that the response of annual U.S. State Department reports of human

rights abuses of foreign countries to the interaction of alliance with the U.S. government and UNSC

membership is very similar to that of news coverage, which is consistent with our interpretation that

the interaction eﬀect captures the inﬂuence of U.S. government manipulation. In contrast, reports

from Amnesty International, a non-government organization, do not respond to the interaction of

UNSC and U.S. alliance. This goes against the alternative explanation that our results reﬂect

changes in actual human rights practice. Second, we show that the interaction term of U.S. alliance

and UNSC membership has no eﬀect on a large number of institutional measures that may be

correlated with human rights practices. This again goes against the alternative explanation that

the results on news coverage and U.S. State Department reports are driven by changes in actual

human rights behavior. Fourth, when we estimate the interaction eﬀect for each newspaper, we ﬁnd

that there is no correlation between the degree of distortion and reader preferences across papers.

While this should be interpreted very cautiously because there are only ﬁve newspapers, the results

are consistent with our interpretation that the interaction eﬀects do not capture reader preferences.

In contrast, we ﬁnd that the magnitude of distortion is positively correlated with newspaper quality

and reputation, which is straightforward to reconcile with our framework and that of Besley and

Prat (2006). Finally, we show that Council membership has no eﬀect on news coverage after the

Cold War for any country. Amongst other explanations, this is consistent with the belief that the

U.S. government’s motivation for manipulating news of Cold War allies declined with the end of

the Cold War.

For policymakers, our results conﬁrm the case evidence that systematic government-driven dis-

tortions can exist for independently owned and highly competitive media outlets in a democratic

regime. At the same time, one may view the fact that our eﬀects dissipate with the end of the Cold

War as reassuring – i.e., it may be diﬃcult for systematic manipulation to last indeﬁnitely in such

contexts. We speculate on this more in the conclusion.

In our focus on the government’s inﬂuence of media coverage, our study is most closely related

to Besley and Prat (2006). It also builds directly on the pioneering work of Besley and Prat (2006),

Durante and Knight (2012), Prat and Stromberg (2005, 2011) and Stromberg (1999, 2004a, 2004b)

by adapting the frameworks developed in these papers and applying them to a novel empirical

context. To the best of our knowledge, our paper is the ﬁrst to provide rigorous evidence that

government distortion has systematically existed in the United States. In doing so, we add to

the important empirical literature on the determinants of news coverage. Our study diﬀers from

previous studies in focusing on government-driven distortions in a democratic regime. In this sense,

we are related to two recent studies about the eﬀect of partisan-controlled media in Italy under

Berlusconi (Durante and Knight, 2012) and the United States during 1869-1928 (Gentzkow, Petek,

Shapiro, and Sinkinson, 2012). The former study ﬁnds evidence of government inﬂuence on the

media in the modern Italian context, while the latter study ﬁnds that the government in power has

little inﬂuence over news composition in the historical U.S. context.

Second, we add to the small but growing number of political economy studies that explore the

causes and consequences of U.S. government foreign policy. In our focus on the Cold War era, our

study is most closely related to Dube, Kaplan, and Naidu’s (2011) study of U.S. covert actions on

U.S. ﬁrm stock prices, and Berger, Easterly, and Satyanath’s (2009) and Berger, Easterly, Nunn,

and Satyanath’s (2010) studies of U.S. Cold War policies on trade. It broadens the scope of this

literature by examining the eﬀect of U.S. foreign policy on the American public. Finally, our use of

United Nations Security Council membership as a proxy for a country’s importance to the United

States borrow’s from Kuziemko and Werker’s (2006) insight that Council membership plays an

important role in the behavior of the U.S. government. Similarly, our use of voting in the UN

General Assembly as a proxy of U.S. alliance follows Alesina and Dollar (2000), which ﬁnd that

voting with the U.S. in the General Assembly is positively correlated with U.S. foreign aid receipts.

This paper is organized as follows. Section 2 discusses the background. Section 3 presents the

conceptual framework. Section 4 presents the empirical strategy. Section 5 describes the data.

Section 6 presents the main empirical results. Section 7 presents additional, more speculative,

results. Section 8 oﬀers concluding remarks.

2 Background

2.1 “White Propaganda” During the Cold War

The main period of our study, 1976-1988, was characterized by a commitment to ﬁght commu-

nism on the part of the American government, which climaxed during the Reagan administration

(1980-88). As with all of the Cold War, rivalry between the two superpowers was expressed through

military coalitions, propaganda and proxy wars (e.g., the Soviet war in Afghanistan 1979-89). The

Cold War ended during 1989-91 when the Berlin Wall fell and the U.S.S.R. dissolved. For the pur-

poses of our paper, we loosely interpret 1989 as the end of the Cold War. At this time, the strenuous

competition between the United States and the U.S.S.R. for the alliance of smaller countries ended.

An important feature of the Cold War in the United States was the focus on the superior

morality of the West. The U.S. government and news media often described its allies as “good”

Kuziemko and Werker (2006) ﬁnds that Council membership in years that are strategically important for the

U.S. government results in higher U.S. aid. They cleverly measure importance with the number of articles in the

New York Times about the Security Council.

and the Eastern Bloc and its allies as “evil”. Recently declassiﬁed ﬁles stored in the U.S. National

Security Archives document both the method and the motives for the U.S. government to inﬂuence

the press coverage of the human rights practices of its political allies.

The government needed public support for its political actions, which included public approval

of its political allies. Given the focus on morality, it followed that U.S. allies should have better

human rights abuses than Eastern Bloc allies.

For the most part, U.S. government support for its

Cold War allies with poor human rights abuses ended with the Cold War. Internal memos show

that the executive branch believed that one of the ways to shape public opinion against opponents

was to exaggerate human rights abuses in those countries and emphasize that amongst other things,

they were “evil”, “forced conscription” or engaged in the “persecution of the church”. Conversely,

the government attempted to increase support for political allies by calling them “freedom ﬁghters”,

“religious” or simply “good” (Jacobwitz, 1985b).

The task of inﬂuencing press coverage was oﬃcially delegated to the Oﬃce of Public Diplomacy

(OPD) during the Reagan administration. The OPD was part of the State Department and worked

closely with the National Security Council (NSC). Its explicit purpose was to inﬂuence public and

congressional opinion to garner support for the President’s strong anti-communist agenda in a

“public action” program (Parry and Kornblub, 1988). The memo speciﬁes that audiences for the

information campaign include the U.S. media (Jacobwitz, 1985b).

“..we can and must go over the heads of our Marxist opponents directly to the Amer-

ican people. Our targets would be within the United States... the general public [and]

media.” – Kate Semerad, an external relationship oﬃcial at the Agency for International

Development (AID) in 1983.

Government methods for inﬂuencing the media can be broadly categorized into two groups. First,

the government can attempt to directly manipulate news reports by exerting pressure on editorial

boards or incentivizing journalists. The OPD monitored news reports by the American media and

would directly confront journalists and editors in order to convince them to change the reports

For studies on U.S. government favoritism of human rights reports of its Cold War allies, see studies such as

Carleton and Stohl (1985), Mitchell and McCormick (1988), Poe and Vazquez (2001), and Qian and Yanagizawa-

Drott (2009).

In the case of the The New York Times, which published an international version under the title of The International

Herald Tribune, manipulation could also aﬀect the opinion of foreign readers. Also, inﬂuencing the press could also

aﬀect congressional opinion, whose favor was often necessary for legislative purposes (Blanton and Blanton, 2007).

(Schultz, 2001). Upon the appearance of news reports that did not conform to the wishes of the

OPD, oﬃcials could press the owners and editorial boards to change their journalists in the ﬁeld. The

OPD also dealt directly with journalists using a carrot-and-stick strategy. For example, uncoopera-

tive journalists became the targets of character assassination meant to induce skepticism about the

information they reported and were sometimes even forcibly removed from foreign countries from

which they were reporting.

In contrast, journalists seen as cooperative to the administration’s

agenda were rewarded with increased access to government information. For example, an OPD

memo stated that certain favorable correspondents had “open invitations for personal brieﬁngs”

(Cohen, 2001).

Second, the government can manipulate the supply of information and provide disinformation.

Information can be disseminated through the numerous government aﬃliated publicity events and

publications. One such publication is the Country Reports on Human Rights Practices, which we will

discuss later in the paper. In a letter to House Speaker Patrick Buchanan, the Deputy Director for

Public Diplomacy for Latin American and the Caribbean (SLDP), Jonathan Miller, described how

the OPD was carrying out “white propaganda” operations. This included writing opinion articles

under false names and placing them in leading newspapers such as the Wall Street Journal (Hamilton

and Inouye, 1987; Miller, 2001). Similar opinion editorials were planted in the New York Times

and the Washington Post (Brooks, 1987). The OPD paid extra attention to prominent journalists.

In general, the OPD ﬂooded the media, academic institutions and other interested groups with

information. For example, in 1982, the OPD booked more than 1,500 speaking engagements with

editorial boards, radio, and television interviewers, distributed materials to 1,600 college libraries,

520 political science faculties, 122 editorial writers, and 107 religious groups (Parry and Kornblub,

1988).

2.2 The United Nations

The United Nations (UN), a source of much of the diplomatic inﬂuence and the principal outlet

for the foreign relations initiatives of non-superpower countries, was especially important during the

One famous case was the removal of New York Times reporter Raymond Bonner from El Salvador after his

unfavorable reporting of the massacre by the Salvadoran government. The U.S government pressured the NYT to

recall Bonner (Parry and Kornblub, 1988). Other outlets such as the Wall Street Journal subsequently published

articles criticizing the NYT for publishing Bonner’s reports.

Blanton and Blanton (2007) provides an overview of all the actions taken by the Oﬃce for Public Diplomacy

(OPD) during the Reagan Administration (1980-88). For detailed accounts of when the media allows the government

to distort reports, see Bennet, Bennet and Livingston (2007) and Thomas (2006).

Cold War.

Two of the ﬁve principal organs of the UN are the General Assembly and the Security

Council (UNSC). During the period of our study, there were approximately 150 member countries,

of which more than two-thirds were developing countries. The General Assembly votes on many

resolutions brought forth by sponsoring states. Most resolutions, while symbolic of the sense of the

international community, are not enforceable as a legal or practical matter. The General Assembly

does, however, have authority to make ﬁnal decisions in areas such as the UN budget, and in case

of a split vote in the Council when no veto is exercised, the issue goes for a vote in the General

Assembly.

The Security Council is comprised of ﬁfteen member states. Hence, it is signiﬁcantly smaller

than the General Assembly. Council members have more power than General Assembly members

because the Council can make decisions which are binding for all UN member states, including

economic sanctions and the use of armed force (Chapter Seven of the UN Charter). There are

ten temporary seats that are held for two-year terms, each one beginning on January 1st. Five

are replaced each year. The members are elected by regional groups and conﬁrmed by the UN

General Assembly. New members are typically announced the year before the term begins.

There

are ﬁve permanent members (P5): China, France, Russia, the United Kingdom and the United

States. These members hold veto power for blocking adoption of a resolution. Experts vary in their

assessment of the power of rotating members over important issues during our period of study. On

the one hand, rotating members cannot overturn vetoes and some political scientists argue that

they have limited real power (e.g., O’Neill, 1996). On the other hand, studies such as Voeten (2001)

argue that P5 countries prefer multilateral agreements, which, in turn, gives much power to rotating

members. For example, deadlocks on the Council can only occur if there is no veto and nine of the

ten deadlocks that have ever occurred in the history of the UN occurred during the Cold War.

For a detailed discussion of the history and institutions of the United Nations, see Malone (2004).

Africa elects three members; Latin America and the Caribbean, Asian, and Western European and others blocs

choose two members each; and the Eastern European bloc chooses one member. Also, one of these members is an

Arab country, alternately from the Asian or African bloc. Members cannot serve consecutive terms, but are not

limited in the number of terms they can serve in total. There is often intense competition for these seats (Malone,

2000).

1956 Suez Crisis; 1956 Soviet Invasion of Hungary (Hungarian Revolution); 1958 Lebanon Crisis; 1960 Congo

Crisis; 1967 Six Days War; 1980 Soviet invasion of Afghanistan; 1980 Israeli-Palestinian Conﬂict; 1981 South African

occupation of Namibia (South West Africa); 1982 Israeli Occupation of the Golan Heights (Golan Heights Law); 1997

Israeli-Palestinian conﬂict (East Jerusalem and Israeli-occupied territories).

3 Model

In this section, we develop a framework of how alliance with the U.S. government and member-

ship on the United Nations Security Council can interact to aﬀect U.S. news coverage of human

rights abuses of foreign countries. Speciﬁcally, our framework studies the incentives of the govern-

ment to distort media coverage of state repression in foreign countries. In our model, domestic

voters care in part about the foreign policy that the U.S. government pursues, but voters cannot

directly and fully evaluate the foreign policy (preferences) of the incumbent. Voters partly base

their inferences on the behavior of allied foreign countries that vote with the United States in the

UN. News reports about human rights violations of the allies serve as indicators of U.S. foreign

policy, which aﬀects the probability that the incumbent U.S. government will be voted out of the

oﬃce.

Before presenting the formal model, we sketch the basic intuition behind it. In our model, “worse”

countries are more likely to commit human rights violations and are more likely to vote with the

United States if the U.S. government’s foreign policy is bad. U.S. voters observe voting behavior

and read about human rights violations to make their inferences about the U.S. government’s type.

There are two groups of voters. The ﬁrst group is interested in and reads the news about all

foreign countries. As a result, voters in the ﬁrst group make inferences based on the behavior of

all countries. The second group is interested only in the countries that are currently on the UNSC.

We do not formally model the reason for this. This assumption is motivated by the fact that the

Council discusses more important issues and/or has more power over these issues. Alternatively,

it could simply be because being on the Security Council acts as a focal point for readers with

limited interest in foreign policy. The second group solely bases its inferences on the news coverage

of Council members.

In our model, obtaining a seat on the Security Council generates two eﬀects on news coverage.

The ﬁrst is a demand eﬀect. As a country becomes a member of the Council, more people are

interested in reading about it. In the absence of government interference, newspapers would then

increase their coverage of human rights abuses in these countries. The second eﬀect is a distortion

eﬀect that comes from the incentives of the government to manipulate the media. We show that if the

number of countries not on the Council is much larger than the number of countries on the Council

(e.g., the General Assembly), it is much cheaper for the U.S. government to manipulate public

opinion by suppressing news about Security Council countries than non-Security Council countries.

Because voters in the ﬁrst group based their inferences on voting and human right violations of all

countries, distorting news coverage about one of them has little eﬀect on the posterior beliefs of this

group when the total number of countries is large. In contrast, the voters in the second group base

their inferences on the voting behavior of a relatively small number of countries on the Council, and

the distortions in the coverage of one country has a large eﬀect on the voters’ posterior beliefs about

the U.S. government’s type. As a result, when the country enters the Council, it is optimal for the

U.S. government to signiﬁcantly intensify its distortion of news coverage. Moreover, this eﬀect is

monotone. The closer the foreign country is aligned with the United States, the more severe the

distortion will be.

Methodologically, our approach combines recent theories of endogenous news coverage developed

by Prat and Stromberg (2005, 2011) and Stromberg (1999, 2004a, 2004b) with a theory of media

manipulation by Besley and Prat (2006). That voters in our model try to infer the quality of the

government policies from news reports and that newspaper coverage aﬀects the posterior beliefs of

the voters about the quality of those policies is very similar to Durante and Knight (2012).

The

ultimate goal of our model is to derive testable implications to guide the empirical investigation of

whether there is government distortion in the context that we study.

3.1 Setup

Our model of media coverage is a close adaptation of Prat and Stromberg (2011, section 5), and

we mostly follow their notation. We consider a simple two-period model with no discounting. An

incumbent U.S. government has type θ

, which is the same in both periods. In the end of period

one, U.S. voters decide whether to keep the incumbent or replace it with a challenger. Both the

incumbent’s and the challenger’s types are drawn from a uniform distribution which takes values

on [0, 1] . Neither type is observed directly by voters. Voters make inferences about the incumbent’s

type from the U.S. government’s voting record in the UN General Assembly and from news coverage

of U.S. allies in the UN. Higher θ

corresponds to the U.S. government being “better” from the

point of the of domestic voters.

Durante and Knight (2012) studies the optimal choice of news outlet based on their ideological leaning. We

abstract from the diﬀerences in ideology and focus on the incentives of the government to manipulate news coverage.

There are four main groups of players in our model: the U.S. government, foreign governments,

U.S. voters and the U.S. media. We discuss each of these groups in turn.

3.1.1 U.S. government

The U.S. government cares about two objectives: i) passing the foreign policy issue that it favors

in the UN, and ii) rents associated with being in oﬃce. It has two instruments at its disposal to

achieve its goals. The ﬁrst instrument is its vote on a given issue. The second is the distortion

of newspapers to get favorable coverage. We assume that the U.S. government votes sincerely and

focuses on the incentives to distort the news coverage.

3.1.2 Foreign governments

We assume that there are N + 1 foreign countries and each country n can be one of the two

types, either “bad” (θ

= 0) or “good” (θ

= 1). For concreteness, we assume that each country’s

type is drawn ex-ante independently and is equally likely to take either type. In the UN, the U.S.

and foreign countries vote on a large number of resolutions. Each issue, x

, takes a value which is

being drawn from a uniform distribution on [0, 1] . The payoﬀ for a country of type k ∈ {0, 1, US}

if the resolution j is passed is



− (x

− θ

)



. (1)

The payoﬀ if a resolution is not passed is normalized to zero.

Under the assumption that countries vote sincerely, this convenient formalization means that

the U.S. votes with a country of type θ

= 0 with probability 1 − θ

and with a country of

type θ

= 1 with probability θ

. Since we assume that the number of resolutions is large, the

observed frequency of voting with country k reveals the degree of alliance with the United States,

US,k

= |θ

− θ

|, perfectly.

Out of N + 1 foreign countries, one is randomly selected to be on the Council. Without any

loss of generality, we denote this country by n = N + 1. In this model, the only distinction between

being and not being on the Council is that more U.S. voters will be interested in reading the news

about countries on the Council.

It is possible to extend this model to account for strategic voting by the U.S. and show that the main insights

continue to hold in that case. This extension is available upon request.

One can make the case that, in general, more information will be revealed about a country’s type when it is on the

Security Council, for example, because policy deliberations generally reﬂect more important issues. Such mechanisms

will generally strengthen the eﬀect we consider in this model.

Foreign governments of type θ

= 0 are repressive and commit human rights abuses. Govern-

ments of type θ

= 1 are not repressive and do not abuse human rights.

3.1.3 U.S. Voters

U.S. voters do not observe foreign governments’ types or the U.S. government’s type directly.

Voters read newspapers and they become informed about the type of the foreign government and

how closely it is allied with the United States with some probability. We assume that there are two

groups of voters. Group 1 reads news about all foreign countries. Group 2 reads only about foreign

countries that are “powerful”, i.e., on the UN Security Council.

Let m

and m

be the strictly

positive fraction of voters in groups 1 and 2, respectively.

In the next section, we describe how voters become informed. Based on media coverage, some

fraction s of the population becomes informed about human rights violations of foreign governments

and how frequently that government voted with the United States. Both pieces of information,

therefore, perfectly reveal the type of the incumbent, θ

. A fraction, 1 − s, remains uninformed,

and their posterior beliefs about the incumbent type remains unchanged, with the expected value

of 1/2.

The expected utility from keeping the incumbent in oﬃce in period 2 is E [θ

], where E is the

expectation given the information that each voter has. The utility from selecting the challenger is

E [θ

] + δ, where E [θ

] is the expected challenger’s type and δ is an idiosyncratic characteristic

of the challenger. We assume that δ is uniformly distributed on [−1/2, 1/2] .

For brevity, we focus on sincere voting. Since θ

is drawn from the same distribution as θ

the uninformed voters vote for the incumbent if −δ ≥ 0, which occurs with probability 1/2. The

informed agents vote for the U.S. government if θ

− 1/2 − δ ≥ 0. The δ to satisfy this condition

is realized with probability θ

. Therefore, the probability of re-election of the incumbent, µ, is

µ = sθ

+ (1 − s)

. (2)

We study how voters become informed in the following section.

As in Prat and Stromberg (2011), one can allow for an arbitrary number of groups and issues that are policy-

relevant. We focus on foreign news given the nature of the data in the empirical section, but the main implications

of the model are potentially applicable to domestic news as well.

3.1.4 Mass Media

We now consider the media market. The main idea is that bigger news coverage about any

country increases the fraction of voters informed about it. We follow the arguments of Prat and

Stromberg (2005, 2011) and Stromberg (1999, 2004a, 2004b) . Let q

be the amount of news coverage

for country n. A reader buys the newspaper based on the amount of coverage and on idiosyncratic

characteristics. The reader knows that if he buys a newspaper with q

stories about country n,

he will ﬁnd the news to be entertaining and the information to be relevant with probability ρ (q

where ρ is an increasing function. If the reader ﬁnds the information entertaining, he obtains 1 unit

of utility. Let ε

be the reader’s exogenous idiosyncratic valuation of newspapers, which is uniformly

distributed on [0, 1] . If the price of the newspaper is p, the reader buys it if

ρ(q

) − ε

≥ p.

For simplicity, we assume that ρ (q) = q and the parameters of the model are such that in equilib-

rium, q ≤ 1. Thus, if the newspaper has q

news about country n, the probability that a reader is

interested in country n is max {q − p, 0} . Once a reader buys the newspaper, he learns that θ

= 1

and the degree of alliance of that country with the United States, A

US,n

, which is suﬃcient for

inferring the true type of the U.S. government, θ

As in Prat and Stromberg (2005, 2011), we assume that newspapers operate an increasing

returns to scale technology, since there are costs of gathering news and writing a story which is

independent of the number of newspaper copies sold. We assume that news cannot be fabricated,

and it is therefore impossible to write about human rights violations for countries for which θ

= 1.

For countries that commit human right violations (θ

= 0), newspapers select the optimal amount

of coverage. The cost of publishing q

stories is

+ ˜m (q

− p) d,

where d reﬂect the cost of distribution, γ > 0 is a parameter, and ˜m = m

for a country on the

Security Council and ˜m = m

for any other foreign country. Here, the ﬁrst term is the ﬁxed cost

that is independent of the number of copies and the second term represents constant marginal costs,

d, which are proportional to the demand for newspaper, ˜m (q

− p) . The proﬁt for the newspaper

from publishing q

news is then

Π (q

) = (p − d) ˜m (q

− p) −

. (3)

For simplicity, we assume that there is only one newspaper and that p is exogenous and greater

than d. Both of these assumptions can be relaxed along the lines suggested by Stromberg (2004a,

2004b).

Before we characterize the equilibrium with news manipulation, it will be informative to describe

the equilibrium when the government does not interfere. In this case, the optimal news coverage ˆq

maximizes (3) for each n for which θ

= 0. It is straightforward to see that ˆq

satisﬁes

ˆq











0 if θ

= 1

(p − d) m

if θ

= 0 and n ≤ N

(p − d) (m

+ m

) if θ

= 0 and n = N + 1

. (4)

This implies that news coverage of human rights abuses is higher if the country joins the Council.

To make sure that ˆq

−p are well deﬁned probabilities when θ

= 0, we assume that (p, d, m

, m

, γ)

jointly satisfy

(p − d) m

− p > 0,

(p − d) (m

+ m

) < 1.

To ﬁnd the probability with which the U.S. incumbent retains power, we need to ﬁnd the number

of informed voters. We start with group 2. In the context of our model, if the country on the Security

Council does not commit any human rights violations, then no news coverage is available and all

voters in group 2 retain their prior beliefs that the probability that the U.S. government is of type

1 is 0.5.

If the country committed human rights violations, then fraction

(p − d) (m

+ m

) − p

In principle, the lack of news about country n can be an informative signal about country n, since in our model,

all news is assumed to be bad news. For simplicity, we assume that readers update their beliefs about country n (and

about the U.S. government) only if they read the news about this country and if they ﬁnd the news entertaining. In

a richer model in which good news can be generated about good countries, this issue does not arise and the analysis

of that model is very similar to the one presented in this section. Since our empirical strategy allows us to identify

of voters in group 2 will see the news and learn that the true type of the U.S. government is θ

The fraction 1 −



(p − d) (m

+ m

) − p



remains uninformed and retain the prior that the U.S.

government is of type 1 with probability 0.5.

Next, we turn to voters in group 1. A voter learns the type of the U.S. government if she spots

news coverage for at least one country. Suppose that out of the N countries,

N commit human

rights violations and the country on the security council is also type θ

N+1

= 0. Then the probability

that she does not spot any news coverage is and retains her prior belief is

(1 − max {ρ (ˆq

) − p, 0}) × ... × (1 − max {ρ (ˆq

N+1

) − p, 0})



1 −



(p − d) m

− p





1 −



(p − d) (m

+ m

) − p



Then the total share of informed voters who learn the type of the U.S. government, s, is

s =

1 −



1 −



(p − d) m

− p





1 −



(p − d) (m

+ m

) − p





(p − d) (m

+ m

) − p



Note that this pins down the probability of the U.S. incumbent retaining power, as given by (2).

3.1.5 Equilibrium with manipulation

Next, we turn to describing the equilibrium with manipulation. We keep the basic structure of

the game the same as before. We introduce one more stage, following the logic of Besley and Prat

(2006), in which the U.S. government can oﬀer a transfer T (∆

) to the newspaper and suppress

∆

news from publication.

The newspaper can either reject the transfer and publish its proﬁt-

maximizing news quantity or accept it and publish at most q

. We keep the rest of the model as

above.

Let s



}

N+1

n=1



be the fraction of voters who are informed if country n has q

reports of human

rights abuses. Suppose that R is the value of the incumbent to remain in power, which is strictly

only bad news, we focus on this simpler set up.

As in Besley and Prat (2006), the bribe that the U.S. government pays to a newspaper is not necessarily a

monetary transfer, but can be diﬀerent forms of non-pecuniary beneﬁts that aﬀect newspaper proﬁts, such as oﬀering

exclusive interviews with the incumbent or leaking valuable political information to the newspaper.

positive. Then the incumbent solves

max

}

N +1

n=1





}

N+1

n=1





1 − s



}

N+1

n=1





R −

N+1

n=1

T (ˆq

− q

) (5)

subject to

T (ˆq

− q

) + Π (q

) ≥ Π (ˆq

) for all n

and

≤ ˆq

Here, the ﬁrst constraint is a best response for the newspaper that agrees to suppress ∆

= ˆq

−q

news only if its proﬁts from doing so exceed the proﬁts from rejecting the oﬀer. The second constraint

ensures that the newspaper cannot publish more news than the newspaper had originally planned

and, in particular, that no human rights violation stories can be fabricated for countries θ

= 1.

This problem is, in general, not well behaved. Function s is neither concave nor convex, which

makes the analysis harder. The problem simpliﬁes when we focus on the empirically relevant case

when the number of countries is large. In this case, it is easy to show that it is (approximately)

not optimal to distort countries that are not on the Security Council. The reason for this is as

follows. Note that the probability that a voter in group 1 is not informed about the type of the

U.S. government is (1 − max {ρ (q

) − p, 0}) × ... × (1 − max {ρ (q

N+1

) − p, 0}) . When N is large,

so is the number of countries which violate human rights,

N. If the probability that newspapers

report human rights violation for such countries is positive, max {ρ (q

) − p, 0} > 0, the fraction

of uninformed types in group 1 becomes very small. The only way to substantially change the

fraction of informed is to signiﬁcantly suppress news for a large number of countries, which becomes

prohibitively costly for a large N.

Formally, let {q

∗

}

N+1

n=1

be the equilibrium quantities of news that are a solution to (5). We have

the following result, which we formally prove in the Appendix.

Lemma 1 Suppose N is large. Then q

∗

≈ ˆq

for n = 1, ..., N

Next, we focus on the distortion of news for a foreign country on the Council. Let s



}

N+1

n=1



and s

N+1

) be the fraction of informed citizens in groups 1 and 2. When q

∗

N+1

is interior, the

ﬁrst order condition for q

∗

N+1



∂

∂q

N+1



∗

}

N+1

n=1



∂

∂q

N+1



∗

N+1





−



R + (p − d) (m

+ m

) = q

∗

N+1

For the reasons explained in the proof of Lemma 1, as N → ∞,

∂

∂q

N +1



∗

}

N+1

n=1



→ 0

and therefore for large N , we can ignore this term. Since s

N+1

) = m

N+1

− p) , the above

expression becomes

∗

N+1

= (p − d) (m

+ m

) +



−



Since q

∗

N+1

≤ ˆq

N+1

, this condition holds only for



−



≤ 0. For



−



> 0, the optimal

news suppression is zero, q

∗

N+1

− ˆq

N+1

= 0.

This result allows us to compare news coverage of a country on the Council and an identical

country not on the security council. If this country is of type 1, there is obviously no news coverage

of human rights violations in any case. If the country is of type 0, the diﬀerence in coverage ∆ is

given by

∆ = (p − d) m

| {z }

demand eﬀect

− max



− θ



, 0



| {z }

distortion eﬀect

This formula shows that news coverage is determined by two eﬀects when a country gets on

the Security Council. The “demand eﬀect” leads to an increase in coverage since more people want

to read about the country on the Security Council. The “distortion eﬀect” leads to a decrease in

coverage for allies that violate human rights. Moreover, the closer the United States is allied to

the foreign country (i.e., the lower θ

), the stronger this eﬀect will be. Since distortion occurs if

≤

and the foreign country on the Security Council is of type 0, which happens with joint

probability 1/4, in expectation the distortion eﬀect is positive. We summarize these ﬁndings in the

theorem

Theorem 2 For a repressive country not allied with the United States, news coverage of its human

rights violations increases when it enters the Security Council. The magnitude of the increase

declines with the degree of alliance. If the beneﬁt of being in power, R, is suﬃciently large, then

news coverage falls for close allies when they enter the Council.

The empirical analysis investigates whether these relationships are present in the data.

4 Empirical Strategy

The relationship between news coverage, U.S. alliance and Council membership can be charac-

terized as the following:

= β(A

× C

) + αC

+ θX

+ γ

+ δ

+ ε

, (6)

where the outcome variable, news coverage of human rights abuses, in country i in year t, Y

is a function of: the interaction of alliance to the United States, A

, and membership on the

Security Council, C

; the uninteracted term for Council membership; a vector of country-year

speciﬁc controls, X

; year ﬁxed eﬀects, δ

; and country ﬁxed eﬀects, γ

. The standard errors are

clustered at the country level to adjust for serially correlated shocks within countries. The country

ﬁxed eﬀects control for all time-invariant diﬀerences across countries. Year ﬁxed eﬀects control for

changes over time that aﬀect all countries similarly. X

includes a vector of country-year controls,

which will be motivated and discussed later as they become relevant. Note that our measure of

alliance is time-invariant and collinear with country ﬁxed eﬀects. Therefore, we do not control for

the uninteracted alliance term in the baseline regression.

Our strategy is conceptually similar to a diﬀerences-in-diﬀerences (DD) strategy. We compare

outcomes for countries when they are on the Council to when they are not, between countries that

are strongly allied to the United States to those that are less allied. α is the association of Council

membership and news coverage for foreign countries that are not allied to the United States at all,

= 0. β is the diﬀerential association of Council membership and news coverage between countries

that are not allied at all, A

= 0, and countries that are “perfectly” allied, A

= 1. α + β is the

“net” or “total” association between news coverage and Council membership for countries that are

perfectly allied with the United States. In the context of our conceptual framework, ﬁnding

β < 0

will be consistent with the presence of government distortion (i.e., the news is partly determined by

U.S. government distortion), and that ˆα > 1 will be consistent with the presence of reader demand

eﬀects (i.e., the news is partly determined by reader demand). These two eﬀects are not mutually

exclusive and can co-exist. Finding that ˆα + A

β < 0 means that for a country that is allied with

the United States by A or more, Council membership will reduce news coverage.

Interpreting the association between the interaction eﬀect and news coverage as causal requires

the assumption that Council membership does not diﬀerentially aﬀect allies in some way that will

inﬂuence news coverage through channels other than U.S. government distortion or reader demand.

Speciﬁcally, for the interaction term to overstate the true degree of government distortion, the

omitted factor needs to reduce the increase in news coverage according to the level of political

alliance with the United States. For example, if improvement in human rights practices when

entering the Council is positively correlated with alliance, then the interpretation of our estimates

will be confounded. We will carefully consider this and other robustness concerns after we present

the main results.

5 Data

This paper uses data that are constructed from numerous publicly available sources. For brevity,

we only describe the data for the main analysis in this section. Other data will be discussed as they

become relevant.

News coverage of human rights violations is measured as the number of newspaper articles about

human rights abuse in a given country. Following the deﬁnitions used by Freedom House and the

Political Terror Scale project, we deﬁne human rights as physical violence committed by the state

onto civilians. We calculate the number of articles based on a search of the text of articles in

the ProQuest Historical and National Newspapers database. We search for articles containing the

country’s name, the phrase “human rights” and require at least one of the words or phrases that

fall under the UN Declaration for Human Rights (and that are therefore also commonly used in

news articles on human rights abuse). These include “torture”, “violations”, “abuse”, “extrajudicial”,

“execution”, “arbitrary arrests”, “imprisonment” and “disappearances”. Our measure of human rights

coverage is the total number of articles that results from the search per country per year. The

newspapers we examine are The New York Times (NYT), The Washington Post, The Wall Street

Journal (WSJ, only available 1976-91), The Chicago Tribune (only available 1976-86) and The Los

Angeles Times (L.A. Times).

These are the only newspapers with high circulation for which we

could conduct a full text search for the main period of our study. All of the newspapers in our

In an earlier version of the paper, we also used news from the Christian Science Monitor. Since we want to focus

in newspapers with large circulation, we have dropped this much smaller newspaper from the sample. All of the

results are similar with its inclusion. Please see the earlier version for those results.

sample are in the top ten of the highest circulation newspapers in the United States. Our measure

includes both articles written by journalists employed by newspapers and stories picked up from

newswires and other sources, although the newspapers in our sample, and in particular the NYT

and Washington Post, were known for original international news reporting.

This does not aﬀect

the interpretation of the results, but for completeness, we will also examine the impact on articles

from newswires after we present the main results.

We proxy for the degree of alliance with the United States during the Cold War using the mean

fraction of votes that a country votes in agreement with the United States on issues for which

the United States votes in opposition to the Soviet Union in the UN General Assembly, A

, where

∈ [0, 1].

Our measure of alliance includes abstentions.

Our main measure of alliance is the

fraction of votes a country voted with the United States averaged over the period 1985-88. This

period provides us with the highest number of divided votes and therefore the best measure of

alliance during this period. We use a time-invariant measure of alliance because it is less likely to

be an outcome of changing U.S. favoritism than a time-varying measure and because using voting

patterns from years where there were very few divided issues produces a very noisy measure of

alliance.

Data on Council membership are collected from The United Nations Security Council Member-

ship Roster.

This is a time-varying dummy variable for whether a country is a rotating member

The source of the story is often embedded within an article. Therefore, we were not able to accurately and

systematically distinguish between articles written by diﬀerent sources.

Each year, there are approximately 100-150 resolutions in the Assembly, of which the United States and U.S.S.R.

disagree on approximately 70-90. We do not examine voting patterns in the Council because most issues are discussed

prior to being put onto the agenda. Therefore, the sample of issues voted on are not representative of the actual

issues being deliberated by Council members.

Excluding them does not signiﬁcantly change either the measure of alliance or the regression results. For brevity,

we do not report those results in the paper.

Figure A.1 plots the fraction of divided votes over time. It shows that as Cold War tensions escalated in the

1980s, the fraction of divided votes increased from approximately 30% during the late 1970s to almost 70% in the late

1980s. Also plotted are the fraction of votes with the United States averaged over all the divided votes each year.

Using the measure of alliance presented in the paper, the top three allies of the United States and the fraction of

divided issues they voted with the United States are: Turkey (0.4), Belize (0.28) and Costa Rica (0.27). The three

countries that are least allied are Mongolia (0), Lao PDR (0), and Czech Republic (0). Our estimates are robust to

changing the measure of alliance to be the average of votes during periods between 1981 and 1989, when there were

many divided votes. The magnitude of the estimates vary slightly across diﬀerent deﬁnitions, but the results are

always qualitatively similar. For brevity, we do not report estimates with these alternative measures in the paper.

They are available upon request.

See http://www.un.org/sc/list_eng5.asp for a list of all countries that were ever members and the years of their

memberships. 46 countries in the sample were on the Council as a rotating member at least once during this time.

They are listed in Appendix Table A.1. 21 countries were on the Council at least twice, among which ﬁve countries

were on the Council three times.

of the UN Security Council, C

We will also use reports on human rights practices in our analysis. These indices are provided

by the Political Terror Scale (PTS) project for the years 1976-2005 and measure the annual extent

of human rights violations according to two sources: the U.S. Department of State (USSD) Country

Reports on Human Rights Practices, and the Amnesty International Annual Report. The PTS uses

a ﬁve point coding scheme, where a PTS value of ﬁve indicates the most severe abuse, and a PTS

value of one means that the country is under a secure rule of law and people are not imprisoned for

their views. The scale is based on an earlier one developed by Freedom House. As we are interested

in countries that commit human rights violations, we restrict our analysis to countries that have an

Amnesty International PTS value above one in at least one year.

The ﬁnal sample of countries excludes former Soviet Republics that did not have membership

in the United Nations before 1991 and South Africa, which was excluded from UN activities due to

the UN’s opposition to apartheid. The ﬁve permanent members of the Council are also excluded

since they cannot experience any variation in Council membership.

The main analysis focuses on the Cold War years, 1976-88. The sample begins in 1976 because

of the limitation of the PTS data. The sample ends in 1988, the year prior to the fall of the Berlin

Wall, which we interpret as the “beginning of the end” of the Cold War era. Our estimates are

qualitatively similar if we include 1989-1991.

The main sample comprises of an unbalanced panel

of 91 countries. The panel is unbalanced due to the fact that the number of countries in the United

Nations increases over time. After we present the main results, we will show that our estimates are

not driven by selection of countries into the United Nations, and we will also show the results from

using data from the post-Cold War period.

Table 1 presents the descriptive statistics. Panel A shows the Cold War period. Panel B shows

the post-Cold War period. We focus our discussion on the former. The data show that the average

country in our sample votes with the United States on 8.9% of divided issues. Approximately 7% of

the country-year observations are Security Council members. The sample average for human rights

A PTS value of two implies (among other things) there is “a limited amount of imprisonment for nonviolent

political activity... A few persons are aﬀected; torture and beating are exceptional... and political murder is rare.”

The CIRI Human Rights Data Project, like the PTS Project, reads the reports by Amnesty and the State Department

and provides a score. However, the CIRI indices only begin in 1981. They also diﬀer from PTS in that they attempt

to provide disaggregated indices for the type of human rights. This means that while the two indices are correlated

(approximately 0.65-0.73), they are not directly comparable. See Wood and Gibney (2010) for a detailed discussion.

These results are available upon request.

practices is “medium” since the PTS index ranges from one to ﬁve and the mean PTS in the sample

is close to three. The U.S. State Department reports countries as having better human rights, on

average, than does Amnesty International.

The statistics also show that approximately 60% of the country-year observations have at least

one news article published on human rights in a U.S. newspaper. The U.S. newspapers in our sample

published almost eight articles about the human rights abuses of the average foreign country in our

sample each year. The Washington Post, New York Times and Los Angeles Times published three

to ten times as many articles as the Wall Street Journal or the Chicago Tribune. We will discuss

newswire and U.K. newspapers later in the paper.

6 Results

6.1 Main Results

In this section, we present the estimates of equation (6). Table 2 presents our main results on

news coverage. The sample means of the dependent variables are presented at the top of the table.

Our main measure of news coverage is the log of the number of news articles about country i’s

human rights abuses during year t across all newspapers.

In panel A column (1), we control for

the uninteracted U.S. alliance term instead of country ﬁxed eﬀects to examine the coeﬃcient for this

variable. When we do this, the interaction eﬀect is negative, but statistically insigniﬁcant. The main

eﬀect of Council membership is positive, but statistically insigniﬁcant. The main eﬀect of alliance

is positive and almost signiﬁcant at the 10% level, which implies that alliance and news coverage

is positively correlated for countries not on the Council. However, this should be interpreted very

cautiously as only suggestive because U.S. alliance is almost certainly endogenous to a large number

of other factors.

Column (2) replaces the control for U.S. alliance with year ﬁxed eﬀects. This greatly increases

the precision of the estimates for the interaction eﬀect and the uninteracted Council eﬀect without

causing much change to the magnitudes of the estimates. The interaction eﬀect is statistically

signiﬁcant at the 10% level and the uninteracted Council eﬀect is signiﬁcant at the 1% level. In

If there are zero articles, we calculate this as the log of 0.1 to maximize sample size. To check that our results

do not rely on this transformation, we will later show the estimates for the number of articles (without taking logs)

and the log number of articles on a sample restricted to observations where there is at least one article.

Note that our model, which assumes that readers do not distinguish allies from non-allies, is silent on the

relationship between alliance and news coverage for non-Council members.

column (3), we replace the control for the uninteracted Council eﬀect with the interaction of the

Council dummy with the full vector of year ﬁxed eﬀects to address the possibility that the importance

of Council members to the U.S. government or the prominence of the Council to the American public

changes over time. Adding this control increase the precision of the interaction eﬀect to the 1%

signiﬁcance level.

In column (4), we additionally control for the interaction of U.S. alliance and the full vector of

year ﬁxed eﬀects to address the possibility that the relationship between the U.S. government and

its allies changed over time. Note that by interacting Council membership and U.S. alliance with

year ﬁxed eﬀects, we are allowing the inﬂuences of each to be fully ﬂexible over time. Thus, our

main interaction eﬀect is only driven by variation in news coverage that systematically varies with

alliance and Council membership that is not captured by this rigorous set of controls. Column (4)

is our baseline speciﬁcation.

Interpreted within our framework, the positive estimates of the Council main eﬀect in columns

(1)-(2) show that reader demand eﬀects are present in our context. The negative interaction eﬀects

in columns (1)-(4) show that government distortion is also present. To assess the magnitude of

the estimated distortion implied by the coeﬃcients in column (4), we compare the eﬀect of Council

membership for a country that is on the 75th percentile of the distribution of U.S. alliance that

votes with the United States on approximately 10.6% of divided issues on average (e.g., Ecuador,

Egypt) to a country that is on the 25th percentile of the distribution of alliance that votes with

the United States on approximately 4.3% of divided issues on average (e.g., Mali, the Republic of

Congo). This is shown at the bottom of the table. We ﬁnd that Council membership of the stronger

ally will result in 42% (−6.62 × (0.106 − 0.043) = 0.42) less coverage relative to membership of the

weaker ally.

Another way to assess the magnitude of the estimates is to ask how strong alliance needs to

be for Council membership to reduce news coverage, and how many countries in the sample were

suﬃciently allied to experience this reduction. For this back-of-the envelope calculation, we use the

estimates in column (2), which show that the uninteracted coeﬃcient for Council membership is

0.43 and the interaction coeﬃcient is -5.04. This implies that a country that votes with the United

States on 8.5% or more of divided issues will experience a reduction in news coverage when entering

the Council (5.04/0.43 = 0.085), which applies to 39% of the countries in our sample.

6.2 Sensitivity

6.2.1 Alternative Measures of News Coverage and Sample Restrictions

In column (5), we examine the number of news articles (without logs) as the dependent variable.

The estimates are similar in sign and statistically signiﬁcant at the 10% level. The diﬀerential eﬀect

on the 75th percentile and 25th percentile allies shown at the bottom of the table imply that

Council membership will result in approximately eight fewer stories for the stronger ally relative to

the weaker one.

In Column (6), we examine a dummy variable that equals one if there is at least one news

article about country i’s human rights abuses in year t (in any newspaper). This reveals whether

the main results are driven by changes in news coverage for countries that are typically not covered

by U.S. newspapers (i.e., the extensive margin), or changes in the number of stories for countries

that already receive some coverage (i.e., the intensive margin). The estimate is similar in sign, but

is not statistically signiﬁcant at conventional levels.

In column (7), we return to examining the main outcome variable, the log of the number of all

news articles, but restrict the sample to country-year observations that have at least one article.

The estimated interaction eﬀect is nearly identical to our main estimates in magnitude, sign and

precision. In column (8), we restrict the sample to countries that are covered at least once during

the sample period. Again, the estimates are nearly identical to the baseline. The estimates in

columns (5)-(8) show that our estimates are more precisely estimated for countries that typically

receive some news coverage of human rights abuses, which may simply reﬂect the fact that these

are the countries with governments that commit human rights abuses.

In column (9), we check that our estimates are not solely driven by the countries with the worst

or best human rights practices by dropping the ﬁve percent of countries that have the worst practices

and the ﬁve percent with the best practices. Speciﬁcally, we drop countries with highest or lowest

5% average (over the sample period) PTS scores according to either the U.S. State Department or

Amnesty PTS scores. The estimates in column (9) show that our main results are very robust to

excluding the countries with the best and worst human rights behavior from the sample.

In column (10), we restrict the sample to the years of the Reagan Administration (1980-1988),

from which most of the case evidence we presented earlier are from. As expected, the estimates are

very similar to the main results.

In column (11), we restrict our sample to countries that were on the UNSC at least once during

our sample period. The estimates are similar to the baseline, which shows that the baseline estimates

are not driven by countries that were never on the Council.

In column (12), we estimate an alternative speciﬁcation to examine whether the relationship

between alliance and UNSC membership is monotone across alliance levels. For this exercise, we

divide countries into ﬁve equally sized groups according to how closely they are allied to the United

States. We then interact each of the ﬁve alliance dummy variables with UNSC membership. Since

we include all ﬁve interactions, we exclude the UNSC membership main eﬀect, which is collinear.

We exclude the interactions of UNSC and year ﬁxed eﬀects and U.S. alliance and year ﬁxed eﬀects.

The estimates show that the coeﬃcients are indeed monotonically declining with alliance. The

pattern across coeﬃcients is consistent with our model. The interaction terms for the weakest and

strongest alliance groups are signiﬁcant at the 1% level.

6.2.2 Timing of the Eﬀect

Interpreting the association between the interaction of UNSC membership and U.S. alliance

and news coverage as a causal relationship assumes that conditional on the baseline controls, there

are no other factors that are simultaneously correlated with UNSC membership and the degree of

political alliance with the United States (i.e., omitted variables) that can also aﬀect news coverage.

Speciﬁcally, for the interaction term to overstate the true degree of government distortion, the

omitted factor needs to reduce the increase in news coverage from Council membership according to

the level of political alliance with the United States. For very strongly allied countries, this omitted

factor needs to cause Council membership to result in less coverage. In the remaining robustness

section, we provide evidence that our estimates are unlikely to be driven by such omitted variables

in two exercises.

First, we show that the main results are unlikely to be driven by spurious correlations. The

most direct way to do so is to estimate the eﬀect for each year prior to and after a country is on the

UNSC. To investigate this, we estimate an equation similar to our baseline, except that instead of

a dummy variable indicating Council membership, we create seven dummy variables for two years

before Council membership is announced, one year before it is announced, the year it is announced,

the ﬁrst year of Council membership, the second year of Council membership, one year afterwards

and two years afterwards.

We control for these seven dummy variables and their interactions with

U.S. alliance, in addition to the baseline controls of the interaction of U.S. alliance with year ﬁxed

eﬀects, the interaction of Council membership with year ﬁxed eﬀects, and country and year ﬁxed

eﬀects.

For comparison purposes, we conduct this analysis for both the Cold War and post-Cold War

sample. Given that our empirical strategy applies to the Cold War, ﬁnding that the coeﬃcients

are similar in the years leading up to and after Council membership, but diverge during Council

years between the Cold War and post-Cold War samples would lend credibility to our interpretation

that our main results indeed capture changes that occur that depend on Cold War alliance when

a country is on the Council. Since Council memberships are announced one year in advance and

the announcement often results in news coverage of the country that includes discussion about its

human rights practices, the eﬀects could begin the year before Council membership.

The coeﬃcients and standard errors of the interaction variables for the Cold War and post-Cold

War periods are presented in Appendix Table A.2, and are plotted on Figure 2a.

The ﬁgure shows

that the interaction terms for the years leading up to and after Council membership are similar

during the two time periods, but diverge the year before a country’s entry onto the Council once its

election is announced, is most pronounced for the ﬁrst year of Council membership, and converges

after exiting the Council. During the Cold War, the interaction terms during Council membership

are negative, while they are zero during the post-Cold War period. These results are consistent

with the previous results that our estimates are particular to the Cold War era. The divergence

that occurs close before entry onto and after exit from the Council also means that our main results

are unlikely to be driven by spurious trends.

Figure 2b plots the year-by-year interaction coeﬃcients for each news paper. For brevity, we

only plot the interaction coeﬃcients for the Cold War era. The timing of the eﬀect is broadly similar

Since our Cold War sample only include thirteen years, the number of observations are obviously not balanced

across the pre and post years (i.e., the number observations declines as we examine more years before and after

membership). However, the patterns exhibited by our results are very similar regardless of the number of pre-

and post-Council years we examine. Hence, we present the estimates with many pre and post years for illustrative

purposes. Alternative results are available upon request.

Note that in the main analysis, the Council membership dummy variable takes a value of zero for the year of

the election. Thus, if the announcement results in news coverage of human rights abuses, our main results will be

attenuated.

The coeﬃcients of the interaction terms and their 95% conﬁdence intervals for the Cold War period are plotted

in Appendix Figure A.4a.

across papers. Thus, the timing of the eﬀect on all articles is not driven by any one newspaper.

Note that the ﬁnding that there is no eﬀect in the post-Cold War era has several potential

interpretations. It could simply be the case that our measure of alliance is no longer relevant after

the Cold War ended. Alternatively, the U.S. government may value human rights more after the Cold

War period and be unwilling to ally itself with governments that have bad human rights practices.

Both interpretations are consistent with the main of the paper that Cold War news coverage was

inﬂuenced by the U.S. government. They are also consistent with the anecdotal evidence. For

example, the president of Zaire (renamed the Democratic Republic of Congo in 1997), Mobutu Sese

Seko (in oﬃce 1965-1997) was a strong supporter of the United States during the Cold War and

had been repeatedly criticized for human rights abuses. However, during a state visit to the United

States in 1983, United States president Ronald Reagan responded to these criticism by stating

publicly that Mobutu was a “voice of good sense and good will”. Immediately after the Cold War

ended, the State Department began to criticize Zaire’s human rights violations. In 1993 Mobutu

was denied a visa for visiting the United States and said “I am the latest victim of the Cold War,

no longer needed by the United States. The lesson is that my support for American policy [now]

counts for nothing” (Gbadolite, 2011).

6.2.3 Additional Controls

Second, we address the concern of simultaneity bias by controlling for the factors that are most

likely to inﬂuence news coverage and be correlated with Council membership. Table 3 column (1)

restates our baseline for comparison purposes. Column (2) controls for region-year ﬁxed eﬀects to

address the possibility that reader interests and U.S. policy objectives may shift geographically over

time. We use the Hadenius and Teorell (2005) deﬁnition, which divides the world into ten regions

according to geo-political characteristics. Column (3) controls for political characteristics of foreign

countries that may inﬂuence U.S. readers’ interests in these countries, the strategic value of their

alliance to the United States and a country’s ability to obtain Council membership. We use the

Freedom House indices of Civil Liberty and Political Rights and the Polity2 index for constraints

on the executive.

For each variable, we calculate the average for each country over our sample

period. This maximizes the sample size and lessens the concern that contemporaneous measures

The Freedom House indices range from 1 (most free) to 7 (least free). Polity2 is an index that measures the

autocracy of the executive. It ranges from -10 to 10, where higher values reﬂect more democratic governments and

is reported by the Polity IV Project. These variables are all reported by the Quality of Governance (QoG) data.

can be outcomes of Council membership and alliance. To allow the inﬂuences of these variables to

vary over time, we control for the interactions of each variable with the full vector of year ﬁxed

eﬀects. In column (4), we control for average human rights abuse levels interacted with the full

set of year ﬁxed eﬀects. In column (5), we control for average real GDP (as reported by the Penn

World Table) interacted with year ﬁxed eﬀects to address the possibility that a country’s wealth is

correlated with its alliance to the United States, Council membership and news coverage. In column

(6), we control for all of the additional variables in one regression. The interaction eﬀect is always

negative, large in magnitude and statistically signiﬁcant at the 1% level.

In column (7), we address the possibility of spurious trends by controlling for country-speciﬁc

linear time trends. The estimated interaction eﬀect is nearly identical to the baseline result and

signiﬁcant at the 1% level.

In column (8), we control for three lags of the dependent variable to address the possibility that

there may be some mechanical persistence in coverage over time. For example, if there is a ﬁxed cost

to reporting news of a foreign country (e.g., setting up a foreign oﬃce or hiring reporters that speak

a foreign language), then “setting up shop” to report on events in a foreign country in a given year

will reduce the average cost of reporting the following years and can thereby lead to more coverage

the following years. The estimates are very robust to these controls, but should be interpreted only

as suggestive since controlling for lagged dependent variables can introduce the Nickel (1981) bias.

The robustness of our results to additional controls, together with the ﬁnding that the timing

of the reduced form eﬀects correspond to the timing of Council election and exit, imply that the

main ﬁndings are unlikely to be driven by omitted variables.

6.2.4 Additional Sensitivity Checks

Here, we check that our estimates are not driven by the selection of countries into the United

Nations over time. The main sample includes 103 countries over a period of thirteen years (1976-

88). 93 countries are observed for at least 5 years of the sample. 66 countries are observed for ten

or more years. 25 countries are observed for all thirteen years. Columns (9)-(11) show that the

coeﬃcients are similar in magnitudes and in signs when we restrict the sample to countries that are

observed for ﬁve or more years, ten or more years, or all thirteen years. The coeﬃcients are always

negative, large in magnitude and statistically signiﬁcant at the 1% level. Thus, the main results are

not driven by selection into the United Nations over time.

In addition to the exercises presented in the paper, we also checked that our main estimates

are not driven by outliers or particularly inﬂuential variables. We calculate the DF betas of each

observation to examine how inﬂuential each observation is to our regression estimate, and then ﬁnd

that our estimates are very similar when we drop the top ﬁve, ten, etc. inﬂuential observations.

These estimates are available upon request.

6.3 Mechanisms

The main results are consistent with the conceptual framework presented in section 3 that news

coverage is inﬂuenced by government distortion, as well as by reader demand. In this section,

we provide additional evidence consistent with the presence of government distortion, as well as

evidence that contradicts the main alternative explanation.

6.3.1 U.S. State Department Reports

Since our preferred explanation for the negative interaction eﬀect in the main analysis is that the

degree of government suppression of news coverage of human rights abuses increases with alliance,

we examine whether reports of human rights abuses made by the U.S. government decrease with

alliance when a foreign country enters the Council. While it is beyond the scope of our analysis

to quantitatively determine the contribution of the diﬀerent policy instruments used by the U.S.

government to distort the news (recall the discussion in Section 2), ﬁnding that oﬃcial government

reports respond to U.S. alliance and Council membership in a similar way would provide evidence

that our empirical strategy does indeed capture U.S. government objectives.

We examine the U.S. State Department’s Country Reports on the level of human rights practices

of each country. This is an annual publication submitted to Congress and open to the public,

including journalists.

That journalists are aware of this publication is consistent with the spike

in the number articles about human rights printed in newspapers the day after the reports are

announced.

As the content is entirely determined by the government, it is one of many instruments

The United States is the only country that systematically releases its reports to the public. The way in which it

gathers information is not transparent. However, it is generally assumed that the reports are based on information

from government intelligence and diplomatic apparatuses. The wording of the reports also suggest that the information

is mostly based on these sources.

Appendix Figure A.2 plots the average number of articles over time for each day before and after the Country

Reports are released. The release dates vary year to year. There is a spike in the number of articles the day following

the release. However, the total number written in the days following the release of the reports are less than one-tenth

of all of the articles written during an average years. This is consistent with the belief that although Country Reports

can serve as a reference for journalists writing about human rights, their release does not determine the timing of

most news articles about human rights.

by which the government can inﬂuence the media. Our hypothesis implies that Country Reports

should favor UNSC members that are allies relative to those that are less allied to the United States.

Finding this relationship in the data would lend additional credibility to the interpretation of our

empirical results.

A potential concern from examining State Department reports of human rights practices is

that the reports will capture both government distortion and actual human rights practices. Thus,

observing that these reports vary with alliance and Council membership could mean that either the

U.S. government distorts reports for strong allies on the Council or that strong allies on the Council

improve human rights practices more than less allied countries. To address this, we benchmark the

State Department country reports to those by Amnesty International. Since Amnesty is a non-

government organization, it should not systematically bias its reports based on U.S. government

objectives. As we discussed in the section on Data, both reports are quantiﬁed by the Political

Terror Scale (PTS), where lower scores reﬂect better practices.

In Table 4, we ﬁrst examine each PTS score separately. Columns (1) and (2) show that the

signs of the interaction term, the uninteracted Council term and the uninteracted U.S. alliance term

are diﬀerent for the State Department and Amnesty PTS scores. However, none are statistically

signiﬁcant. In column (3), we examine the diﬀerence between the State Department and Amnesty

PTS scores as the dependent variable. The signs are consistent with those shown in columns (1)

and (2), but the estimates are much more precise. The negative sign for U.S. alliance implies that

relative to Amnesty, the State Department reports allies that are not on the Council as having better

human rights practices. The positive sign on the uninteracted Council membership term implies

that relative to Amnesty, the State Department reports a country that never votes with the U.S. as

having worse human rights practices. The negative interaction term implies that the stronger the

alliance, the more the State Department will favor it with reports of good human rights practices

(relative to Amnesty) when a country enters the Council.

In column (4), we introduce the baseline controls. The interaction eﬀect is still positive and

statistically signiﬁcant at the 1% level. In column (5), we exclude countries with the top or bottom

5% PTS scores according to either Amnesty or the State Department. In column (6), we restrict

the sample to the Reagan era. In column (7), we restrict the sample to countries that were on the

Council at least once during our sample period. The estimates are very robust.

In column (8), we divide the countries into ﬁve equally sized groups of alliance and interact a

dummy indicating that a country is in each group with the Council membership dummy. These

estimates are imprecise, but the coeﬃcients suggest a pattern that is consistent with our inter-

pretation. The coeﬃcient is more negative for more allied countries. For the most strongly allied

group, the coeﬃcient is statistically signiﬁcant at the 1% level. As with the analogous results on

news coverage, these estimates suggest that the degree of U.S. government distortion of a foreign

country’s human rights abuses is roughly monotonically increasing in alliance.

As with the main results on news coverage, we can examine whether the timing of the reduced

form results for the diﬀerence in U.S. State Department and Amnesty PTS scores correspond to

entry and exit from the Council. We estimate a similar equation as before, where we control for

the seven dummy variables for the years since Council membership and their interactions with U.S.

alliance. Figure 3a plots the interacted coeﬃcients for the years since Council membership dummy

variables for the Cold War and post-Cold War sample.

As with news coverage, the eﬀects are

similar between the Cold War and post-Cold War periods for the years prior to and after Council

membership, while for the years on the Council, the coeﬃcients diverge. Moreover, when we examine

State Department and Amnesty PTS during the Cold War separately, Figure 3b shows that the two

scores are similar for the years prior and after Council membership but divergent for the years on

the Council.

The results in this section support the interpretation of our main results.

6.3.2 Alternative Explanation

The main alternative explanation for our results is the possibility that strongly allied countries

improve actual human rights practices (relative to weaker allies) when they enter the Council. This

seems unlikely given that Table 4 shows that there is no eﬀect on human rights practices as reported

by Amnesty International. However, to investigate this possibility more thoroughly, we also examine

institutional outcomes that are potential correlates of human rights practices. We use all of the

measures that are available to researchers today and that are reported by non-government agencies.

These include the Civil Liberties and Political Rights indices reported by Freedom House, the Polity

2 Index for constraints on the executive and the incidence of civil conﬂict that results in 25 or more

battle deaths as reported by UCDP/PRIO. For Polity2, we follow the literature and use a dummy

The estimates and their 95% conﬁdence intervals for the Cold War period are shown in Appendix Figures A.4b.

that equals one if the index is less than zero to indicate that a country is an autocracy.

Table 5 begins with the Amnesty PTS score.

Columns (3)-(10) show the estimates for the

other outcomes. For each outcome, we show both a speciﬁcation that shows the uninteracted Council

eﬀect as well as the baseline speciﬁcation that controls for the interaction of year ﬁxed eﬀects with

the Council dummy and with U.S. alliance. The interaction eﬀect is statistically insigniﬁcant for all

outcomes. To help compare the magnitudes of the eﬀects across the outcomes, we standardize the

estimates. At the bottom of the table, we present the standard deviation change in the dependent

variable that results from a one standard deviation change in the interaction term. The standardized

eﬀects are similarly small in magnitude across outcome variables. In terms of absolute value, the

standardized eﬀect for the baseline estimates in the evenly numbered columns range from 0.01 to

0.025. In contrast, a one standard deviation change in the interaction term causes a much larger

– a 0.1 standard deviation – change in the log of news articles on human rights abuses (this is not

shown in the tables).

In the last column, we examine the ﬁrst principal component of all of the institutional measures.

As before, we ﬁnd no eﬀect. These estimates are inconsistent with the alternative interpretation

that our estimates are driven by improvements in the relative human rights practices of strongly

allied countries when they enter the Council.

7 Additional Results

This section provides speculative results that can potentially help to extend our understanding

on the determinants of news coverage.

7.1 Distortion, Newspaper Quality and Circulation

A natural question that follows from the main analysis, which focuses on the total number of

news articles across papers, is whether the degree of distortion in each newspaper is correlated with

the quality of foreign news reporting or a newspaper’s reputation. For example, Besley and Prat

(2006) argues that in the presence of government manipulation, the government will focus its eﬀorts

on newspapers with good reputations. This may be because distorting the coverage of high quality

These variables are provided by the Quality of Government (QoG) dataset. In addition to what is presented

here, we examine a large array of other variables reported by the QoG dataset and ﬁnd no eﬀect on any of them.

These results are available upon request.

The estimates are slightly diﬀerent from Table 4 column (2) because the speciﬁcation is diﬀerent.

newspapers will have a larger impact on reader perceptions because readers place more trust in high

quality newspapers. In the context of our study and model, it can also be because readers who are

interested in foreign news may be more likely to read newspapers with a good reputation for foreign

news coverage.

To explore this, we examine the correlations of the estimated coeﬃcients across for each newspa-

per with a proxy for the quality of foreign journalism: the number of Pulitzer Prizes in International

Reporting for each newspaper during 1976-2005.

Table 6 columns (1)-(10) present the estimated

eﬀects on the New York Times, Washington Post, Wall Street Journal, The Chicago Tribune and

The L.A. Times. For each paper, we show both the speciﬁcation with the uninteracted Council

eﬀect as well as the baseline speciﬁcation. We will focus our discussion on the interaction term in

the baseline speciﬁcation shown in the even numbered columns. The estimated interaction eﬀect is

negative for all papers, but varies in magnitude and precision.

Next, we use a bivariate regression to estimate the correlations between the estimated eﬀects for

each paper and the quality measure. Figure 4a plots the respective regression line and data points.

Consistent with our framework, it shows that the degree of distortion (the absolute value of the

interaction coeﬃcient) is larger for newspapers with more Pulitzers.

7.2 Distortion and Reader Preferences

Given that news is determined both by reader demand and government distortion, another ques-

tion that arises from our results is whether the degree of distortion is aligned with reader preferences

across papers. For example, it may be easier for a strongly anti-communist government to distort

newspapers with more conservative readers. To explore this, we examine the correlations of the esti-

mated coeﬃcients with a ranking of reader attitude according to the 2008 Mondo Conservativeness

Rating.

As before, we use a bivariate regression to estimate the correlations between the estimated eﬀects

It is straightforward to extend the framework presented in Section 3 to incorporate newspaper quality eﬀects.

This is available upon request.

The data are shown in Appendix Table A.3. The number of Pulitzer Prizes are from

http://www.pulitzer.org/bycat/International-Reporting. Our results are similar if we use the number of prizes from

larger or smaller windows of time. Similarly, the qualitative point is unchanged if we use prizes from all categories.

These results are available upon request.

Note that there are fewer observations for the Chicago Tribune because it is not available for text analysis for

all years in the sample. Our main results examining all articles are similar if we omit the Tribune from our sample.

Thus, they are not driven by the selection of newspapers into the ProQuest database.

The data are shown in Appendix Table A.3.

for each paper shown in Table 6 and the conservativeness ranking. Figure 4b plots the respective

regression lines and data points. It suggests that government distortion was larger in magnitude

is less conservative newspapers.

This is consistent with the model in that the U.S. government

will attempt to distort news where it can aﬀect the opinions of the highest number of U.S. voters,

if readers of conservative newspapers are more likely to support the incumbent government despite

the human rights abuses of its allies (e.g., conservatives may value the objectives of the Reagan era

policy more than other voters).

Note that the estimates of the correlations across papers should be interpreted loosely as merely

suggestive for several reasons. First, we observe very few newspapers. Second, the bivariate cor-

relations do not take into account the fact that the degrees of distortion are estimates. For these

two reasons, we do not present the standard errors of the bivariate correlations. Finally, the Mondo

conservativeness measure is from many years after the period of our study and may not accurately

represent the views of readers during the Cold War.

7.3 Newswires and U.K. Newspapers

Since a signiﬁcant amount of foreign news in newspapers is originally reported by newswires

such as the Associated Press, we examine the eﬀect on newswires in Table 6 columns (11)-(12).

The estimates for both the interacted and uninteracted terms are similar to our baseline estimate

in sign; they are large in magnitude and statistically signiﬁcant at the 1% level.

In order to use all of the available data, we also collect data on the number of news stories about

human rights abuses published in a United Kingdom newspaper, The Guardian. Columns (13)-(14)

show that there is no eﬀect on coverage in this U.K. newspaper. While there are many potential

explanations for the diﬀerence in coverage between U.S. and U.K. newspapers, these results are

consistent with the two mechanisms highlighted in our model, since the British government has

plausibly less leverage over U.S. newspapers and U.K. readers are likely to have diﬀerent interests

from American readers.

We also ﬁnd that the degree of distortion is uncorrelated with the slant measure from Gentzkow and Shapiro

(2006). These are not presented for brevity and available upon request.

8 Conclusion

This paper explores the possibility that the government can systematically distort news coverage

from independently owned media outlets in the United States. Using data from the latter part of

the Cold War, we ﬁnd that membership on the Security Council increased news coverage of the

human rights abuses of foreign countries when they are not politically allied to the United States.

In contrast, for countries that are strongly allied to the United States, membership reduced news

coverage of bad behavior. We argue that these results are consistent with news coverage being

driven by two forces: government distortion and reader interests. The government distortions that

we detect in the data dissipate with the end of the Cold War. This could be due to the inability of

the U.S. government to sustain manipulation, or to a shift in political alliances after the Cold War

that causes our measure of Cold War alliance to no longer be meaningful.

These results provide novel and rigorous evidence that government distortion can systematically

exist (albeit for a ﬁnite period of time) in a highly competitive media market amongst independently

owned media. For policymakers and practitioners, our results produce mixed forebodings. On the

one hand, they conﬁrm anecdotal evidence that there is systematic government distortion in U.S.

media content and are consistent with the fear that government distortion can impede the media’s

ability to monitor the government on behalf of its readers. On the other hand, the fact that our

results dissipate after the Cold War suggests that perhaps government distortion would not have

been sustainable over time. It is beyond the scope of our study to be conclusive on this point.

The results and limitations of our study suggest several avenues for future research. We need to

determine whether systematic government distortion of independently owned media in democracies

has existed outside of our context. Another closely related subject of research is to understand

the conditions under which a democratically elected government can systematically distort the

news. Our ﬁnding that there is distortion during the Cold War together with the recent work of

Gentzkow, Petek, Shapiro, and Sinkinson (2012), which ﬁnds that the government in power has

little eﬀect on news composition in the historical U.S. context, suggests that there is variation in

government distortion even within one political regime. Better understanding of the determinants

of a democratically elected government’s ability to distort the news is an important avenue of future

research.

Caution is needed for extrapolating our estimates to other countries or time periods. The

estimates we obtain are speciﬁc to the context of our study. Nevertheless, it is important to

note that the United States has one of the largest and most competitive media markets in the

world (Djankov and Shleifer, 2003). Given the argument that government distortion decreases with

market competition and independent ownership, the degree of government distortion in the United

States can potentially be interpreted as the lower bound of the extent of government distortion in

other democratic regimes.

References

Alesina, A., and D. Dollar (2000): “Who Gives Foreign Aid to Whom and Why?,” Journal of

Economic Growth, 5(1), pp. 33–63.

Bennet, Lance, L. R., and S. Livingston (2007): When the Press Fails: Political Power and

the News Media from Iraq to Katrina. University of Chicago Press, Chicago.

Bennett, W., R. Lawrence, and S. Livingston (2008): When the Press Fails: Political Power

and the News Media from Iraq to Katrina, Studies in Communication, Media, and Pub. University

of Chicago Press.

Berger, D., W. Easterly, N. Nunn, and S. Satyanath (2012): “Commercial Imperialism?

Political Inﬂuence and Trade During the Cold War,” The American Economic Review.

Berger, D., W. Easterly, and S. Satyanath (2009): “Was Lenin Right? US Political Inﬂuence

and International Trade during the Cold War Era,” Discussion paper, Harvard University Working

Paper.

Besley, T., and A. Prat (2006): “Handcuﬀs for the Grabbing Hand? Media Capture and

Government Accountability,” The American Economic Review, 96(3), pp. 720–736.

Blanton, S. L., and R. G. Blanton (2007): “What Attracts Foreign Investors? An Examination

of Human Rights and Foreign Direct Investment,” The Journal of Politics, 69(1), pp. 143–155.

Brooks, J. (1987): “Letter to The Honorable Dante B. Fascell (on the Investigation of the Oﬃce

of Public Diplomacy),” Discussion paper.

Carleton, D., and M. Stohl (1985): “Foreign Policy of Human Rights: Rhetoric and Reality

from Jimmy Carter to Ronald Reagan,” Human Rights Quarterly, 7(2), pp. 205–229.

Chiang, C.-F., and B. Knight (2011): “Media Bias and Inﬂuence: Evidence from Newspaper

Endorsements,” Review of Economic Studies, 78(3), 795–820.

Cohen, J. (2001): “The Return of Otto Reich: Will government propagandist join Bush Adminis-

tration?,” Discussion paper, Fairness and Accuracy in Reporting.

Cook, T. (1998): Governing with the News: The News Media as a Political Institution, Studies in

Communication, Media, and Pub. University of Chicago Press.

DellaVigna, S., and E. Kaplan (2007): “The Fox News Eﬀect: Media Bias and Voting,” Quar-

terly Journal of Economics, 122(3), pp. 1187–1234.

Djankov, Simeon, M. C. N. T., and A. Shleifer (2003): “Who Owns the Media?,” Journal of

Law and Economics, 46(2), pp. 341–81.

Dube, A., E. Kaplan, and S. Naidu (2011): “Coups, Corporations, and Classiﬁed Information,”

The Quarterly Journal of Economics, 126(3), pp. 1375–1409.

Durante, R., and B. Knight (2012): “Partisan Control, Media Bias, And Viewer Responses:

Evidence From Berlusconi’S Italy,” Journal of the European Economic Association, 10(3), 451–

481.

Enikolopov, Ruben, P. M., and E. Zhuravskaya (2011): “Media and political persuasion:

Evidence from Russia,” The American Economic Review, 101(7), pp. 3253–3285.

Gbadolite, A. Z. (2001): “Leaving Fire in His Wake,” TIME Magazine.

Gentzkow, M. (2006): “Television and Voter Turnout,” The Quarterly Journal of Economics,

121(3), pp. 931–972.

Gentzkow, M., N. Petek, J. M. Shapiro, and M. Sinkinson (2012): “Do Newspapers Serve

the State? Incumbent Party Inﬂuence on the US Press, 1869-1928,” NBER Working Papers

18164, National Bureau of Economic Research, Inc.

Gentzkow, M., and J. M. Shapiro (2006): “What Drives Media Slant? Evidence from US

Daily Newspapers,” Discussion paper, National Bureau of Economic Research Working Paper

No. 12707.

(2008): “Preschool Television Viewing and Adolescent Test Scores: Historical Evidence

from the Coleman study,” The Quarterly Journal of Economics, 123(1), pp. 279–323.

Gerber, A., D. Karlan, and D. Bergan (2009): “Does the Media Matter? A Field Experi-

ment Measuring the Eﬀect of Newspapers on Voting Behavior and Political Opinions,” American

Economic Journal: Applied Economics, 1(2), pp. 35–52.

Graber, D. A. (2006): Mass media and American politics, Mass Media and American Politics.

CQ Press.

Hadenius, A., and J. Teorell (2005): “Cultural and Economic Prerequisites of Democracy:

Reassessing Recent Evidence.,” Studies in Comparative International Development, 39(4), 87–

106.

Hamilton, L. H., and D. K. Inouye (????): “Report of the Congressional Committees Investigat-

ing the Iran-Contra Aﬀair with Supplemental, Minority, and Additional Views, 100th Congress,

1st Session,” Discussion paper.

Jacobwitz, J. (1985a): “Duties of TDY Military Personnel (Memo to Ambassador Reich),” Dis-

cussion paper, National Security Archives.

(1985b): “Public Diplomacy Action Plan. Support for the White House Educational Cam-

paign,” Discussion paper, National Security Archives.

Jensen, R., and E. Oster (2009): “The Power of TV: Cable Television and Women’s Status in

India,” The Quarterly Journal of Economics, 124(3), pp. 1057–1094.

Kuziemko, I., and E. D. Werker (2006): “How Much Is a Seat on the Security Council Worth?

Foreign Aid and Bribery at the United Nations,” Journal of Political Economy, 114(5), pp. 905–

930.

La Ferrara, E., A. Chong, and S. Duryea (????): “Soap Operas and Fertility: Evidence from

Brazil,” Discussion paper, Working Paper, Department of Economics, University of Bocconi,

Milan, Italy.

Malone, D. M. (2000): “Eyes on the Prize: The Quest for Nonpermanent Seats on the UN Security

Council,” Global Governance, 6(1), pp. 3–23.

(2004): The UN Security Council: From the Cold War to the 21st Century, A project of

the International Peace Academy. Lynne Rienner.

Miller, J. (2001): “White Propaganda, An Operation (Memo to Mr. Pat Buchanan),” Discussion

paper, National Security Archives.

Mitchell, N. J., and J. M. McCormick (1988): “Economic and Political Explanations of Human

Rights Violations,” World Politics, 40(4), 476–498.

Olken, B. A. (2009): “Do television and radio destroy social capital? Evidence from Indonesian

villages,” American Economic Journal: Applied Economics, 1(4), 1–33.

O’Neill, B. (1996): “Power and Satisfaction in the United Nations Security Council,” The Journal

of Conﬂict Resolution, 40(2), pp. 219–237.

Paluck, E. L. (2009): “Reducing intergroup prejudice and conﬂict using the media: a ﬁeld exper-

iment in Rwanda,” Journal of Personality and Social Psychology, 96(3), pp. 574–587.

Parry, R., and P. Kornblub (1988): “Iran-Contra’s Untold Story,” Foreign Policy, (72), pp.

3–30.

Poe, Steven C., C. S. C., and T. C. Vazquez (2001): “How Are These Pictures Diﬀerent? A

Quantitative Comparison of the US State Department and Amnesty International Human Rights

Reports, 1976-1995,” Human Rights Quarterly, 23(3), pp. 650–677.

Prat, A., and D. Stromberg (2005): “Commercial television and voter information,” Discussion

paper, CEPR Discussion Paper No. 4989.

(2011): “The Political Economy of Mass Media,” Discussion paper, Mimeo, LSE and IIES.

Qian, N., and D. Yanagizawa-Drott (2009): “The Strategic Determinants of US Human Rights

Reporting: Evidence from the Cold War,” Journal of the European Economic Association, 7(2-3),

pp. 446–457.

Schultz, G. P. (1984): “New Coverage of Central America (Memo to The President),” Discussion

paper, National Security Archives.

Stromberg, D. (1999): The Politics of Public Spending. Princeton University.

(2004a): “Mass Media Competition, Political Competition, and Public Policy,” Review of

Economic Studies, 71(1), pp. 265–284.

(2004b): “Radio’s Impact on Public Spending,” The Quarterly Journal of Economics,

119(1), pp. 189–221.

Thomas, H. (2006): Watchdogs of Democracy? The Waning Press Corps and How It Has Failed

the Public. Scribner, New York.

Voeten, E. (2001): “Outside Options and the Logic of Security Council Action Source,” The

American Political Science Review, 95(4), pp. 845–858.

Wood, R. M., and M. Gibney (2010): “The Political Terror Scale (PTS): a Re-introduction and

a Comparison to CIRI,” Human Rights Quarterly, 32(2), pp. 367–400.

Appendix: proof of Lemma 1

Proof. Let q

∗

(N) be the solution to (5) and ˆq

(N) be deﬁned by (4) for a given N. We will show

that ˆq

(N) − q

∗

(N) → 0 a.s. as N → ∞, which is the formal sense in which lemma 1 holds.

Let

N(N) be the number of countries that commit human rights violations. By the law of large

numbers, the fraction of countries which commit human right violations,

N(N)/N → 1/2 a.s., as

N → ∞. For the rest of the proof, we drop the explicit conditioning on N, but q

∗

, q

N are all

understood to be functions of N.

The proof goes in two steps. First, we will show that out of countries that commit human rights

violations, it is not optimal to distort at least half of them by more than some small δ. The cost of

suppressing ˆq

− q

∗

news is

T (ˆq

− q

∗

) ≥ (p − d) m

(ˆq

− q

∗

) − γ

(ˆq

)

−

∗

)

= (ˆq

− q

∗

)

(p − d) m

−

(ˆq

+ q

∗

)

= (ˆq

− q

∗

)

(p − d) m

−

(2ˆq

+ (q

∗

− ˆq

))

(ˆq

− q

∗

)

where in the third equality we used the fact that ˆq

solves (4). Pick some δ > 0. If ˆq

− q

∗

> δ,

then T (ˆq

− q

∗

) ≥

. Suppose there are at least

N/2 countries which committed human rights

violations and for which the equilibrium distortion is ˆq

− q

∗

> δ. Then the total cost of bribing,

N+1

n=1

T (ˆq

− q

) ≥

Nγ

. Since the total beneﬁt of remaining in power is R, if N (and hence

is suﬃciently large, then

Nδ

> R and the sum of bribes exceeds the total beneﬁt of remaining in

power, which is suboptimal. Therefore, for at least

N/2 countries in equilibrium ˆq

− q

∗

< δ.

Now we can show that it is not optimal to distort any country not on the UNSC by more than

δ. Without any loss of generality, we show that it is not optimal to do so for country 1 because the

optimal news distortion for this country should satisfy





∗

}

N+1

n=1





1 − s



∗

}

N+1

n=1





R − T (ˆq

− q

∗

)

≥





ˆq

, {q

∗

}

N+1

n=2





1 − s



ˆq

, {q

∗

}

N+1

n=2





where the right hand side represents beneﬁts of not distorting country 1 at all, and the left hand

side is the beneﬁt from distorting. Alternatively, it can be written as



∗

}

N+1

n=1



− s



ˆq

, {q

∗

}

N+1

n=2

i



−



R (7)

≥ T (ˆq

− q

∗

) .

For any combination of news coverage {q

} , the fraction of uninformed agents in group 1 is

given by



}

N+1

n=1



= [1 − (1 − max {q

− p, 0}) × ... × (1 − max {q

N+1

− p, 0})] m

+ (q

N+1

− p) (m

+ m

) .

Pick δ <

(p − d) m

. From the ﬁrst step, if N is suﬃciently large, there are

N/2 countries for

which 1 ≥ q

∗

− p ≥ ˆq − p − δ > 0 where ˆq =

(p − d) m

. Therefore, for any q

, s



, {q

∗

}

N+1

n=2



≥

1 − (1 − (ˆq − p − δ))

N/2

which goes to m

N → ∞. This implies that s



∗

}

N+1

n=1



−



ˆq

, {q

∗

}

N+1

n=2



→ 1 and hence from (7) ˆq

− q

∗

→ 0 as N → ∞.

Table 1: Descriptive Statistics

Variable Obs Mean

Std. Dev.

Obs Mean

Std. Dev.

U.S. Alliance 1001 0.089 0.071 1296 0.092 0.073

UNSC Dummy Variable 1001 0.068 0.252 1296 0.059 0.236

USSD PTS 1001 2.677 0.953 1296 3.025 1.066

Amnesty PTS 1001 3.031 0.978 1296 3.005 1.082

USSD-Amnesty PTS 1001 -0.354 0.748 1296 0.019 0.697

HR News Dummy, All Newspapers

1001 0.602 0.490 1296 0.771 0.420

HR News, All Newspapers

1001 7.778 19.70 1296 11.78 26.26

HR W Post 1001 2.080 5.488 1296 4.326 10.81

HR NYT 1001 2.542 6.063 1296 2.449 5.365

HR WSJ 1001 0.244 0.825 1296 1.064 2.764

HR L.A. Times 1001 2.283 6.179 1296 3.938 9.356

HR C. Tribune (1976-1986) 812 0.776 2.737

HR Newswires 1001 6.030 19.66 1296 14.81 33.53

HR The Guardian (1976-2003)

1001 0.436 1.247 1115 0.704 2.058

A. Cold War 1976-88

B. Post Cold War 1992-2005

The observations are at the country-year level.

Table 2: Security Council Membership, U.S. Alliance and News Coverage of Human Rights Abuses

Ln HR Ln HR Ln HR Ln HR HR HR Dummy Ln HR Ln HR Ln HR Ln HR Ln HR Ln HR

Sample: Full Full Full

Full Full Full

At least 1

HR Article

this Year

At least 1

HR Article

during

1976-88

Omit top

and bottom

5% Avg.

PTS

Reagan

Years

(1980-88)

Ever On

UNSC

Full

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

Dependent Var. Mean 1.16 1.16 1.16 1.16 7.78 0.60 1.94 1.36 1.18 1.17 1.14 1.16

U.S. Alliance x UNSC -6.88 -5.04 -6.29 -6.62 -134.01 -2.15 -6.69 -7.85 -7.15 -5.87 -6.12

(5.662) (2.935) (2.301) (2.367) (79.706) (1.673) (2.597) (2.357) (2.480) (2.330) (2.387)

UNSC 0.69 0.43

(0.467) (0.226)

U.S. Alliance 3.08

(1.767)

U.S. Alliance, Least Allied Quintile x UNSC 0.64

(0.319)

U.S. Alliance, 2nd Quintile x UNSC 0.00

(0.123)

U.S. Alliance, 3rd Quintile x UNSC 0.18

(0.206)

U.S. Alliance, 4th Quintile x UNSC -0.13

(0.178)

U.S. Alliance, Most Allied Quintile x UNSC -0.15

(0.070)

Controls:

Year FE Y Y Y Y Y Y Y Y Y Y Y Y

Country FE N Y Y Y Y Y Y Y Y Y Y Y

UNSC x Year FE N N Y Y Y Y Y Y Y Y Y N

U.S. Alliance x Year FE N N N Y Y Y Y Y Y Y Y N

Observations 1,001 1,001 1,001 1,001 1,001 1,001 603 859 836 766 660 1,001

R-squared 0.039 0.754 0.758 0.757 0.589 0.564 0.707 0.725 0.755 0.813 0.780 0.755

Effect of UNSC for at 75th vs. 25th

Percentile Ally

-0.43 -0.32 -0.40 -0.42 -8.44 -0.14 -0.42 -0.49 -0.45 -0.37 -0.39

The standard errors are presented in the parentheses. They are clustered at the country level. At the bottom of the table, we report the difference in the estimated effect of UN Security

Council membership for countries on the 75th and 25th of the U.S. Alliance distribution. Unless if otherwise stated, the sample comprises a panel of countries for the years 1976-1988.

The number of observations varies across columns due to the availability in the dependent variables.

Dependent Variable: The Number of Human Rights News Articles

Table 3: Security Council Membership, U.S. Alliance and News Coverage of Human Rights Abuses – Robustness to Controls

Sample: Full Full Full Full Full Full Full Full

In Sample

for 5+

Years

In Sample

for 10+

Years

In Sample

for 13

Years

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

U.S. Alliance x UNSC -6.63 -7.23 -8.03 -6.71 -9.15 -10.41 -5.38 -6.05 -6.76 -7.23 -18.45

(2.368) (4.001) (4.020) (2.347) (3.450) (3.702) (2.619) (2.110) (2.406) (2.625) (5.580)

Controls

Region FE x Year FE N Y N N N Y N N N N N

Governance Indicators x Year FE N N Y N N Y N N N N N

Avg. Amnesty PTS x Year FE N N N Y N Y N N N N N

Avg GDP x Year FE N N N N Y Y N N N N N

Country-Year Trends N N N N N N Y N N N N

Ln HR News Lagged 1, 2, 3 years N N N N N N N Y N N N

Observations 1,001 979 961 1,001 944 926 1,001 600 979 775 160

R-squared 0.761 0.771 0.776 0.768 0.779 0.799 0.809 0.810 0.756 0.747 0.722

Dependent Variable: Ln Human Rights News Articles

All regressions include country fixed effects, year fixed effects and the interaction of year fixed effects with Council membership and U.S. alliance.

In columns (3) and (6), the governance indicators are the average of the Freedom House Civil Liberties index, the Freedom House Political Rights

index, and the Polity2 index; each index is interacted with year fixed effects. The standard errors are presented in the parentheses. They are

clustered at the country level. Unless if otherwise stated, the sample comprises a panel of countries for the years 1976-1988. The number of

observations varies across columns due to the different availability of the control variables.

Table 4: Security Council Membership, U.S. Alliance and Human Rights Practices as reported by the U.S. State Department and Amnesty

International

USSD Amnesty

USSD -

Amnesty

USSD -

Amnesty

USSD -

Amnesty

USSD -

Amnesty

USSD -

Amnesty

USSD -

Amnesty

Sample: Full Full Full Full

Omit top and

bottom 5%

Avg. PTS

Reagan

Years

(1980-88)

Ever On

UNSC

Full

(1) (2) (3) (4) (5) (6) (7) (8)

Dependent Var. Mean 2.68 3.03 -0.35 -0.35 -0.38 -0.27 -0.36 -0.35

U.S. Alliance x UNSC -3.11 2.37 -5.48 -5.40 -5.42 -5.68 -4.63

(4.280) (3.338) (2.592) (2.382) (2.118) (2.357) (2.475)

UNSC 0.10 -0.30 0.40

(0.314) (0.263) (0.208)

U.S. Alliance -1.78 0.61 -2.39

(1.174) (1.209) (0.509)

U.S. Alliance, Least Allied Quintile x UNSC 0.47

(0.329)

U.S. Alliance, 2nd Quintile x UNSC 0.06

(0.099)

U.S. Alliance, 3rd Quintile x UNSC 0.11

(0.243)

U.S. Alliance, 4th Quintile x UNSC -0.02

(0.167)

U.S. Alliance, Most Allied Quintile x UNSC -0.70

(0.071)

Controls:

Year FE Y Y Y Y Y Y Y Y

Country FE N N N Y Y Y Y Y

UNSC x Year FE N N N Y Y Y Y N

U.S. Alliance x Year FE N N N Y Y Y Y N

Observations 1,001 1,001 1,001 1,001 836 766 660 1,001

R-squared 0.050 0.009 0.148 0.381 0.370 0.413 0.355 0.370

Effect of UNSC for at 75th vs. 25th Percentile Ally -0.20 0.15 -0.35 -0.34 -0.34 -0.36 -0.29

The standard errors are presented in the parentheses. They are clustered at the country level. At the bottom of the table, we report the difference in the

estimated effect of UN Security Council membership for countries on the 75th and 25th of the U.S. Alliance distribution. Unless if otherwise stated, the

sample comprises a panel of countries for the years 1976-1988.

Dependent Variable: PTS Index for Human Rights Practices

Table 5: Security Council Membership, U.S. Alliance and Institutional Outcomes

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

Dep. Var. Mean 3.03 3.03 5.09 5.09 5.08 5.08 -4.03 -4.03 0.16 0.16 0.00 0.00

U.S. Alliance x UNSC 1.32 0.96 -0.85 -0.61 -4.06 -0.44 -10.58 -2.45 -0.43 -0.51 -3.63 -0.36

(2.820) (2.894) (3.785) (3.781) (3.906) (3.792) (17.917) (16.535) (0.627) (0.575) (3.930) (3.975)

UNSC -0.12 -0.10 0.07 0.29 0.04 0.07

(0.235) (0.212) (0.305) (1.106) (0.038) (0.282)

Controls:

Country FE Y Y Y Y Y Y Y Y Y Y Y Y

Year FE Y Y Y Y Y Y Y Y Y Y Y Y

UNSC x Year FE N Y N Y N Y N Y N Y N Y

U.S. Alliance x Year FE N Y N Y N Y N Y N Y N Y

Observations 1001 1001 906 906 906 906 943 943 836 836 750 750

R-squared 0.62 0.62 0.84 0.84 0.83 0.83 0.81 0.81 0.86 0.86 0.85 0.86

Standardized Effect, U.S. Alliance x UNSC 0.027 0.019 -0.011 -0.011 -0.041 -0.008 -0.032 -0.017 -0.020 -0.025 -0.040 -0.006

The standard errors are presented in the parentheses. They are clustered at the country level. At the bottom of the table, the standardized effect is the standard

deviation change in the dependent variable as a result of a one standard deviation change in U.S. Alliance x UNSC. The sample comprises a panel of countries for

the years 1976-1988. The number of observations varies across columns due to the different availability of the dependent variables.

Freedom House

Civil Liberties

Index

Freedom House

Political Rights

Index

UCDP/PRIO Civil

Conflict Dummy

1st Principal

Component of All

Outcomes

Amnesty

International PTS

Index

Polity 2 Index

Dependent Variable

Table 6: Security Council Membership, U.S. Alliance and News Coverage of Human Rights Abuses Across Newspapers

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)

Dep. Var. Mean, level 2.08 2.08 2.54 2.54 0.24 0.24 2.28 2.28 0.78 0.78 6.03 6.03 0.44 0.44

U.S. Alliance x UNSC -4.23 -4.78 -4.99 -6.35 -3.16 -3.67 -1.77 -3.50 -3.00 -4.55 -4.31 -5.71 -0.52 -2.02

(1.618) (2.038) (2.334) (1.839) (1.683) (1.624) (3.079) (2.455) (2.278) (2.578) (3.802) (3.812) (1.880) (1.729)

UNSC 0.26 0.38 0.30 0.24 0.24 0.37 0.11

(0.151) (0.174) (0.147) (0.254) (0.198) (0.308) (0.155)

Controls:

Country FE Y Y Y Y Y Y Y Y Y Y Y Y Y Y

Year FE Y Y Y Y Y Y Y Y Y Y Y Y Y Y

UNSC x Year FE N Y N Y N Y N Y N Y N Y N Y

U.S. Alliance x Year FE N Y N Y N Y N Y N Y N Y N Y

Observations 1,001 1,001 1,001 1,001 1,001 1,001 1,001 1,001 812 812 1,001 1,001 1,001 1,001

R-squared 0.672 0.681 0.677 0.688 0.435 0.471 0.657 0.670 0.561 0.577 0.693 0.701 0.469 0.5

Effect of UNSC for at 75th

vs. 25th Percentile Ally

-0.91 -1.02 -1.07 -1.36 -0.68 -0.79 -0.38 -0.75 -0.64 -0.97 -0.92 -1.22 -0.11 -0.43

The standard errors are presented in the parentheses. They are clustered at the country level. At the bottom of the table, we report the difference in the estimated effect of UN

Security Council membership for countries on the 75th and 25th of the U.S. Alliance distribution. The sample comprises a panel of countries for the years 1976-1988. The

number of observations varies across columns due to the availability in the dependent variables.

NY Times

Wall St. Journal

LA Times

Chicago Tribune

Newswires

The Guardian

(U.K.)

Washington Post

Dependent Variable: Ln Human Rights News Articles

Figure 1: Coeﬃcients of Years Since Security Council Membership Dummy Variables × U.S. Alliance

on News Coverage

(a) Ln Number of Articles on Human Rights, Cold War and Post-Cold War Periods

(b) Ln Number of Articles on Human Rights by Paper, Cold War

Figure 2: Coeﬃcients of Years Since Security Council Membership Dummy Variables × U.S. Alliance

on Human Rights Practices

(a) USSD - Amnesty PTS, Cold War and Post-Cold War Periods

(b) USSD PTS and Amnesty PTS, Cold War Only

Figure 3: The Correlates of Government Distortion

(a) Estimated Distortion and Quality as measured by the number of Pulitzer Prizes in

International Reporting 1976-2005

(b) Estimated Distortion and Reader Attitude as measured by the Mondo Conservativeness

Rating in 2008

Table A.1: UN Security Council Members

Year Country Year Country Year Country Year Country

1976 Benin 1983 Zaire/DRC 1992 Cape Verde 1999 Argentina

1976 Guyana 1983 Guyana 1992 Ecuador 1999 Brazil

1976 Libya 1983 Jordan 1992 Hungary 1999 Gabon

1976 Pakistan 1983 Malta 1992 India 1999 Gambia, The

1976 Panama 1983 Nicaragua 1992 Venezuela 1999 Malaysia

1976 Romania 1983 Pakistan 1992 Zimbabwe 2000 Argentina

1976 Tanzania 1983 Poland 1993 Brazil 2000 Bangladesh

1977 Benin 1983 Togo 1993 Cape Verde 2000 Jamaica

1977 India 1983 Zimbabwe 1993 Djibouti 2000 Malaysia

1977 Libya 1984 Burkina Faso 1993 Hungary 2000 Mali

1977 Mauritius 1984 Egypt, Arab Rep. 1993 Pakistan 2000 Tunisia

1977 Pakistan 1984 India 1993 Venezuela 2001 Bangladesh

1977 Panama 1984 Malta 1994 Argentina 2001 Colombia

1977 Romania 1984 Nicaragua 1994 Brazil 2001 Jamaica

1977 Venezuela 1984 Pakistan 1994 Djibouti 2001 Mali

1978 Bolivia 1984 Peru 1994 Nigeria 2001 Mauritius

1978 Gabon 1984 Zimbabwe 1994 Oman 2001 Tunisia

1978 India 1985 Burkina Faso 1994 Pakistan 2002 Bulgaria

1978 Mauritius 1985 Egypt, Arab Rep. 1994 Rwanda 2002 Cameroon

1978 Nigeria 1985 India 1995 Argentina 2002 Colombia

1978 Venezuela 1985 Madagascar 1995 Botswana 2002 Guinea

1979 Bangladesh 1985 Peru 1995 Honduras 2002 Mauritius

1979 Bolivia 1985 Thailand 1995 Indonesia 2002 Mexico

1979 Gabon 1985 Trinidad and Tobago 1995 Nigeria 2002 Syrian Arab Republic

1979 Jamaica 1986 Bulgaria 1995 Oman 2003 Angola

1979 Nigeria 1986 Congo, Rep. 1995 Rwanda 2003 Bulgaria

1979 Zambia 1986 Ghana 1996 Botswana 2003 Cameroon

1980 Bangladesh 1986 Madagascar 1996 Chile 2003 Chile

1980 Jamaica 1986 Thailand 1996 Egypt 2003 Guiinea

1980 Mexico 1986 Trinidad and Tobago 1996 Guinea-Bissau 2003 Mexico

1980 Niger 1986 Venezuela 1996 Honduras 2003 Pakistan

1980 Philippines 1987 Argentina 1996 Indonesia 2003 Syrian Arab Republic

1980 Tunisia 1987 Bulgaria 1996 Poland 2004 Algeria

1980 Zambia 1987 Congo, Rep. 1997 Chile 2004 Angola

1981 Mexico 1987 Ghana 1997 Costa Rica 2004 Benin

1981 Niger 1987 Venezuela 1997 Egypt 2004 Brazil

1981 Panama 1987 Zambia 1997 Guinea-Bissau 2004 Chile

1981 Philippines 1988 Algeria 1997 Kenya 2004 Pakistan

1981 Tunisia 1988 Argentina 1997 Poland 2004 Phillipines

1981 Uganda 1988 Brazil 1998 Brazil 2004 Romania

1982 Zaire/DRC 1988 Nepal 1998 Costa Rica 2005 Algeria

1982 Guyana 1988 Senegal 1998 Gabon 2005 Benin

1982 Jordan 1988 Yugoslavia/FYR 1998 Gambia, The 2005 Brazil

1982 Panama 1988 Zambia 1998 Kenya 2005 Phillipines

1982 Poland 2005 Romania

1982 Togo

1982 Uganda

Cold War 1976-1988

Post Cold War 1992-2002

Table A.2: Year-by-Year Estimates

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18)

coef se coef se coef se coef se coef se coef se coef se coef se coef se

U.S. Alliance x Years

since UNSC Dummy

x 2 Years before

2.640

(2.098) 0.491 (2.042) -2.149 (1.999) -3.850 (2.759) -0.849 (2.500) -0.415 (1.670) -1.914 (1.027) -5.042 (2.023) 0.125 (1.696)

x 1 Year before

1.427

(2.387) 0.449 (2.227) -0.978 (2.130) -0.272 (2.380) -1.388 (1.736) 0.731 (2.001) -0.576 (1.355) -0.431 (2.213) 1.063 (2.169)

x Election year

2.573

(3.246) 3.691 (3.366) 1.118 (3.468) -4.149 (3.702) -5.019 (2.232) 0.0407 (3.752) -4.382 (1.577) -2.000 (3.204) -2.893 (3.024)

x 1st Year on Council

1.128

(3.876) -3.287 (2.202) -4.415 (3.045) -11.84 (2.889) -9.445 (2.945) -8.242 (3.096) -5.545 (2.012) -7.743 (3.182) -6.753 (4.162)

x 2nd Year on Council

2.474

(3.345) -4.493 (3.083) -6.968 (3.048) -4.963 (4.097) -7.029 (3.392) -1.515 (2.434) -3.646 (1.898) -1.268 (3.328) -3.412 (1.896)

x 1 Year after

-0.160 (2.350) 0.0532

(2.507) 0.213 (2.687) -5.665 (4.852) -5.792 (3.443) -2.710 (1.984) -0.457 (1.454) -2.665 (3.376) -1.398 (2.524)

x 2 Years after

0.329

(2.921) 0.127 (3.062) -0.203 (4.829) -4.907 (3.345) -6.504 (3.014) -1.484 (1.694) -1.576 (1.427) 0.492 (3.024) -2.817 (1.710)

Observations

1,001 1,001 1,001 1,001 1,001 1,001 1,001 1,001 812

R-squared

0.632 0.679 0.389 0.765 0.694 0.684 0.482 0.676 0.585

U.S. Alliance x Years

since UNSC Dummy

x 2 Years before

-0.0418 (1.642) -0.178 (1.470) -0.137 (1.758) 0.415

(1.826) -0.474 (1.707) 1.905 (1.374) 0.0777 (1.121) 1.574 (1.981)

x 1 Year before

2.154

(1.391) 0.870 (1.526) -1.284 (1.423) 1.406 (1.574) -1.316 (1.791) 1.634 (1.451) -1.136 (0.955) 1.876 (1.752)

x Election year

0.391

(1.819) 0.191 (1.312) -0.200 (1.412) 1.146 (1.876) -0.400 (1.411) 0.704 (1.739) -0.0673 (1.227) 1.645 (1.734)

x 1st Year on Council

-2.052 (1.765) 1.214

(1.292) 3.266 (1.736) -1.180 (1.879) -0.589 (1.996) 0.478 (1.755) 1.919 (1.352) -1.359 (1.492)

x 2nd Year on Council

1.369

(2.510) 0.213 (2.097) -1.156 (1.672) -0.106 (2.031) 0.826 (1.844) 1.900 (2.254) -1.124 (1.557) -1.497 (1.811)

x 1 Year after

-0.985 (1.790) -1.051 (1.949) -0.0654 (1.634) -1.826 (2.247) -1.954 (1.744) 0.573

(2.493) -0.687 (1.187) 1.324 (2.025)

x 2 Years after

-0.191 (2.165) -1.875 (2.058) -1.684 (2.051) -2.912 (2.554) -2.509 (2.342) -0.938 (2.495) 1.160

(1.737) -0.852 (1.772)

1,296 1,296 1,296 1,296 1,296 1,296 1,296 1,296

0.676 0.753 0.209 0.806 0.679 0.716 0.586 0.752

All regressions also control for the number of years since Council membership dummy variables (2 years before, 1 year before, election year, first year on Council, second year on Council, first year after,

second year after), year and country fixed effects, the interaction of Council membership dummy with year fixed effects and the interaction of U.S. alliance with year fixed effects. Standard errors are

presented in parenthese. They are clustered at the country level.

B. Post Cold War 1992-2005

Amnesty PTS

USSD PTS

USSD - Amnesty

PTS

Ln All HR

A. Cold War 1976-1988

Dependent Variable

Ln Tribune

Ln NYT

Ln WP

Ln WSJ

Ln LAT

Table A.3: Pulitzer Prizes and Reader Attitude Ratings

Newspaper

# of Pulitzer Prizes in International

News Reporting

Mondo Conservativeness

Ranking (2008)

NY Times 12 4

Wash. Post 6 3

WSJ 4 2

Chicago Tri. 2 1

LA Times

2 5

Notes: Coefficient of UNSC x U.S. Alliance is from the baseline specification. See Table 5.

Figure A.1: The Fraction of Divided Votes and Votes with the United States in the UN General

Assembly

Figure A.2: The Number of Newspaper Articles on Human Rights Abuses for each Day since the

Release of U.S. State Department and Amnesty Reports (Averages across years, 1976-1988)

Figure A.3: The Eﬀects of U.S. Alliance × Years Since Council Membership Dummy Variables with

95% Conﬁdence Intervals, Cold War (1976-88)

(a) News Coverage (Ln Number of Articles on Human Rights)

(b) Human Rights Practices (USSD PTS - Amnesty PTS)