New Explorations into International Relations: Democracy, Foreign Investment, Terrorism, and Conflict

By Seung-Whan Choi

This book addresses a range of issues surrounding the search for scientific truths in the study of international conflict and international political economy. Unlike empirical studies in other disciplines, says Seung-Whan Choi, many political studies seem more competent at presenting theoretical conjecture and hypotheses than they are at performing rigorous empirical analyses. When we study global issues like democratic institutions, flows of foreign direct investment, international terrorism, civil wars, and international conflict, we often uncritically adopt established theoretical frameworks and research designs. The natural assumption is that well-known and widely cited studies, once ingrained within the tradition of the discipline, should not be challenged or refuted.

However, do such noted research areas reflect scientific truth? Choi looks closely at ten widely cited empirical studies that represent well-known research programs in international relations. His discussions address such statistical and theoretical issues as endogeneity bias, model specification error, fixed effects, theoretical predictability, outliers, normality of regression residuals, and choice of estimation techniques. In addition, scientific progress made by remarkable discoveries usually results from finding a new way of thinking about long-held scientific truths, therefore Choi also demonstrates how one may search for novel ideas at minimal cost by developing new research designs with original data.

Here is a valuable resource for students, scholars, and policy makers who want to quickly grasp the evolutionary pattern of scientific research on democracy, foreign investment, terrorism, and conflict; build their research designs and choose appropriate statistical techniques; and identify their own agendas for the production of cutting-edge research.

• Language: English
• Publisher: University of Georgia Press
• Release date: Mar 15, 2016
• ISBN: 9780820349060

Seung-Whan Choi

SEUNG-WHAN CHOI is an associate professor of political science at the University of Illinois at Chicago.

Book preview

NEW EXPLORATIONS INTO INTERNATIONAL RELATIONS

SERIES EDITORS

William W. Keller

Professor of International Affairs, Center for International Trade and Security, University of Georgia

Scott A. Jones

Director of Export Control Programs, Center for International Trade and Security, University of Georgia

SERIES ADVISORY BOARD

Pauline H. Baker

The Fund for Peace

Eliot Cohen

Paul H. Nitze School of Advanced International Studies, Johns Hopkins University

Eric Einhorn

Center for Public Policy and Administration, University of Massachusetts, Amherst

John J. Hamre

The Center for Strategic and International Studies

Josef Joffe

Hoover Institution, Institute for International Studies, Stanford University

Lawrence J. Korb

Center for American Progress

William J. Long

Sam Nunn School of International Affairs, Georgia Institute of Technology

Jessica Tuchman Mathews

Carnegie Endowment for International Peace

Scott D. Sagan

Center for International Security and Cooperation, Stanford University

Lawrence Scheinman

Monterey Institute of International Studies, CNS-WDC

David Shambaugh

The Elliott School of International Affairs, George Washington University

Jessica Stern

FXB Center, Harvard School of Public Health

New Explorations into International Relations

Democracy, Foreign Investment, Terrorism, and Conflict

Seung-Whan Choi

Capitalist Peace, Democratic Peace, and International War first appeared as Re-Evaluating Capitalist and Democratic Peace Models in International Studies Quarterly (2011, pp. 759–69). This material is reproduced with permission of John Wiley and Sons, Inc.

© 2016 by the University of Georgia Press

Athens, Georgia 30602

www.ugapress.org

All rights reserved Set in Minion Pro by Graphic Composition, Inc., Bogart, Georgia Printed and bound by Sheridan Books, Inc. The paper in this book meets the guidelines for permanence and durability of the Committee on Production Guidelines for Book Longevity of the Council on Library Resources.

Most University of Georgia Press titles are

available from popular e-book vendors.

Printed in the United States of America

20 19 18 17 16 P 5 4 3 2 1

Library of Congress Cataloging-in-Publication Data

Names: Choi, Seung-Whan, author.

Title: New explorations into international relations : democracy, foreign investment, terrorism, and conflict/Seung-Whan Choi.

Description: Athens, Georgia : The University of Georgia Press, 2016. | Series: Studies in security and international affairs | Includes bibliographical references and index.

Identifiers: LCCN 2015023652| ISBN 9780820349077 (hardcover : alk. paper) | ISBN 9780820349084 (pbk. : alk. paper) | ISBN 9780820349060 (ebook)

Subjects: LCSH: International relations. |

World politics. | Democracy. | Investments, Foreign. | Terrorism.

Classification: LCC JZ1305.c46 2016 |

DDC 327—dc23

LC record available at http://lccn.loc.gov/2015023652

To Shali Luo, with everlasting love and appreciation

Science is perhaps the only human activity in which errors are systematically criticized and, in time, corrected.

KARL POPPER

The only way that science can make progress is by showing that theories are wrong.

DAVID L. GOODSTEIN

CONTENTS

List of Tables and Figures

Preface

Introduction

PART I: Double Take

CHAPTER 1. Democracy, Ethnicity, Religion, and Civil War: Endogeneity Bias

CHAPTER 2. Capitalist Peace, Democratic Peace, and International War: Model Specification Errors

CHAPTER 3. A Reanalysis of the Selectorate Model: Fixed Effects, Heteroskedasticity, and Autocorrelation

CHAPTER 4. Examining the Predictability of the Selectorate Theory: Which Aspect of Democracy Explains Better, the Winning Coalition or Civil Liberties?

CHAPTER 5. Democracy, Foreign Direct Investment, and Outliers

CHAPTER 6. Explaining the Foreign Direct Investment-Democracy Controversy: Normality of Regression Residuals

CHAPTER 7. Terrorism and Zero-Inflated Negative Binomial Regression: A Mismatch between Theory and Statistical Model

CHAPTER 8. Democracy and Transnational Terrorism Revisited: Rule of Law

PART II: Searching for New Ideas and Empirical Evidence

CHAPTER 9. Old Habits Die Hard: Leaders’ Prior Military Experience, Repression, and Civil War

CHAPTER 10. Democracy, Status Quo, and Military Manpower Systems

CHAPTER 11. Selectorate Theory, Democracy, and Terrorism: Null Results

CHAPTER 12. The Political Economy of Foreign Direct Investment: Democracy, Economic Crisis, and Domestic Audience Benefits

CHAPTER 13. The United States’ Use of Military Force and Terrorism

Conclusion

Notes

References

Index

TABLES AND FIGURES

Table 1.1. Simultaneous Analysis of Civil War Onset and Democracy

Table 1.2. Simultaneous Analysis of Civil War Onset, Ethnicity, N*, Polarization, Exclusion, and Religion

Table 1.3. Simultaneous Analysis of Ethnic War Onset

Table 1.4. Simultaneous Analysis of Ethnic War Onset: Robustness Tests

Table 2.1. Logit Regression of Liberal Variables on International Conflict

Table 2.2. Logit Regression of Liberal Variables on International Conflict: Politically Relevant Dyads

Appendix Table 2.1. Logit Regression of Liberal Variables on Militarized Interstate Disputes: Replications

Appendix Table 2.2. When Peace Years Are Included

Table 3.1. Reanalysis of the Selectorate Model

Table 4.1. Reanalysis of the Selectorate Model and Democracy

Table 4.2. Comparison of the Coefficient Sizes of W and Civil Liberties

Table 4.3. Examination of the Coefficient Sizes of W and Civil Liberties

Table 5.1. Descriptive Statistics without Outliers

Table 5.2. Descriptive Statistics with Outliers

Table 5.3. The Effect of Democratic Institutions on Inflows of FDI: Replications

Table 5.4. The Effect of Democratic Institutions on Inflows of FDI: China and Botswana Dummies

Table 5.5. The Effect of Democratic Institutions on Inflows of FDI: Robust Regression

Table 5.6. Li and Resnick’s Model 1: Measuring the Dependent Variable in Jensen’s FDI/GDP Ratios

Appendix Table 5.1. Li and Resnick’s Model without PROPERTY RIGHTS PROTECTION

Table 6.1. Univariate Normality Analysis of Observations: Shapiro-Francia Test

Table 6.2. Univariate Normality Analysis of Regression Residuals: Shapiro-Francia Test

Table 6.3. Heteroskedasticity and Autocorrelation in the Seven Regression Models

Table 6.4. The Effect of Democracy on Inflows of FDI: Prais-Winsten Regression

Table 6.5. Substantive Effects of FDI Inflows

Table 6.6. The Effect of Democracy on Inflows of FDI: Fixed Effects

Table 6.7. The Effect of Democracy on Inflows of FDI: Dynamic Panel-Data Estimation

Table 6.8. The Effect of Democracy on Inflows of FDI: Robust Regression

Table 8.1. Effects of the Rule of Law on Transnational Terrorist Incidents within Countries, 1984–1997

Table 8.2. Effects of Contract-Intensive Economy and the Rule of Law on Transnational Terrorist Incidents within Countries

Table 8.3. Marginal Effects of Statistically Significant Variables

Table 8.4. Marginal Effects of Statistically Significant Variables: Robustness Tests

Table 8.5. Effects of the Rule of Law on Transnational Terrorist Casualties within Countries, 1984–1997

Table 9.1. Leaders’ Prior Military Experience, Political Terror, and Civil War Onset

Table 9.2. Leaders’ Prior Military Experience, Physical Integrity Rights, and Civil War Onset

Table 9.3. Leaders’ Prior Military Experience, Repression, and Civil War Onset: Generalized Estimating Equations

Table 9.4. Leaders’ Prior Military Experience, Repression, and Civil War Onset: Rare Event Logit

Appendix Table 9.1. Leaders’ Prior Military Experience, Repression, and Civil War Onset: More Controls at the First Stage

Table 10.1. Predicting the Likelihood of Conscription, 1886–1992

Table 11.1. Size of the Winning Coalition, Executive Constraints, and Terrorism, 1970–2000

Table 11.2. Substantive Effects of Statistically Significant Variables

Table 11.3. Size of the Winning Coalition, Executive Constraints, and Terrorism, 1970–2000: Fixed Effects

Table 12.1. The Effect of Domestic Audience Costs on FDI Inflows as Economic Conditions Change

Table 13.1. The Effect of U.S. Military Interventions on Terrorism, 1970–2005

Table 13.2. Two-Step Analysis of U.S. Military Interventions and Terrorism, 1970–2005

Table 13.3. Domestic versus International Terrorism, 1970–2005

Table 13.4. Terrorism within versus Terrorism outside the Intervened Country, 1970–2005

Appendix Table 13.1. The Effect of Each Type of U.S. Military Mission on Terrorism, 1970–2005

Figure 2.1. An Interaction Effect between GDPPC (Low) and Contiguity: Politically Relevant Dyads

Figure 5.1. OLS Regression Lines

Figure 5.2. Partial Regression Plot: Li and Resnick’s Model

Figure 5.3. Partial Regression Plot: Jensen’s Model

Figure 12.1. Marginal Effect of Audience Costs on FDI Inflows as Economic Conditions Change

PREFACE

As political scientists, we purport to explain and predict important events in the contemporary world and to offer policy recommendations regarding those events. Achieving these goals requires empirical research that is thorough and rigorous; however, we are, of course, fallible and unknowingly fail to obtain scientific truth from time to time—there is always room for improvement. As such, replication projects are an efficient means to disseminate and increase scientific knowledge because they build upon the previously established work of fellow scientists. Ideally, we could improve the quality of even the most authoritative studies of political science in an effort to advance scientific progress and provide better policy recommendations while at the same time testing new research ideas with simple but innovative methods. In such vein, this book sets out to gather improved scientific knowledge by testing theories of classic international relations while engaging in original research with new data in the hopes that such a dual approach may lead to scientific breakthroughs in the future and change the political world for the better.

This book can be a valuable resource for future generations of researchers—including advanced undergraduate and graduate students—by shortening the trial and error processes of their own analyses and by providing examples of well-written empirical papers. While future researchers read through each of the replication or original research chapters in this book, I hope that they will learn how to avoid the kinds of obstacles that I encountered myself, as explained below. This book seeks to quickly connect their statistical skills with their political topics, to perform rigorous empirical research, and to engage in replication and original research projects on important political questions. In addition, I hope that the reader will have a better understanding of political science in general and of international relations in particular after being exposed to the diverse and controversial issue areas covered in this book. As a collection of replications and original research on politically important topics, there is no published book comparable where the reader can—in a single volume—grasp the four salient issue areas of the contemporary political world: democracy, foreign investment, terrorism, and conflict.

The writing of this book was a long intellectual journey, involving the resolution of several obstacles. The first was to find a way to apply econometrics and statistical techniques to political questions. I learned my empirical skills from the econometricians and statisticians at the Department of Economics and the Department of Statistics at the University of Missouri-Columbia (MU). Thus, my graduate training provided a wonderful opportunity to acquire highly advanced statistical methods but offered virtually no exposure to quantitative political literature. Furthermore, the fact that the MU Department of Political Science was then dominated by qualitative scholars did not aid me in the struggle to connect my statistics knowledge with traditional political science. At the same time, I was laboring with several theoretically important questions that would later become the centerpiece of my dissertation project. Fortunately, those troubles quickly faded when MU hired Patrick James—my dissertation supervisor—one and a half years before my graduation day. With his caring advice and support, I was able to connect the dots in my dissertation and earn my doctorate within twelve months.

The second obstacle I dealt with was learning how to appropriately conduct rigorous research. Although I felt like I had mastered the state-of-the art theories of econometrics and statistics, textbooks did not teach me how to play with real world data and statistical programming. My earlier data analyses in graduate school were often careless, but I eventually acquired sufficient hands-on experience learning how to begin and end rigorous research on various economic issues. Due to my strong background in econometrics, I was hired by MU’s Department of Economics and later by the Missouri State Government’s Department of Social Services Research and Evaluation Unit. Working under meticulous economists these several years, I was further instructed on methods of conducting rigorous data analysis, such as identifying the correct estimation methods for different economic problems, performing addition tests for robustness, and double-checking potential errors.

The third obstacle I faced was determining how to train political science graduate students in statistics at the University of Illinois at Chicago (UIC). Although I was hired as an international relations specialist, I was asked to teach two consecutive statistical methods courses. However, because many students expressed an interest in qualitative methods over quantitative ones, they were not all that eager to learn statistical theories and techniques. However, when I started to assign replication projects, and to dramatically reduce my lectures on statistical proofs and properties, I found that they learned better, and more, because they could work on a research topic of replication that they chose themselves. Even many qualitative students developed keen research interests in quantitative fields because they were, through replication projects, exposed to major empirical studies in urban politics, American politics, and international relations.

While doing the research for this book, I have received valuable feedback and help from many people. In particular, by reading through the entire manuscript, Nora Willi has provided me with indispensable research assistance. Special thanks are due to Anahit Gomtsian, Patricia Hajek, and Joshua Pakter who kindly reviewed draft portions of the book. I am especially indebted to David Carment, Constantine P. Danopoulos, Paul F. Diehl, Douglas M. Gibler, Patrick James, John R. Oneal, James A. Piazza, Jeffrey Pickering, Bruce Russett, George Tsebelis, Douglas Van Belle, and John Vasquez, who not only greatly influenced my thinking on the various political and economic issues of this book, but who have also given me very helpful advice on academic life in general. I also owe thanks to Dennis R. Judd, Evan McKenzie, and Dick W. Simpson in the UIC Department of Political Science for providing me with a graduate assistant. Last, but certainly not least, Walter Biggins and Beth Snead at the University of Georgia Press deserve credit for their continued guidance on how best to navigate the jungle of book publication.

Introduction

Gray’s Anatomy, a textbook originally written by Henry Gray in 1858, is widely regarded as the most influential work on the subject of human anatomy. This work describes the morphology of the human body and details the process of dissection for the purpose of aiding the scientific study of the structure, position, and interrelation of its various parts. Every accomplished physician, especially surgeons and doctors working in certain diagnostic specialties such as histopathology and radiology, must possess a thorough working knowledge of human anatomy. Just as understanding the process of dissection of the human body is an essential step toward understanding the functions of specific organs and structures in the body, the dissection or interrogation of major findings in empirical political research must be the first step toward an increase in our collective scientific knowledge. Marcus Cicero, the famous Roman political theorist, once said, by doubting we arrive at the truth. With this aphorism in mind, the eight chapters in part 1 of this book question, or dissect, several influential theories and empirical findings that heretofore have been widely accepted in the discipline of political science in general and in the field of international relations in particular. In addition, because the scientific progress made by remarkable discoveries usually results from finding a new way of thinking about what we have long taken for granted as scientific truth, the five chapters of part 2 explore completely new research ideas and seek to identify undiscovered empirical regularities. The expectation is that this book will point us toward fresh findings and insights in the four salient issue areas of international relations: democracy, foreign direct investment, terrorism, and conflict.

Unlike Marcus Cicero, contemporary political scientists tend to shy away from voicing doubt and raising critical questions. In contrast, researchers from some other fields of social science have done much to improve the rigor of their scientific analyses. For example, in their book Betrayers of Truth: Fraud and Deceit in the Halls of Science, Broad and Wade (1983) showed no hesitation in pointing to certain studies in which scientists fudged data or misrepresented results. In 1986, the American Economic Review (AER)¹ published Dewald, Thursby, and Anderson’s NSF-funded article examining the role of replication in empirical economic papers that had been previously accepted by the Journal of Money, Credit, and Banking for publication. The main findings of the AER article suggest that inadvertent errors in published empirical articles are a commonplace rather than a rare occurrence (587–88). In their own reexamination of 1,148 estimated results reported in forty-nine articles in two prestigious psychology journals, Wicherts, Bakker, and Molenaar (2011) found the reluctance to share data to be associated with weaker evidence . . . and a higher prevalence of apparent errors in the reporting of statistical results. Undoubtedly, publications like these caused some controversy; however, they also significantly contributed to the increased rigor in their respective empirical research traditions (G. King 1995). In particular, their insistence that the replication of research is indispensable to the soundness of all scientific investigation has been adopted as an axiom within their respective scholarly communities.²

Unlike empirical research in economics and psychology, political studies have not yet been subjected to this same degree of rigorous scrutiny; more importantly, many of those engaged in political scientific research may be unfamiliar with its presumably rigorous empirical tradition. The growing popularity of empirical political research is based on the belief that any number of political researchers should be able to obtain the same objective empirical evidence. But what would happen if what appears to be objective evidence is, in fact, a statistical artifact or the result of a flawed research design? It would be undesirable, of course, if the replication results were inconsistent with the claims of previous research; therefore, it is necessary to investigate the possible causes of discrepancy, to offer improved estimation methods, and to report new findings. In the attempt to make a meaningful contribution to the scientific advancement of the discipline as a whole, a researcher should consider developing his or her own agenda by engaging in a replication of a previous study. Furthermore, because many potential researchers lack sufficient institutional support and cannot afford the luxury of collecting a costly dataset for each new research project, they should consider an alternative way in which a replication project can offer the opportunity to advance a body of scientific knowledge at minimal cost. This book (1) reanalyzes several authoritative empirical political studies in the attempt to illustrate the ease with which erroneous statistical conclusions may be unknowingly drawn due to a lack of rigorous research traditions and the insufficient scrutiny of previous studies, and (2) demonstrates how one may search for novel ideas at minimal cost by developing new research designs with original data. In so doing, this book offers original findings and insights as well as relevant policy implications regarding the causes and effects of the outbreak of civil war, international conflict, acts of terrorism, democratic governance, and inflows of foreign direct investment.

By demonstrating the way erroneous data analysis can drastically diminish the validity of published empirical studies, this book hopes to give the next generation of scholars the opportunity and encouragement to improve their empirical research so that it can aid scientific progress and, even, revolution. The accurate analysis of data, using standardized statistical methods, is a critical step in determining the soundness of empirical work and, therefore, in avoiding erroneous inferences and conclusions. For example, ordinary least squares (OLS) regression models make a number of assumptions about independent variables, dependent variables, and the relationships between them; yet, if a researcher ignores or violates the assumptions of this statistical model for some unknown reason, his or her estimated results will be biased and inconsistent, and, thus, the findings will be misleading. For example, it would be incorrect to choose an OLS estimator when the mean of the dependent variable is, perhaps due to the presence of outliers, not a linear combination of the parameters (regression coefficients) and the independent variables. Furthermore, OLS regression models are used under the assumption that each variable on the right-hand side of the equation exerts an independent effect on the dependent variable; accordingly, it would be inappropriate to employ a single OLS equation as opposed to a simultaneous equations model when, due to the presence of reverse causality, that assumption does not hold up. Simply put, accurate statistical analysis is always essential to the soundness of empirical political research, as books and journal articles that lack an empirical foundation are of little use to the advancement of scientific knowledge in the discipline of political science.

Naturally, political scientists will begin an investigation by posing an important research question (e.g., do ethnic and religious antagonisms cause the outbreak of civil war?) that they will treat as the object of their statistical data analysis. They will then develop a theory regarding the question and, based on that theory, will draw certain hypotheses or predictions that can be tested against a set of data that they or someone else has collected. The outcomes of these tests may or may not support the initial theory upon which the hypotheses and predictions were based. Yet, in comparison with empiricists in other disciplines, political scientists appear to be more competent in terms of producing theoretical conjectures/arguments and drawing hypotheses than they are at conducting rigorous empirical analyses. It may be that the relatively late infusion of statistical methods into this discipline is responsible for its emphasis on theory and its weak progress in terms of rigorous data analysis; in extreme cases, there are some political scientists who consider statistical technique an alien language, giving little credit to the interpretation of estimated coefficients and standard errors. An unfortunate trend in political science has allowed researchers who offer persuasive theoretical discussions and speculations, but who conduct less than adequate data analyses, a better chance to publish manuscripts than those researchers whose focus is primarily on the accuracy of their statistical inferences. If we seek, as we well should, the development of political science as a serious discipline, then it is imperative that all researchers be fluent in the language of statistics as well as the language of theory building.

There are those political researchers who believe that a paper should be published on the basis of a strong overall argument or on the strength of the theoretical reasoning behind its empirical modeling. These researchers tend to treat the data analysis component as an afterthought, failing to perform a careful examination of their estimated results, let alone of the validity of their primary data source. On the other hand, when a paper has a somewhat loose theoretical development but a rigorous statistical analysis, its publication is dubious, as some political scientists dismiss estimated coefficients as meaningless. As a result, those book manuscripts and journal papers that are primarily data oriented have a slimmer chance of appearing in the most prestigious scholarly outlets than those oriented toward theory building. While other social scientific disciplines, such as economics and sociology, have been moving toward a more serious consideration of heavily empirical works even with little or no theory, political science maintains its bias against the publication of quantitative studies with a weak theory. Paradoxically, the effect of this bias has been to allow the publication of political science papers with subpar, often flawed, statistical analyses; this is an injustice to the discipline as a whole. It may also indicate that the statistical skills of reviewers—especially at those stages of review where they do not have access to the data and statistical programs—are not sophisticated enough to locate these kinds of empirical flaws, thereby continuing the trend of publication described above.

Once an empirical study is widely circulated and cited, its findings tend to acquire a nearly law-like status, which, once established, is rarely challenged. Furthermore, as replication projects are much less likely to be published than original works, political scientists have less incentive to closely examine the research designs and estimated results of previously published studies. This creates an undesirable research culture that does a disservice to the discipline of political science by dis-incentivizing the interrogation of faulty findings. It is well known that data-driven research progresses scientific knowledge by demonstrating the existence of previously unknown phenomena, by illuminating new relationships between existing phenomena, or by discovering that some widely shared understanding is either incomplete or entirely wrong. In the sixteenth century, for example, scholars such as Galileo Galilei would have been prosecuted had they spoken the scientific truth about heliocentrism versus the well-established theory of geocentrism. Or consider the supersession of Newton’s two-hundred-year-old theory of mechanics by Albert Einstein’s theory of relativity at the beginning of the twentieth century; it is also worth noting that even though the theory of relativity is now considered to be a cornerstone of modern physics, not all of Einstein’s ideas and concepts were eagerly accepted in his time. In general, it appears that contemporary academics have forgotten Karl Popper’s (1963, 216) insight in Conjectures and Refutations: The Growth of Scientific Knowledge that science is . . . perhaps the only [human activity]—in which errors are systematically criticized and, in time, corrected. This is why we can say that, in science, we often learn from our mistakes. It seems, though, that contemporary academic culture views being critical as stepping on the toes of other scholars rather than as an essential part of scientific inquiry. Many political scientists are apprehensive about getting their hands dirty; this attitude impedes the scientific advancement of the discipline.

Of course, political empiricists had their moment in the mid-1990s when replication projects were accepted and encouraged as a legitimate and cost-effective way to publish for academic achievement and career advancement. During this period, a large number of empirical political studies were scrutinized, faulty estimations were discussed publicly, and improved models were suggested. Indeed, Gary King (1995, 445) emphasized that the most common and scientifically productive method of building on existing research is to replicate an existing finding—to follow the precise path taken by a previous researcher, and then improve on the data or methodology in one way or another. However, the emphasis on replication has faded away or at least drastically diminished since then; qualitative and even quantitative scholars ridicule and marginalize replicated works.³ Replicated studies are too often rejected for publication on the grounds that they fail to provide stand-alone research, even in cases where they clearly demonstrate a refutation of the major findings from previously published books or journals. This serious and all too common bias against replications deprives researchers of the opportunity to learn from one another’s mistakes and help each other proceed to a higher stage of scientific discovery.

As replication projects usually rely on the research design of published studies, they are frequently charged with a lack of stand-alone research. However, it should be noted that while replication projects do use data from previously published works, they usually collect original data with reference to a new and theoretically interesting variable. The charge that replication studies produce no stand-alone research is ironic in the sense that most empirical research already relies on publicly available data sources (perhaps excepting the main variable of contribution). Stand-alone researchers claim to be doing original work, but their data often comes from collections previously published by private and government agencies; furthermore, their estimation methods rarely differ from those of existing studies. While researchers from prestigious institutions are provided with ample material support and are, therefore, more capable of collecting original data, those researchers who are less connected regularly struggle to secure research time and funding. For such researchers, replication is an ideal scientific project with a minimal cost. For example, the availability of the data and research design of John Oneal and Bruce Russett’s (2005) democratic peace research program has encouraged the proliferation of work on international conflict. Similarly, Oneal and Russett’s statistical model has been widely replicated for years, thus minimizing the possibility of coding errors or faulty model building; ultimately, it has proven to be highly reliable compared to other conflict models.

Replication studies are rarely published in top-tier journals because this would open up a ‘cheap’ way for authors to have their work published . . . and every Tom, Dick, and Harriet . . . could potentially seek to replicate some study, just to get published (Ishiyama 2014, 82). The appropriate response to this criticism would be to question if it is a good idea to publish empirical studies with incorrect or even fabricated data analysis, the findings of which, without rigorous scrutiny, are then allowed to become principles or laws of political research. This practice is as insulting to the scientific community as it is injurious to the accumulation of knowledge. Empirical studies without sound empirical results are like swimming pools without water during hot summer days or fancy sport cars without gasoline during the Fédération Internationale de l’Automobile World Endurance Championship. Those who remain skeptical of the legitimacy of replication projects should ask themselves this question: which of the above approaches—fabrication or replication—is really the cheap way for authors to publish and get promoted? While the former, in fact, makes no scholarly contribution and should—as is commonplace in other disciplines—be swiftly retracted (Tabuchi 2014), the latter contributes to the degree of scientific rigor by pointing out mistakes or questionable methodological choices in published research and also reports new substantive findings on the subject in question.

Perhaps one reason that replication projects are discouraged has to do with the interests of the empiricists themselves. That is, the authors of the original studies may be the loudest voices objecting to replicated pieces that are critical of their published work, as they are unlikely to be comfortable with a project that scrutinizes their statistical models or estimation methods. When a replication study points out a flaw in an empirical model, its originator is likely to resist on the grounds that the replication lacks theoretical innovation. Of course, this is an odd defense considering that the stated goals of such a critique are precisely to demonstrate how the main findings of the published study are unsubstantiated and to suggest an improved or corrected empirical research strategy.

Although the lack of publication opportunities discourages the proliferation of replication projects, some journals do, in fact, require that researchers post their replication data and program files online (see James 2003; Lupia and Elman 2014). This requirement is intended to increase the transparency of empirical research. However, there is a danger that subsequent researchers will use these materials simply to reproduce identical models without performing a careful examination of the estimation methods and measurements; in this case they are prone to repeat mistakes, making the open source policy a moot point.

Although replication projects do continue to appear in some publication outlets, these are highly infrequent. This book argues, therefore, that journal editors and book publishers must provide a greater opportunity for researchers to engage in the critical reexamination of published works. Scholarly exchange requires the ability to constructively criticize and publish improved or corrected versions of original works. This is the only way to increase our discipline’s cumulative knowledge and to navigate the path toward scientific discovery, and perhaps even scientific revolution. As Patrick James (2003, 85) correctly points out, replication is a scientific ideal, but it also turns out to be good for scholars in practical, even career-oriented ways.

The body of this book is twofold. Part 1 consists of eight chapters that scrutinize ten empirical political studies by Bueno de Mesquita et al. (2003), Fearon and Laitin (2003), Gartzke (2007), Hoffman, Shelton, and Cleven (2013), Jensen (2003), Li (2005), Li and Resnick (2003), Piazza (2011), Santifort-Jordan and Sandler (2014), and Savun and Phillips (2009). Most of these ten works represent well-known research programs in political science and have been widely cited over the last decade. Fearon and Laitin’s (2003) article is a must-read for anyone who researches civil war; Gartzke’s (2007) piece makes one of the most interesting contributions to the international conflict literature; Bueno de Mesquita et al.’s (2003) book is extensively recognized for rigorous formal modeling and a wide range of empirical applications in studies of democracy and public policy; the two articles by Li and Resnick (2003) and Jensen (2003) are pioneering research in the area of foreign direct investment; the four articles by Hoffman, Shelton, and Cleven (2013), Piazza (2011), Santifort-Jordan and Sandler (2014), and Savun and Phillips (2009) employ zero-inflated negative binomial regression, an increasingly popular estimation technique, to the study of terrorist activity; and Li’s (2005) piece is one of the most frequently cited studies of terrorism. Each of the eight chapters addresses important statistical and theoretical issues including endogeneity bias, model specification error, fixed effects, theoretical predictability, outliers, and normality of regression residuals. The new statistical analyses in these chapters indicate that the findings and conclusions of the original ten studies are unreliable because the estimated results are likely to be biased and inconsistent. Simply put, these chapters reveal the danger that presents itself when the statistical estimations of empirical studies are insufficiently rigorous.

Part 2 comprises five chapters that aim to locate novel causal factors through the development of original research programs with new datasets in the same research areas as part 1 (i.e., democracy, foreign investment, terrorism, and conflict). Although part 2 includes some discussion on statistical issues, its main purpose is to explore new research ideas. The four research areas may appear to be unconnected at a glance, but each of the five chapters addresses the question of how democracy is related to each of these salient political phenomena. For example, chapter 13 examines the question of whether the foreign policy behavior of the U.S. government—which is considered one of the most democratic regimes worldwide—leads to an increase or decrease in the rate of international terrorism. Because democracy is an essential part of our political lives in the contemporary world, and because it remains a major focus of research and contention in the discipline of political science, understanding why and how it affects or is affected by different political events (such as acts of the U.S. military, international terrorist groups, and multinational corporations) is crucial and should be carefully analyzed. In so doing, each of the five chapters draws some important policy implications, especially concerning the relation of democracy to foreign direct investment and terrorist activity, with some attention also paid to the relation of democracy to civil and international conflict. Accordingly, part 2 is designed to help the reader understand the general patterns and related policy implications of the most controversial political issue areas in the context of democratic governance.

Together, parts 1 and 2 lead the reader to new discoveries and insights regarding four contentious research topics: the causes and effects of civil and international war; growing terrorist threats; democratic governance; and inflows of foreign direct investment. This book intends to benefit students, scholars, and policy makers who are interested in quickly grasping the evolutionary pattern of scientific research on these important political issues, in building their research designs and choosing appropriate statistical techniques, and in identifying their own agendas for the production of cutting-edge research in the above research areas. Most previously published books focus on only one of these research topics. Although these books have their own merits, they fail to provide an overall picture of the various critical issues of the contemporary political world. There are also several edited volumes that may include more than one topic; however, because they are written by several authors, they come from differing perspectives and employ very dissimilar research approaches. In search of scientific knowledge in these four related and crucial research areas, this book offers a single, coherent theme with respect to the interrogation of authoritative empirical studies and in terms of exploring new ideas with original research designs.

Following this introduction, chapter 1 reevaluates Fearon and Laitin’s (2003) civil war model. Because Fearon and Laitin’s study fails to account for endogeneity problems in their single equation logit model, the estimated results are biased at best and inaccurate at worst. More specifically, the use of an erroneous estimation technique leads Fearon and Laitin to conclude that, contrary to popular belief, democracy, ethnicity, and religion are not causes of civil conflict. However, a reexamination, using a simultaneous equations model to correct for the endogeneity bias, provides evidence that these three variables are indeed important determinants of civil violence. While well-established democracies and religiously diversified countries tend to experience fewer civil wars, ethnically diverse countries are more vulnerable to this type of violence.

Chapter 2 reexamines Gartzke’s (2007) capitalist peace model. After replicating Oneal and Russett’s (1997, 1999a) democratic peace model, Gartzke’s study contends that capitalism, and not democracy, leads to peace. Additional research is needed to corroborate, extend, and even refute the findings reported here (180). In response to this open invitation, chapter 2 reevaluates Gartzke’s capitalist peace model along with Oneal and Russett’s democratic peace model. The chapter finds that while the capitalist peace model suffers from misspecification, observation omission, and sample selection bias, the democratic peace model commits measurement error. After correcting these four errors, chapter 2 demonstrates that capitalism