The Quantitative Peace

While the IR people are away, the Americanists will play.

Editor's note: This is a guest post by Joshua N. Zingher, a doctoral candidate in the Department of Political Science at Binghamton University. The content of this post is based on Josh's paper, ‘An Analysis of the Changing Social Bases of America's Political Parties: 1952-2008’, appearing in Electoral Studies. This piece is cross-posted at the London School of Economics' USAPP Blog, which can be found here. 

 

Demographic changes mean that traditional Republican constituencies are shrinking as the Democrats’ grow.

It is difficult to discuss electoral politics in the United States without talking in terms of social groups.  Journalistic accounts of party competition often stress the important role that specific constituencies play (such as the “evangelical” or “Latino” vote) in determining the outcome of presidential elections.  A great deal of political science research has shown that these social cleavages structure party competition.  However, the political importance of these social cleavages can change, which can lead to a reshuffling of the social bases of the party coalitions.  In addition, demographic changes also have the potential to dramatically reshape the political landscape, as new groups enter the electorate and the ratio of existing groups’ changes.  I am interested in how changes in the cleavage structure and changes in the ratio of groups work in conjunction to transform electoral coalitions.  In new research, I investigated two related questions: what group memberships influence individual level voting behavior and how has the number of votes each political relevant group contributes to the party coalitions changed over time?

 

Identifying the Group Bases of Party Support 

So what groups are politically relevant?  To answer this question, I conduct a multivariate statistical analysis utilizing nearly 60 years of American National Election Survey (ANES) spanning 1952 to 2008.  I utilized group memberships as independent variables and vote choice as the dependent variable.  The battery of group memberships included variables for class, race, ethnicity, region, education, birth cohort and age. I define politically relevant group memberships as those that exert a statistically significant influence on individual level vote choice.  I ran a separate regression for each decade, which allowed me to assess whether the effect of specific group memberships have changed over time.  The substantive results of these models are displayed in Figure 1 below.

 

Figure 1 – Predicted effect of group membership on vote choice in presidential elections

 

 

The results of the analysis displayed in Figure 1 demonstrate that the following group memberships lead individuals to be consistently and statistically significantly more likely to support one party over the other: 

 

Table 1: The Core Democratic and Republican Social Groups 

Democratic Groups	Republican Groups
African Americans	Whites
Latinos	Protestants
Catholics	Weekly Church Attendees
Union Members	Southern Whites
Non-Religious	Men
College Graduates
Women Jews

 

This analysis highlights the continuity and change that characterizes the parties’ social bases.  Some groups have been stable members of a party coalition for the last 60 years.  African Americans, Catholics, union members and Jews have all been consistent members of the Democratic coalition while whites and Protestants have reliably favored the Republicans. 

Yet, other groups’ political leanings have changed considerably.  Whites in the Southern United States were once reliably Democratic voters; today they are one of the most solidly Republican constituencies.  The divide between the religiously devout and the secular has also become more politically important in recent elections.  Non-religious voters have become increasingly likely to vote for Democratic candidates while weekly church attendees have become more likely to support Republicans.  Similarly, a partisan “gender gap” emerged in the 1980s.  Men have become significantly more likely to vote for Republican candidates while women have become more likely to support Democrats.  In sum, many traditionally politically important social cleavages have endured (e.g. black/white) while other cleavages have become politically important only in recent decades (secular/religious or female/male).

           

Group Contributions and the Importance of Demographic Changes

The preceding analysis established what group memberships significantly influence vote choice, however, simply understanding what group memberships shape individual level voting behavior paints an incomplete picture of electoral change.  Winning an election is not about having core groups of loyal supporters.  Rather, parties must be able to obtain a majority of votes in order to win elections.  Thus, any process that leads to a party gaining more or less from a specific social group is potentially politically consequential.  Figures 2 and 3 show my calculations for each group’s contribution to the Democrats’ and Republicans’ party coalitions, as measured as a proportion of the party’s total votes. There have been considerable changes in the size of group contributions to the party coalitions—many of which highlight the important role that demographic shifts play in reshaping the party coalitions. 

 

Figure 2 - Group contribution as a proportion of Democrats’ total votes: 1952-2008

 

 

Figure 3 - Group contribution as a Proportion of Republicans’ total votes: 1952-2008

 

The Democratic Party has been obtaining an increasing proportion of its support from groups that are growing as proportion of the electorate, such as Latinos, college graduates and the non-religious. In addition, increasing turnout among Latinos and African Americans has also led to the size of these groups’ contributions to the Democratic coalition.  The Republican Party is being forced to confront the opposite scenario—many Republican groups, whites, Protestants and weekly church attendees, are making up a continuously shrinking proportion of the electorate.  Thus far, the Republican Party has been able to offset demographic changes via high turnout among their core groups as well as through winning an increasingly large majority of the Southern white vote.

So what are the conclusions can be drawn from this analysis? Firstly, the group sources of party support are dynamic, the political relevance of certain social cleavages ebb and flow across decades, which is evidenced by the changing relationship between group membership and vote choice.  Secondly, demographic changes can be politically consequential.  The Republican Party is currently in the undesirable position of having to find new sources of electoral support in order to remain viable in future elections.  The Republican Party’s need to find new sources of electoral support has come about not as a result of the loss of electoral support among loyal groups—rather, the Republican Party must find new sources of electoral support simply because traditionally Republican constituencies are making up a shrinking proportion of the electorate.     

 

 

 

Bipartisanship and the Foreign Policy Bureaucracy

Editor's note: A version of this piece is cross-posted at the LSE USAPP blog, and can be found here. This post is based on my recent article, The International and Domestic Sources of Bipartisanship in U.S. Foreign Policy, in Political Research Quarterly.

---

The idea that foreign policy and national security issues are somehow exempt from the partisan rancor that often characterizes domestic politics is perhaps best encapsulated by the old adage, “politics stops at the water’s edge.” In American foreign policy, nowhere is this spirit better exemplified than by President Franklin Roosevelt’s appointment of Henry L. Stimson as Secretary of War. Since World War II, Stimson and other prominent bureaucratic appointees—such as Robert Lovett and John McCloy—have come to represent this ideal. This phenomenon of bipartisan cooperation that is held to have characterized the early Cold War period has been popularly referred to as the “Cold War Consensus”, or as “Establishment” bipartisanship.

However, following the Vietnam War pronouncements of the collapse of this bipartisan consensus became commonplace. Similar claims followed the end of the Cold War, and more recently, the terrorist attacks of September 11, 2001. Indeed, such assertions have become common in discussions of American foreign policy, with the broader substantive implication being that the conduct of American foreign policy has become increasingly erratic and subject to the kind of partisan political battles that have long afflicted domestic politics. 

Some of my recent work evaluates just how closely these stylized truths match up with reality. Earlier accounts of the breakdown of bipartisanship in foreign policymaking relied largely on anecdotal accounts or individual’s assertions with little in the way of systematic evidence (see here, here, and here). Some work by political scientists has done better in approaching the concepts of bipartisanship and consensus in a more systematic way (see here, here, here, and here). Still, there are some gaps that my work tries to address. First, earlier qualitative works often focused on elite bureaucratic policymakers (like Stimson), while quantitative studies have tended to focus on congressional voting patterns or public opinion. Second, foreign policy scholars often view bipartisan cooperation through the lens of foreign threats and crises, ignoring or downplaying more basic domestic political pressures that may drive bipartisanship. My work aims to apply a more systematic form of analysis to the subject of bipartisanship in presidential appointments, while also assessing the relative importance of domestic and international factors in affecting bipartisan cooperation.

Using data on over 1,000 individual appointees to the foreign policy bureaucracy—agencies like the State Department and the Department of Defense—I identified those individual appointees who were members of the opposition party (e.g., Republican serving in a Democratic administration). Figure 1 shows the number of bipartisan appointees in a given year between 1948 and 2011. While this is only a starting point, it does not lend much support to the conventional narrative regarding the apparent collapse of bipartisanship following either Vietnam or the end of the Cold War. In fact, the early Cold War period sees some of the lowest counts observed in the data.

Figure 1 — Annual count of bipartisan appointees, 1948–2011.

However, Figure 1 does not provide us with any insights into the effects of a wider range of variables that might be influencing bipartisanship. Accordingly, I modeled the probability of observing a bipartisan appointment in a given year as a function of these different forces. Briefly stated, I find no evidence to support the idea that bipartisanship has undergone a structural shift following either Vietnam or the end of the Cold War. In fact, as is displayed in Figure 2, each party’s probability of making a bipartisan appointment has converged when comparing the pre- and post-Vietnam periods. Though it may be contrary to our expectations, this finding makes sense when we consider the domestic political environment of the early Cold War period. The McCarthy hearings, as well as the long absence of Republicans from the White House, help to explain the scarcity of bipartisan appointments during some of the early Cold War years.

 

Figure 2 — Predicted probability of a bipartisan appointment by party and time period

 

 

Similarly, I find no evidence that US conflict involvement promotes bipartisanship. On the contrary, my findings indicate conflict involvement actually has a negative effect on bipartisanship. As Figure 3 shows, the predicted probability of observing a bipartisan appointment decreases as the duration of US war involvement increases. Furthermore, these results actually match the findings of previous studies. Using different metrics, an earlier study by Peter Trubowitz and Nicole Mellow also finds a negative correlation between international crises and bipartisanship in Congressional voting. This negative relationships is likely due to the increasingly contentious and politicized nature of particularly long military engagements.

 

Figure 3 — Predicted probability of a bipartisan appointment as a function of war duration.

 

Lastly, I find that domestic political conditions play an important part in promoting bipartisanship. I argue that bipartisan appointments can serve to compensate presidents for lower levels of partisan support in Congress and aid in the construction of moderate coalitions. However, this dynamic only works when the parties are close and there are moderate legislators in Congress with whom the president and his appointees can work. But as the parties in Congress have become more polarized, moderate legislators have become increasingly scarce. Figure 4 illustrates this dynamic. At low levels of polarization, a decrease in the number of seats that the president’s party controls in Congress leads to an increase in the probability of seeing a bipartisan appointment. As polarization increases the size of this effect becomes smaller, eventually becoming negative.

 

Figure 4 — Marginal effect of a decrease in presidential support in Congress on the probability of a bipartisan appointment

Note: Gray box denotes area of statistical insignificance.

 

Overall, these findings cast doubt on some of the conventional wisdoms surrounding American foreign policy, pointing instead to the importance of the domestic political environment in promoting bipartisanship. Accordingly, my research suggests that we should expect to see less bipartisan cooperation through presidential appointments moving forward. The fact that polarization unfolds over several years at a time also suggests that this trend is unlikely to reverse itself any time soon. President Obama’s retention of Robert Gates, and his appointment of Chuck Hagel as Secretary of Defense, would seem to provide some evidence to the contrary. However, while these appointments are surely important, they are not necessarily indicative of broader trends.

 

Troop Deployment Research and Poland

Interstate conflict is a rare phenomenon. Since 1816, the Correlates of War project only counts 96 different wars occurring over an almost 200 year period.  Given Russia's incursion into the Crimean peninsula, it is of little surprise that IR scholar blogging activity has been rampant the past week; we may just have conflict 97 just around the corner.

It is of little surprise that Ukraine is on the forefront of our discussions across twitter, blogs, and other forms of social media.  It is an intellectually ripe ground for explanation, prediction, and is an important teaching tool for me.  In the classroom, the ongoing crisis is engaging for students.  Their preferred lenses to view the conflict appears to be a mix between neorealist power politics and bargaining models of conflicts; the students found constructivist and liberal explanations to be somewhat less satisfying. Though, there was a healthy discussion about whether or not Russia should count as a democracy. 

Beyond Ukraine, my preperation for conference season is holding my attention.  One manuscript I am presently working on, while keeping an eye on the Ukrainian-Russian crisis, is a collaborative piece with Michael Flynn and Julie VanDusky-Allen.  In a follow up to another work that we have, we are presenting an article at the Midwest Political Science Association that looks as the regional contexts US troop deployments provide for foreign policy decisions of states by employing Lake's 2011 work on contractual hierarchies. He argues that sovereignty is a fungible good that can be traded away in various domains for various side-payments. In the case of the United States, we often provide security services to states (in the form of troops, bases, weapons, etc.) in exchange for more control over their foreign policy decision-making.  Previous research, including our own, shows that for the average state, when troops are deployed to their country, they tend to spend less on their own defense.  In the classic guns-versus-butter framework of government spending, a state granted additional resources can supplement its own defense spending and shift its spending priorities to other social (butter) issues.

However, despite this on-average effect, we see two additional wrinkles to this process.  First, NATO states do not follow this trend. NATO countries tend to respond to new troop deployments with further spending, not reduced spending.  Second, states in the midst of a regional concentrations of US troops, across the board, will increase their spending.  Today, the Obama administration announced that it would be sending military assets as well as 300 troops to Poland. While this move is almost certainly more to assure Poland than it is to threaten Russia, the current crisis will undoubtedly see jumps in military spending by most countries in the region. 

 

Kane's 2006 work on troop deployments is the source for troop deployment data and Michael and I used the Department of Defense's published reports to fill in the years from 2003-2011 in our recent CMPS article. Through 2011, the US had a nominal number of US military personnel in Poland, but the first time the US officially "stationed forces" in Poland happened about 18 months ago in June 2012.  The small number of US military personnel in Poland from 1980-2011 is a relatively insignificant number as the US, currently, has some level of troops in almost every country. These military personnel represent and conduct several kinds of activity and do not represent a deployment that projects US power or capability to coerce.  The announcement by the US today is a significant increase in the size of US troops in Poland beyond previous deployments and represents an operational, fighting force.  It is likely that the number of troops in Poland will stay well above 300 no matter how the Crimean dispute resolves. Ultimately, the troops represent a security commitment to Poland's defense by the United States, and given the regional activity by Russia, the US would be hard pressed to drawn down our forces in Poland.  If anything, I expect the long term number to be higher, perhaps requiring the US to redeploy forces elsewhere in Europe. It will be hard for the US, or John Kerry for that matter, to forget Poland. 

 

 

 

Mapping Torture Allegations Using ITT

In conjunction with a project on state capacity and repression, I've recently been exploring the specific allegation data from Courtenay Conrad and Will Moore's Ill-Treatment & Torture (ITT) Data Collection Project.  ITT's specific allegation data are interestingly different from existing human rights data sets like the one (full disclosure) I co-direct, the CIRI Human Rights Data Project.  Whereas CIRI provides a yearly country score based on a standards-based ordinal coding of the level of government respect for its citizens' right to be free from torture and ill-treatment, the ITT data provide information on all allegations of government torture or ill-treatment made by Amesty International.  As such, the unit of observation for the ITT data is the individual Amnesty International torture allegation. Further, for each allegation, the ITT specific allegation data contain information about the type of abuse, the number of victims, the agents responsible for carrying out the abusive act(s), and the victim type, as well as many other topics.  For more information, the codebook can be found HERE, and a forthcoming Journal of Peace Research article introducing the data, written by Courtenay Conrad, Jillienne Haglund, and Will Moore, can be found HERE.

Figure 1. Total AI Torture Allegations, 1995 - 2005

As is my wont, I decided to get to know the data by making a few maps.  In doing so, I stumbled on some potentially interesting patterns.  First, the map above shows the total number of torture allegations that Amnesty International made against each state over the time period from 1995 to 2005, i.e. the time period covered by the ITT data.  As one might suspect given the research on Amnesty International's information efforts (e.g. Ron, Ramos, and Rodgers 2005; Hendrix and Wong 2013), the number of allegations does not neatly match up with the level of torture in each state, as AI's strategic considerations lead it to focus its attention more on some states rather than others. This problem can be partly remedied by taking into account the approximate size of the affected population in each allegation, which is provided by ITT. However, as Conrad, Haglund, and Moore point out, due to the strategic bias in Amnesty's reporting as well as the impossibility of observing all torture events, even this method is likely to result in a biased undercount of the number of torture events that actually took place in a given country over this time period. There are several ways of attempting to deal with coverage bias in such measures; indeed, Patrick Brandt and John Beieler each recently posted on ways to deal with similar issues in GDELT. For the purposes of this blog post, the maps below will simply display the various attributes of torture allegations as a proportion of the total allegations made against the state. With that in mind, you may want to remember to refer to the map above regularly if some states seem to perform particularly poorly in one area or another, as it could be the result of having been the target of relatively few allegations overall (for instance, see Mongolia).

Figure 2. Type of Torture as a Proportion of Total AI Torture Allegations, 1995 - 2005

(Click the image to enlarge)

The series of maps displayed in Figure 2 show the proportion of total AI torture allegations made up by the four types of abuse distinguished in ITT: scarring torture, stealth torture, ill-treatment, and unstated torture.  It should be noted that abuses are not mutually exclusive; that is, a single AI allegation may contain information on several types of abuse.  As such, the proportions for any individual state will often not sum to 1.  Overall, it appears that stealth torture, i.e. torture that is more difficult to detect after the fact because it leaves no scars on the body, is, on average, less common than the other types of torture and ill-treatment.  Indeed, the average state received only about 22 allegations of stealth torture from 1995-2005, whereas the average state was accused of ill-treatment approximately 48 times, scarring torture about 56 times, and unstated torture around 39 times.

Matching up with work done by Darius Rejali, highly developed states, and particularly highly developed democracies, appear to be more likely to engage in higher proportions of stealth torture than states at lower levels of development. The United States, Canada, Norway, Sweden, Finland, Belgium, Chile, South Africa, and South Korea have all been the targets of fairly high proportions of stealth torture allegations. That said, stealth torture also appears to be relatively more common among Asian states, particularly China, Kazakhstan, Uzbekistan, Myanmar (Burma), Laos, and Vietnam. On the other hand, scarring torture makes up a much lower proportion of allegations in these states. Instead, the proportion of allegations made up by scarring torture is much higher in many Sub-Saharan African states, like Angola, Mozambique, Central African Republic, Niger and Ghana, as well as in Eastern European states, such as Bulgaria, Romania, Hungary, and Slovakia. Finally, it seems we know the least about the type of torture used in North African and Middle Eastern states, with unstated torture accounting for a large proportion of allegations in Algeria, Libya, Egypt, Saudi Arabia, Syria, and Iran.

Figure 3. Agency of Control as a Proportion of Total AI Torture Allegations, 1995 - 2005

(Click the image to enlarge)

The ITT data also provide information on the agency of control, i.e. the "domestic institution and/or agent(s) ... responsible for a given allegation of torture" (Conrad and Moore 2011: 12).  Figure 3 maps the various agencies of control as a proportion of each state's allegations.  As with the various types of torture and ill-treatment, agencies of control are not mutually exclusive; a single torture allegation may name several government agencies as responsible for engaging in abuse.  

Across Figure 3, a few interesting patterns stand out.  While Canada, Australia, and European states are the targets of a relatively moderate number of torture allegations overall, it would appear that the majority of those allegations target torture and ill-treatment carried out by the police. On the other hand, prison abuse appears to account for a relatively high proportion of allegations in the United States, China, and Brazil; unsurprisingly, these states rank 1st, 2nd, and 4th, respectively, in total prison populations. The map of immigration abuse shows that these allegations primarily apply to highly developed democracies, like the U.S., Germany, Italy, Australia, South Korea, and Japan, possibly due to the high demand for admittance into those countries. Abuse by intelligence agents appears to account for a relatively high proportion of allegations in Middle Eastern and North African states, like Egypt, Jordan, Israel, Yemen, and the United Arab Emirates. Intelligence abuse also accounts for a high proportion of allegations in South Korea and Mongolia, although those states were subject to relatively few allegations overall. Finally, the states for which we see high proportions of military and paramilitary abuse align fairly well with the states highlighted in Figure 4 below, which shows the states in which armed conflicts were fought between 1995 and 2005 according to the PRIO Conflict Site data developed by Johann Hallberg.

Figure 4.  Locations of Armed Conflict, 1995 - 2005

(Conflict Sites in Red)

Of course, all of the patterns that I see in these maps could disappear upon controlling for other factors, and there are surely many patterns that I have missed. Primarily, I wanted to draw attention to the wealth of information present in the ITT data. I have only focused on the number of allegations, the type of abuse, and the agency of control; the ITT also contains information about the identities and numbers of victims, whether the abuse occurred within the state or abroad, whether the abuse was investigated, and many other factors. Overall, these data should open the door to many studies of phenomena that we previously could not investigate crossnationally, and I hope we see more projects like the ITT in the future.

 

*NOTE: All maps in this post were made using Weidmann, Kuse, and Gleditsch's CShapes GIS Vector Dataset, Version 0.4-2, in ArcMap 10.0.

The Dark Knight Rises and the Economic Impact of Bane's Occupation of Gotham

I started working on this post a long time ago and, for whatever reason, never got around to finishing it. So please keep in mind that this was largely written shortly after the film first came out. I should also disclose at the outset that this post will contain spoilers, so if there is some sort of unbelievably powerful force that has kept you from seeing this fantastic movie, please be warned. For those still interested, there's more after the jump.

Educational Attainment and Race in Alabama

A part of my responsibilities as a post-doc here at Alabama entails helping faculty with seeking out external funding for projects. One of these projects has involved developing a program to promote educational outreach throughout the state. In the course of this work I've been digging through some data on educational attainment rates in Alabama. In terms of learning about the political and economic characteristics of my new state, this has been an informative exercise. I also think it's helpful in illustrating the benefits of digging through your data, and helps to shed some light on the benefits of considering the distribution of your data more carefully, and so I thought I'd write a quick post on the subject. For the purposes of this post I'll be drawing on data collected from the American Community Survey's 2006–2010 five year population estimates. 

The mean percentage of Alabama's population holding a bachelor's degree or higher for this time period is 21.7%.  Compare this figure with that of my home state—New York's overall rate is 32.1%. This is a fairly large disparity, but further disparities emerge when we begin to disaggregate the data by race. The graph below shows the mean value of this same variable across the black and white populations of both states. 

From this we can see that the total values mask some important differences between groups within each state. In both cases the college attainment rate of the white population is slightly higher than the overall mean value. Alternatively, the attainment for the black population is significantly lower than the overall mean value in each state. Even New York, which on its face appears to have a significantly higher overall college attainment rate, is not much different from Alabama when we compare black college attainment rates with Alabama's values. 

Even when disaggregating the data in this way, we often still lean on the mean value pretty heavily. The distribution of the data around the mean can be incredibly informative as well, but we don't often pay as much attention to this as we should. Below I present a quick illustrative example of why this can matter.

Looking at Alabama's data again, I have drawn on county-level college attainment data for both groups.

Looking back at the first graph we can see that the statewide mean white college attainment rate was about 24%, and the statewide black attainment rate is roughly 14%. Looking at the county-level distributions can help us to get a better grasp on educational attainment rates throughout the state. In the case of white attainment, we can see that the distribution is slightly skewed towards the low end, but the lowest value in the distribution is around 8–9%. Alternatively, the distribution of black college attainment rates at the county level is more heavily skewed, and the lowest value in the distribution is considerably lower than for the white attainment rates, as shown in the preceding graphs where the X-axes are held constant to help make the figures more comparable. In fact, .30% is the lowest black attainment rate recorded in the data. 

Looking at the distribution of the data in addition to the mean values for these different groups can provide some additional insights into the substantive issues we're interested in. In this case, it's already apparent from considering the mean values that there are disparities in the college attainment rates of the black and white segments of the state's population. However, these values may actually underestimate the magnitude of those disparities. Though the distribution for white attainment rates is still skewed, it's floor is considerably higher than that of black attainment rates. In fact, by my count there were 14 counties in which the recorded black attainment rate was lower than the lowest recorded white attainment rate. The maximum for each group is roughly the same, however. Areas around Tuscaloosa, Birmingham, and Huntsville have fairly high college attainment rates. But these areas are highly unrepresentative when considering the state more broadly. In fact, although the mean statewide black attainment rate is 14%, the majority of counties fall well below this figure. And while most counties also fall below the mean on white attainment rates, there are, comparatively, more counties falling above the mean. All of this suggests that the mean statewide black college attainment figures are more heavily influenced by outliers in the data than are white attainment rates.  

Lastly, looking at these plots reveals another important feature of the distributions—the variance. The variance in white college attainment rates is 66.2. The variance in black college attainment rates is nearly half that, at only 37.5. This tells us that not only are black college attainment rates lower than white attainment rates, but they are more uniformly so. To put it differently, there is greater dispersion in white college attainment rates, while black college attainment rates are more tightly clustered around lower values. This is made apparent by looking at the two preceding graphs. Though both are skewed towards the lower end of the X-axis, white college attainment figures are less tightly clustered here, as I mentioned above.

This is an issue that I'll come back to in the future, but for now it seemed like this was a useful example for illustrating the importance of familiarizing yourself with the data that you're working with, and how relying less on simple mean values can reveal interesting features of the data you're working with. Bear Braumoeller has an interesting Political Analysis piece about examining the variance, which is worth checking out. More substantively, I also think that this exercise helps to more fully draw out the disparities that may exist between social groups. Relying on simple indicators of central tendency can be useful in some contexts when we want to quickly convey bits of information, but these figures can also be misleading. At the very least, they don't always tell us the full story.

 

On Hegemons and Trade (Part I): Origins

I am currently in the process of entirely rewriting and redesigning one of my earliest academic papers that I want to update with the context of better data and methods in evaluating the hypotheses I initially proposed. That is, I am writing the same paper a second time.  As part of this process, I am revisiting some classic works on the topic of Hegemonic Stability Theory. Consequently, this is part one of a multi-stage blog post series.  Part 1 of the project deals with the origins of Hegemonic Stability Theory, Part 2 will dig into the advancements in the theory in the 1980s and 1990s, and Part 3 discusses the empirical record for the topic.  

*****

Hegemonic Stability Theory (HST) contains a set of different authors who propose that a state that is dominant internationally has the capacity to make particular rules, norms, and institutions that guide the course of international relations.  The theories often articulate that a unipolar system is unique and the enabling condition of unipolarity cannot, or is unlikely to, manifest in multipolar systems.  The outcomes of interest for theorists of global power preponderance theories vary with popular topics including trade, war, and the creation and maintenance of international regimes.  For the context of this series of posts, I am focusing on the trade-related theory of hegemony as it has been developed in political science and economics.

The concept of hegemony itself is a loaded term that has many different meanings for different scholars and schools of thought.  Rather than give a precise definition now, I will generally refer to it as a state that has global dominance in military, economic, and financial power. Throughout this post, I have included the CINC scores for major powers from 1816-2007.  Naturally, CINC scores are imperfect for several different reasons depending on the context in which they are used. In the case of hegemony, they may be best apt for gauging some level of military power, but are insufficient to capture the other aspects of global hegemony that are important. For the rest of this series, I will be using hegemon, unipolarity, and preponderance of power as synonyms.  Having a more refined sense of when we know we have hegemony and when we do not is a task better suited for the third part of this series. 

The idea that the hegemon under unipolarity is afforded special privileges that other states do not have is not unique to scholarship in the latter half of the 20th century (for example, Hirschman's 1945 book), nor the 20th century alone, but much of the principle tenants of the theory solidify under scholars examining the topic.  Specifically, the 1970s offered two major advancements in the theory in the work done by Kindleberger and Krasner. The former developing a thread of HST that is best described as "Leadership Theory" within HST and the latter offering a Structural Theory of HST. 

Kindleberger's (1979) foundational work, The World in Depression, is an ambitious scholarly undertaking to identify the key international and domestic causes of the great depression. Fundamentally, for Kindleberger, stability in the international system is provided by a dominant power that can maintain the rules of the system and provide the incentives necessary to encourage states to behave in a way that is conducive to liberalization and integration.  The interwar period offers a unique period of transition as there are two possible actors that could fulfill this role: Great Britain and the United States.  Great Britain has the willingness to be the leader, but lacks the capability to do so (Great Britain acts in the appropriate manner in the 1920s, but, according to Kindleberger, it becomes evident in 1931 that the British are unable to provide the necessary leadership to maintain stability). The United States has the capability, but lacks the domestic willingness to do what is necessary to prevent a race towards retrenchment in globalization.  Specifically, Kindleberger highlights five necessary behaviors for a hegemon to encourage stability (p. 289):

1. maintaining a relatively open market for distress goods;
2. providing countercyclical, or at least sable, long-term lending;
3. policing a relatively stable system of exchange rates;
4. ensuring the coordination of macroeconomic policies
5. acting as a lender of last resort by discounting or otherwise providing liquidity in financial crisis.

He follows this list with the explicit statment of "These functions, I believe, must be organized and carried out by a single coutnry that assumes responsibilty for the system" (Kindleberger 1973, p. 283). After going through each of these five points, Kindleberger comments on the contemporary hegemonic decline of the 1970s and prospects for US leadership in the 1980s and 1990s. 

 

 

Krasner's branch of HST comes from his 1976 article titled "State Power and the Structure of International Trade." The structural view of HST asserts that it is not a function of willingness that determines liberalizing behavior in regards to trade, but it is a product of a state's position in the international system.  Fundamentally, and contrary to Kindleberger, capability breeds willingness.  Explicitly, a state in a privileged position in the international system ought to seek global liberalization of trade policies as such policies necessarily advantage its strategic global position.  It becomes part of the state's national interest to encourage free trade.  Beyond having a national willingness for trade liberalization, it would necessarily have the ability to encourage the system as well. Krasner argues that the hegemon has an array of positive and negative incentives, including offering access to competitive markets, providing cheap goods, encouraging other states to outcompete and ruin the business recalcitrant states, and withholding aid/grants/loans (p. 322-323). 

Thus, for Krasner, the 19th century and post-WWII bouts of liberalization were products of hegemonic dominance while the interwar period was marked by uneven multipolarity.  For Kindleberger, leadership marks the 19th and latter half of the 20th century, while the interwar period lacked a state willing and capable of providing hegemonic leadership.  The two different veins of HST create a possible conundrum for empirical scholarship as deriving testable implications may rely on definitional issues of hegemony.  Since, as social scientists, we are interested in causal determinants of behavior, it becomes hard to pinpoint the effects of HST if part of knowing whether there is a hegemon in the first place is that a state is acting hegemonic.  Finding other measures to predict hegemonic behavior becomes part of the task. In the structural sense, we should expect a dominance of power to lead to a willingness for hegemonic behavior; however, for Kindleberger and leadership theory, we may not be able to predict willingness for hegemony until we see hegemony.  

 

A few notes. First, both authors are writing in a time where the decline of the United States seemed evident to scholars and non-scholars alike.  Powerful states, especially the winners of World War II, had created powerful institutions that seemed to guide world politics to a degree that transcended mere state-base perceptions of International Relations.  Naturally, with a perceived decline of the United States in the 1970s and 1980s, trying to figure out how the state of international affairs would change, warp, or collapse was important.  Would hegemonic institutions endure? Would a leader that was incentive-compatible with the United States emerge and lead to a peaceful transition?  Would Japan/Germany become the new global financial centers?  The debates surrounding hegemonic stability theory escalated from this point and remained healthy throughout the 1980s.  With the collapse of the Soviet Union, questions about US decline halted for the unipolar moment and other questions of hegemony emerged.  We will turn to the expansions of both authors' work in the second installment of this discussion.

 

 

A second note: Krasner does allow for non-hegemonic configurations to encourage liberalization. A common misconception is that hegemony is conducive to trade, but other configurations are not, In the original 1976 piece, he argues "Let us consider a system composed of a large number of small, highly developed states.  Such a  system is likely to lead to an open international trading structure. The aggregate income and economic growth of each state are increased by an open system..." (p. 321). Thus, in a multipolar world dominated by highly industrialized states, a push for liberalization may emerge.  Of course, he goes on to argue that a system composed of a small number of large, asymmetrically developed states would lead to protectionist behavior and norms.  This creates the following table of possible system configurations on page 323: 

Finally, in Krasner's original article, he concludes, after examining qualitative and quantitative evidence of trade-related behavior under different configurations of world power, that a simple measure of power is insufficient to understand hegemonic pressures on liberalization. Instead, periods of increased integration occur after shocks to the system encourage a change in behavior by the dominant state. This clarification of temporal shocks does become become important when we evaluate attempts to test Hegemonic Stability Theory.

Teaching Introductory Graduate Methods: Some Ideas About Understanding Sampling Distributions

Everyone has their own way of teaching methods, but in this entry I thought I’d share a brief exercise that I thought worked well. This code was developed when I was the teaching assistant for a second-semester methods class focusing on regression. Specifically, this was a section of the class from the first few weeks that dealt with two things: first, showing/repeating the idea of a sampling distribution, and second, showing some introductory code on loops and graphics.

The code uses the ggplot2 package for R. This software, and this package, have (in my experience) a somewhat steep learning curve so starting slow is important, even in the second semester. Equally, the concept of a sampling distribution is something that takes a while to click with most people who don’t have a statistical background. So what I’m trying to do with this code is show some basic graphics, and also introduce sampling distributions by ‘bringing one to you on a plate’.

            The first part of the code simulates ten rolls of a (pseudo) six-sided die:

cenlt<-vector(length=10)

for (i in 1:10)
{

            x<-runif(10,1,6)

            cenlt[i]<-mean(x)

}

cenlt

Here we create a vector, called cenlt (named for the central limit theorem), of length ten. We don’t give R any instructions about what to put in this vector, so if we inspect it (typing “cenlt”) it will display ‘FALSE’ ten times. We leave this empty vector to one side. Then we create a new imaginary conveyor belt, telling R to sit beside the conveyor belt and perform the same operation every time the belt rolls past, which it will do ten individual times. 

            The operation R will perform is to create ‘x’, which is a set of ten numbers, each randomly sampled from a uniform distribution between 1 and 6. We then take the mean of this set, and store it as a new element of cenlt. When we’ve done this ten times, then, cenlt will be full. We then display cenlt. We can actually display it a little bit more attractively if we graph it: 

library(ggplot2)

p1<-ggplot()

p1<-p1+geom_histogram(aes(x=cenlt),binwidth=0.2,
linetype=1, alpha=.3, color="black", fill="red")

p1<-p1+ ylab("Count") +
xlab("Sample Mean") + xlim(c(0,6)) + ylim(c(0,6))

p1<-p1+opts(title="Distribution of Sample
Means")

p1<-p1+geom_segment(aes(x = 3.5, y =0, xend =
3.5, yend =6))

p1

            The first line creates a new object called p1, which we set to be a graph object. Then we take it, and add another layer. The second line tells it to fill the graph with the shapes necessary to make a histogram: aesthetically, this means an x-axis defined by the values of cenlt. We ask for the shapes to have a solid black outline (linetype = 1, color = “black”) and to be semi-transparent and red (alpha = 0.3, fill = “red”). We then add a line showing the expected mean (geom_segment) as well as other details. 

            This code was then followed by a discussion of what each element of ‘cenlt’ represented. Each was a mean. So the mean of cenlt was a ‘mean of means’, and so cenlt has its own distribution: the distribution of means of ‘x’, and if we consider each x to be a sample, then seeing cenlt as the sampling distribution is relatively straightforward. The code can then be tweaked to make ‘x’ bigger, or to run ‘x’ more times, and discussing each of these options helps elucidate the idea of a sampling distribution. Changing the distribution away from uniform can also be useful. We also got some good discussions out of manipulating the basic parameters of the graph (the axis labels, the colors and binwidth of the histogram), and then later by switching the geoms to different options. All in all, this was a useful exercise that led to some good discussions, and helped introduce the ideas of sampling distributions, loops, and statistical graphics.

 

Human Rights, Foreign Direct Investment, & INGOs

Over at IPE@UNC, Lindsay Tello discusses the article on foreign direct investment and human rights shaming that Colin Barry, Michael Flynn, and I have in this month’s International Studies Quarterly.  Here’s the abstract:

Non-state actors, such as international non-governmental organizations (INGOs) and multi- national corporations (MNCs), have attained an increasingly prominent role in modern world affairs. While previous research has focused on these actors’ respective interactions with states, little attention has been paid to their interactions with each other. In this paper, we examine the extent to which the decisions of private actors seeking to invest abroad are affected by the reputational costs of doing business in countries publicly targeted by human rights activists. We find that “naming and shaming” by human rights INGOs tends to reduce the amount of foreign direct investment received by developing  states, providing  evidence that INGO activities affect the behavior of MNCs. An additional implication of our findings is that shaming by INGOs can impose real costs on targeted states in the form of lost investment.

I really appreciate Lindsay’s plug of the paper, and I would like to take the opportunity to share a few thoughts about where we are going from here.

Our paper is an attempt to advance three different literatures: the nascent literature on interactions between non-state actors, the longstanding literature on the relationship between government respect for physical integrity rights and foreign direct investment, and the growing literature on the effects of human rights INGO activities.  Lindsay focuses particularly on the interaction between non-state actors, i.e. MNCs and INGOs, and like her, I believe that this is an important direction for future research.  One thing we see as among the next steps is to look at more disaggregated FDI data.  As Nate Jensen has discussed recently, collecting these data can be incredibly difficult; luckily for me, my co-author, Colin Barry, has already engaged in the hard work of collecting data on 10,738 foreign subsidiaries of 77 major companies from 1993-2010.  We hope to use these data to understand which types of companies are most susceptible to the influence to INGO shaming.

While the study of interactions between non-state actors is still in its early stages, there is a long-standing literature that examines the degree to which state human rights practices and foreign direct investment are interrelated.  While we believe this paper represents a substantial step forward in delineating the separate mechanisms that connect human rights practices to the receipt of foreign direct investment, there is still more to be done.  Our focus in this paper was on how the external perception of the governments a company does business with may affect that company’s investment decisions. However, Lindsay points to another crucial question: How do labor standards (and shaming of individual companies for their exploitation of those labor standards) affect investment decisions?  We are hoping to get at this question in a couple of different ways.  First, Colin and I are currently working with David Cingranelli on a dataset that contains information on seven different labor rights in law and practice for almost every country in the world from 1994 to the present.  We hope to utilize these data in future studies to determine how different types of labor standards affect investment decisions.  Second, we hope to eventually pair data on INGO shaming of individual corporations for their labor standards with disaggregated FDI data, allowing us to get a more direct understanding of the interaction between MNCs and INGOs.

Finally, while recent research has drawn attention to the ability of human rights INGOs to improve human rights practices, we understand far less about the costs that INGO “naming and shaming” can impose on states to encourage them to engage in these improvements.  As such, we see this paper as a step toward understanding those costs, but we still have many ideas about other ways in which INGO shaming is capable of imposing costs on target states.  According to Keck and Sikkink’s “boomerang model” and Murdie and Davis’ findings, INGO shaming is more effective when that shaming is echoed by third party states. As such, we are currently working on a project that looks at the influence of human rights INGOs on US foreign policy; our early results suggest that INGO shaming influences both foreign policy outcomes and statements made by individual policy-makers, such as members of Congress.

Of course, these are just some thoughts about where we would like to take our work.  Where would you like to see this go?  Give us some feedback in the comments.

Identifying Cities with the Safest Drivers

In May 2013 Allstate released its "America's Best Drivers Report", which identifies cities with residents who have the longest years between accidents. Boise is #2 on the list, so it caught my attention. The data are based on Allstate insurance claims from January 2010 to December 2011.  

Although the report is interesting, I think the name of the report is misleading (which is clear when you read how Allstate discusses the report). The report may not necessarily identify where the safest drivers are. Instead it may identify cities where environmental conditions make it safer to drive. If that is the case, the drivers are not necessarily better in these cities, they are just driving in better conditions, leading to more years between accidents. Also, it is possible that the report be identifying cities where accidents are more likely go unreported, which would make it seem like drivers in those cities have more years between accidents. 

An alternative way to determine where the safest drivers are is to predict the average time between accidents in these cities based on environmental conditions and likelihood of reporting an accident, and then compare those predictions with the actual averages. If a city did worse than predicted, then perhaps those drivers are less safe. If a city did better than average, those cities may be more safe.

I collected data on factors that influence accident reporting (which include whether an accident will occur, and if so, if it will be reported). Then I made a series of predictions for each city and compared those predictions with the actual data. I identified four cities that do worse than predicted and four cities that performed better than predicted. All the other cities performed as expected.

Likelihood of Reporting Variables

I have identified two potential reasons why motorists might not report an accident. First, not all states require motors to automatically report accidents to the police. In most states motorists only have to report an accident if there is a certain amount of property damage (or if someone dies as a result of the accident). For the minimum reporting variable, I collected data on the minimum amount of property damage needed in to report an accident for the state a city was located in. For states with no minimum, I set the value to $5000.

I expected the sign of the minimum reporting variable to be positive. In places with higher minimum property damages requirements, minor accidents are less likely to be reported, increasing the measured years between accidents. Hence the higher the amount of minimum property damage to report in a city, the higher the average years between accidents in those cities. 

Income may also have a positive or a negative effect on whether someone reports an accident, which would decrease or increase the average time between accidents.

Individuals with higher income are more likely to have insurance, and hence more likely to report an insurance claim when in an accident. This would reduce the measured years between accidents. For the income variable, I took the log of median household income for the state each city was located in 2005. The sign of this variable could be negative: cities with higher average income have less years between accidents.

On the other hand, someone with higher income may be less likely to report an accident because they can pay for the damages. This would increase the measured years between accidents. So I expected that the sign of this variable could also be positive: cities with higher average income have more years between accidents.

Environmental Conditions

Besides reporting requirements, several environmental factors that may lead to more collisions. More densely populated cities may create more opportunities for drivers to get into accidents. I collected data on population density of these cities and logged the values. I expected the sign of this variable to be negative: individuals in more densely populated cities have less years between accidents.

Weather conditions could also lead to more accidents. For this measure, I took the average rainfall in the cities in the data set. I expected the sign of this variable to be negative: as rainfall increases in a city, the time between accidents should decrease.

More bike friendly cities may have less accidents on average because there there are less drivers on the road. This would increase the time between accidents. For this variable I used data from Bicylcing magazine's Top 50 bike friendly cities in the US. I expected the sign of this variable to be positive: individuals who live in more bike friendly cities should have more time between accidents.

Teenagers may be more likely to get into accidents than adults. So as the minimum driving age decreases, the time between accidents should decrease. I collected data on the minimum driving age in each state a city was located in. I expected the sign of this variable to be positive: in cities with higher minimum driving age, time between accidents should increase.

Higher speed limits may also lead to less measured time between accidents. Although most studies suggest that higher speed limits do not increase the number of collisions, they may increase the number of severe collisions. If higher speed limits may lead to more severe collisions, this will reduce te time between accidents because more severe collisions are more likely to be reported. For this variable I collected data speed limits in the states the cities are in. I expected the sign of this variable to be negative: higher speed limits lead to less time between accidents.

The Results

The results in the table below suggest that individuals in cities with higher median income have less time between accidents. This may provide support for the argument that wealthier people are more likely to have insurance, so they are more likely to report a claim. The table also demonstrates that as population density and rainfall increases in a city, time between accidents in that city decreases. This suggests that environmental factors play a role in frequency of accidents. Higher speed limits also decrease the time between reported accidents. 

Now to the predictions. The table below presents the predictions for each state and compares them with the actual values. Four cities did better than average: Fort Collins, CO, Brownsville, TX, Madison, WI, and Visalia, CA. And four cities did worse than average: Anchorage, AK, Portland, OR, San Francisco, CA, and Glendale, CA. But for the majority of cities, they performed as expected giving the living conditions in those cities. 

So the take away message: drivers in most cities perform as expected. There are several cities where drivers have long years between accidents, but they drive in better conditions. And there are several cities where drivers have very few years between accidents, but they drive in more hazardous conditions. Drivers should be judged on the probability that they would get into an accident given the conditions they live in. That way it is easier to determine whether they are being safer than expected or more reckless than expected.