How do you learn about cause and effect without randomized experiments? We provide a dramatically improved method of inferring causality even when randomization is impossible. We also invented a computational approach that makes our technique approximately one million times faster than previous approaches, meaning it will usually take only a few minutes to run for most commonly sized data sets, and opening up new ways of conducting empirical analyses. The result is an easy-to-understand statistical technique, along with free and easy-to-use open source software, that researchers can use to learn about cause and effect without experiments. Correlation is still not causation, but we show how observational correlations, when used appropriately, can teach us a great deal about causation.

Although randomization is obviously impossible for numerous crucial questions in political science, the social sciences, the sciences, medicine, education, government, public policy, and business, we often have no choice but to draw some type of causal conclusion in these areas. For example, no one has ever conducted a randomized experiment on tobacco consumption in humans, but we know that smoking kills half a million people a year in the US alone. Similarly, on one has ever randomized incumbency status to members of congress, and yet we have learned that voters are more likely to cast their ballots for incumbents than challengers. Much hard work has gone into solving these problems; our article offers a way to make such generalizations easier, faster, and more reliably.

Our approach builds on a highly intuitive technique called matching. The idea is sophisticated, but the basics are simple. Suppose we want to know whether renovating school buildings improves educational outcomes. As generous as NSF grants are, no researcher will be able to randomly assign new buildings to school districts. So instead, we can compare school districts which choose to build new buildings with those that do not, but we will have to avoid the bias that would occur if those districts differ in other ways. For example, districts with new buildings might also be areas with lots of wealth and parental involvement in their children’s education. So a simple version of matching in this case would be to find a sample of school districts which recently built new buildings. Then for each one, find a very similar (“matched”) district that chose not to renovate its buildings. We can define “very similar” according to wealth, parental involvement, the conditions of the school buildings, teaching styles, and many other factors. If we can find these matched pairs, then we might be able to approximate a randomized experiment from purely observational data.

In practice, no two school districts are exactly the same, and so a large number of matching methods have been developed, each which works best in different circumstances at making the treated group (the one with the new schools) as similar as possible to the control group (those without the new schools). In contrast, our American Journal of Political Science article, now available here for early view, introduces new methods for matching that result in the optimal matched sample, without having to choose a matching method.

In particular, every matching method begins with a data set and prunes non-matches in order to try to improve the similarity between the remaining data in the treated and control groups. For any number of pruned observations, there exists a subsample of the original data that has highest level of similarity, but the number of possible subsets that need to be checked to find the optimal one is gargantuan (usually larger than the number of elementary particles in the universe) and so previous matching methods try to approximate this optimum. Our algorithm quickly identifies the best possible subset, without running any matching algorithm. Given these criteria, we have proven mathematically that no matching method in existence, or which could be invented, can produce higher levels of similarity between the treated and control groups than our algorithm. The algorithm is so fast that we are able to recommend computing it for all possible numbers of observations pruned and presenting the full results for the analyst to choose, without tradeoffs or compromises. Please see our software at http://j.mp/MatchingFrontier.

Written by Gary King of Harvard University, Christopher Lucas of Harvard University, and Richard Nielsen of Massachusetts Institute of Technology. Their article “The Balance-Sample Size Frontier in Matching Methods for Causal Inference” will be published in a forthcoming issue of the American Journal of Political Science and is currently available for Early View.

In the following blog post, the authors summarize the forthcoming American Journal of Political Science article titled “Gone with the Wind: Federalism and the Strategic Location of Air Polluters”:

States have strong incentives to encourage major air polluters to locate near their downwind borders. This allows them to reap the benefits of economic development while making their neighbors shoulder the resulting environmental and health costs. Although there are provisions in the U.S. Clean Air Act designed to mitigate interstate air pollution, they have historically been ineffective. And, there is some evidence that air pollution levels are heightened near downwind state borders. However, prior research has produced little evidence that this is the result of government behavior, such as weaker enforcement of environmental laws.

The solution to this paradox may lie in industrial location itself.  We argue that both state governments and firms have incentives to strategically locate polluting facilities near a state’s downwind borders.  We test this idea with a method that is novel to political science—point pattern analysis. This technique is common in fields like epidemiology, which models where disease cases emerge relative to the broader population at risk. We use the method to analyze where major air polluters emerge in latitude and longitude in the continental 48 states. We compare air polluters to a control group of other industrial facilities: large quantity generators of hazardous waste. Hazardous waste polluters are a good comparison group because they have similar locational needs as air polluters, but the wind cannot carry away the pollution they produce. Hence, a key reason we might observe any difference in the location of these sites is because of attempts to export air pollution to downwind neighbors. As an illustration of our data, the above figure shows air polluters and hazardous waste polluters in the state of Georgia.

We find that major air polluters are significantly more likely to be located near a state’s downwind border than hazardous waste generators.  Our baseline model suggests that moving from a site on a state’s downwind border to the site that is farthest upwind diminishes the odds of a major air polluter relative to a hazardous waste generator by a substantial 22.4%.  This effect is even stronger for facilities that produce highly toxic air emissions.

Our results also indicate that the pattern of industrial location varies systematically across states.  States with stronger environmental programs, for instance, evidence weaker propensity to have air polluting firms locate near downwind borders, while states that make greater use of “smokestack chasing” economic development incentives show the opposite configuration.  This suggests that air polluter location is responsive to public policy.

About the Authors: This author summary was written by Jamie Monogan of University of Georgia, David Konisky of Indiana University, and Neal Woods of University of South Carolina. Their article “Gone with the Wind: Federalism and the Strategic Location of Air Polluters” will be published in a forthcoming issue of the American Journal of Political Science and is currently available for Early View.

In the following blog post, the authors summarize the forthcoming American Journal of Political Science article titled “Informed Preferences? The Impact of Unions on Workers’ Policy Views“. This post originally appeared on the MPSA blog and is reposted here with permission. 

What is the impact of labor unions in shaping the political preferences of workers? More specifically, to what extent can we trace the anti-trade sentiment we are now seeing among many U.S. workers to the influence of their unions? Due to their shrinking memberships, unions are often dismissed as a spent force in contemporary politics. Yet such a view overlooks an important point: Even after years of membership decline, unions still represent a sizable share of many electorates: a quarter of all workers in Britain, a third in Italy, and over half the workforce in countries such as Norway or Belgium. Even in the U.S., union members still account for about 11 percent of the workforce—a conservative figure that excludes non-members working under union agreements and family members whose livelihoods often depend on a unionized wage earner. A key question is whether and how unions’ access to a large swath of the electorate translates into political influence.

To understand the role that unions play in shaping the political views of their members is, however, a challenging empirical task.  Even in instances where union members are found to hold systematically distinct views from non-members, the cause is not an obvious one: Is it the result of a union’s direct influence on workers’ policy views, or is it that workers who choose to join a union differ from non-members to begin with?

Our article seeks to provide answers to these questions by utilizing an original survey of more than 4,000 American workers that was administered across 12 industries selected to provide the full range of exposure to various aspects of globalization. The dataset provides sizable samples of both union members and non-members within each industry, allowing for comparisons with a meaningful control group. Another key advantage is the availability of detailed information on pertinent aspects that are often missing from standard surveys: the exact union to which the worker belongs, the intensity of communication initiated by their unions on a range of policy issues, and the degree to which a worker is aware of her union’s policy positions. To assess how accurately workers know their union’s position, we devise a new measure of a union’s position on trade policy (what we call the “protectionism score”). This measure is based on a union’s ‘revealed preference’, namely through their lobbying activity and official statements given on a wide range of trade bills.

Using this data, we begin to explore whether unions indeed serve as information providers for their members. The evidence decidedly shows that they are. (See Figure 1;  unions are sorted along the vertical axis by their protectionism score.) As the left-panel indicates, unions indeed discuss the issue of trade with their members, and, not surprisingly, the more protectionist the union, the more likely it is to impart such information to their members. We can also see that the intensity of the communication is associated with how familiar members are with their union’s position (center panel). Members of protectionist unions not only tend to express greater familiarity with their union’s stance on trade, but also to correctly describe their union as protectionist (right panel).

Figure 1

In short, unions clearly operate as information providers. Moreover, workers also seem to ‘get’ their union’s message. Yet to what extent do members tend to be influenced by this message and adopt their union’s position? As an initial step, we examine the association between members’ own attitudes on trade and the protectionism score of their unions. We find a clear alignment between the two. Notably, such positive association is not found among non-unionized workers who are employed in the same industry (see Figure 2). Still, the key empirical task here is to assess the possibility of a self-selection process: If workers decide to join a union because of the union’s policy position, the findings could simply be an outcome of a reverse causal process. We address this possibility by exploiting two sources of variation, as detailed below.

Figure 2

First, we leverage the fact that there are state-level differences in laws with respect to union membership, or so-called the “Right-to-Work” (RTW) laws. In states that adopt the RTW provision, labor unions cannot legally require workers to pay union dues as a condition for employment. Union membership in RTW states therefore depends much more on individual workers’ own discretion, and is less a function of an institutional requirement to become members. This difference allows us to test the self-selection question: If self-selection accounts for members’ preferences, we should expect to see unions have much more of an influence on members in states in which membership is entirely voluntary.

Our analysis provides very little support for the self-selection account. We find that union membership itself is associated with a clear effect on the trade policy views of workers, and that this effect is conditional on how strong the union’s position is with respect to trade. Significantly, we find no systematic difference between members in RTW and non-RTW states. This is clearly inconsistent with the selection mechanism being the prominent factor.

Second, we wanted to assess what happens to the views of members when their union takes a different stance on a given policy: do workers follow suit? If so, that would be a strong indication of the unions’ influence. To do so, we exploit the case of a sudden and dramatic shift in the United Auto Workers’ (UAW) position toward a major trade liberalization deal. For many years, the UAW had been strongly opposed to the signing of a trade agreement with Korea, which, according to its statement, would be “worsening our lopsided auto trade deficit and threatening jobs of tens of thousands of American workers.” Yet after a set of changes had been incorporated into the agreement, the UAW announced a reversal of its position. “We believe an agreement was achieved that will protect current American auto jobs, [and] that will grow more American auto jobs,” its statement now read.

How did this shift in the UAW’s position influence the views of the autoworkers on trade? Figure 3 below provides a striking answer: While union members working in the auto industry had been significantly more protectionist than non-members before the shift, the level of support for trade restrictions significantly decreased after the UAW endorsed the free trade agreement. Notably, this change in attitudes toward trade liberalization had not been observed among non-members working in the same auto industry. This finding remains intact even when we control for potential confounding factors.

Figure 3

Taken together, our findings provide compelling evidence that unions exert influence on their members in a clear and systematic manner. The analysis points to the important role of unions as information providers. While previous studies have highlighted the role of unions as the “voice of workers,” via campaign contributions or lobbying, we have demonstrated an underexplored yet important source of union influence: communication with, and dissemination of information to members. This influence can be significant, given unions’ breadth and reach. In other words, unions should be thought of not only as institutions that channel the political preferences of their members. Instead, they should also be understood as institutions that shape and influence the views of many workers.

About the Authors: Sung Eun Kim is an Assistant Professor of Political Science at the National University of Singapore.  Prof. Yotam Margalit is an Associate Professor at the Political Science Department in Tel-Aviv University. He is also a Senior Fellow at the Israel Democracy Institute. Their paper “Informed Preferences: the Impact of Unions on Worker’s Policy Views” was awarded the Best Paper by an Emerging Scholar at the 2016 MPSA Conference.

In the following blog post, the authors summarize the American Journal of Political Science article titled “Inside Irredentism: A Global Empirical Analysis” (AJPS 61:1):

Russia justified its 2014 annexation of Crimea by pointing to the failure of the Ukrainian government to safeguard the rights of the ethnic Russians. For Russia, this created a legitimate demand for unification, but others regarded this argument as thinly veiled realpolitik. This is a classic case of irredentism, defined as a state’s use of military force to advance a claim over territory in a neighboring state by creating more congruence between the ethnic group and the state. Unlike secession– a decision to leave one state and to establish another one on grounds of an ethnic difference–irredentism is a governmental decision that appropriates territory from one state and adds it to another on the basis of shared ethnicity.

Although less common than secession, there are many examples of irredentism, since numerous ethnic groups have kin groups in adjacent countries. The mismatch between cultural and political boundaries is pervasive, but the international community does not condone irredentism. To mitigate international censure, irredentist states often frame their actions as humanitarian intervention to protect discriminated ethnic kin. Hitler’s irredentist action towards the Sudetenland in Czechoslovakia, Armenia’s intervention in Nagorno-Karabakh, Serbia’s invasion in Croatia, and Russia’s annexation of Crimea all used similar rationalizations – blending realpolitik with jus bellum iustum (just war) logic and humanitarian justifications.

Irredentism poses significant danger to international order and human security. While previous studies offer various reasons for irredentism, Inside Irredentism tests several rival arguments and proposes a novel explanation using new data on all actual and potential irredentist actions between states around the world from 1946 to 2014. The analysis indicates that irredentism is more likely when there is an ethnic group that has a large demographic majority but relatively weak economic standing when compared to other ethnic groups in the country. This fosters grievances within the dominant ethnic group. These grievances are more likely to find an outlet (and are less likely to be countered by other groups) under “winner-take-all” electoral systems in more ethnically homogeneous irredentist states. This creates an opportunity for the kin group to pursue an explicitly irredentist foreign policy agenda. Where these grievances and opportunities intersect, irredentism is most likely because political elites, responding to their electoral incentives, seek to divert attention away from status inconsistency through the promise of ethnic unification.

Although claims of “protecting discriminated ethnic brethren” may help mobilize citizens to rally around the flag and justify actions, differences in discrimination appear to play little role in explaining irredentism, except rhetorically. The study shows that irredentist foreign policy is most likely when economic and political interests are at play in the kin state. Ethnicity is sometimes viewed as a problem in itself, but our results suggest that its effect on irredentism depends on how it maps on to economic competition and the political system. Policymakers would therefore do well to address the domestic sources of irredentism in the initiating state, in addition to the conditions of the co-ethnic enclave in the target state.

About the Authors: David S. Siroky is an Associate Professor in the School of Politics and Global Studies at Arizona State University and Christopher W. Hale is an Assistant Professor in the Department of Political Science at The University of Alabama. Their article “Inside Irredentism: A Global Empirical Analysis” was published in the January 2017 issue of the American Journal of Political Science.