Abstract
There is a growing multidisciplinary effort to understand the political causes and consequences of the US deportation system amid rapidly changing policies and significant data limitations. I contribute to this timely work by considering how interactions between an increasingly empowered executive branch and a politically polarized Congress shape deportation policymaking. I apply a policy feedback approach, a theoretical lens that analyzes policies and polities as co-constitutive, and construct a novel dataset to computationally analyze how Congressmembers publicly responded on Twitter to the Trump administration’s implementation and suspension of the family separation policy in 2018, during which migrant parents were prosecuted and detained separately from their children. Findings suggest that Republican and Democratic Congressmembers are unlikely to respond to executive assertions of power on deportation, even when policies are highly unpopular among the public. This raises concerns about shared governance on deportation policy and its implications for migration and democracy.
Empirically and systematically studying the growing deportation system in the US is a pressing priority for social scientists. However, this is often a difficult endeavor because of rapidly changing deportation and immigration policies and significant data limitations. Despite these challenges, research on the deportation system is flourishing. For example, a growing body of research examines the politics of the deportation system, namely how deportation policy shapes experiences of citizenship and non-citizenship (Asad 2023; Rocha et al. 2015; Valdivia 2025, this issue) and the process of deportation policy agenda-setting (Flores 2018; Hopkins 2010; Jones and Martin 2017).
I contribute to this burgeoning and timely multidisciplinary effort to understand the political causes and consequences of the US deportation system by applying a policy feedback approach to consider how interactions between an increasingly empowered executive branch and a politically polarized Congress shape contemporary deportation policy. A policy feedback approach offers a theoretical lens to analyze policies and polities as co-constitutive (Mettler and Soss 2004; Pierson 1993; Skocpol 1992) and has been used to study policy domains such as social welfare policy (Campbell 2003; Hacker 2002; Mettler 2005). This approach can also be used to study the politics of deportation policymaking.
I examine how Congressmembers publicly reacted on Twitter (now known as X) to the Trump administration’s implementation and suspension of the family separation policy in 2018, which authorized federal authorities to prosecute migrants and separately detain parents and children while they awaited criminal proceedings and deportation. I construct a novel dataset of approximately 830,000 tweets generated by members of the 115th US Congress during their term (2017–2019) and analyze these data using computational text analysis methods and a quasi-experimental design. I find that, irrespective of party affiliation, there was no change in Congressmembers’ likelihood of issuing an online public statement about deportation and immigration after the implementation of the family separation policy. However, Congressmembers were less likely to issue online public statements about deportation and immigration after the policy was suspended. I also find that when Congressmembers discussed deportation and immigration during the family separation policy, there was no measurable change in politically polarizing rhetoric, as measured by use of moralizing language, though their tweets about deportation and immigration were highly moralized compared to their tweets on other topics throughout their entire two-year term. The findings confirm that legislators behave as strategic communicators on online social media platforms, as with more traditional mediums of political communication. More important, the findings suggest that, far from serving as a check on executive overreach, Congressmembers are unlikely to respond to executive assertions of power on deportation policy, even when policies are highly unpopular among the public. This raises concerns about shared governance on deportation policy and its potential implications for migration and democracy.
THE DEPORTATION SYSTEM AND POLICY FEEDBACK
Over the past four decades, social scientists have advanced understanding of the policy feedback process, or how “public policies are not merely products of politics but also shape the political arena and the possibilities for further policymaking” (Campbell 2012, 334). Policy feedback research differs from studies of political process, which focus on ascertaining how policies are produced (that is, policy as a dependent variable), and from policy analysis, which focuses on evaluating the outcomes of policies (that is, policy as an independent variable) (Mettler and SoRelle 2014; Mettler and Soss 2004; Pierson 1993). Instead, this multidisciplinary research engages public policies as living institutions that are not only continuously shaped by, but also actively define, the polity (that is, policy as independent and dependent variables) (Campbell 2012; Mettler and SoRelle 2014; Pierson 1993).
Within the context of the United States, research on policy feedback has primarily examined the role of political elites, interest groups, and the public in the development and implementation of social welfare policies, such as social security, health care, and Aid to Families with Dependent Children, and, in turn, how these policies affect the political behavior of these actors (Campbell 2003; Hacker 2002; Mettler 2005; Pierson 1994; Skocpol 1992; Soss and Schram 2007). In addition to comprehensively identifying the diverse individuals and organizations involved in social welfare policymaking (for example, elected legislators, caseworkers, labor unions, businesses) and tracing their interactions, this research has helped to explain the evolution of social welfare policy over time and to elucidate the obstacles to and possibilities for change (for reviews, see Campbell 2012; Mettler and SoRelle 2014; Mettler and Soss 2004). Specifically, Suzanne Mettler and Mallory SoRelle (2014) identify four important findings from policy feedback research on American social welfare policy: Policies define citizenship or inclusion in the polity; there is a reciprocal relationship between policies and governance; policies produce interest groups whose advocacy enables policies to persist; and the process of defining and mobilizing support for a policy agenda is dynamic and continual. Taken together, this research provides a useful mapping of social welfare policy that connects and galvanizes researchers and practitioners who seek to improve not just social welfare policy but also the very process that generates and governs these policies.
A policy feedback approach may be similarly useful for organizing and supporting burgeoning research on deportation policymaking. The contemporary deportation system in the US is composed of a vast and growing array of immigration and criminal laws and procedures, developed and implemented by diverse actors at the federal and local levels—all of which collectively define and stratify the experiences of migrants to the US, including their initial journey, their navigation of daily life once they arrive, and their removal (and remigration) if and when they are deported (see Patler and Jones 2025, this issue). The deportation policies that are produced also often fail to produce intended effects and even backfire (Massey et al. 2002, 2014, 2016; Rocha et al. 2014). Given the scale and complexity of the deportation system, a policy feedback approach, with its emphasis on mapping diverse actors and elucidating dynamic processes, may provide an especially helpful theoretical framework for studying deportation policymaking.
There is growing research that examines the politics of the deportation system. This research, which is varied in foci, data, and methods, and which has parallels to policy feedback research on social welfare policy, can generally be organized into two analytic streams: research that analyzes how deportation policy shapes experiences of citizenship and non-citizenship, and research that analyzes the process of deportation policy agenda-setting. Studies from the first analytic stream elaborate the quotidian and extraordinary ways in which deportation policies define citizenship and non-citizenship and the resulting implications for migrants and others living in the US. For example, Asad L. Asad (2023) illustrates how the deportation system and its growing surveillance apparatus require citizen and noncitizen Latino migrants to regularly navigate forms of institutional inclusion that vary in benefits and coercion while Carolina Valdivia (2025, this issue) highlights the particularly precarious plight of previously deported undocumented migrants who have returned to the US and their efforts to navigate conditions of what she refers to as hyper-illegality. In a complementary line of research, Rene R. Rocha and colleagues (2015) show that the effects of immigration enforcement on civic inclusion are not limited to migrants but also affect native-born citizens, such that local immigration enforcement can decrease political trust among native-born Latinos while increasing political trust among non-Hispanic Whites. Additional studies show that deportation policies can activate feelings of linked fate among native-born Latinos (Maltby et al. 2020) and incentivize their withdrawal from electoral politics while simultaneously mobilizing them to engage in collective action (Walker et al. 2019).
Meanwhile, studies from the second analytic stream clarify the process of setting and mobilizing support for a deportation policy agenda, particularly how political elites and the mass public interact on issues about migration. Several studies show that political elites are instrumental in mobilizing public support for deportation and other types of restrictionist policies because they can legitimize views already held by some members of the public and provide compelling framings of target populations, social problems, and policy solutions (Newton 2008; Wallace and Zepeda-Millán 2020; Roe 1989; Schneider and Ingram 1993; see also Edelman 1974). Yet, this research also shows that there are important limitations to political elites’ influence. Namely, the effectiveness of elite cues is typically conditioned by political partisanship and local context (Hopkins 2010; Jones and Martin 2017), and any effects are usually short-lived, given that public attitudes about migration tend to remain relatively fixed (Flores 2014, 2017, 2018).
I contribute to this growing research on the politics of the deportation system by applying a policy feedback approach to consider how interactions between the two political branches of the US government—the executive branch and Congress—may shape contemporary deportation policy. As I elaborate in this article, interactions between the executive branch and Congress are important when it comes to understanding the development and implementation of deportation policy because these two branches possess unique governing power on issues of immigration, and because current trends pose serious challenges for shared governance and effective policymaking. My focus on how the executive branch and Congress interact on deportation policymaking also complements burgeoning research that examines the role of judicial decision-making in the deportation system (Asad 2019; Blasingame et al. 2024; Kocher 2019; Law 2010).
In the US, the executive branch and Congress assume primary responsibilities on immigration policy. In general, the plenary power doctrine provides Congress with “unfettered discretion to set forth immigration rules absent direct constitutional restraints” and caselaw has largely extended this judicial deference to the executive branch on matters of immigration (Koh 2020, 956). Legal scholars observe that this judicial deference has enabled the executive branch to exert much more influence on deportation compared to other policy domains. Because many constitutional protections ensuring checks and balances and judicial oversight and review do not apply in immigration law, the executive branch, which includes the president, appointed advisers, and fifteen executive departments and their associated federal agencies, have uniquely wide latitude for shaping deportation policy (Cox and Rodríguez 2009; Koh 2020; Legomsky 1984). Tools that the executive branch can wield include policymaking (for example, issuing executive orders) and administrative decision-making regarding organization, staffing, and operations of executive departments and federal agencies such as the Department of Homeland Security (DHS), Customs and Border Protection (CBP), and Immigration and Customs Enforcement (ICE) (Chen 2017; Kagan 2001; Koh 2020). Despite the executive branch’s unusual policymaking power as it relates to deportation, Congress nevertheless retains distinct powers for enacting deportation policies and checking executive overreach, including the ability to change immigration laws, set agency budgets, initiate oversight and investigations, and, most important for this study, publicly justify or criticize actions (Koh 2020).
However, two distinct yet reinforcing trends currently challenge this balance of governing power between the executive branch and Congress on deportation policy. First, the executive branch’s power in shaping deportation policy has increased in recent years as the deportation system has grown and executive administrations—both Republican and Democratic—have been increasingly willing to exercise such power (Donato and Amuedo-Dorantes 2020; Koh 2020). Second, as executive administrations have exercised more power on deportation policy in recent years, Congress has grown more politically polarized, rendering Congressmembers with fewer incentives to check executive overreach and uphold effective governance—especially when it means having to act against their party’s interests (Fennelly et al. 2015; Hacker and Pierson 2019; Layman et al. 2006). These two trends are not unrelated; for example, legal scholars observe that the uptick in executive policymaking on deportation stems in part from executive administrations increasingly viewing legislative reform on immigration as being politically infeasible (Cox and Rodríguez 2009; Koh 2020; see also Donato and Amuedo-Dorantes 2020). Taken together, these two trends significantly distort the current deportation policymaking process and pose broader concerns about policy stability and coherence, shared governance, and democratic responsiveness (Chen 2017; Cox and Rodríguez 2015; Koh 2020).
As such, the current context calls for social scientists who are concerned about deportation policy to closely examine interactions between the executive branch and Congress, particularly when and how Congress responds to executive assertions of power. Toward this end, I apply the policy feedback approach to examine the reactions of the 115th US Congress to the Trump administration’s separation of migrant families at the southern border in 2018. In the next section, I discuss this empirical case, including the opportunities and challenges it presents for addressing data limitations that have long hindered research on the deportation system.
CONGRESS AND THE 2018 FAMILY SEPARATION POLICY
On May 7, 2018, the US Department of Justice (DOJ) announced the implementation of the Trump administration’s family separation policy (Sessions 2018a; Dickerson 2022; see also Alcaraz et al. 2024). The announcement authorized federal authorities to prosecute migrants attempting to cross the southern border and to separately detain parents and children in CBP facilities while they awaited criminal proceedings and deportation. The Trump administration’s separation of migrant families was part of a broader zero tolerance approach to border security (Sessions 2018b; Trump 2018a), which originates from the George W. Bush administration’s response to the 9/11 attacks and relies on the use of criminal prosecution to deter unauthorized entry by migrants at the southern border (Dickerson 2022; Patler and Jones 2025, this issue). Although previous executive administrations—both Republican and Democratic—have considered separating families to deter migration, the practice was not adopted as part of a federal approach to immigration and deportation until 2018 (Dickerson 2022). The family separation policy, which affected 3,924 children between January 20, 2017, and January 20, 2021, based on estimates from the DHS (2023) faced swift and strong public disapproval (Arango and Cockrel 2018; Everett and Caygle 2018; YouGov 2018), and was quickly suspended by President Trump through an executive order issued on June 20, 2018 (Trump 2018b).1
I argue that the Trump administration’s separation of migrant families in 2018 serves as a useful empirical case for examining when and how Congress responds to executive assertions of power on deportation policy, particularly during a period of heightened political polarization. There are three main reasons. First, the separation of migrant families exemplifies both an assertion of executive power over deportation policymaking and the limitations of that power. That is, it illustrates the extraordinary political power over deportation that is increasingly consolidated in the executive branch, but also demonstrates the heavy political liabilities that executive administrations incur when wielding this power (Cox and Rodríguez 2009; Koh 2020). This dynamic example of executive power in practice is useful for empirical purposes, as social science research on deportation has typically examined static instances of executive power. Second, the short time frame in which the separation of migrant families was authorized and then subsequently halted by the Trump administration—forty-five days in total—provides a unique window for observing how deportation policy immediately affects governance. Specifically, the separation of migrant families can provide insights into political learning, or how policies shape the political attitudes and behaviors of elected legislators, which has been challenging for policy feedback researchers to empirically observe (Mettler and SoRelle 2014; Pierson 1993). Third, given that “feedback research has focused on historical periods in which partisan cleavages were far less prominent than they are today,” this empirical case can test existing knowledge about policy feedback and provide updates for the current era of elite political polarization (Hacker and Pierson 2019, 13).
In this article, I focus on empirically elaborating one type of readily observable policy feedback during the Trump administration’s separation of migrant families in 2018: how Congressmembers publicly responded on Twitter to the implementation and suspension of the family separation policy. This approach has some analytical advantages. Namely, this approach leverages new and relatively accessible digital data to advance knowledge about the deportation system, which social scientists have traditionally studied using survey, interview, and administrative data that are more difficult to obtain. Indeed, researchers are increasingly turning to digital data to study immigration and deportation by, for example, analyzing social media data to examine attitudes about immigration among political elites and the public (Flores 2017; Mendelsohn et al. 2021) and using Google Trends data to measure public awareness of deportation threat (Alsan and Yang 2024; Johnson et al. 2024).
However, an important limitation of this approach is that I am unable to determine Congressmembers’ motives for their behavior—for example, whether specific actions taken by the Trump administration caused them to issue a public statement on Twitter. As such, I am unable to make any causal inferences. Another limitation is that although tweeting is an increasingly important form of political behavior among Congressmembers (Barberá et al. 2019; Golbeck et al. 2010; Hemphill et al. 2013), tweeting does not constitute policymaking, and I am unable to make any inferences about Congressmembers’ legislative activity based solely on their tweets. Nevertheless, analyzing Congressmembers’ public statements during this time frame can provide much-needed empirical insights about congressional political behavior during executive assertions of power on deportation, which can be further elaborated by future research that examines more empirical cases.
Should we expect Congressmembers to publicly react to executive assertions of power on deportation policy, and if so, how might we expect them to react? Extensive and long-standing research in political communication demonstrates that Congressmembers are strategic communicators focused on reelection (Fenno Jr. 1978; Grimmer 2013; Mayhew 1974). Increasingly, Congressmembers rely on social media platforms such as Twitter to engage in strategic communication. A growing number of studies show that Congressmembers use social media to share their work with their constituencies and engage in credit-taking, as well as to gauge the sentiment of their constituents and the broader public to inform the positions they take and the policy priorities they set (Barberá et al. 2019; Box-Steffensmeier and Moses 2021; Golbeck et al. 2010; Hemphill et al. 2013; Lee et al. 2025). In contemporary American politics, then, online engagement is an increasingly prevalent type of congressional political behavior.
A growing body of research on political polarization and social media also shows that political elites, such as Congressmembers, increasingly engage in politically polarizing discourse that reinforces in-group solidarity and out-group antipathy along party lines (Hacker and Pierson 2019; Layman et al. 2006), and that this tendency is further amplified on social media (Rafail et al. 2024; Cinelli et al. 2021; see also Barberá et al. 2020). A recent study finds that since the 1970s, political elites have used more politically polarizing rhetoric over time to discuss the topic of immigration (Card et al. 2022). A hallmark of politically polarizing rhetoric in contemporary American politics is the use of moralizing language, or language that frames political arguments using moral values that emphasize cooperation (that is, other-regarding behavior) and therefore casts opponents as not only uncooperative but also immoral (Jung and Clifford 2025). Studies show that elite political discourse is increasingly moralized, in part because moralizing language is highly effective in mobilizing co-partisans and attacking members of other parties (Clifford et al. 2015; Jung 2023; Wang and Inbar 2021, 2022). Moralizing language serves as a particularly effective tool for Congressmembers when they need to act expeditiously during moments of political tumult to either justify their party’s political actions or criticize those of the opposite party.
Given that previous research finds that Congressmembers tend to behave as strategic communicators and increasingly rely on politically polarizing rhetoric based on moralizing language to pursue political goals, I hypothesize that Congressmembers—both Republicans and Democrats—increased their discussion of deportation and immigration on Twitter immediately following the Trump administration’s implementation of the family separation policy, albeit for different reasons. Republicans who supported the policy, or at least viewed it as a political victory, likely sought to publicly align themselves with the administration and the policy, whereas Democrats who opposed the policy, or at least viewed it as a political liability, likely sought to publicly criticize the administration and the policy. I also hypothesize that Congressmembers across party lines increased their discussion of deportation and immigration on Twitter immediately following the suspension of the policy. Here, Republicans likely sought to justify the administration’s actions in order to avoid the appearance of suffering a political loss, whereas Democrats likely sought to highlight their opposing party’s apparent political misstep. With both events, Congressmembers’ use of moralizing language in their tweets about deportation and immigration serves as a gauge of political polarization during this time frame. I hypothesize that Congressmembers were more politically polarized, and used more moralizing language after each executive action on the family separation policy, as they likely viewed the widely publicized actions as moments of political tumult in which they needed to quickly mobilize their party.
DATA AND METHODS
To examine Congressmembers’ public statements on Twitter during the Trump administration’s implementation and suspension of the family separation policy, I constructed a novel dataset from Twitter and analyzed these data using computational text analysis methods and a quasi-experimental research design. The analysis consisted of four main tasks: constructing a tweet corpus, identifying tweets about deportation and immigration, coding tweets for moralizing language, and measuring trends in congressional tweeting behavior immediately after the start and end of the family separation policy.
The 115th US Congress Tweet Corpus
I constructed a dataset, or corpus, of tweets generated by members of the 115th US Congress during their term, which began on January 3, 2017, and ended on January 3, 2019. This corpus consists of approximately 830,000 tweets. Table 1 summarizes this corpus, describing the Congressmembers and their tweets. All except eight Congressmembers had a Twitter account, which provides further evidence that social media is an important tool for Congressmembers in contemporary American politics (Barberà et al. 2019; Golbeck et al. 2010; Hemphill et al. 2013). In terms of tweets, slightly more than half (58 percent) were written by Democratic Congressmembers, who tended to tweet more on average compared to Republican Congressmembers (approximately 1,884 tweets per Democrat versus 1,104 tweets per Republican).2 The tweets were also overwhelmingly generated by Congressmembers who are incumbents (87 percent) and males (75 percent), though this is largely consistent with their overrepresentation in Congress more generally. Appendix A provides additional information about this corpus, including detailed information on how I constructed this corpus.
Descriptive Statistics of 115th US Congress Tweet Corpus
I also preprocessed the text in this corpus. Because previous research shows that preprocessing decisions, and the order in which they are implemented, can have important downstream implications for text analysis (Denny and Spirling 2018; Grimmer et al. 2022), I describe my preprocessing steps here. I first implemented a set of steps to make the tweet text machine-readable: For example, I set all text to lowercase; removed URLs, emojis, and punctuation; replaced the ampersand symbol with the word “and”; and removed white spaces around text. Notably, I retained hashtags to bolster my ability to identify immigration-related tweets, though I was unable to analyze the hashtags themselves in later analyses because hashtags are not recognized as everyday language and therefore not included in common computational text analysis tools. I then worked to identify tweets written in English since the text analysis tools I use perform best with English language text. I used the textcat package in R (Hornik et al. 2013a, 2024b), which supports n-gram-based language categorization, to detect tweets written in English and Spanish. Approximately 98 percent of tweets were detected as containing English. I removed non-English tweets, resulting in a slightly smaller corpus that still retains representation of all members of the 115th Congress who had active Twitter accounts during this term.
The preprocessed corpus contains 820,267 tweets from 548 members of the 115th Congress. For the text analyses, I further removed certain words from tweets to focus the analyses. For example, I removed numbers, stopwords, or frequently occurring words that are not substantively meaningful (for example, I, this, that), and extremely rare words that appear less than ten times in the entire corpus and may therefore introduce misleading signals in the analyses. I also performed procedures to tokenize text, or split tweets into individual words, so that they could be readily measured using computational text analysis tools.
Identifying Deportation and Immigration-Related Tweets
To measure trends in Congressmembers’ discussion of deportation and immigration on Twitter during the family separation policy, it is necessary to identify tweets in the corpus that mention these topics. Given the size and heterogeneity of the corpus, I used computational text analysis methods to identify relevant tweets. I completed a preliminary analysis to compare a simple dictionary-based approach and Structural Topic Modeling (Roberts et al. 2013, 2014). I found that the dictionary-based approach was more appropriate for identifying deportation- and immigration-related tweets for this corpus because these tweets occur relatively infrequently and do not cluster in a single topic about immigration but instead appear across diverse topics. As such, I proceeded with a dictionary-based approach to identify relevant tweets. I compiled a dictionary of words and phrases related to deportation and immigration by drawing on previous research on American public opinion on immigration (Levy and Wright 2020; Wright and Levy 2019) and immigration discourse on Twitter (Rowe et al. 2021; Stocking et al. 2018), as well as feedback from fellow contributors to this issue. This deportation and immigration dictionary, which consists of fifty-six words and phrases, is available for use in future research and provided in appendix B.
Coding Tweets for Moralizing Language
Trends in Congressmembers’ use of moralizing language on Twitter during the family separation policy can serve as a gauge of political polarization. To examine this, I coded deportation- and immigration-related tweets, as well as tweets about other topics, using the extended Moral Foundations Dictionary (eMFD) (Hopp et al. 2020), which is based on Moral Foundations Theory (Haidt and Graham 2007; Graham et al. 2009). This social psychological theory posits that there are five main moral values: care, fairness, loyalty, authority, and sanctity. Care and fairness emphasize individual rights and welfare and tend to be associated with political liberalism, with care centered on concerns about nurture and protection and fairness centered on concerns about reciprocity and justice (Graham et al. 2009). Loyalty, authority, and sanctity emphasize group cohesion and tend to be associated with political conservatism, with loyalty focused on fulfilling obligations, authority focused on respecting hierarchical roles, and sanctity focused on maintaining cultural boundaries (Graham et al. 2009). Each value can be expressed in positive or negative terms. In political discourse about deportation, for example, care can be positively invoked to emphasize the need for compassion for migrants or negatively invoked to emphasize the need to protect citizens from potential harm posed by migrants. Although there are many other value systems, research shows that the Moral Foundations may be most relevant for identifying moralizing language, especially in American political discourse, because these values tend to be most moralized among Americans (Jung and Clifford 2025).
The eMFD is a validated tool that quantifies the presence and valence of the five moral values in everyday text. The eMFD consists of a set of 3,270 words, in which each word has a defined level of association with each of the moral values—that is, the probability that the word, when used, is associated with a particular value on a scale of 0 (no association) to 1 (perfect association)—and a defined sentiment of association with a particular moral value on a scale of −1 (most negative) to 1 (most positive). For example, the words betrayal and loyalty are both strongly associated with the value of loyalty (level = 0.48 and 0.47, respectively). However, betrayal has a negative sentiment (−0.41) while loyalty has a positive sentiment (0.55).
I used the eMFD to code all deportation- and immigration-related tweets, as well as tweets about other topics, for moralizing language. For each tweet, I determined whether any of the words appeared in the eMFD. I then measured each tweet’s level and sentiment of association with the five moral values—care, fairness, loyalty, authority, and sanctity—as the mean of all words in the tweet that appeared in the eMFD. Overall, 98.45 percent of all tweets in the corpus contained at least one word from the eMFD. Within a given tweet, an average of 56.23 percent of words appeared in the eMFD (SD = 17.21 percent). In table 2, I provide examples of two deportation- and immigration-related tweets that were strongly associated with each of the five moral foundations: one with overall positive sentiment and one with overall negative sentiment.
Examples of Tweets Highly Associated with the Five Moral Foundations
Measuring Trends in Congressional Tweeting Behavior
To measure trends in Congressmembers’ tweeting behavior during the family separation policy, I used interrupted time series (ITS) analysis. ITS analysis is appropriate for this study because I am evaluating a complex intervention, defined as a set of time points of interest that are interrelated and occur closely in time (the implementation and suspension of the family separation policy), rather than a single intervention occurring at one point in time. Compared to other quasi-experimental approaches such as regression discontinuity in time designs (Hausman and Rapson 2018), ITS analysis provides greater flexibility for addressing the challenges associated with evaluating complex interventions (Bernal et al. 2017, 2018). ITS has also previously been used in studies examining Twitter behavior (Jaidka et al. 2019; Endres et al. 2021).
I observed congressional tweeting behavior between April 1, 2018, and August 1, 2018, which covers approximately one month before the Trump administration implemented the family separation policy (May 7, 2018) and one month after the administration suspended the policy through executive order (June 20, 2018). I used this relatively short time frame because political discourse on Twitter transpires rapidly and therefore it was likely that Congressmembers responded quickly to the family separation policy on Twitter if they responded at all. The unit of analysis is at the tweet-level and observed on a weekly basis. Congressmembers generated 159,787 tweets during this period, which make up 19 percent of all tweets in the corpus. Of these tweets, 12,580 tweets were identified as discussing deportation and immigration.3 I measured two dimensions of congressional tweeting behavior during this time frame, which serve as my dependent variables: the proportion of weekly tweets that were about deportation and immigration, and the average proportion of moralizing language or eMFD words in weekly tweets about deportation and immigration. I used generalized least squares (GLS) models to account for potential temporal trends in the data. I estimated separate models for each dependent variable, with both time points of interest included in all models. I also estimated separate models by party to examine partisan differences (six models in total). Formally, the models can be expressed as follows:
where B0 is the baseline level of the outcome (either the proportion of weekly tweets that were about deportation and immigration or the average proportion of eMFD words in weekly tweets about deportation and immigration), B1 is the change in outcome associated with a time-unit increase, B2 is the change in the level of the outcome following the intervention (either the start or end of the family separation policy), B3 is the difference between the pre-intervention and post-intervention slopes of the outcome, T is the number of weeks elapsed since the start of the study, and Xt is a dummy variable indicating intervention status at time t.
RESULTS
Tweets About Deportation and Immigration by the 115th US Congress
In general, members of the 115th US Congress discussed deportation and immigration relatively infrequently on Twitter during their term between 2017 and 2019. Using a dictionary-based approach, I identified 42,716 tweets about deportation and immigration, which constitute approximately 5 percent of all tweets in the corpus. Given that immigration was a top policy priority for the Trump administration, as evidenced by the many presidential orders on immigration enforcement that were issued (Donato and Amuedo-Dorantes 2020; Immigration Policy Tracking Project 2025; Waslin 2020), it is striking that Congressmembers rarely discussed deportation and immigration on Twitter, especially since the president’s party held majorities in both chambers of Congress and Republican Congressmembers therefore could have leveraged this advantage to advance the administration’s immigration policy agenda.
When deportation and immigration were discussed on Twitter, Democrats were considerably more likely to tweet about these topics compared to Republicans, generating three-fourths of all deportation- and immigration-related tweets (n = 32,207) in the corpus, despite constituting less than half of all members in Congress. The average number of deportation- and immigration-related tweets for Democrats was about 128 tweets, compared to 33 tweets on average for Republicans. This sizable partisan disparity in tweeting was driven by a small number of highly active Democrats, some of whom, such as Representative Pramila Jaypal and Representative Joaquin Castro, often appeared in traditional news media as outspoken critics of the Trump administration during this period. Table 3 describes the top five Congressmembers from each party who tweeted about deportation and immigration and provides examples of their tweets using the word border. As might be expected, Republicans typically invoked border to voice support for the Trump administration’s policies, whereas Democrats used the same word to criticize these policies.
Top Five Congressmembers Tweeting About Deportation and Immigration, 2017–2019
Democratic and Republican Congressmembers also tended to use different words to discuss deportation and immigration on Twitter. Appendix table C.1 shows the most distinctive words and hashtags in deportation- and immigration-related tweets among Democrats and Republicans, which are measured using term frequency-inverse document frequency (TF-IDF) values. TF-IDF is a natural language processing measure that quantifies how important a word is to a corpus of text by taking into account the frequency with which a word appears in a document—in this case, a tweet—and the overall corpus (see Grimmer et al. 2022; Jurafsky and Martin 2024). As the table shows, there was no overlap in distinctive words and hashtags in deportation- and immigration-related tweets between Democrats and Republicans. Whereas Democratic discourse on deportation and immigration on Twitter was defined by words and hashtags such as #keepfamiliestogether, #standwithdreamers, and #refugeeswelcome, Republican discourse on these topics was defined by words and hashtags such as #securetheborder, native, and #fairimmigration. This indicates strong linguistic polarization among Democrats and Republicans in discourse about deportation and immigration and underscores the importance of using a well-defined dictionary developed by domain experts to appropriately analyze this distinct type of discourse.
In addition, Congressmembers’ tweets about deportation and immigration contained more moralizing language on average compared to their tweets about other topics. I calculated the mean level and sentiment of each of the five moral values in each Congressmember’s tweets on these topics and their tweets about other topics, and then conducted paired t-tests on each of these ten variables. Table 4 shows that there are significant and strong effects of tweet topic, such that tweets about immigration were strongly and negatively associated with each of the five moral values. This finding is consistent with previous research showing that political elites increasingly use politically polarizing rhetoric involving moralizing language to advance goals on immigration (Card et al. 2022; Simonsen and Bonikowski 2022), and further attests to generally high levels of political polarization in Congress during this time period.
Level and Sentiment of the Five Moral Foundations in Immigration and Non-Immigration Tweets
Trends in Congressional Tweeting Behavior During the Family Separation Policy
Although deportation and immigration were relatively rare topics of discussion on Twitter for members of the 115th Congress during their two-year term, they were frequently discussed during the forty-five days in which the Trump administration implemented and suspended its family separation policy. Figure 1 shows the weekly number of deportation- and immigration-related tweets by Republican and Democratic Congressmembers between April 1 and August 1, 2018, with the first vertical line indicating the implementation of the family separation policy (May 7) and the second vertical line indicating the suspension of the policy (June 20). The figure reveals an acute spike in deportation- and immigration-related tweets between the start and end of the family separation policy for Congressmembers of both parties, with the weekly number of relevant tweets for each party increasing rapidly for several weeks post-implementation and then finally decreasing about a week before President Trump issued an executive order to halt family separations. Although Democrats generally tweeted more about deportation and immigration during the study period, trendlines are remarkably parallel for Democrats and Republicans. This suggests that during the family separation policy, Congressmembers made more public statements on Twitter about deportation and immigration than usual.
Weekly Number of Deportation- and Immigration-Related Tweets by Democratic and Republican Congressmembers, April 1 to August 1, 2018
Source: Author’s compilation.
Note: The first vertical line indicates the implementation of the Trump administration’s family separation policy (May 7, 2018), and the second vertical line indicates the suspension of the policy (June 20, 2018).
Results from the GLS models provided in table 5 help to clarify these trends. There is no evidence of an association between Congressmembers’ proportion of weekly tweets that discuss deportation and immigration and the start of the family separation policy. However, there is evidence that Congressmembers were less likely to tweet about deportation and immigration after the end of the policy (coefficient = −0.14, p-value = 0.04). In the party-specific models, there is evidence of a post-suspension decrease in relevant tweets for Democrats only (coefficient = −0.18, p-value = 0.02). These findings run counter to my hypotheses that Congressmembers—both Republicans and Democrats—would be politically motivated to increase their discussion of deportation and immigration on Twitter following the implementation and suspension of the family separation policy. The lack of a post-implementation increase in relevant tweets for both parties suggests that there may be a potential lag between executive assertions of power on deportation and congressional responses, particularly when Congressmembers view an action as politically unviable and unpopular, irrespective of their own partisan considerations. The significant post-suspension decrease in relevant tweets among Democrats is somewhat surprising, as it suggests that Congressmembers may be more apt to move on from, rather than capitalize on, executive missteps in deportation policymaking when they are not the party in power, especially when they view the policy as being unpopular. Indeed, the family separation policy was not only widely publicized but also met with immediate and strong public disapproval (Arango and Cockrel 2018; Everett and Caygle 2018; YouGov 2018). Congressmembers, regardless of their party affiliation, may have therefore sought to distance themselves from this highly unpopular executive action—even after the action was suspended.
Generalized Least Squares Models Predicting Weekly Proportion of Tweets About Deportation and Immigration Among Congressmembers, April 1 to August 1, 2018
I also hypothesized that Congressmembers likely viewed the executive actions on the family separation policy as generating political tumult and increased their use of moralizing language during this time frame to mobilize their own party and attack the opposing party. There is preliminary evidence partially supporting this hypothesis based on a basic quantification of Congressmembers’ use of moralizing language in deportation- and immigration-related tweets. During the family separation policy, Democratic tweets on deportation and immigration used more moralizing language and more negative sentiment, while Republican tweets actually used less moralizing language. Figure 2 shows the average level and sentiment of the five moral values—care, fairness, loyalty, authority, and sanctity—in Congressmembers’ weekly tweets on deportation and immigration during the study period, with the first vertical line indicating the implementation of the family separation policy (May 7) and the second vertical line indicating the suspension of the policy (June 20). The top half of the figure, corresponding to level of association with moral values, shows that compared to Republicans, Democrats generally used more moralizing language when tweeting about immigration and deportation during the study period. However, during the family separation policy, the parties diverged in their use of moralizing language, with Democrats referring more to care, fairness, authority, and sanctity in their tweets and Republicans referring less to these same values in their tweets. Interestingly, Democrats used not only individualizing values that are typically associated with political liberalism (care and fairness) during the family separation policy, but also binding values that are typically associated with political conservatism (authority and sanctity).
Average Weekly Level and Sentiment of the Five Moral Foundations in Deportation- and Immigration-Related Tweets by Democratic and Republican Congressmembers, April 1 to August 1, 2018
Source: Author’s compilation.
Note: The first vertical line indicates the implementation of the Trump administration’s family separation policy (May 7, 2018) and the second vertical line indicates the suspension of the policy (June 20, 2018).
The second half of figure 2, corresponding to sentiment of association with moral values, shows that Congressmembers from both parties expressed the five values with negative sentiment (for example, care as protection from harm instead of compassion) during the study period. However, during the family separation policy, Democratic tweets were somewhat more negative, perhaps because Democrats sought to mobilize their party in a reactive manner (for example, highlighting the harms they believed the policy was inflicting). Republican tweets, on the other hand, remained relatively unchanged in terms of sentiment during the family separation policy, though there were more positive expressions of loyalty after the end of the policy.
Results from the GLS models in table 6 temper these preliminary findings. In the general and party-specific models, there was no evidence of an association between Congressmembers’ proportion of weekly deportation- and immigration-related tweets that use moralizing language and either the start or end of the family separation policy. This lack of an association runs counter to my hypothesis that Congressmembers from both parties would immediately and noticeably increase their use of moralizing language in order to mobilize their own party to respond to the political tumult generated by the executive actions on migrant families. However, it is quite possible that no change was detected in Congressmembers’ use of moralizing language immediately after the implementation and suspension of the family separation policy because, as previously discussed, their tweets about deportation and immigration were already highly moralized relative to their tweets about other topics (see table 4).
Generalized Least Squares Models Predicting Average Proportion of Moralized Words in Weekly Tweets About Deportation and Immigration Among Congressmembers, April 1 to August 1, 2018
Results from an additional set of models potentially suggest an unexpected way that Congressmembers may use moralizing language during moments of political tumult involving deportation policy. Table 7 presents results from GLS models estimating the association between Congressmembers’ use of moralizing language in weekly tweets that are not about deportation and immigration and the start and end of the family separation policy. There is some evidence that immediately after the implementation of the policy, Congressmembers were more likely to use moralizing language in tweets about topics unrelated to deportation and immigration (coefficient = 0.01, p-value = 0.08). The party-specific models provide some evidence of a post-implementation increase in moralizing language in Democratic tweets about other topics (coefficient = 0.01, p-value = 0.08). For example, many Democrats used moralizing language in the week after the start of the family separation policy to criticize the 2018 Farm Bill, which at the time proposed new work requirements for Supplemental Nutrition Assistance Program recipients. These results potentially suggest that during periods of political upheaval generated by specific executive actions on deportation, Congressmembers may not necessarily increase their use of moralizing language about the topic at hand, but instead draw on moralizing language to discuss other issues as a way to mobilize their party more generally, especially when their party is not in power. However, these results, which suggest an apparent spillover of moralizing discourse from deportation and immigration to other issues during politically tumultuous moments, are quite tentative and should be examined more closely in future work.
Generalized Least Squares Models Predicting Average Proportion of Moralized Words in Weekly Tweets About Other Topics Among Congressmembers, April 1 to August 1, 2018
Role of Retweets
The trends I identified in congressional tweeting behavior during the family separation policy may be misleading if this behavior was systematically different by political party. Namely, it is possible that Republicans tweeted less about deportation and immigration during this time frame and instead relied on retweeting President Trump at the time, who frequently tweeted about immigration in this period. I examined retweets and found little evidence supporting this. Table C.2 shows that there was no partisan difference in likelihood of retweeting in general nor on the specific topic of deportation and immigration. In fact, Democrats were somewhat more likely to make use of retweets when discussing immigration (0.23 of all tweets) compared to other topics (0.21). Table C.3 also shows that President Trump was not the most frequently retweeted account by Republicans on the topic of deportation and immigration, as might be expected; the top retweeted accounts among Republicans were @FoxNews and @HouseHomeland (account at the time for the DHS). Together, this suggests that it is unlikely that the trends I observed in congressional tweeting behavior during the family separation policy were driven by Republicans’ greater reliance on retweeting the executive.
DISCUSSION AND CONCLUSION
In this study, I examined how Congressmembers publicly responded on Twitter to the Trump administration’s family separation policy in 2018. I applied a policy feedback approach (Mettler and Soss 2004; Pierson 1993; Skocpol 1992) and constructed a novel dataset of tweets to computationally analyze Congressmembers’ public statements on Twitter immediately following the implementation and suspension of the family separation policy. My analysis shows that there was no change in Congressmembers’ likelihood to issue online public statements about deportation and immigration after the implementation of the widely publicized and highly unpopular policy, though they were less likely to make such statements after the policy was suspended. When Congressmembers publicly discussed deportation and immigration during the family separation policy, there was no measurable change in their rhetoric in terms of moralizing language, though their tweets on deportation and immigration were highly moralized compared to their tweets on other topics throughout their two-year term. It seems that during the family separation policy, Congressmembers from both parties generally did not exercise their ability to check executive overreach and instead chose to distance themselves from a policy they viewed as politically unviable and unpopular.
These findings provide helpful and timely insights into how an increasingly empowered executive branch and a politically polarized Congress interact to shape contemporary deportation policy. In the US, statutory frameworks call for shared governance on deportation policy between Congress and the executive branch, which is distinct from how other policy areas are typically governed (Cox and Rodríguez 2009; Koh 2020). As executive power on deportation has expanded over time, it has become increasingly incumbent for Congress to check executive overreach (Donato and Amuedo-Dorantes 2020; Koh 2020). The findings in this study cast doubt on the ability of Congress to fulfill this role. That Congressmembers are unwilling to respond to executive assertions of power on deportation when their party is in power may not be surprising, given that elite political polarization decreases incentives for Congressmembers to act against their own party (Hacker and Pierson 2019). That is, it is not surprising that Republicans did not publicly respond in a measurable way to the Trump administration’s family separation policy. However, it is surprising to observe this same behavior among Democrats, who we would expect, based on the literature on elite political polarization, to have strong political incentives to criticize these executive actions. Although a small number of Democrats vocally criticized and even engaged in protest against the family separation policy (Arango and Cockrel 2018), the study shows that the vast majority of Democrats were not more likely to publicly discuss deportation and immigration on Twitter after the start of the policy and were even less likely to discuss these topics after the end of the policy. These findings are especially concerning, given that these executive actions were widely publicized and highly unpopular (Everett and Caygle 2018; YouGov 2018), which would seemingly compel, or at least allow, Congressmembers—of either party—to respond. This study reveals that when the executive branch asserts power on deportation, Congressmembers may be just as likely to avoid responding as they are to justify or criticize these actions, which has the effect of allowing the boundaries of executive power to expand even as specific actions—such as the family separation policy—may succeed or fail.
This study has limitations. Namely, while I analyzed when and what Congressmembers posted on Twitter, I am unable to determine why they posted—for example, whether they were motivated to post in response to specific actions taken by the Trump administration and public outcry. Additional data on Congressmembers’ interactions with the Trump administration and the public during this time may help to test causal links. This is an important limitation of this study, which reflects general limitations of observational data and long-standing data and methodological challenges that hamper policy feedback research (Campbell 2012; Pierson 1993). I am also unable to rule out the possibility that Congressmembers’ tweeting behavior was influenced by other deportation- and immigration-related events that occurred during the same period, though I attempted to mitigate this by defining a short study period.
Future research should continue to elaborate the relationship between deportation policy and governance by broadening analysis to include more individuals and organizations who assume important roles in federal deportation policymaking alongside the president, Congress, and the judiciary. For example, more research is needed to understand federal agencies and their internal and external dynamics with regard to deportation policy (Chen 2017; Cox and Rodríguez 2015; Rocha et al. 2014). These include federal agencies with obvious purview over deportation policy, such as the DHS, CBP, and ICE, but may also include agencies that appear less directly relevant, such as the Department of the Treasury and the Department of Commerce (including the US Census Bureau). Relatedly, there is also a need for more research on the individuals and organizations that constitute what Ian G. Peacock (2025, this issue) calls the “subfederal deportation system.” For example, sheriffs often exercise considerable discretion in enforcing deportation policy at the county level, and understanding when and why they exercise discretion can surface new and useful insights (Peacock 2025, this issue; Pedroza 2019). In addition, there is ample opportunity to understand the role of nongovernmental interest groups in deportation policymaking and how these groups collaborate and compete to wield influence on deportation policy over time (Wong 2006). As the deportation system expands and evolves, it will be necessary for social scientists to broaden their efforts to analyze more actors and actions, while also connecting these efforts into a unified account of this increasingly pervasive and consequential system.
The first months of the second Trump administration indicate that executive action on deportation, which has increased steadily over the past several executive administrations, is likely to accelerate and expand (Congressional Research Service 2025). This development, paired with continued elite political polarization, will likely make interventions by Congress simultaneously more necessary and more difficult. Moving forward, social science scholarship will be crucial for tracking and explaining the increasingly tight coupling of deportation policy and governance in the US and its implications for migration and democracy.
APPENDIX A. DEVELOPMENT OF THE 115TH US CONGRESS TWEET CORPUS
I constructed a dataset of all tweets generated by the 115th US Congress by compiling and linking data across three sources. First, I collected data on the members of the 115th Congress, which commenced on January 3, 2017, and ended on January 3, 2019, from the official congressional website (https://congress.gov). For this Congress, there were a total of 561 members, consisting of 115 Senators, 440 Representatives, and 6 nonvoting delegates. Notably, there was considerable turnover, due in part to this term commencing shortly after the 2016 election, with 42 members serving only part of this term. This Congress was also distinct in that both the Senate and House of Representatives had Republican majorities. I used this data source to identify members as well as to construct several member-level covariates, namely party, incumbency status, and gender.
To collect Twitter handles, or account names, for this Congress, I obtained a list of Congressmember handles used in a previous study (Schwemmer et al. 2020), and manually added and updated handles for Congressmembers who had missing or outdated handles in this list, preferring verified to unverified accounts. Of the 561 members of the 115th Congress, I was able to identify Twitter accounts for 553 members. The eight members without Twitter accounts likely either did not have accounts or deleted them by the time I began this study.4
Finally, I collected tweets generated by members of the 115th Congress via the Twitter Application Programming Interface (API) with the rtweet package in R (Kearney 2018). Using the Twitter API, I collected the 3,200 most recent tweets associated with each identified Congressmember handle.5 The nature of the API pulls made it so that I was only able to collect tweets that were present on that date, meaning that I may be missing tweets that were created but deleted before that date, which is an acknowledged limitation of this dataset. I collected tweets for 520 Congressmembers shortly after the end of the 115th congressional term on January 11, 2019. Twelve Congressmembers who were not reelected had already deleted their accounts by this date; for these Congressmembers, I relied on tweets collected earlier, on October 26, 2018. I collected tweets for 21 additional Congressmembers on February 19, 2019, March 14, 2019, and March 15, 2019. I was ultimately able to collect 830,158 tweets for 548 members of the 115th Congress, with 5 members having no tweets generated during this term despite having accounts.6 For each Congressmember, I retained tweets generated during the time they served as part of the 115th Congress (whether for a full or partial term).
The 115th US Congress Tweet Corpus I have assembled for this study is a novel and useful dataset, given its comprehensive coverage of social media activity among Congressmembers for the entirety of a congressional session, as well as its overlap with the first two years of the first Trump administration. These types of social media datasets are becoming rare, as platforms such as Twitter (now known as X) increasingly restrict research access to platform data.
APPENDIX B. DEPORTATION AND IMMIGRATION DICTIONARY
The following is a list of words and terms I used to identify tweets discussing deportation and immigration in the 115th US Congress Tweet Corpus.
APPENDIX C: ADDITIONAL TABLES
Most Distinctive Words and Hashtags in Deportation- and Immigration-Related Tweets Identified with TF-IDF Values
Tweets and Retweets by Party and Topic
Top Five Twitter Accounts Retweeted by 115th Congress on Topic of Deportation and Immigration
FOOTNOTES
↵1. Although it is difficult to definitively ascertain the number of migrant families who were affected by the 2018 family separation policy due to significant data limitations, a separate report (Dickerson 2018) concludes with a similar figure, estimating that a minimum of 4,335 migrant children were separated from their parents during the Trump administration as of June 18, 2018.
↵2. Members of other parties (Independent and Libertarian) tweeted quite prolifically given their small numbers in Congress; the average number of tweets for members of other parties during this time period was approximately 2,076.
↵3. I also removed a small number of Congressmember tweets about deportation and immigration (n = 66) to ensure that the data for the models did not contain any missing values.
↵4. The eight members include Madeleine Bordallo (Democrat, Delegate for Guam), Jason Chaffetz (Republican, Representative for Utah District 3), Trent Franks (Republican, Representative for Arizona District 8), Tim Murphy (Republican, Representative for Pennsylvania District 18), Collin Peterson (Democrat, Representative for Minnesota District 7), Mike Pompeo (Republican, Representative for Kansas District 4), Tom Price (Republican, Representative for Georgia District 6), and Jeff Sessions (Republican, Senator for Alabama).
↵5. At the time of this study, 3,200 was the maximum number of tweets that could be collected from a single account, as set by the Twitter API.
↵6. These five members include Rob Bishop (Republican, Representative for Utah District 1), Evan Jenkins (Republican, Representative for West Virginia District 3), Brenda Jones (Democrat, Representative for Michigan District 13), Jon Kyl (Republican, Senator for Arizona), and Mick Mulvaney (Republican, Representative for South Carolina District 5).
- © 2025 Russell Sage Foundation. Law, Tina. 2025. “Unpacking the Politics of the US Deportation System.” RSF: The Russell Sage Foundation Journal of the Social Sciences 11(4): 49–75. https://doi.org/10.7758/RSF.2025.11.4.03. Thank you to Em Bello-Pardo for his engagement with this research project, including help with data collection and analysis. Thanks also to the issue editors, fellow contributors, and anonymous reviewers for helpful comments on earlier drafts. This research was supported in part by a collaborative research seed grant from the Summer Institutes in Computational Social Science, Russell Sage Foundation (regranted from grant number G-6821, ROR: https://ror.org/02yh9se80), and Alfred P. Sloan Foundation. Direct correspondence to: Tina Law, at law{at}ucdavis.edu, University of California, Davis, Social Sciences and Humanities, Department of Sociology, One Shields Drive, Room 1298, Davis, CA 95616, United States.
Open Access Policy: RSF: The Russell Sage Foundation Journal of the Social Sciences is an open access journal. This article is published under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
REFERENCES
In this issue
Jump to section
- Article
- Abstract
- THE DEPORTATION SYSTEM AND POLICY FEEDBACK
- CONGRESS AND THE 2018 FAMILY SEPARATION POLICY
- DATA AND METHODS
- RESULTS
- DISCUSSION AND CONCLUSION
- APPENDIX A. DEVELOPMENT OF THE 115TH US CONGRESS TWEET CORPUS
- APPENDIX B. DEPORTATION AND IMMIGRATION DICTIONARY
- APPENDIX C: ADDITIONAL TABLES
- FOOTNOTES
- REFERENCES
- Figures & Data
- Info & Metrics
- References
Related Articles
Cited By...
- No citing articles found.