Overcoming the Political Exclusion of Migrants: Theory and Experimental Evidence from India*
Nikhar Gaikwad Assistant Professor Columbia University Department of Political Science 420 W 118 St, Mail Code 3320 New York, NY 10027 [email protected] ORCID ID: 0000-0002-9660-222X
Gareth Nellis Assistant Professor University of California, San Diego Department of Political Science Social Sciences Building 301 9500 Gilman Drive, Mail Code 0521 La Jolla, CA 92093 [email protected] ORCID ID: 0000-0002-8620-8640
April 2021
*We are grateful to the Institute of Social and Economic Research and Policy at Columbia University and the University of California, San Diego for funding; to MORSEL PLC for fieldwork; and to our NGO partner for their assistance with the intervention. For comments and advice, our thanks to Claire Adida, Allison Carnegie, Richard Clark, Jasper Cooper, Fabien Cottier, Thad Dunning, Susan Hyde, Hyeran Jo, Kimuli Kasara, Mayya Komisarchik, John Marshall, Carlo Prato, Kenneth Scheve, Jack Snyder, Michael Weaver, and workshop participants at Columbia University, the University of California, San Diego, the University of Rochester, the 2019 Conference on South Asia at Madison, the 2020 International Political Economy Society Annual Conference, the 2020 Global Research in International Political Economy Webinar, and Pia Raffler’s replication seminar at Harvard University. For outstanding research assistance, we are indebted to Heba Abbas, Thomas Brailey, Aaron Christenson, Aura Gonzalez, Karminder Malhotra, Bidisha Mandal, Rochan Mathur, Amritanshu Patnaik, Seungyup Sin, Nitin Teotia, and Shane Xuan. Overcoming the Political Exclusion of Migrants: Theory and Experimental Evidence from India
Abstract: Migrants are politically marginalized in cities of the developing world, participating in destination-area elections less than local-born residents. We theorize three reasons for this shortfall: migrants’ socio-economic links to origin regions; bureaucratic obstacles to enrollment that disproportionately burden newcomers; and ostracism by anti-migrant politicians. We randomized a door-to-door drive to facilitate voter registration among internal migrants to two
Indian cities. Ties to origin regions do not predict willingness to become registered locally.
Meanwhile, assistance in navigating the electoral bureaucracy increased migrant registration rates by 24 percentage points and substantially boosted next-election turnout. An additional treatment arm informed politicians about the drive in a subset of localities; rather than ignoring new migrant voters, elites amplified campaign efforts in response. We conclude that onerous registration requirements impede the political incorporation, and thus the wellbeing, of migrant communities in fast-urbanizing settings. The findings also matter for assimilating naturalized yet politically excluded cross-border immigrants.
1 Countries witnessing rapid economic development frequently struggle to assimilate new migrants into cities. As the population of Britain’s towns doubled in size during the industrial revolution, Friedrich Engels (2010) described a burgeoning urban proletariat “cast out and ignored by the class in power” (114) and living in a “state of dilapidation, discomfort, and misery” (viii). During the Great Migration in the United States, African Americans escaping Jim
Crow laws met with “unwritten, mercurial, [and] opaque” resentment in northern cities; they were pushed to the margins, leading Martin Luther King, Jr. to lament that “Chicago has not turned out to be the New Jerusalem” (Wilkerson 2010: 386).
Internal migrants, who number one seventh of the world’s population, face similar challenges across much of today’s Global South (Bell and Charles-Edwards 2013). Political exclusion is commonplace (Bhavnani and Lacina 2015; Thachil 2020; Weiner 1978). Evidence suggests that those who shift from the countryside to cities participate in destination-area politics at lower rates than local-born residents.1 This shortfall matters normatively, cutting against democracy’s promise of equal representation. It is also of practical consequence, as social groups not exercising suffrage experience state neglect (Fujiwara 2015).
What accounts for migrants’ under-representation in politics? We theorize three mechanisms by which mobility induces political marginalization. The first centers on migrants’ enduring economic and social ties to their origin regions. Migrants who maintain close links to
1 In this paper, “migrant” refers to internal migrants who share citizenship and voting rights with local-born residents in destination areas. “Destination area” refers to the jurisdiction to which migrants move, while “origin area” refers to the village, town, or city from which they have relocated.
2 “home” may be unwilling to refocus their political activities, opting to remain detached from political life in destination areas. Second, bureaucratic obstacles associated with participation in host regions—above all, the hassle costs of updating voter registration and navigating electoral bureaucracies—militate against engaging there. Whereas governments normally assume responsibility for initiating the registration process in advanced industrialized states, we find that sixteen of the twenty most populous low- and middle-income democracies place the onus on citizens to initiate enrollment (see Supplementary Information A). Last, ostracism by local-born residents and their elite representatives could impede migrant integration (Dancygier 2010).
“Sons of the soil” parties that vilify newcomers have sprung up in Mumbai, Karachi, and in parts of South Africa and Malaysia in recent decades (Bhavnani and Lacina 2018). It stands to reason that migrants meeting with broad-based indifference or hostility will foresee few benefits to sinking their energies on politics in host communities.
We study the role played by these factors in undermining migrants’ political incorporation, which we conceptualize to comprise both citizen-side political engagement and elites’ readiness to include citizens in their electoral coalitions. To do so, we fielded a large randomized controlled trial in India, a leading case for evaluating the political under-representation of internal migrants. Our focus is on rural-to-urban migrants, and the reasons why such individuals struggle to incorporate politically in countries whose demographics are being transformed by high economic growth. We evaluate a door-to-door campaign to facilitate voter registration among migrants to two cities: Delhi, the national capital, and
Lucknow, which is the capital of India’s largest state and is emblematic of a class of mid-tier cities increasingly attractive to jobseekers (Thachil 2017). Partnering with an NGO in advance of
3 the 2019 national parliamentary elections, we recruited 2,306 migrants who lacked local voter registration documents. Half of those who expressed an interest in registering to vote in the city were then offered intensive assistance in applying for a voter identification card that enabled them to cast a ballot locally in the upcoming polls. In addition, we built a cluster-level experiment on top of the individual-level design. Its purpose was to inform politicians in a randomly chosen subset of neighborhoods that the registration drive had taken place, and thus to test whether attaining registered status paved the way to migrants’ full-fledged political incorporation.
Previewing the results, there is little to suggest that migrants’ ongoing links to their former places of residence prevent them from incorporating politically at their destinations—our first theoretical conjecture. Asked whether they wished to register locally, 98 percent of eligible respondents replied “yes,” indicating that voluntary disengagement is rare in our sample. This is striking since interviewed migrants reported significant social and economic ties to their prior hometowns. By contrast, there is clear evidence to support our second theoretical claim: that bureaucratic obstacles to registering to vote hinder migrants’ electoral participation. Alleviating these constraints—by providing at-home assistance in completing and submitting voter registration documents—increased migrant registration rates by 24 percentage points and next-election turnout by 20 percentage points. It also shifted downstream outcomes, raising political interest and perceptions of local political accountability.
Does elite non-responsiveness further undermine migrants’ local political incorporation, as our final theoretical proposition predicts? The eagerness of migrants to accept registration assistance suggests that anticipation of ostracism is not a major determinant of exclusion on the
4 demand side. Yet, we go further to assess experimentally whether the basis for such perceptions exists. If city politicians are constrained by the anti-migrant preferences of urban electorates, then learning about the mass registration of migrant voters locally should fail to influence their campaign strategies. Against this expectation, we find that electioneering increased in the vicinity of polling stations listed in our communications to candidates. As migrants found a place on local electoral rolls and politicians learned as much, candidates began soliciting migrant support. Overall, we conclude that stringent registration requirements—rather than “opting out” or expectations of ostracism by local political machines—drive the political incorporation gap between migrants and local-born residents in the fast-growing cities we study.
Our study breaks new ground. We study a “patronage democracy” where access to government benefits is intimately bound to individual voting behavior. That migrants do not always assert their political participation in their primary places of residence, despite possessing the full constitutional rights to do so, poses a significant puzzle.
Recent studies suggest that registration drives can in some cases be effective tools for spurring enrollment among unregistered citizens (e.g. Nickerson 2015; Harris et al Forthcoming).
Yet to date there has been negligible theoretical or empirical work on the roadblocks to political access encountered by migrants in the developing world, or, indeed, by movers writ large. Going beyond prior studies, our novel cluster-level experiment evaluates the “supply side” of political incorporation—investigating whether politicians are responsive to news of the enfranchisement of a previously disempowered population group. Observational studies on enfranchisement’s effects have generated mixed findings (cf Paglayan 2021). Our layered research design enables us to experimentally probe equilibrium dynamics as new groups of voters enter the electoral fray,
5 and politicians update their campaign strategies in reaction. To the extent that these strategies entail the targeting of individualized benefits and local public goods, our study can further advance understanding of the sources of urban deprivation (Auerbach 2019).
Finally, as we document, naturalized cross-border immigrants consistently vote at lower rates than native-born citizens in wealthier democracies. Across countries in the Organisation for
Economic Co-operation and Development, turnout rates averaged 80 percent for native-born citizens compared to 74 percent for foreign-born naturalized citizens between 2008 and 2016
(OECD 2019: 128). An analogous 10 to 12 percentage point disparity existed in the United
States throughout the 2000s (Wang 2013). Analyzing data from the 2014 World Values Survey, which covered 52 countries, we find that 82 percent of native-born citizens report that they
“always voted” or “usually voted,” compared to 71 percent of foreign-born citizens (see Online
Appendix A). Civil society organizations advocating on behalf of Hispanic communities in the
United States and Muslim communities in Europe, for example, have underscored the importance of making these groups’ voices heard in the political arena.2 Naturalized immigrant groups may therefore also gain from interventions like the one evaluated here (Braconnier et al
2017; Pons and Liegey 2019).
Conceptualizing Migrant Political Incorporation
Before presenting our theory, we introduce the concept of political incorporation and highlight its relevance for migrants worldwide.
2 See, for example, the work of Voto Latino (bit.ly/3bBZjsC).
6 Critical to our conceptualization is the dual contribution of demand- and supply-side inputs in bringing about migrant incorporation.3 On the demand side, incorporation requires migrants’ full and active participation in destination-area politics. Behaviorally, it presumes that migrants first register to vote in city-based elections—a prerequisite for subsequent participation in formal political institutions—and then turn out to cast a ballot, the fundamental democratic act. Attitudinally, it entails shifts pertaining to political interest, perceptions of political accountability, efficacy, and trust.4 Yet citizen action alone is insufficient to bring about full political incorporation, on our perspective. For that to happen, political elites must reciprocate by treating migrants equally vis-à-vis local-born residents, and by acknowledging them to be bona
3 We parallel the “multidimensional view of integration” advanced by Sobolewska et al (2017), who note the panoply of cultural, social, and economic factors influencing the environment in which migrants engage politically.
4 The attitudinal components of incorporation may be spelled out further. First, healthy democracy rests on a watchful and informed electorate; thus we conceptualize politically incorporated migrants to be those who take an interest in politics at the local and national levels.
Second, enfranchised individuals may come to perceive political elites as more subject to citizen control, and, consequently, as disincentivized to engage in corruption or mismanagement. For this reason, migrants’ assessment of local political accountability matters for incorporation.
Third, incorporation implies enhanced political efficacy: the sense that “people like me” have influence over the government. And fourth, a high degree of trust in political institutions, reflecting openness to giving up some personal autonomy to the state, is intrinsic to the notion of political incorporation.
7 fide members of local electorates. We minimally expect incorporation to include outreach to migrant citizens during campaign seasons, inclusion of migrants in local networks of clientelistic exchange, and sensitivity to problems faced by migrant communities. Paying attention to these supply-side actions is essential, not least because politicians’ failure to include migrants in local political coalitions may demotivate migrants from taking steps to participate in the first place.
Our conceptualization maps onto the empirical record. In the realm of domestic migration, there is significant evidence that mobility is associated with lower political incorporation globally. Changing place of residence in Costa Rica “disrupts” voting and is associated with an eight to nine percentage point reduction in turnout propensity
(Alfaro-Redondo 2016: 73). Focusing on Turkish municipalities, Akarca and Tansel (2015) estimate a strong negative province-level relationship between in-migration and electoral participation. Gay (2012) shows for the United States that utilization of a randomly assigned housing-relocation voucher reduced the probability of voting in national elections by seven percentage points. Qualitatively, scholars have substantiated internal migrants’ relative political disengagement in Nigeria, Colombia, Ukraine, and Myanmar (see Supplementary Information
B). Ample evidence also attests to politicians’ exclusionary behaviors—that is, the core supply-side impediment to incorporation. Côté and Mitchell (2016: 662-6) marshal case study evidence from Côte d’Ivoire and Indonesia, where, according to observers, politicians have weaponized sons-of-the-soil narratives to “turn the politics of resentment to their electoral advantage.” Weiner (1978: 9) describes elites’ anti-migrant stances in the cities of Eastern
Europe following the collapse of the Hapsburg Empire.
8 In short, the historical and comparative case literature bears out the two-sided barriers to political incorporation for migrants. Our next task is to explain what drives variation in this phenomenon.
Theorizing the Migrant-Local Participation Gap
We now theorize the key citizen- and politician-side constraints to migrants’ political incorporation. Regarding citizens’ constraints, we develop two main hypotheses—one centered on voluntaristic detachment, the other on bureaucratic hurdles—explaining non-incorporation.
These hypotheses hew to a cost-benefit analysis of political engagement, which posits that citizens participate politically when the expected benefits of doing so exceed the expected costs
(Downs 1957).5 Overlooked in standard models, however, are costs and benefits that fall uniquely on those who move. Regarding politicians’ constraints, our third hypothesis homes in on the electoral pressures felt by candidates to eschew migrants. Such neglectful treatment by elites has the potential to feed back into migrants’ decisions to engage politically in the city.
5 Beyond this rationalist paradigm, others have pinpointed the influences of habit formation
(Meredith 2009), life-cycle timing (Ferwerda et al 2020), and economic self-investments (Peters et al 2019) on individual political engagement. Voicu and Comşa (2014) view immigrants’ voting propensity as a product of socialization in both destination and origin areas; by seeing political participation as a socially embedded act, their culturalist approach elucidates group-level differences in migrant engagement whereas our focus is on individual-level constraints.
9 Voluntary Detachment
Migrants may voluntarily decline to engage in destination-region politics because they see greater advantages to remaining politically involved in their home regions. Unlike local-born residents, migrants possess a choice about where to exercise their political participation: they can do so either in their region of origin—their default option—or in the place they settle. Thus, one engagement repertoire for migrants entails delinking their place of residence (for clarity of exposition, “the city”) from their place of political participation (“the village”).
A delinked engagement strategy holds out several attractions for migrants. First, social and emotional attachments weigh against withdrawing from politics in origin areas. Political interest develops during the early, formative years of individuals’ lives, and does so in specific locales—places where social pressures to participate may be felt more intensely. At the same time, migrants often find themselves socially isolated at their destinations. Among rural-to-urban migrants in China, for example, “non-kin social ties between migrants and local urban residents are limited [and] non-resident ties still make up the majority of migrant networks” (Yue et al
2013: 1720). This is true of cross-border immigrants too.6 Such dislocation effects are likely to be increasing in the cultural distance between migrants and non-migrants. Thus, social relationships—that are strong in origin areas and weaker in destinations—may prolong origin-area political participation, even after migrants move away.
6 Because “migrants cluster in ethnic communities” and have “limited contact with the host society,” their “connection to the political life in their host country is at best limited” (Careja and
Emmenegger 2012: 880-1), whereas their attachment to country-of-origin politics remains strong
(Alarian and Goodman 2017: 140). See also Fouka (2019).
10 Second, economic motivations could dissuade migrants from shifting their locus of political participation. Rationally, migrants with significant material assets (e.g. property) in their former homes will seek to maintain a political voice there post-migration—to ward off expropriation and other economic threats. Intangible assets provoke similar calculations.
Notably, migrants may be reluctant to sever long-nurtured clientelistic relationships with origin-area politicians. These relationships profit election-minded politicians, too, who have accordingly been known to bus migrant workers back to villages at election time and to offer gifts and additional benefits to woo migrants back during polls.7 Analogously in the international domain, Wellman (2021) finds that ruling parties encourage diaspora voting when they perceive electoral advantages to doing so.
Summing up, there are compelling reasons for migrants to wish to anchor their political participation in origin regions, rather than to transfer it to their point of destination, offering an explanation for the migrant/local incorporation gap. This leads us to hypothesize that migrants who are more socially and economically attached to their places of origin will be more likely to remain politically detached in their new places of residence.
Bureaucratic Hassle Costs
A second explanation for migrants’ political disengagement emphasizes the high administrative barriers migrants face in registering to vote in destination regions. Registration involves gathering and copying paperwork, and completing and submitting forms. Citizens routinely
7 “MP Polls: Migration of Voters Big Worry for Parties in Bundelkhand Region.” November 24,
2018. Economic Times.
11 procrastinate on time-consuming bureaucratic tasks. Randomized trials of voter registration campaigns in France, Kenya, and the United States demonstrate that enrollment is sometimes—though not always—sensitive to convenience costs for the average citizen
(Nickerson 2015; Braconnier et al 2017; Harris et al Forthcoming).8 Missing from consideration in this literature have been the particular challenges migrants confront as they grapple with electoral bureaucracies.
Why might migrants be asymmetrically hard hit by bureaucratic registration hurdles?
Migrants’ difficulties may be the unintended consequence of systems designed with non-movers in mind. First, newcomers lack familiarity with local government procedures, rules, and regulations. The everyday knowledge required to navigate local bureaucracies (e.g. knowing the location of the nearest ward office) is likely to be second nature to local-born residents but opaque to outsiders. Second, in multilingual contexts, migrants from peripheral regions often speak a different language or dialect from those living in urban centers. Migrants who struggle with foreign-language forms will be handicapped in their bid to register to vote. Third, registrants must typically provide supporting documents along with their application. Here, too, migrants lag—particularly those living in informal settlements without title deeds or formal utilities. Finally, in many contexts, migrants are shouldered with a “double-registration burden”: they are required to deregister to vote in their prior place of residence before reregistering in their destination region. Locals do not have to jump through this additional hoop.
8 To contextualize our study, Supplementary Information C provides a systematic review of field-experimental studies of the impacts of voter-registration assistance published to date.
12 Alternatively, bureaucratic elites may deliberately erect barriers to thwart migrants seeking to register. Bureaucratic bias against culturally distinct outsiders—or against marginalized groups over-represented in the migrant pool—can amplify enrollment costs (White et al 2014). Bureaucrats may recognize migrants and treat them in a demeaning or dismissive way during in-person interactions at government offices. When processing documents submitted remotely, bureaucrats can use proxies such as ethnic naming conventions or addresses (along with information about neighborhood demography) to detect migrant status. Bureaucrats can then drag their feet, refuse advice, call for supplemental evidentiary documents, or deny migrant petitions on spurious grounds.
Engaging with the state is demanding for almost any class of citizens. The foregoing discussion highlights the theoretical reasons why migrants may be levied with a registration surcharge, both due to the side-effects of procedures created without thought to “voters on the move,” as well as willful attempts to suppress migrant incorporation by bureaucrats. A testable implication is that programming crafted to ease the voter registration costs borne by movers will have pronounced, positive effects on their subsequent political incorporation.
Political Ostracism
Our final theoretical perspective attributes migrants’ political exclusion to ostracism and anti-migrant backlash by urban elites. Across domains, local-born residents foresee that the arrival of newcomers will heighten labor market competition, strain public services, and “dilute” the ethno-cultural fabric of urban society (Scheve and Slaughter 2001; Gaikwad and Nellis
2017). The upshot is that locals, along with their elected representatives, are prone to exhibit
13 anti-migrant attitudes. This can manifest as passive indifference: migrant entry is permitted but nothing more is done to smooth migrants’ integration or to deliver them services to which they are entitled. It can also materialize as active antagonism—for instance, through the passage of voter suppression laws.9 These strategies are especially likely to arise under conditions of local economic scarcity (Dancygier 2010). In short, politicians can respond with apathy to enfranchised migrants, or in certain circumstances, by stifling migrant turnout.
Migrants may disengage politically in these environments. Those who expect to be sidelined will foresee little gain from taking part in city politics. Migrants fearful of police harassment, additional taxes, and pogroms may prefer to live in the “shadow of the state.”
Ethnographic accounts attest to these concerns. For example, Jones (2020: 93) finds that migrants in Guariba, Brazil “do not belong and are not entitled to make demands” on the government because “civic ostracism places them on the margins of local politics [and] they have internalized the exclusion that they experience at the hands of permanent residents.” How general these experiences are remains to be established.
There is also a subtler logic by which politicians’ uncertain beliefs about migrants’ preferences can produce exclusion. In places where anti-migrant sentiment is muted, it seems intuitive that parties would help migrants register and participate. Where parties can be confident of migrants’ probable vote choice, such party-led drives to enlist migrant voters make rational
9 A historical example of migrant-targeted voter suppression is provided by the French Second
Republic, which in May 1850 instituted a local residency requirement that disenfranchised city-based workers from the countryside and “drove republican politics underground” (Berenson
1984: 169).
14 sense. Yet, more commonly, migrants are unknown quantities in local politicians’ eyes.
Ethno-cultural dissimilarities between politicians and migrants make their partisan leanings hard to discern; the tendency for low-income migrants to reside in dense, heterogeneous informal settlements renders their political preferences less legible to urban elites; and politicians may presume ex ante that migrants’ home attachments will depress demand for city-based registration. Meanwhile, the costs of shepherding migrants through the process are high.
Running migrant-focused registration drives thus carries two risks for politicians and parties: (a) newly registered migrants may not turn out to vote, meaning that scarce resources have been squandered; and (b) migrants may accept registration help but then go on to vote for a competitor. To the extent that the political preferences of local-born residents are more transparent to elites, therefore, it makes sense to focus recruitment and mobilization efforts on locals.
If it is the case that ostracism by politicians is what drives migrant non-incorporation, a testable implication follows: even upon learning that migrants are registered to vote locally, politicians will decline to bring migrants into their local electoral coalitions. This is the third hypothesis we set up our empirical approach to evaluate.
Migration and Voting in India
We now describe a study setting—India—conducive to a rigorous test of our three theoretical claims. India has 1.4 billion citizens, 900 million eligible voters, and an estimated 325 million internal migrants, comprising 29 percent of the country’s population (Government of India
2010). At 34 percent, India’s current level of urbanization is low by international standards.
15 However, the country’s urban population is projected to grow to 590 million people by 2030, up from 290 million in 2001, and the preponderance of all new jobs generated over the next decade will be city-based (Sankhe et al 2010).
India’s rural-to-urban migrants constitute a disadvantaged population category. Online
Appendix B analyzes nationally representative survey data, revealing that migrant-engaged households are poorer than non-migrant households overall, and are more likely to belong to marginalized ethnic communities—differences that underline the integration obstacles that internal migrants face.10 A United Nations report notes that “a holistic approach is yet to be put in place that can address the challenges associated with internal migration in India” (UNESCO
2012: 2). New migrants encounter discrimination in accessing government services (Gaikwad and Nellis 2021), migrant slums lack basic facilities (Auerbach 2019), and demands for “locals only” employment quotas and discriminatory language stipulations have been prevalent (Weiner
1978).
The migrant/local participation gap is stark in India. Subnational states with more migrants evince lower voter turnout (Tata Institute of Social Sciences 2015: 30). Survey data from five states suggest that between 60 and 83 percent of domestic migrants did not cast a ballot in at least one national, state, or local election after moving (Tata Institute of Social Sciences
10 According to official data, 60 percent of migrants moving from neighboring states and 84 percent of migrants moving from non-neighboring states do not share a common language with local-born residents in destination areas; among urban-destination migrants, 61 percent of women cite marriage while 56 percent of men cite employment as the primary reason for migration (Kone et al 2018; Government of India 2010).
16 2015: 32). Representative surveys collected after the 2014 national election revealed turnout to be 69 percent in rural constituencies, 63 percent in smaller cities and towns, and 57 percent in large metropolitan areas (Kumar and Banerjee 2017: 83).11 Micro-level studies in Delhi identified low turnout among urban-based migrants as a prime contributor to this rural-urban turnout divide: in 2014, only 65 percent of recent migrants to Delhi possessed a voter ID card allowing them to vote in city elections, while the overall average for Delhi residents was 85 percent.12 Similarly, Thachil (2017) finds in a sample of Delhi construction workers that only one in five migrants had ever voted in the city’s elections. A survey of India’s five largest cities conducted after the 2019 general elections revealed that 91 percent of urban-based migrants in their twenties reported that they were not registered to vote in the city.13 The de facto
11 In Online Appendix B.4, we present evidence that the urban-rural participation gap persists after controlling for socio-economic status. That rural turnout exceeds urban turnout in India deepens the puzzle of low rates of migrant participation in destination cities insofar as socialization in high-turnout origins areas is not evident in migrants’ city-based political behaviors.
12 See Kumar and Banerjee (2017). In a 2018 survey of 6,884 slum dwellers in Mumbai, 61 percent of respondents who were born out-of-state reported having a local voter ID card, while the corresponding figure for in-state respondents was 71 percent. 25 percent of migrants reported having an origin-place voter ID card (Gaikwad et al 2021).
13 “91% of Urban Migrants Not Registered as Voters in Cities They Live.” March 6, 2019. Hindu
BusinessLine.
17 disenfranchisement of internal migrants has been dubbed a “serious infirmity in the electoral process of the world’s largest democracy.”14
Case Selection
Our study was fielded in two cities. Delhi is home to 19 million residents and is India’s national capital. It absorbs more migrants than any other metropolitan area. Like other megacities, its urban landscape is dotted with jhuggi jhopri (“slum hut”) dwellings, constructed from plastic, corrugated iron, and wood. Estimates suggest that 40 percent of Delhi’s population comprises migrants from other Indian states.15 Lucknow, meanwhile, has 2.8 million residents. It is the capital of Uttar Pradesh, India’s largest state. Most migrants to Lucknow are attracted by higher wages and originate from economically backward districts of the state (Bose and Rai 2014:
53-5).
Delhi and Lucknow were chosen for three reasons. First, we heeded the call from scholars of urban politics to study not only megacities but also small and medium-sized urbanities, which, as Auerbach et al (2018: 262) emphasize, are the world’s “most quickly expanding urban centers.” The comparison of Delhi, a Tier I Indian city, and Lucknow, a Tier II city, adds crucial generalizability in this regard. Second, and at the same time, we sought a controlled comparison, opting for two locations that were broadly similar in terms of the locally dominant language used (Hindi), historical and colonial legacies, national electoral preferences,
14 “The Migrants Indian Democracy Forgot.” February 7, 2019. Diplomat.
15 “Delhi Has Highest Share of Inter-State Migrants.” July 28, 2019. Hindustan Times.
18 socio-religious structures, and the primary sending regions of their migrant populations.16 Last, we built on prior migration work exploiting the paired comparison of Delhi and Lucknow (see
Thachil 2017).
Local Context
To contextualize our study, we document the formal voting registration process in India, and provide preliminary insights—from prior research and qualitative fieldwork—illuminating the difficulties migrants encounter in this regard.
Voter Registration Process
India uses simple plurality rules to elect 543 Members of Parliament (MPs) to single-member districts. Each citizen in our sample is further represented by an elected Member of the
Legislative Assembly (MLA, a state-level position), as well as an elected municipal corporator.
Our investigation was timed to coincide with the 2019 Indian national (MP) elections.
Citizens are required to initiate the registration process, which can be done online or on paper.
The process, which is mandated by the federal Election Commission of India and is the same nationally, has four components:
● Form-filling. Registrants must complete the “Application for Inclusion of Name in
Electoral Roll” (Form 6, reproduced in Online Appendix C). Migrants must also annul
16 Note that although the main sending regions for these cities are also predominantly
Hindi-speaking, highly localized differences in dialects make newcomers’ migration status easily linguistically visible to local-born residents.
19 their registration in their prior place of residence and submit the deletion slip (Form 7), or
they must declare that they were not previously registered elsewhere.
● Documentation. Citizens must provide proof of local residence (for example, a locally
addressed electric/gas bill, ration card, driver’s license, or bank passbook), passport size
photographs, and proof of age (e.g., birth certificate). If the applicant is a tenant, her
landlord is usually expected to sign an affidavit confirming current occupancy.
● Submission. Applications are submitted to the registration office of the Assembly
Constituency in which the citizen resides. Alternatively, applicants can hand over the
documents to a Booth Level Officer (BLO) during brief voter registration drives
conducted annually by local election offices.
● Verification. Local voter registration offices process the forms. If in order, a BLO will
then pay an in-person visit to the applicant’s given address to verify that the submitted
photograph matches the applicant. A voter identification card is mailed by post to the
applicant following approval, and their name is entered on the local voter roll. Rejections
occur either when documents are judged incomplete or improperly filled, or when the
applicant is not found during the BLO’s visit.
Barriers to Registration
Inertia, corruption, and classism mar India’s voter registration system. Peisakhin (2012) found the median processing time for new voter registrations to be 331 days for slum residents in
Delhi. To maximize rents, “officials at election registration offices did almost everything in their power to indirectly encourage applicants … to turn to middlemen for assistance” (Peisakhin
20 2012: 139). Non-bribe payers were additionally harassed, being asked to supply documents not required by law.
Our own qualitative interviews back these claims, while also highlighting the peculiar problems migrants face in registering to vote.17 BLOs betrayed animus toward migrants in interviews, referring to them as “troublemakers.” According to one officer:
Petty crimes in the area have shot up since the recent influx of migrants. They use voter IDs to get loans and then abscond. I would perform numerous background checks on a prospective tenant who is a migrant since all his ID proofs will carry my address, and it is me who stands to get bogged down by all the police paperwork in the event of an untoward incident. At the voter office, we are advised to exercise caution in our dealings with migrants.18
Local elites were skeptical about migrants’ motives for registering. “It wouldn’t be a stretch to say that in the case of migrants, the primary motive for obtaining voter IDs is not the right to vote itself;” rather, they care only about the “potential benefits,” including a “claim to a government plot” if the slum is demolished, as well as “healthcare and education benefits.”19
Pradhans (local community leaders) flagged landlords’ apprehensions about migrants obtaining local voter ID cards:
Landlords frequently refuse to sign the mandatory undertaking required by a tenant while filling registration Form 6. Further, landlords have, in the past, dragged the election
17 Confidential, semi-structured qualitative interviews with migrants and local elites were conducted outside the sample employed for the randomized trial described below. Individuals were approached based on snowball sampling methods. The purpose of this field research was to provide a textured understanding of the local context, and to cast light on causal mechanisms.
18 Interview: BLO, Karol Bagh Assembly Constituency, Delhi, August 22, 2019.
19 Interview: BLO, New Delhi Assembly Constituency, Delhi, August 13, 2019.
21 office to court for registering migrants without their approval. Subsequently, BLOs have been as wary as the landlords themselves in dealing with tenants in bastis [slum colonies].20
Migrants harshly criticized the election management system. “The voter office is jolted out of its inactivity only days before the elections. I suspect bastis are not even a priority for them. This year BLOs arrived at Ambedkar Camp … clueless and ill-prepared.”21 According to another: “A few years ago, I had approached the Election Officer at the Dwarka voter registration office only to be told that the concerned BLO had resigned, and there was nobody assigned to our basti at the moment.”22 These accounts testify to migrants’ predicaments in registering to vote in destination cities.
Indian parties have, at times, engaged in voter registration drives to bring on board new supporters. Yet parties’ record of migrant outreach is patchier. Registration costs are high for parties since Indian electoral law requires that registration applications be filed before candidate nominations are declared, at a time when party organizations are typically shuttered. Knowing whether and in what way migrants will vote is challenging: India is multi-ethnic, multi-lingual, has sprawling migrant-dominant slum settlements, and voters and candidates routinely switch parties. Online Appendix D analyzes representative data from the 2015 Delhi state assembly elections showing that recent internal migrants were 13.6 percentage points less likely to have been targeted for voter outreach by the city’s political parties compared to Delhi’s long-term
20 Interview: Pradhan, Punjabi Bagh, Delhi, September 7, 2019.
21 Interview: Resident, Karol Bagh, Delhi, September 12, 2019.
22 Interview: Pradhan, Sagarpur, Delhi, September 22, 2019.
22 residents. Overall, India’s urban migrants evince low rates of registration under the status quo.
Parties’ engagement with migrants appears to fall short of their engagement with local-born residents.
Research Design
We implemented a large, multi-level field experiment to shed light on the reasons why migrants often do not incorporate politically into destination cities, and the degree to which such disengagement is remediable.
Sampling, Recruitment, and Baseline Survey
Sampling and the administration of the baseline survey proceeded in four stages.
1. Site selection. We first generated a list of 150 migrant-dominated settlements, which is to
our knowledge the most comprehensive list of such settlements for Delhi and Lucknow.
To make this initial frame of potential sites, we relied on census data, schedules of
informal settlements produced by city governments, and information gleaned from
respondents at migrant-dominated labor chawks (markets). Scoping teams of 30
researchers assessed the suitability of potential sites over a period of nine months. During
visits to settlements, they surveyed residents regarding the possession of local voter ID
cards. The study includes those settlements where informants reported the greatest
numbers of unregistered internal migrants residing.
2. Individual-level screening. Within selected neighborhoods, enumerators employed
interval sampling to choose potential households to interview. Informed consent was
23 requested from the adult household respondent with the next upcoming birthday. Only
one subject was sampled per household, therefore. Subjects stating that they were neither
born in the city nor had a voter ID card enabling them to vote there were deemed eligible.
Thus our full sample is intended to be representative of unregistered migrants in sampled
colonies.
3. Short baseline on omnibus sample. Eligible subjects were asked about basic
demographics, their past voting behavior, as well as the extent of their social, economic,
and political connections to their home villages or towns. At the conclusion of this
module, we asked whether they wished to obtain a local voter ID card. The survey ended
if a respondent replied “no.” These initial questions, as well as the overall take-up rate,
are used to test the plausibility of the voluntary detachment hypothesis.
4. Long baseline on experimental sample. Respondents replying “yes” entered the
experimental sample. We posed a larger set of demographic and attitudinal questions to
this group. Descriptive statistics are provided in Appendices H and I. These subjects were
then randomized into different treatment conditions.23
Representativeness
We benchmark our experimental sample against nationally representative survey data collected in the second round of the Indian Human Development Survey (IHDS-II). Comparing household
23 Note, therefore, that the “experimental” sample is a subset of the “omnibus” sample, comprising those individuals who wished to register to vote in their destination cities. The conclusions from our randomized study thus hold for this population of interest.
24 characteristics in IHDS-II by migrant/non-migrant status in both rural and urban areas, we highlight three key respects in which our sample generalizes. First, migrants overall evidence lower political participation rates than non-migrants. Second, marginalized caste groups and the poor are over-represented both in our sample and in the wider internal migrant population. A third point of note concerns connectivity to patronage networks. Conceivably, our sampling criteria, which hinge on migrants’ stated willingness to register, select for migrants who are comparatively excluded from clientelistic channels in home regions. In fact, the IHDS-II benchmark demonstrates that rural migrant-sending households are less enmeshed in client-patron relationships than others, potentially spurring migrants’ decision to “move to opportunity” to begin with.24
Treatment 1: Individual-Level Registration Drive (Migrant-Targeted)
Our primary intervention (“T1”) operated at the level of individual migrants. Its purpose is to test our second hypothesis regarding bureaucratic hassle costs. Simple randomization was used to assign individuals who completed the long baseline survey to T1 with 50 percent probability.
Remaining subjects were assigned to a pure control condition.
The intervention was an intensive door-to-door facilitation campaign to help migrants obtain a local voter identification card (see Online Appendix C for further details). This card would enable them to participate in the forthcoming national elections in the city where they were living (either Delhi or Lucknow). To begin, a worker trained with the help of our NGO
24 According to the IHDS-II sample, approximately nine percent of rural households engaged in seasonal migration over the past five years.
25 partner visited migrants at their place of residence and presented their credentials. The worker described the benefits of holding a local voter ID card, the process for getting one, and the type of assistance the worker could provide. The worker asked whether the migrant had the necessary supporting documents at hand.25 If they did, the migrant was asked to gather those documents in time for a future visit, and a time was set. Migrants who lacked these documents were instructed on how to get them.
At the follow-up visit, the worker helped the migrant to complete the required forms online using an internet-enabled computer tablet. At the end of the meeting, workers ensured that the form, along with uploaded photographs of the required documents, were submitted to the local registration office. The worker then tracked the progress of the applications. Where problems arose—often, for example, BLOs were unable to track down the applicant at their listed residence—the worker intervened to try to fix the issue.
25 Our analysis of the nationally representative IHDS-II data reveals that migrant households were substantially less likely than non-migrant households to possess officially acceptable proof-of-residence and photo ID documentation (see Online Appendix B.5). This finding was buttressed in our ethnographic interviews. According to one interview respondent, “At the time of applying, migrants need to attach an ID proof to the form. Some do not even possess the bare minimum proof, and have to bribe the BLOs.” Interview: Pradhan, Indira Colony, Delhi,
September 2, 2019.
26 Treatment 2: Cluster-Level Information Dissemination Campaign (Politician-Targeted)
Our second treatment arm—henceforth, “T2”—operated at a cluster level and was intended to help adjudicate the plausibility of our third theoretical claim concerning politician ostracism.
Using GIS software and publicly available information on the locations of polling stations in
Delhi and Lucknow, we identified the 87 polling stations in closest proximity to our geo-located sample of experimental subjects. Each subject was tagged to the nearest one of these polling stations. We then divided the 87 polling stations into four blocks, defined by city and whether the number of experimental subjects assigned to that polling station was above or below the city’s median number of experimental subjects per polling station. Last, we randomly assigned polling stations within each randomization block to either the T2 intervention or to the T2 control condition. Randomization to T2 was thus fully independent of the T1 trial.
Between two and four weeks before election day—during India’s month-long campaign season and following parties’ selection of candidates—we sent individually tailored letters,
WhatsApp messages, and emails to four types of politicians tied to each polling booth: the incumbent MP; officially declared MP candidates; incumbent MLAs; and incumbent municipal corporators. Politicians’ contact details were gleaned from public databases. The messages were written in both English and Hindi. They informed the politicians that a voter registration drive had been recently carried out among internal migrant communities in their area (see Online
Appendix E for examples of the communications).26 The purpose of the treatment was to
26 The communications specify that migrants have moved from other parts of India and possess full constitutional rights to vote in destination areas.
27 disseminate information about the registration drives that had taken place, and thus to positively update politicians’ beliefs about the average registration rate of migrants in those localities.
Outcomes
Outcomes were measured in an endline survey conducted approximately two months after the elections were held. The study timeline is shown in Online Appendix F, and the survey question wordings are given in Supplementary Information G. Methods used to create indexed outcomes are described in Online Appendix G.
Analysis
For estimations involving T1, we run ordinary least squares regressions of the following form:
(Equation 1) Yi = a + b*T1i + X’iλ + ui where, i indexes subjects, Y is the dependent variable of interest, T1 is a dummy taking 1 if the subject was assigned to the registration facilitation campaign and 0 if assigned to control, and X is a vector of baseline covariates, included to improve statistical precision. All specifications control for gender, age, religion, caste, education, income, married status, length of residence in the city, and homeownership status.27 Where specified, we also included pre-treatment measures of outcomes. T1 analyses use Huber-White robust standard errors.
For the T2 analysis, we employ weighted least squares regression:
s (Equation 2) Yij = a + b*T2ij + X’iλ + δ + uij
27 The income covariate was winsorized to address significant outliers (see Online Appendix H).
28 Here, Y is the outcome of interest, T2 is a binary indicator taking 1 if individual i in cluster j was assigned to the T2 treatment and 0 if assigned to T2 control, and X is a vector of controls—the same set used for the T1 estimation. δs are block fixed effects. We use cluster-robust standard errors, clustering at the polling booth level, which is the unit of randomization. As pre-specified, we re-weight individuals such that each cluster contributes equally to the estimation.
The intent-to-treat (ITT) effect is the quantity of interest across all experimental analyses.
In tests of balance and attrition we statistically find no threats to internal validity (see Online
Appendix I). We successfully re-interviewed 91 percent of baseline subjects at endline. Our study was pre-registered at the Evidence in Governance and Politics registry (#20191007AA).28
Ethics
Our experimental design was informed by ethical best practices. To begin, we selected a sample size that was no larger than we deemed necessary to detect meaningfully sized treatment effects.
Further, the sites in which we worked were not electorally competitive—the national ruling party secured strong victories across all study constituencies in the preceding election and in the election we investigated—implying that there was a negligible risk of affecting aggregate electoral outcomes.
Next, our non-partisan intervention partner had already conducted extensive voter registration campaigns among urban migrants. It is a not-for-profit organization working in the fields of slum rehabilitation, housing rights for migrant workers, as well as slum sanitation and public health. Public advocacy for migrants has been at the core of its activities, meaning that
28 See https://osf.io/7vtqh.
29 such drives—conducted both by our partner and other civil society organizations—would have occurred in our absence.29 This was a multi-stakeholder project. As academic partners, our role was to scientifically evaluate the efficacy of the NGO’s efforts in helping empower migrants to take up their constitutional rights to participate politically in destination cities.
Regarding context, we relied on the community-level embeddedness of our partner organization and field team (a) to be confident ex ante that the intervention was contextually and culturally appropriate; (b) to ensure that the registration intervention, as well as common knowledge of it, would not invite pushback—based on our partner’s relationships with local elites and past experience running and publicizing such campaigns; and (c) to have in place a set of protocols capable of providing immediate feedback on potential unanticipated events on the ground. In addition, we sought the permission of the Election Commission of India as well as community leaders in each slum colony prior to beginning work there.30 We further vetted the design with Delhi- and Lucknow-based NGOs and academics not involved in the study.
The costs to study participants were quite minimal: the voluntary completion of two surveys and the time required to initiate registration if the subject so wished. There were several likely benefits. Directly, subjects receiving the registration intervention were able to exercise suffrage locally. Our prior research provided a strong basis for believing that political incorporation would yield material benefits for newly enfranchised urban migrants, in the form
29 T1 control group participants remained free to submit applications for voter ID cards at any time, and many did, attesting to the routineness of this process.
30 The Election Commission has itself promoted migrant voter registration in their place of
“ordinary residence” (Tata Institute of Social Sciences 2015).
30 of increased supply of essential constituency services by local politicians (Gaikwad and Nellis
2021).31 Apart from gains to study participants themselves, there was (and remains) a pressing need to learn what works to raise turnout among hitherto politically marginalized groups. The unique incorporation challenges facing internal migrants in fast-urbanizing democracies have largely eluded attention, even as this population balloons in size. Our study helps establish an evidence base for policy initiatives that can be deployed by governments and NGOs working in this space.32
Results
What causes migrants’ political exclusion? Turning to our data, we first characterize the sample.
Subjects’ average age was 29; 54 percent were female, and 65 percent had attended primary school. 77 percent of respondents had relocated to the city from villages, eight percent from towns, and the remainder from other small or large cities. We asked the profession of the main working member of each interviewed household. Most household heads were said to be engaged
31 In the Indian context, Bussell (2019) and Kruks-Wisener (2018) similarly underscore the significance of constituency service and “claims-making” for citizen welfare.
32 The study design was approved by the Institutional Review Boards of Columbia University
(#IRB-AAAR7603), the University of California, San Diego (#181336S), and MORSEL
Research and Development (#IRB/1807/MRD). For additional discussion of study ethics see
Online Appendix L.
31 in unskilled/semi-skilled (33 percent) or skilled (26 percent) production, followed by commerce
(16 percent). Subjects had spent 11 of the past 12 months residing in host cities, on average.33
Voluntary Detachment
Recall, our first hypothesis zeroed in on socio-economic ties to home regions as an explanation for political non-incorporation. Baseline data, plotted in Figure 1, underscore the prima facie plausibility of these claims. Subjects retained strong linkages to their hometowns. Socially, a large majority (72 percent) weakly or strongly agreed that they felt “more at home” in their previous place of residence—suggesting the possibility that they continue to feel alienated in their destination cities. This was true even though 75 percent of migrants lived more than
100kms away from their home district, and 35 percent lived more than 500kms away, making regular visits impractical. A significant minority had material stakes in origin areas: 28 percent personally owned land or property and 47 percent reported benefiting from government schemes there. Simply put, migrants in our sample proved highly attached to their origin regions.
[INSERT FIGURE 1 ABOUT HERE]
Under the theory of voluntaristic detachment, such socio-economic bonds to “home” should predict ongoing political participation there. To investigate this, we regress indicators for
(a) whether the migrant possessed an origin-area voter ID card, and (b) whether they had previously left the city to vote in their hometown, on the socio-economic characteristics just described as well as one extra migrant attribute. The results are displayed in Table 1. Greater
33 See Supplementary Information F for summary statistics.
32 subjective attachment to origin regions, current receipt of government schemes, and owning hometown property all emerge as robust, positive correlates of origin-area political engagement post-migration. Longer-term residence in the city and greater geographic distance to home are negatively associated with those outcomes.
[INSERT TABLE 1 ABOUT HERE]
On first inspection, the patterns highlighted in Table 1 suggest that persistent socio-economic attachments to home may induce migrants to hold onto origin-area engagement, as theories of social-group pressure, economic interests, and clientelism imply. However, we now proceed to examine our formal test of the extent to which these factors depress demand for political incorporation among migrants. At the end of the short baseline survey that was fielded on the omnibus sample (intended to be representative of city-based unregistered migrants), we asked eligible migrants whether they were interested in taking steps to register to vote locally—the main eligibility criterion for entering the experimental sample. Of the 2,350 subjects in the omnibus sample, an overwhelming majority—98 percent—replied “yes.” In other words, notwithstanding their material and social connections to hometowns, almost all favored concentrating their political activities in destination areas. We infer that such factors are not sufficiently potent in and of themselves to reduce migrants’ willingness to incorporate politically into the city. There is, then, sizable pent-up demand for incorporation.
While perhaps surprising in light of Figure 1, migrants’ stated desire to integrate politically sits with their positive estimations of their economic situation in the city. A majority—65 percent of respondents—rated their current employment opportunities as “good” or
“very good,” something that only 23 percent said of their opportunities in their place of origin at
33 the time they left. Only three percent stated that their incomes had gotten worse since migrating, whereas the rest claimed their incomes had stayed the same (26 percent) or gotten better (71 percent). Consistently, 96 percent of subjects expressed an intent to reside in their destination cities permanently. At the same time, many subjects reported a low sense of political efficacy.
We asked to what extent they agreed that “people like me don’t have any influence on the government in [Delhi/Lucknow].” Forty-seven percent agreed. It appears, then, that city-based political incorporation is seen as desirable—if not essential—by a population that has come to be economically and residentially ingrained in the city, even as individuals continue to maintain links elsewhere.
Bureaucratic Hassle Costs
Voluntary detachment is inadequate to account for the low rates of voting and political participation among internal migrants in the Indian cities we study. We thus turn to our second candidate explanation for migrant exclusion—onerous voter registration procedures—which we test using our T1 experimental manipulation.
Table 2 presents the key T1 experimental results for our primary political outcomes.
Column 1 shows that the intervention had a sizable first-stage impact. Note, only migrants lacking locally valid voter ID cards at baseline were eligible to participate in the study. By endline, 16 percent of control group subjects had gone on to obtain a local voter registration document. According to our main specification (Table 2, column 1), providing simple assistance with this procedure boosted registration rates by 24 percentage points, over and above the control group mean. This equates to a 147 percent proportional increase. The estimated effect is
34 substantively large and statistically significant (p<0.001), supplying strong evidence that the real or perceived costs of local voter registration prevent large swathes of India’s migrants from appearing on city voter rolls.34 Providing simple at-home assistance can go a long way toward remedying migrants’ deficient representation in urban electorates.
[INSERT TABLE 2 ABOUT HERE]
The right-hand columns of Table 2 quantify effects for turnout: whether subjects voted locally in India’s 2019 national election, and their assessments of how likely they were to vote in future subnational elections. Assignment to the registration campaign increased the probability of turning out to vote by 20 percentage points (p<0.001), totaling a 114 percent increase relative to the control group average (Table 2, column 2).35 These are large effects. Benchmarking them against theoretically established drivers of participation, we note that Panda (2019) finds each additional year of education is associated with a 2.4 percentage point reduction in the probability of voting in Indian national elections, while below-the-poverty-line status corresponds to a 3.2 percentage point increase in turnout likelihood. The impact of our bureaucratic-assistance intervention outstrips these associations in magnitude. The intervention also caused subjects to
34 The burdensome nature of this process is substantiated by administrative data on the tracked history of case-files of migrants in the T1 treatment condition (see Online Appendix J).
35 The proportion of control group subjects who reported having voted in the city slightly exceeds the proportion who said they were registered to vote locally. A potential explanation is that individuals who voted in a village in the vicinity of the city interpreted this to be city-based voting. Note, in Supplementary Information H.1, we find no evidence that stronger home-attachment characteristics diminish the efficacy of the registration intervention.
35 state that they would be more likely to vote in the city’s next state election (Table 2, column 3).
This hints at the potential for a longer-term impact—persistence that might also be engendered by the habit-forming effects of first-time participation in city elections (cf Meredith 2009).
We next examine how the treatment shaped subjects’ political consciousness. To begin, we combine two measures of political interest—one capturing attention to city politics and the other to national politics—to generate a Z-score index. The results for the overall “Interest” outcome are shown in Table 3, column 1. Assignment to the intervention caused a 0.091 standard deviation increase in political interest (p=0.010).36 Our survey asked respondents the extent to which they agreed that politicians are accountable to citizens. The control group average for this ordinal variable suggests that most citizens “somewhat agree” that politicians are accountable
(Table 3, column 2). We find that assignment to treatment raised perceptions of accountability by four percentage points (p=0.003). Possessing the documentation needed to vote thus enhanced citizens’ beliefs that good types of politicians can be rewarded at the ballot box and bad types punished.
[INSERT TABLE 3 ABOUT HERE]
By contrast, we do not observe any increase in citizens’ sense of political efficacy (Table
3, column 3) or of political trust, measured as an index that combines trust in national, state, and municipal governments as well as political parties (Table 3, column 4). It may be that these null estimated effects reflect the reality that, while voting raises awareness that politicians are
36 Supplementary Information H.2 demonstrates that T1 caused a rise in attention to both city and state/national politics; regarding the latter, this was a national-level election featuring candidates bearing both national state party labels.
36 vulnerable to being ousted (accountability), it does not change certain fundamentals: that an individual citizen has limited sway over the government of a vast polity (efficacy), and that core democratic institutions will not immediately be transformed by an additional migrant registering to vote (trust).37
We have thus far presented the main effects of our T1 experimental intervention. An important question is whether the campaign worked symmetrically across different classes of migrants, or whether it impacted migrant subgroups differentially. We test five moderators, all but one of which focus on salient dimensions of social and economic marginalization in the
Indian setting:
● Muslims, who comprise 14 percent of the country’s population and 24 percent of our
study sample, have historically been targets of discrimination and violence (Gaikwad and
Nellis 2017).
● Scheduled Castes (SCs) and Scheduled Tribes (STs) are on the lowest rungs of India’s
rigid caste hierarchy. Social bias against these groups is widespread. Combined, SCs and
STs make up 38 percent of the sample, surpassing their share of the national population
(25 percent).
37 We ran two exploratory analyses. We find that T1 significantly increases rates of politician contacting and non-electoral participation (Supplementary Information H.3). A further analysis
(Supplementary Information I) suggests that migrants’ party-wise preferences roughly approximated those of immediately surrounding local populations.
37 ● While India enshrines the right to primary education in its constitution, in practice many
citizens do not receive even basic schooling. 35 percent of respondents in our sample had
not attended primary school.
● We partition subjects at the sample median of monthly household income. At INR 10,000
(USD 131), this number is low, highlighting the poverty of many urban migrants.
● Last, we look for differential effects by whether the respondent is a long-term migrant;
that is, whether their number of years in the city exceeds the sample-median value.38
[INSERT TABLE 4 ABOUT HERE]
Table 4 displays our two sets of pre-registered tests of heterogeneity for the two primary political outcomes: receipt of a local voter ID card, and city-based voting in the 2019 election.
The interaction coefficients are interpretable as the differences in T1’s effect across the groups defined by each dichotomous moderator. The estimates in column 1 paint a consistent picture. In terms of first-stage effects, the intervention proved more impactful for relatively advantaged migrant subpopulations. Controlling for migrants’ other background characteristics and their interactions with the treatment, we find that the impact of the intervention was eight percentage points larger for those with primary education (SE=0.041). The gulf in treatment efficacy is wider still for Muslims and SCs/STs. The estimated impact of the intervention was 11 percentage point lower for Muslim migrants than for non-Muslims (SE=0.048). Likewise, the effect of T1 on SCs/STs is 11 percentage points lower than it was for non-SCs/STs (SE=0.042). That said, the
38 Supplementary Information H.4 finds no evidence of heterogeneity according to subjects’ fluency in Hindi, nor does the main effect of T1 attenuate after including this variable as a control (Supplementary Information H.5).
38 heterogeneous effects for turnout in column 2 are substantially smaller, and only statistically significant for the SC/ST moderator.
A fuller explanation for the variability in treatment effects on registration awaits future research. Yet our qualitative interviews with local brokers and community representatives in several sampled slum colonies indicates that the actions of local electoral gatekeepers may form part of the story.39 A common refrain heard was that minorities, when made to establish their credentials, were held to a double standard:
Muslim migrants from West Bengal must constantly deal with persecution owing to the authorities’ suspicion of them being Bangladeshi. This may have been aggravated due to the government’s NRC [National Register of Citizens] policy.40
A second interviewee stressed that “[t]here is wide variation in documentation requirements for forward castes vis-à-vis minorities [Muslims and SCs/STs]. In case of the former, verbal confirmation often suffices, while the latter are subjected to onerous formalities.”41 The extra burden placed on minorities may be a product of individual-level discrimination by officials, or may represent more systemic discrimination. As one respondent commented, “BLOs and AEROs
39 Some respondents alleged widespread voter suppression, e.g., “[e]xclusion of names from lists just days prior to the election is pretty much the norm. People from my neighborhood, which is majority Muslim, have seen their names disappear this time even though they had voted in the last state election and the 2014 Lok Sabha election. Initially I thought this was limited to Muslim localities, but turns out the story is very similar in Valmiki [SC] neighbourhoods too.” Interview:
Community volunteer, East Seemapuri, Delhi, November 20, 2019.
40 Interview: Community organizer, Okhla, Delhi, November 17, 2019.
41 Interview: Dalit leader and party worker, West Patel Nagar, Delhi, November 12, 2019.
39 [Assistant Electoral Registration Officers] are very risk averse professionally. They are doubly apprehensive and place more stringent documentation requirements on Muslim migrants relative to other demographics, fearing reprimand from senior officials.”42
Ostracism by Political Elites
We have shown that subsidizing the costs of voter registration fosters migrants’ engagement in urban politics. Thus, there is considerable evidence that bureaucratic hurdles forestall migrant registration. We now adjudicate our final theoretical explanation for migrants’ political exclusion: ostracizing behavior on the part of political elites. Put simply, if elected elites on the
“supply side” of politics resist or ignore even registered migrants, then migrants might reasonably find it unprofitable to situate their political affiliations in cities in the first place—or will see little reason to go on investing time and effort into city-based participation going forward.
To assess whether such exclusionary behavior by politicians occurs, we exploit our T2 clustered experimental design, examining five endline variables that measure exposure to the
2019 Lok Sabha campaign, as well as the Z-score index that combines them.43 The raw responses indicate the campaign was quite hard-fought. 79 percent of subjects either strongly or somewhat
42 Interview: Community volunteer, East Seemapuri, Delhi, November 20, 2019.
43 We opted for five endline measures because our priors on precisely which campaign modalities would change in response to the news of increases in migrants’ local registration levels were diffuse; thus, we employ a holistic set of indicators of local campaign activity rather than just one.
40 agreed that politicians and party workers had campaigned hard to win the votes in the respondent’s neighborhood; 40 percent said that at least one incumbent politician or MP candidate had visited their basti; and two percent reported having been offered a gift.
Did informing local politicians about the rollout of a migrant-focused voter registration drive in the vicinity of a polling booth cause them to direct additional campaign resources there, or were such new voters disregarded? The evidence in Table 5 argues strongly for the first possibility. Looking at the pre-registered analysis on the campaign exposure index (column 1), we see that respondents living near polling booths assigned to the T2 information dissemination condition report campaign exposure 0.10 standard deviations higher than subjects close to polling booths that came under the control condition (p=0.043). Columns 2-6 break down the results for each component measure. This exercise reveals that the positive impact on the overall index comes primarily from the increase in perceived campaign intensity (column 6) and, secondarily, from a rise in the number of gifts offered to treated migrants (column 4). This second result is telling. It suggests that the campaigning windfall felt by subjects was
“pro-migrant” and not due to politicians ramping up pro-local rhetoric in areas with a burgeoning migrant electorate.44 We do not see movement on three of the other index components: basti or doorstep visits by politicians, or on migrant-focused campaign activities. In seeking to expand their electoral reach, it seems, city-based politicians were not tailoring their campaigns in ways that might be off-putting to local-born residents—broadcasting, say, messaging in which migrant
44 Bolstering this interpretation, Supplementary Information J.1 shows no evidence that T2 negatively impacted trust in local institutions and politicians, or migrants’ subjective comfort in destination cities.
41 themes feature prominently. Rather, without noticeably modifying the content of their campaigns, politicians appear to have increased the quantity of campaigning where they knew there to be more migrant voters.
[INSERT TABLE 5 ABOUT HERE]
The fact that urban politicians bolstered campaigning in response to information about migrant-focused voter enrollments is encouraging, pointing to a state of the world in which migrants register to vote and politicians incorporate them more fully into the mainstream political life of the city. It further indicates that localist bias and neglect by urban politicians is unlikely to be a principal explanation for migrants’ low registration rates in city elections.
We earlier described the reasons why parties may be hesitant to involve themselves in the business of registering migrants to vote: the unit costs are high, and migrants’ propensity to vote, as well as their ultimate vote choice, are clouded with uncertainty. The analysis in Table 5 clarifies that once the costs of migrant registration have been borne by other actors, office-seeking politicians and party machines treat migrant voters conventionally: plying them with selective benefits and rallying their support.45
45 Note that the T2 results are not susceptible to research demand effects as subjects were unaware of which T2 treatment arm they had been assigned to. Note also that, because T1 was randomized independently of T2 (recall, T1 was randomized at the individual level), any heightened sensitivity to political campaigning induced by the T1 treatment is statistically balanced in expectation across the T2 treatment and control arms.
42 Conclusion
We offer new theory and evidence to explain the drivers of migrants’ low rates of political incorporation into destination regions. A large, multi-site experiment in India identifies bureaucratic hassle costs, rather than migrants’ voluntary abstention from urban politics or political ostracism, as the salient constraint. Subsidizing these costs, by providing at-home assistance in registering to vote, substantially increases political incorporation: raising registration rates, turnout, political interest, and perceptions of political accountability. Moreover, political elites react to news of the local enfranchisement of migrants by boosting electioneering in the vicinity.
The experimental evidence we adduce comes from two major cities that are jointly home to 22 million people (roughly the population of Scandinavia). The substantive findings, along with the theoretical framework that motivates them, have broad scope for generalizability. For example, a report by the Organization for Security and Co-operation in Europe (Mooney and
Jarrah 2004: iii) notes that internally displaced persons “face obstacles in exercising their right to vote, sharply reducing their influence over the many political, economic and social decisions affecting their lives.” To be sure, the dominant factor explaining political exclusion—whether voluntary detachment, bureaucratic impediments, or ostracism by local elites—may vary contextually.46 For instance, in settings where highly pronounced cultural differences exist
46 Conceivably, the relative weights of each factor could also vary over time within countries. To evaluate temporal external validity, Supplementary Information K tabulates longitudinal data from Delhi showing that the 2019 election we investigate was not notably different from recent prior elections held there regarding migrant/local voting dynamics.
43 between migrants and non-migrants, voluntary detachment may emerge more prominently as a cause of non-incorporation. Still, onerous voter-initiated registration systems, which our study distinguishes, are the rule rather than the exception in the world’s largest low- and middle-income democracies. This suggests that the mechanisms we decode likely operate elsewhere.
Policywise, the findings are relevant for election management bodies in states witnessing explosive urban growth. Rejecting proxy voting for internal migrants as logistically unfeasible, a recent review by the Election Commission of India recommended instead that migrants re-register in their new locations, employing the existing voter-initiated procedure.47 As we document, this may underestimate the multiplex and unusual challenges that migrants confront in navigating such systems. Building capacity to streamline and perhaps automate registration should be a priority for the future.
There are also lessons for migrant advocacy groups. Investing in voter registration drives is worthwhile in territories where citizens are mobile. Our results sound a note of caution, however. The registration facilitation outreach, while effective overall, had a more positive impact among migrants belonging to advantaged ethnic, religious, and educational subgroups.
These lopsided benefits highlight a tradeoff. Easing registration requirements can lift the share of migrants in urban electorates, but worsens representativeness along other axes of social identification. It will be important to engineer interventions that avoid producing such political inequalities.
47 S. Irudaya Rajan and Prashant Singh, “The Disenfranchised Migrants.” August 30, 2019. The
Hindu.
44 The impacts of political incorporation on social and economic integration merit investigation. Inter alia, future research should explore whether registration causes migrants to be more likely to consider cities home, to forge social ties with locals, and to be more willing to pay for city-based public goods. These issues have come to the fore during the Covid-19 pandemic, when the needs of millions of India’s migrants went untended by political elites and local populations (Rajan et al 2020).
We end by underlining how the paper’s findings matter not only for within-country migrants. Low registration rates also characterize immigrant-background communities (Alizade et al Forthcoming). Bass and Casper (2001: 105) write of “naturalization and registration” as
“the first two barriers to voting” for foreign-born citizens in the United States, where interventions like the ones evaluated here have recently become a mainstay of activist efforts to incorporate immigrant communities.48 Our evidence suggests that high-intensity campaigns of this sort can facilitate political incorporation and incentivize parties and politicians to bring immigrant communities into their electoral coalitions. Thereby, political exclusion may be overcome.
48 See, e.g.: “Two civic groups launch campaigns to register newly naturalized citizens to vote.”
June 17, 2020. NBC News.
45 The authors declare the human subjects research in this article was reviewed and approved by the University of California, San Diego, Columbia University, and MORSEL Research and Development, and certificate numbers are provided in the text. The authors affirm that this article adheres to the APSA’s Principles and Guidance on Human Subject Research. The authors declare no ethical issues or conflicts of interest in this research. This research was funded by the University of California, San Diego, and Columbia University. Research documentation and/or data that support the findings of this study are openly available in the APSR Dataverse at https://doi.org/10.7910/DVN/G1JCKK.
46 References
Akarca, Ali, and Tansel, Aysit. 2015. “Impact of Internal Migration on Political Participation in Turkey.” IZA Journal of Migration 4(1): 1-14.
Alarian, Hannah, and Goodman, Sara. 2017. “Dual Citizenship Allowance and Migration Flow: An Origin Story.” Comparative Political Studies 50(1): 133-67.
Alfaro-Redondo, Ronald. 2016. “Divided We Vote—Turnout Decline in Established Democracies: Evidence from Costa Rica.” PhD Dissertation: Department of Political Science, University of Pittsburgh.
Alizade, Jeyhun, Dancygier, Rafaela, and Ditlmann, Ruth. Forthcoming. “National Penalties Reversed: The Local Politics of Citizenship and Politician Responsiveness to Immigrants.” Journal of Politics.
Auerbach, Adam, LeBas, Adrienne, Post, Alison, and Weitz-Shapiro, Rebecca. 2018. “State, Society, and Informality in Cities of the Global South.” Studies in Comparative International Development 53(3): 261-80.
Auerbach, Adam. 2019. Demanding Development: The Politics of Public Goods Provision in India’s Urban Slums. Cambridge: Cambridge University Press.
Bass, Loretta, and Casper, Lynne. 2001. “Impacting the Political Landscape: Who Registers and Votes among Naturalized Americans.” Political Behavior 23(2): 103-30.
Bell, Martin, and Charles-Edwards, Elin. 2013. “Cross-National Comparisons of Internal Migration: An Update on Global Patterns and Trends.” United Nations, Department of Economic and Social Affairs, Population Division: Technical Paper No. 2013/1. bit.ly/3fnE51O.
Berenson, Edward. 1984. Populist Religion and Left-Wing Politics in France, 1830-1852. Princeton: Princeton University Press.
Bhavnani, Rikhil, and Lacina, Bethany. 2015. “The Effects of Rainfall-Induced Migration on Sons of the Soil Violence in India.” World Politics 67(4): 760-94.
Bhavnani, Rikhil, and Lacina, Bethany. 2018. Nativism and Economic Integration Across the Developing World: Collision and Accommodation. Cambridge: Cambridge University Press.
Bose, Probir, and Rai, Ramjee. 2014. “Job Search and Labour Market Conditions of Migrants at the Destination: The Case of Lucknow.” Urban India 34(1): 47-67.
Braconnier, Céline, Dormagen, Jean-Yves, and Pons, Vincent. 2017. “Voter Registration Costs and Disenfranchisement: Experimental Evidence from France.” American Political Science Review 111(3): 584-604.
47 Bussell, Jennifer. 2019. Clients and Constituents: Political Responsiveness in Patronage Democracies. Oxford: Oxford University Press.
Careja, Romana, and Emmenegger, Patrick. 2012. “Making Democratic Citizens: The Effects of Migration Experience on Political Attitudes in Central and Eastern Europe.” Comparative Political Studies 45(7): 875-902.
Côté, Isabelle, and Mitchell, Matthew. 2016. “Elections and ‘Sons of the Soil’ Conflict Dynamics in Africa and Asia.” Democratization 23(4): 657-77.
Dancygier, Rafaela. 2010. Immigration and Conflict in Europe. Cambridge: Cambridge University Press.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row.
Engels, Friedrich. 2010 [1845]. The Condition of the Working-Class in England in 1844. Cambridge: Cambridge University Press.
Ferwerda, Jeremy, Finseraas, Henning, and Bergh, Johannes. 2020. “Voting Rights and Immigrant Incorporation: Evidence from Norway.” British Journal of Political Science 50(2): 713-30.
Fouka, Vasiliki. 2019. “How Do Immigrants Respond to Discrimination? The Case of Germans in the US during World War I.” American Political Science Review 113(2): 405-22.
Fujiwara, T. 2015. “Voting Technology, Political Responsiveness, and Infant Health: Evidence from Brazil.” Econometrica 83(2): 423-64.
Gaikwad, Nikhar, and Nellis, Gareth. 2017. “The Majority-Minority Divide in Attitudes Toward Internal Migration: Evidence from Mumbai.” American Journal of Political Science 61(2): 456-72.
Gaikwad, Nellis, and Nellis, Gareth. 2021. “Do Politicians Discriminate Against Internal Migrants? Evidence from Nationwide Field Experiments in India.” American Journal of Political Science: Early View.
Gaikwad, Nikhar, Nellis, Gareth, and Thomas, Anjali. 2021. “Bureaucratic Hurdles, Political Resistance, and Public Service Access: Evidence from a Field Experiment in India.” Mimeo: Columbia University. bit.ly/3mDmOW5.
Gay, Claudine. 2012. “Moving to Opportunity: The Political Effects of a Housing Mobility Experiment.” Urban Affairs Review 48(2): 147-79.
Government of India. 2010. “Migration in India: 2007–2008.” National Sample Survey Office Report No. 533. bit.ly/3btvhUH.
48 Harris, J. Andrew, Kamindo, Catherine, and van der Windt, Peter. Forthcoming. “Election Administration in Fledgling Democracies: Experimental Evidence from Kenya.” Journal of Politics.
Jones, Terry-Ann. 2020. Sugarcane Labor Migration in Brazil. London: Palgrave.
Kone, Zovanga, Liu, Maggie, Mattoo, Aaditya, Ozden, Caglar, and Sharma, Siddharth. 2018. “Internal Borders and Migration in India.” Journal of Economic Geography 18(4): 729-59.
Kruks-Wisner, Gabrielle. 2018. Claiming the State: Active Citizenship and Social Welfare in Rural India. Cambridge: Cambridge University Press.
Kumar, Sanjay, and Banerjee, Souradeep. 2017. “Low Levels of Electoral Participation in Metropolitan Cities.” Economic and Political Weekly 52(45): 82-6.
Meredith, Marc. 2009. “Persistence in Political Participation.” Quarterly Journal of Political Science 4(3): 187-209.
Mooney, Erin, and Jarrah, Balkees. 2004. “The Voting Rights of Internally Displaced Persons: The OSCE Region.” Mimeo: Brookings Institution-Johns Hopkins SAIS Project on Internal Displacement. brook.gs/3cuw8FT.
Nickerson, David. 2015. “Do Voter Registration Drives Increase Participation? For Whom and When?" Journal of Politics 77(1): 88-101.
Organisation for Economic Co-operation and Development. 2019. “Society at a Glance 2019: OECD Social Indicators.” Paris: OECD Publishing. bit.ly/2PY8fk2.
Panda, Sitakanta. 2019. “Political-Economic Determinants of Electoral Participation in India.” India Review 18(2): 184-219.
Paglayan, Agustina. 2021. “The Non-Democratic Roots of Mass Education: Evidence from 200 Years.” American Political Science Review 115(1): 179-198.
Peisakhin, Leonid, 2012. “Transparency and Corruption: Evidence from India.” Journal of Law and Economics 55(1): 129-49.
Peters, Floris, Schmeets, Hans, and Vink, Maarten. 2019. “Naturalisation and Immigrant Earnings: Why and to Whom Citizenship Matters.” European Journal of Population 36(3): 511-545.
Pons, Vincent, and Liegey, Guillaume. 2019. “Increasing the Electoral Participation of Immigrants: Experimental Evidence from France.” Economic Journal 129(617): 481-508.
49 Rajan, S. Irudaya, Sivakumar, P., and Srinivasan, Aditya. 2020. “The COVID-19 Pandemic and Internal Labour Migration in India: A ‘Crisis of Mobility.’” Indian Journal of Labour Economics 63: 1021-39.
Sankhe, Shirish, Vittal, Ireena, Dobbs, Richard, Mohan, Ajit, Gulati, Ankur, Ablett, Jonathan, Gupta, Shishir, Kim, Alex, Paul, Sudipto, Sanghvi, Aditya, and Sethy, Gurpreet. 2010. India’s Urban Awakening: Building Inclusive Cities, Sustaining Economic Growth. McKinsey Global Institute. mck.co/3mDV6Zy.
Scheve, Kenneth, and Slaughter, Matthew. 2001. “Labor Market Competition and Individual Preferences Over Immigration Policy.” Review of Economics and Statistics 83(1): 133-45.
Sobolewska, Maria, Galandini, Silvia, and Lessard-Phillips, Laurence. 2017. “The Public View of Immigrant Integration: Multidimensional and Consensual. Evidence from Survey Experiments in the UK and the Netherlands.” Journal of Ethnic and Migration Studies 43(1): 58-79.
Tata Institute of Social Sciences. 2015. “Inclusive Elections in India: A Study on Domestic Migration and Issues in Electoral Participation.” Technical Report. bit.ly/3by9Ulq.
Thachil, Tariq. 2017. “Do Rural Migrants Divide Ethnically in the City? Evidence from an Ethnographic Experiment in India” American Journal of Political Science 61(4): 908-26.
Thachil, Tariq. 2020. “Does Police Repression Spur Everyday Cooperation? Evidence from Urban India.” Journal of Politics 82(4): 1474-89.
United Nations Educational, Scientific and Cultural Organization. 2012. “National Workshop on Internal Migration and Human Development in India.” Technical Paper. bit.ly/2KkV4U7.
Voicu, Bogdan, and Comşa, Mircea. 2014. “Immigrants’ Participation in Voting: Exposure, Resilience, and Transferability.” Journal of Ethnic and Migration Studies 40(10): 1572-92.
Wang, Tova. 2013. “Expanding Citizenship: Immigrants and the Vote.” Democracy 28. bit.ly/366GWJx.
Weiner, Myron. 1978. Sons of the Soil: Migration and Ethnic Conflict in India. Princeton: Princeton University Press.
Wellman, Elizabeth. 2021. “Emigrant Inclusion in Home Country Elections: Theory and Evidence from Sub-Saharan Africa.” American Political Science Review 115(1): 82-96.
White, Ariel, Nathan, Noah, and Faller, Julie. 2015. “What Do I Need to Vote? Bureaucratic Discretion and Discrimination by Local Election Officials.” American Political Science Review 109(1): 129-42.
50 Wilkerson, Isabel. 2010. The Warmth of Other Suns: The Epic Story of America’s Great Migration. New York: Random House.
Yue, Zhongshan, Li, Shuzhuo, Jin, Xiaoyi, and Feldman, Marcus. 2013. “The Role of Social Networks in the Integration of Chinese Rural–Urban Migrants: A Migrant–Resident Tie Perspective.” Urban Studies 50(9): 1704-23.
51 Do you or your immediate family To what extent continue to benefit Straight−line to you agree or Do you or your from government distance from Delhi/ disagree: I feel spouse personally Length of residence schemes in your Lucknow to district more at home in my own land or property in city (years). previous place of former residence previous place of in your previous of residence − (kms). residence than I do place of residence? for example, PDS, in Delhi/Lucknow. MGNREGA, or cash transfer schemes? 2000
72.4%
1500 53.2% 47.4% 46.8% 1000 39.8% 42.2% 30.9% 29.9% 27.9% 25.5% 27.6% 500 17.5%
Number of respondents 13.1% 11.6% 10.4% 3.9% 0
<5 >25 No Yes No Yes 5−14 <100 15−25 >1000 100−499 agree agree 500−1000 Very disagreemuchSomewhat disagreeSomewhat Very much
Figure 1: Baseline descriptive statistics of migants’ attachments to their former places of residence.
1 Table 1: [Exploratory] Baseline correlates of migrants’ continued political participation in their former place of residence (“hometown”), estimated using OLS regression. For question wordings, see Figure 1. Robust standard errors in parentheses.
Dependent variable: Has Hometown Voter ID Returned to Vote in Hometown (1) (2) More at home in hometown (0-1) 0.158∗∗∗ 0.088∗∗∗ (0.025) (0.020) Still receives hometown schemes (0/1) 0.108∗∗∗ 0.101∗∗∗ (0.019) (0.017) Owns hometown property (0/1) 0.072∗∗∗ 0.039∗∗ (0.022) (0.020) Length of residence in city (years) −0.003∗∗∗ −0.002∗∗∗ (0.001) (0.001) Straight-line distance to home district (kms) −0.0001∗∗ −0.0001∗∗∗ (0.00003) (0.00002) Constant 0.177∗∗∗ 0.146∗∗∗ (0.027) (0.022) DV values {0,1} {0,1} Observations 2,306 2,306 Adjusted R2 0.056 0.047 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
2 Table 2: [Pre-registered] T1 experimental results for primary political outcomes. Outcomes are whether respondent (1) currently has a voter ID card allowing them to vote in city elections; (2) voted in the city during the 2019 Lok Sabha elections; and (3) intends to vote in the next state elections held in the city. OLS estimates of intent to treat effects. Models include covariates. Robust standard errors in parentheses.
Has City-Based Voted in City Likelihood of Voting Voter ID in 2019 in City in Future (1) (2) (3) T1 treatment 0.236 0.203 0.031 (0.019) (0.019) (0.009) p-value (upper) 0.000 0.000 0.000 Control mean 0.161 0.178 0.856 Observations 2,120 2,120 2,120 Adjusted R2 0.084 0.065 0.011 DV values {0, 1}{0, 1}{0, 0.33, 0.67, 1}
3 Table 3: [Pre-registered] T1 experimental results for additional political outcomes. Outcomes are whether respondent (1) pays attention to news about national, state, and city politics; (2) agrees that elected politicians are accountable to city residents; (3) agrees that citizens have an influence on the government; (4) has trust in the national, state, and city governments as well as political parties. OLS estimates of intent to treat effects. Models include covariates. Robust standard errors in parentheses.
Political Politician Sense of Political Interest Accountability Political Trust Index Perceptions Efficacy Index (1) (2) (3) (4) T1 treatment 0.091 0.039 -0.012 0.027 (0.039) (0.015) (0.018) (0.028) p-value (upper) 0.010 0.003 0.745 0.170 Control mean 0.000 0.697 0.450 0.000 Observations 2,120 2,120 2,120 2,120 Adjusted R2 0.027 0.006 0.003 0.019 DV values [−1.44, 1.56] {0, 0.33, 0.67, 1}{0, 0.33, 0.67, 1} [−1.62, 1.17]
4 Table 4: [Pre-registered] Estimates of heterogeneous effects of T1 treatment. Models do not include additional covariates. All independent variables are dichotomous and are described in the text. Robust standard errors in parentheses.
Dependent variable: Has City-Based Voter ID Voted in City in 2019 (1) (2) T1 x Primary education 0.083∗∗ 0.057 (0.041) (0.042) T1 x Muslim −0.114∗∗ −0.018 (0.048) (0.049) T1 x SC/ST −0.113∗∗∗ −0.081∗ (0.042) (0.042) T1 x High income 0.028 0.041 (0.038) (0.038) T1 x Long-term migrant −0.019 −0.007 (0.038) (0.038) T1 0.248∗∗∗ 0.183∗∗∗ (0.049) (0.049) Primary education −0.058∗∗ −0.059∗∗ (0.025) (0.026) Muslim 0.001 0.004 (0.029) (0.030) SC/ST −0.004 −0.008 (0.025) (0.026) High income 0.038∗ 0.009 (0.022) (0.023) Long-term migrant 0.065∗∗∗ 0.052∗∗ (0.023) (0.024) Constant 0.149∗∗∗ 0.188∗∗∗ (0.028) (0.029) Observations 2,120 2,120 Adjusted R2 0.087 0.059 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
5 Table 5: [Index outcome pre-registered; index component analyses exploratory] T2 experimental results for exposure to campaigning during the 2019 Lok Sabha elections. Campaign exposure index (1) based on whether respondent reports that politicians or party workers (2) visited their basti around the 2019 Lok Sabha election campaign; (3) came to the door to request votes; (4) offered gifts; (5) tried to specifically win votes of recent migrants to the city; (6) campaigned hard to win votes in the basti. Weighted least squares estimates of intent to treat effects. Clusters weighted equally. Models include block fixed effects and individual covariates. Cluster-robust standard errors in parentheses.
Index Components Campaigning Basti Home Visit Number Migrant- Perceived Exposure Visits by by Politician or of Focused Campaign Index Politicians Party Worker Gifts Campaigning Intensity (1) (2) (3) (4) (5) (6) T2 treatment 0.101 0.066 0.036 0.017 0.014 0.073 (0.058) (0.078) (0.038) (0.012) (0.047) (0.031) p-value (upper) 0.043 0.203 0.174 0.073 0.384 0.010 Control mean -0.039 0.559 0.550 0.013 0.425 0.676 Observations 1,969 1,969 1,969 1,969 1,969 1,931 No. of Clusters 87 87 87 87 87 87 Adjusted R2 0.056 0.070 0.047 0.019 0.008 0.021 DV values [−0.96, 3.65] {0,..., 4}{0, 1}{0, 1, 2}{0, 1}{0, 0.33, 0.67, 1}
6 Overcoming the Political Exclusion of Migrants: Theory and Experimental Evidence from India
Nikhar Gaikwad, Columbia University
Gareth Nellis, University of California, San Diego
ONLINE APPENDIX
Note: All references to “Supplementary Information” (“SI”) relate to the document, supplementary-information.pdf available at: https://doi.org/10.7910/DVN/G1JCKK
1 Online Appendix
Contents
A Voter turnout rates among naturalized immigrants versus native-born citizens globally 3
B Indian Human Development Survey-II analysis 3 B.1 Political participation ...... 3 B.2 Politician attachments and receipt of government benefits ...... 6 B.3 Confidence in political institutions ...... 8 B.4 Urban versus rural political participation ...... 8 B.5 Documentation ...... 9 B.6 Demographic profile ...... 10
C T1 further information 11 C.1 Official forms ...... 11 C.2 Implementation of T1 in the field ...... 13
D Political parties and migrant outreach 16
E T2 further information 17
F Study timeline 20
G Indexing 21
H Outliers in ‘income‘ covariate 22
I Internal validity 23 I.1 Balance ...... 23 I.2 Attrition rates by treatment condition ...... 25
J Administrative data on T1 application processing 26
K Deviations from pre-analysis plan 27
L Additional ethical considerations 27
2 A Voter turnout rates among naturalized immigrants versus native-born citizens globally
Table A1: Average voter turnout rates by natives versus naturalized immigrants, according to the sixth wave of the World Values Survey (2010-14). Turnout percentages are calculated using data from all countries where responses to two questions were gathered: “Respondent immigrant?” (1 = yes; 0 = no), and “Vote in elections?” (1 = always/usually; 0 otherwise). Included countries are: Algeria, Azerbaijan, Argentina, Armenia, Brazil, Belarus, Chile, Taiwan, Colombia, Cyprus, Esto- nia, Georgia, Germany, Ghana, Haiti, India, Iraq, Kazakhstan, Jordan, South Korea, Kyrgyzstan, Lebanon, Libya, Malaysia, Mexico, Morocco, Netherlands, New Zealand, Nigeria, Pakistan, Peru, Philippines, Poland, Qatar, Romania, Russia, Rwanda, Singapore, Slovenia, South Africa, Zim- babwe, Sweden, Thailand, Trinidad and Tobago, Tunisia, Turkey, Ukraine, Egypt, United States, Uruguay, Uzbekistan, and Yemen.
Native turnout percent Naturalized immigrant turnout percent 81.82 71.1
B Indian Human Development Survey-II analysis
In this section we characterize India’s migrant population using nationally representative survey data. We do this to shed light on the political behaviors of migrants versus non-migrants, and to assess the extent to which our sample conforms to—or deviates from—the demographic traits of the country’s migrant population at large. Our primary resource for conducting these analyses is the second round of the Indian Human Devel- opment Survey (IHDS-II). To our knowledge, IHDS-II is unique among nationally representative surveys in posing both detailed questions about respondents’ migration history and asking a battery of questions about political participation. Table SI3 gives a full roster of the IHDS-II variables used in the subsequent analyses and notes the manner in which they were recoded, where appropriate. Note, in what follows, we decompose the IHDS-II sample into rural and urban areas for the purposes of most analysis (according to the primary sampling unit’s designation given in the 2011 Census of India). This is based on the presumption that rural areas are overwhelmingly migrant-sending regions whereas urban areas are largely migrant-receiving regions. In rural areas, we examine differences between households that did and did not report having had a member who engaged in seasonal migration during the past five years (a question directly posed on the survey). In urban areas, defining who is a migrant is more complex. Respondents were asked about the migration history of their “families” and when their family first came to their current town or city of residence. We class as internal migrants those reporting that their families arrived from a different Indian district or state within the past 10 years.
B.1 Political participation
Turning to the substantive analysis, we first investigate whether migrants and migrant-sending households participate less in politics than non-migrants. We generate a Political Engagement Index, which sums whether (a) the household respondent reported being a member of a political party, (b) a member of the village panchayat (in rural areas) or ward committee (in urban areas), and (c) whether they reported having attended a panchayat or ward committee meeting within the last year. For both urban (Table A2) and rural (Table A3) samples, we find strong evidence that migrants and migrant-sending households, respectively, are less likely to be politically engaged than households not involved in migration.
3 In Table A4, we demonstrate in the pooled (rural and urban) data that the migrant/non-migrant participation gap persists after controlling for households’ religious affiliation. While IHDS-II does not ask directly about voting, this analysis significantly adds to the body of evidence suggesting that India’s internal migrants are politically disempowered.
Table A2: [Exploratory] Are migrant households less politically engaged than non-migrant house- holds in urban areas? Data are from the IHDS-II household-level recode file. Robust standard errors in parentheses.
Dependent variable: Political Political Ward Attended Engagement Party Committee Ward Committee Index Member Member Meeting (1) (2) (3) (4) Migrant −0.039∗∗∗ −0.035∗∗∗ −0.007 −0.077∗∗∗ (0.007) (0.006) (0.008) (0.017) Constant 0.073∗∗∗ 0.042∗∗∗ 0.022∗∗∗ 0.154∗∗∗ (0.001) (0.002) (0.001) (0.003) Observations 14,500 14,546 14,504 14,516 Adjusted R2 0.001 0.0005 −0.00003 0.001 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
Table A3: [Exploratory] Are migrant-sending households less politically engaged than non-migrant- sending households in rural areas? Data are from the IHDS-II household-level recode file. Robust standard errors in parentheses.
Dependent variable: Political Political Attended Engagement Party Panchayat Panchayat Index Member Member Meeting (1) (2) (3) (4) Migrant-sending household −0.010∗∗ −0.013∗∗∗ −0.015∗∗∗ −0.002 (0.004) (0.003) (0.004) (0.010) Constant 0.153∗∗∗ 0.037∗∗∗ 0.052∗∗∗ 0.370∗∗∗ (0.001) (0.001) (0.001) (0.003) Observations 26,936 27,040 26,948 26,989 Adjusted R2 0.0001 0.0004 0.0003 −0.00004 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
4 Table A4: [Exploratory] Are migrant and migrant-sending households less politically engaged than non-migrant households after controlling for religion? Data are from the IHDS-II household-level recode file. Urban and rural samples are pooled. Robust standard errors in parentheses.
Dependent variable: Political Engagement Index Migrant/migrant-sending household −0.013∗∗∗ (0.004) Muslim −0.020∗∗∗ (0.003) Urban −0.080∗∗∗ (0.002) Constant 0.156∗∗∗ (0.001) Observations 41,436 Adjusted R2 0.038 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
5 B.2 Politician attachments and receipt of government benefits
We next turn to evaluate the extent to which migrants (in urban areas) and migrant-sending households (in rural areas) are embedded in networks of clientelistic exchange operated by local politicians. To be sure, measuring quid pro quo transactions of material benefits for votes is challenging. To understand how likely it is that households are involved in such transactions, we make use of two sets of questions put to respondents in the IHDS-II survey instrument: acquaintance with local politicians, and income received from government schemes. Our reasoning is that households who are not acquainted with local politicians are unlikely to be able to access state benefits via clientelistic mechanisms, and, from another angle, that those not receiving substantial income from government schemes are unlikely to be involved in a substantial votes-for-benefits exchange relationship with local politicians, who have significant sway over how state resources are allocated. The Acquaintance Index is the average of the four component binary acquaintance measures. In Tables A5 we observe that migrants in urban India are similarly acquainted with local politicians as are non-migrants, while Table A6 demonstrates that urban migrants are significantly less likely to receive income from state-run schemes.
Table A5: [Exploratory] Are migrant households less connected to politicians than non-migrant households in urban areas? Data are from the IHDS-II household-level recode file. Robust standard errors in parentheses.
Dependent variable: Acquaintance: Acquaintance: Acquaintance: Politician Acquaintance: Party Worker Acquaintance Politician in Outside Party Worker Outside Index Community Community in Community Community (1) (2) (3) (4) (5) Migrant 0.0005 0.001 0.004 0.004 −0.008 (0.017) (0.016) (0.022) (0.019) (0.022) Constant 0.118∗∗∗ 0.068∗∗∗ 0.145∗∗∗ 0.104∗∗∗ 0.158∗∗∗ (0.002) (0.002) (0.003) (0.003) (0.003) Observations 14,458 14,533 14,496 14,509 14,470 Adjusted R2 −0.0001 −0.0001 −0.0001 −0.0001 −0.0001 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
Table A6: [Exploratory] Do migrant households receive less income from government schemes than non-migrant households in urban areas? Data are from the IHDS-II household-level recode file. Robust standard errors in parentheses.
Dependent variable: Winsorized Benefits Income Migrant −272.713∗∗∗ (80.823) Constant 659.872∗∗∗ (13.059) Observations 14,572 Adjusted R2 0.0005 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
Looking at the rural sample, Table A7 shows that migrant-sending households are substantially less likely to have connections with local politicians by comparison with other households. They also receive somewhat less in income from government schemes (see Table A8).
6 Table A7: [Exploratory] Are migrant-sending households less connected to politicians than non- migrant-sending households in rural areas? Data are from the IHDS-II household-level recode file. Robust standard errors in parentheses.
Dependent variable: Acquaintance: Acquaintance: Acquaintance: Politician Acquaintance: Party Worker Acquaintance Politician in Outside Party Worker Outside Index Community Community in Community Community (1) (2) (3) (4) (5) Migrant-sending household −0.034∗∗∗ −0.024∗∗∗ −0.043∗∗∗ −0.020∗∗∗ −0.046∗∗∗ (0.003) (0.004) (0.006) (0.004) (0.006) Constant 0.093∗∗∗ 0.066∗∗∗ 0.127∗∗∗ 0.063∗∗∗ 0.115∗∗∗ (0.001) (0.002) (0.002) (0.002) (0.002) Observations 26,924 27,026 26,965 26,996 26,931 Adjusted R2 0.002 0.001 0.001 0.001 0.002 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
Table A8: [Exploratory] Do migrant-sending households receive less income from government schemes than non-migrant-sending households in rural areas? Data are from the IHDS-II household- level recode file. Robust standard errors in parentheses.
Dependent variable: Winsorized Benefits Income Migrant-sending household −60.649∗ (34.996) Constant 1,048.111∗∗∗ (11.611) Observations 27,080 Adjusted R2 0.0001 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
7 B.3 Confidence in political institutions
IHDS-II respondents were asked about their confidence in various political institutions. We see that migrant households in urban areas (Table A9) and migrant-sending households in rural areas (Table A10) express less confidence in major political institutions than households not involved in migration. In what follows, the Confidence Index is the average of the three component confidence measures.
Table A9: [Exploratory] Do migrant households express less confidence in political institutions than non-migrant households in urban areas? Data are from the IHDS-II household-level recode file. Robust standard errors in parentheses.
Dependent variable: Confidence Confidence: Confidence: Confidence: Index Ward Committees Politicians State Government (1) (2) (3) (4) Migrant −0.034∗∗ −0.122∗∗∗ −0.016 −0.069 (0.016) (0.042) (0.040) (0.044) Constant 0.447∗∗∗ 1.046∗∗∗ 0.577∗∗∗ 1.061∗∗∗ (0.002) (0.006) (0.005) (0.006) Observations 14,467 14,517 14,536 14,482 Adjusted R2 0.0003 0.001 −0.0001 0.0001 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
Table A10: [Exploratory] Do migrant-sending households express less confidence in political insti- tutions than non-migrant-sending households in rural areas? Data are from the IHDS-II household- level recode file. Robust standard errors in parentheses.
Dependent variable: Confidence Confidence: Confidence: Confidence: Index Panchayats Politicians State Government (1) (2) (3) (4) Migrant-sending household −0.008 −0.038∗∗ −0.031∗∗ 0.021 (0.006) (0.016) (0.014) (0.016) Constant 0.470∗∗∗ 1.112∗∗∗ 0.604∗∗∗ 1.103∗∗∗ (0.002) (0.004) (0.004) (0.005) Observations 26,942 27,009 27,027 26,968 Adjusted R2 0.00004 0.0002 0.0001 0.00003 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
B.4 Urban versus rural political participation
Might migrants’ low participation in cities be a consequence of socialization in rural areas, where participation is lower than in cities? As discussed in the main text, aggregate studies of turnout in India have consistently documented that voter turnout is higher in rural areas. Here, we assess whether this pattern is maintained for the broader metric of political participation in the IHDS-II data, and after controlling for salient demographic attributes that might also impact turnout. Table A11 presents this analysis. Consistent with expectations, there is clear evidence that political engagement is substantially lower in urban areas. Moreover, this pattern persists after controlling for religion, caste, and household wealth (as captured in the IHDS-II asset index).
8 Table A11: [Exploratory] Is political participation lower in urban areas than in rural areas? Data are from the IHDS-II household-level recode file. Robust standard errors in parentheses.
Dependent variable: Political Engagement Index (1) (2) Urban −0.080∗∗∗ −0.104∗∗∗ (0.002) (0.002) Muslim −0.011∗∗∗ (0.003) SC/ST 0.012∗∗∗ (0.002) Asset Index 0.004∗∗∗ (0.0002) Constant 0.152∗∗∗ 0.095∗∗∗ (0.001) (0.003) Observations 41,934 41,924 Adjusted R2 0.036 0.050 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
B.5 Documentation
In order to register to vote, applicants must submit personal identification and proof of residency documen- tation. Are migrants and migrant-sending households less likely to possess such documents? The IHDS-II survey asks detailed questions about documentation among household members, enabling us to test whether documentation rates fall short among households not engaged in migration. Tables A12 and A13 unearth clear evidence of a lag, among migrants in urban areas and migrant-sending households in rural areas respec- tively. Lack of access to documentation steepens the challenge of going through the bureaucratic registration process for migrants, in line with the paper’s central finding.
Table A12: [Exploratory] Are migrant households less likely to possess documents than non-migrant households in urban areas? Data are from the IHDS-II household-level recode file. Robust standard errors in parentheses.
Dependent variable: Has Proof of Residence Has Photo ID (1) (2) Migrant −0.196∗∗∗ −0.131∗∗∗ (0.029) (0.022) Constant 0.866∗∗∗ 0.989∗∗∗ (0.003) (0.001) Observations 14,508 14,515 Adjusted R2 0.006 0.023 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
9 Table A13: [Exploratory] Are migrant-sending households less likely to possess documents than non-migrant-sending households in rural areas? Data are from the IHDS-II household-level recode file. Robust standard errors in parentheses.
Dependent variable: Has Proof of Residence Has Photo ID (1) (2) Migrant-sending household −0.180∗∗∗ −0.014 (0.027) (0.009) Constant 0.868∗∗∗ 0.987∗∗∗ (0.003) (0.001) Observations 14,280 14,287 Adjusted R2 0.006 0.0002 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
B.6 Demographic profile
Finally, we examine the profile of the overall population of households involved in migration compared to households not engaged in migration (either in destination urban areas, as recent migrant households, or in rural sending regions, as migrant-sending households). We find that migrant-engaged households have a similar religious demography to non-migrant households, with comparable shares of Muslims in both groups. Households engaged in migration, however, are more likely to belong to the Scheduled Castes and Tribes and have fewer household assets, on average.
Table A14: [Exploratory] Do migrant- and migrant-sending households differ according to socio- economic status compared to households not engaged in migration? Data are from the IHDS- II household-level recode file. Urban and rural samples are pooled. Robust standard errors in parentheses.
Dependent variable: Muslim SC/ST Asset Index (1) (2) (3) Migrant/migrant-sending household 0.005 0.092∗∗∗ −2.784∗∗∗ (0.006) (0.010) (0.104) Urban 0.061∗∗∗ −0.119∗∗∗ 6.447∗∗∗ (0.004) (0.005) (0.059) Constant 0.095∗∗∗ 0.334∗∗∗ 13.387∗∗∗ (0.002) (0.003) (0.039) Observations 41,651 41,652 41,629 Adjusted R2 0.008 0.019 0.238 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
10 C T1 further information
C.1 Official forms
Address of earlier place of ordinary residence (if applying due to shifting from another constituency) ELECTION COMMISSION OF INDIA House No. Street/Area/Locality FORM-6 Acknowledgement No.______Town/Village (See Rules 13(1) and 26) of Registration of Electors Rule-1960 (To be filled by office) Post Office Pin Code Application for Inclusion of Name in Electoral Roll for First time Voter OR on Shifting District State/UT from One Constituency to Another Constituency. I am aware that making a statement or declaration which is false and which I know or believe to be false or do not believe to be true, is punishable under Section 31 of the Representation of the People Act, 1950 (43 of 1950). To, The Electoral Registration Officer, ………………..………………………………………...Assembly / Parliamentary Consitituency I request that my name be included in the electoral roll for the above Constituency. (Tick appropriate box) SPACE FOR PASTING ONE Place……………...... As a first time voter or due to shifting from another constituency RECENT PASSPORT SIZE Particulars in support of my claim for inclusion in the electoral roll are given below:- PHOTOGRAPH (3.5 CM X Date……………………………………….. Signature of Applicant……….……………………………………….. 3.5 CM) SHOWING Mandatory Particulars Remarks of Field Level Verifying Officer: FRONTAL VIEW OF FULL (a) Name FACE WITHIN THIS BOX (b) Surname(if any)
(c) Name and surname of Relative of Details of action taken Applicant [see item (d)] (To be filled by Electoral Registration Officer of the constituency) (d) Type of Relation Father Mother Husband Wife Other (Tick appropriate box) (e) Age [as on 1st January of current calendar year………………..] Years Months The application of Shri / Shrimati/ Kumari ……………………………………………………………………………………..for inclusion of name in the electoral roll in Form 6 has been accepted/ rejected. Detailed reasons for acceptance [under or in pursuance of rule (f) Date of Birth (in DD/MM/YYYY format)(if known) 18/20/26(4)] or rejection [under or in pursuance of rule 17/20/26(4)] are given below:
(g) Gender of Applicant (Tick appropriate box) Male Female Third Gender Place: (h)Current address where applicant is ordinarily resident House No. Date: Signature of ERO Seal of the ERO Street/Area/Locality Town/Village Post Office Pin Code Intimation of decision taken (to be filled by Electoral Registration Officer of the constituency and to be posted to the applicant on the address as given by the applicant) District State/UT Postage Stamp to House No. The application in Form 6 of Shri/Shrimati/Kumari……………………………………………………………………………………………………………..be affixed by the (i) Permanent address of applicant Electoral Current address where applicant is ordinarily resident House No. Street/Area/Locality Registration Authority at the Street/Area/Locality Town/Village time of dispatch Town/Village Post Office Pin Code Post Office Pin Code District State/UT District State/UT (j)EPIC No. (if issued) Optional Particulars Has been (a) accepted and the name of Shri/Shrimati/Kumari………………………………………………………………………………………………… (k) Disability (if any) Visual impairment Speech & hearing disability Locomotor disability Other (Tick appropriate box) Has been registered at Serial No………………….in Part No…………………….. of AC No……………………………………………………………………. (l) Email id (optional) (b) rejected for the reason………………………………………………………………………………………………………………………………………………………. (m) Mobile No. (optional)
DECLARATION - I hereby declare that to the best of knowledge and belief – Date: Electoral Registration Officer (i) I am a citizen of India and place of my birth is Village/Town…………………………..…….District………………………..……State………………………………… (ii) I am ordinarily resident at the address given at (h) above since ……………………………….……………………………….………(date, month, year). Address….……………………………………………. (iii)I have not applied for the inclusion of my name in the electoral roll for any other constituency. Acknowledgement/Receipt *(iv)My name has not already been included in the electoral roll for this or any other assembly/ parliamentary constituency Acknowledgement Number ______Date ______OR *My name may have been included in the electoral roll for______Constituency in ______Received the application in form 6 of Shri / Smt. / Ms. ______State in which I was ordinarily resident earlier at the address mentioned below and if so, I request that the same may be deleted from that [ Applicant can refer the Acknowledgement No. to check the status of application]. electoral roll. * strike off the option not appropriate Name/Signature of ERO/AERO/BLO
Figure A1: Election Commission of India, Form 6.
11 Remarks of Field Level Verifying Officer:
ELECTION COMMISSION OF INDIA FORM-7 Acknowledgement No.______(See Rules 13(2) and 26) of Registration of Electors Rule-1960 (To be filled by office)
Application for Objecting Inclusion of Name of Other Person / Seeking Deletion of Details of action taken Own Name/Seeking Deletion of Any Other Person’s Name in Electoral Roll due to (To be filled by Electoral Registration Officer of the constituency)
Death/Shifting. The application of Shri / Shrimati/ Kumari ……………………………………………………………………………………..objecting to inclusion/ To, The Electoral Registration Officer, ………………..…………………………………………………………………...Assembly / Parliamentary Constituency I hereby object to the proposed inclusion of the name of the under mentioned person in the electoral roll seeking deletion of name of Shri / Shrimati/ Kumari ………………………………………………………………………………... in the electoral roll in Form 7 has been acepted/rejected. I hereby request that entry relating to name of the person mentioned below is required to be deleted I request that the entry relating to myself is to be deleted from Electoral Roll Detailed reasons for acceptance [under or in pursuance of rule 18/20/26(4)] or rejection [under or in pursuance of rule Particulars in support of my objection/deletion are given below:- 17/20/26(4)] are given below: Particulars of the applicant (a) Name Place:
(b) Surname(if any) Date: Signature of ERO Seal of the ERO
(c) Part No. (d) Serial No. Intimation of decision taken (to be filled by Electoral Registration Officer of the constituency and to be posted to the applicant on the address available in the record) (e) EPIC No. (If issued) Postage Stamp The application in Form 7 of Shri/Shrimati/Kumari……………………………………………………………………………………………………………..to be affixed by Details of person inclusion of whose name is objected to/whose entry is to be deleted: the Electoral Current address where applicant is ordinarily resident House No. Registration (a)Name Authority at the time of dispatch (b)Surname(if any) Street/Area/Locality Town/Village (c) Part No. (d)Serial No. Post Office Pin Code
(e)EPIC No.(If issued) District State/UT (f) Reason(s) for objection/deletion: Has been (a) accepted and the name of Shri/Shrimati/Kumari...... has been
deleted from………………………………….. Part No...... of AC No…………………………….…………….
(b) rejected for the reason……………………………………………………………………………………………………………………………………………………….
Declaration I hereby declare that the facts and particulars mentioned above are true to the best of my knowledge and belief. I Date: Electoral Registration Officer am aware that making a statement or declaration which is false and which I know or believe to be false or do not believe to be true, is punishable under Section 31 of the Representation of the People Act, 1950 (43 of 1950). Address……………………………………………….
Place……………...... Acknowledgement/Receipt
Date……………………………………. Signature of Applicant……….……………………………………….. Acknowledgement Number ______Date ______
Received the application in form 7 of Shri / Smt. / Ms. ______[ Applicant can refer the Acknowledgement No. to check the status of application].
Name/Signature of ERO/AERO/BLO
Figure A2: Election Commission of India, Form 7.
12 C.2 Implementation of T1 in the field
Figure A3: Photographs of field workers assisting T1-assigned migrants in gathering documents and filling in the forms needed to register to vote locally.
13 Figure A4: Pictures of the local election offices where applications to register to vote are evaluated and processed.
14 Figure A5: Successful applicants hold up their newly minted voter ID cards, enabling them to vote locally.
15 D Political parties and migrant outreach
Are political parties in our study cities less likely to engage with migrants than with long-term city residents in typical elections? To address this question, we obtained proprietary data from the Centre for Developing Societies (CSDS), which conducted a representative poll of Delhi citizens following the 2015 state assembly elections there. The survey was unique in that it elicited detailed information about interactions with party workers in the lead-up to the election, broken down by political party. Specifically, the survey instrument asked: “Did any candidate, party worker or canvasser come to your house during the campaign to ask for your vote?” If the respondent answered in the affirmative, they were then asked, “Candidates or workers of which party came to you?” (they were able to list up to two parties). Respondents were additionally asked, “For how many years have you been living in Delhi?” with four response options provided. We employ the definition of recent migrants employed in Banerjee and Kumar (2017): those who have been living in Delhi for less than ten years. We run a simple analysis to estimate whether migrants were less likely to be visited by party canvassers compared to longer-term residents (see Table A15). We find strong evidence that this is what occurred. Migrants were 14 percentage points less likely to be visited by a party worker of any stripe—a difference that is qualitatively large and statistically significant. We also examine differences in visits by the various parties. The effects are negatively signed for visits by canvassers from all the major parties, but are largest and statistically significant for those campaigning on behalf of the BJP. These results lend credence to the claim that political parties are less likely to engage in outreach to migrant citizens, plausibly on account of their greater uncertainty about migrants’ political preferences.
Table A15: Canvassing visits to recent migrants versus long-term city residents in Delhi during the 2015 Delhi state assembly elections. Survey data were obtained from the Centre for the Study of Developing Societies and are described at: bit.ly/3nCrACa. Bivariate OLS analysis with robust standard errors in parentheses.
Dependent variable: Party Canvasser AAP BJP INC Visited Canvasser Canvasser Canvasser (1) (2) (3) (4) Migrant −0.136∗∗∗ −0.066 −0.089∗∗ −0.038 (0.047) (0.044) (0.044) (0.033) Constant 0.632∗∗∗ 0.412∗∗∗ 0.443∗∗∗ 0.187∗∗∗ (0.011) (0.011) (0.011) (0.009) Observations 1,994 2,060 2,060 2,060 Adjusted R2 0.004 0.001 0.001 0.0001 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
16 E T2 further information
Address:
Dear ,
Our NGO, , along with , has been working hard over the past year to help migrants in parts of your constituency to be- come politically empowered. In particular, in the run up to the Lok Sabha elections, we have been running a voter registration drive in the city, helping thousands of migrants from other parts of India to register to vote here for the first time. This will allow them to participate fully in the upcoming elections in this city. Below we show a list and a map of the polling booth areas where we’ve been working to help migrants register. We’re sending this letter to all candidates standing in the forthcoming election in this constituency, as well as to sitting MPs, MLAs, and municipal corporators in this area. We hope these newly empowered migrant citizens will exercise their democratic right to vote here. We wanted to let you know that these citizens are now registered to vote in your constituency! Warm regards,
हमारा एन जी ओ, िपछले एक साल से कड़ी मेहनत कर रहा है तािक आपके िनवा셍चन क्षेत्रᴂ म भारत के िवभन्न थानⴂ/े प्रदशⴂ से आये हुए 핍यिक्तयⴂ को राजनीतक 셂प से सशक्त बननेᴂ म मदद िमल सके। िवशेष 셂प से, लोकसभा चुनावⴂ मᴂ, हम शहर मᴂ मतदाता पजीकरणं अभयान चला रहे ह,ℂ ￭जससे भारत के अꅍय िहसⴂ से आये हुए हजारⴂ 핍यिक्तयⴂ को पहली बार यहां मतदान करने के िलए पजीकरणं करने मᴂ मदद िमली ह।ै यह उनह्ᴂ इस शहर मᴂ आगामी चुनावⴂ मᴂ पूरी तरह से भाग लेने की अनुमत देगा। नीचे हम एक सूची और मतदान कᴂ द्र केे क्षत्रⴂ का एक न啍शा िदखाते हℂ जहाँ हम ऐसे 핍यिक्तयⴂ को पजीकरणं करवाने मᴂ सहायता प्रदान करे रह ह।ℂ हम इस िनवा셍चन क्षेत्रᴂ म आगामी चुनाव मᴂ खड़े होने वाले सभी उ륍मीदवारⴂ, साथ ही साथ इस क्षेत्रᴂ म बठेै सांसदⴂ, िवधायकⴂ और नगर िनगम के पाष셍दⴂ को यह पत्र भेजे रह ह।ℂ हम आपको सुचत करना चाहते हℂ िक ये नागिरक अब आपके िनवा셍चन क्षेत्रᴂ म मतदान करने के िलए पजीकं ृ त हℂ एवं हमᴂ पूण셍 िवश्वास है िक ये नागिरक मतदान के अपने लोकतांित्रक अधकार का प्रयोगगे। करᴂ
Polling booth names (English):
• Polling booth names (Hindi):
•
Figure A6: Example of typed letter mailed to local politicians in the lead up to the 2019 elections in T2 treated clusters. For confidentiality, identifying content is redacted and the referenced map is omitted.
1
17 5/10/2020 Gmail - West Delhi constituency
M Gmail @gmail.com>
Wed, May 1, 2019 at 6:48 AM To:
Dear ,
Our NGO, , has been working hard over the past year to help migrants in parts of your constituency to become politically empowered. In particular, in the run up to the Lok Sabha elections, we have been running a voter registration drive in the city, helping thousands of migrants from other parts of India to register to vote here for the first time. This will allow them to participate fully in the upcoming elections in this city. Below we show a list and a map of the polling booth areas where we've been working to help migrants register. We're sending this message to all candidates standing in the forthcoming election in this constituency, as well as to sitting MPs, MLAs, and municipal corporators in this area. We hope these newly empowered migrant citizens will exercise their democratic right to vote here. We wanted to let you know that these citizens are now registered to vote in your constituency!
Warm regards,
mm � �m, , � � "ffic1" � � � qRt illfcn 3TI%f.i ci fo #1 &fl � $ � "fri / >R�TI �c,q 31m� Rfi;q1·101 cfiTx ..n fa¢ � � "ff�Tm ffi# � m �I ��� � �' � � #, m �� # l-JciC:lcil tj0fl¢xo1 � 'qC1f �t, � � $ 3F Polling booth names (English): • • • • Polling booth names (Hindi): • • • Figure A7: Example of email sent to local politicians in the lead up to the 2019 elections in T2 treated https://mail.google.com/mail/u/2?ik=991 bc3e 7 c3&view=pt&search=all&permthid=thread-f%3A1632337652336252907 &simpl=msg-f%3A16323376523... 1/5 clusters. For confidentiality, identifying content is redacted and the referenced map is omitted. Figure A8: Example of WhatsApp message sent to local politicians in the lead up to the 2019 elections in T2 treated clusters. For confidentiality, identifying content is redacted. 18 87 polling stations; respondents assigned to nearest polling station Delhi Lucknow (42 polling (45 polling stations) stations) Number of Number of Number of Number of respondents respondents respondents respondents in polling in polling in polling in polling station below station above station below station above city median city median city median city median Treatment Control Treatment Control Treatment Control Treatment Control Block 1 Block 2 Block 3 Block 4 Figure A9: Flow diagram of T2 randomization blocks. Clusters (polling stations) are assigned to T2 treatment or control within four blocks. 19 F Study timeline 300 200 T1 intervention T2 intervention Number of cases 100 Delhi elections Lucknow elections Lucknow 0 Jul 19 Oct 18 Nov 18 Dec 18 Jan 19 Feb 19 Mar 19 Apr 19 May 19 Jun 19 Aug 19 Baseline Endline T1 Form Accepted/Rejected Figure A10: Study phasing. Red and green bars show the timing of the surveys. Blue bars show the timing of final voter registration application decisions, as reported in online administrative data. 20 G Indexing Table A16: Index construction. Z-score indexes are constructed by coding component variables such that higher values are more beneficial. Component variables are then centered and standardized using the control group mean. The final index is then the average of the standardized components. Variable type Indexed variable label Component variable labels Method of indexing Outcome Political interest Interest in politics at the city Z-score index level (ordinal); Interest in politics at the national level (ordinal) Outcome Political trust Trust in national government Z-score index (ordinal); Trust in state government (ordinal); Trust in municipal corporation (ordinal); Trust in parties (ordinal) Outcome Contacting city officials Contacting officials (categorical) Sum Outcome Non-electoral participation Non-electoral participation Sum (categorical) Outcome Campaign exposure Basti visits by politicians Z-score index (integer); Home visit by politician or party worker (integer); Number of gifts (integer); Migrant-focused campaigning (binary); Perceived campaign intensity (ordinal) Lagged DV Political trust Trust in national government Z-score index (ordinal); Trust in state government (ordinal); Trust in municipal corporation (ordinal) Lagged DV Politician visits Politician visits to basti, by Sum municipal corporator, MLA, and MP 21 H Outliers in ‘income‘ covariate When cleaning the data we observed several extreme outliers in the income covariate. To minimize the influence of these extreme values, we winsorize the variable, setting all values above the value of the 99th percentile to the value of the 99th percentile itself. The transformed variable is used in all statistical analyses. 150 100 50 Income (000s INR) 0 Original Winsorized Figure A11: Box plots show the distribution of the raw income covariate before and after winsorizing. 22 I Internal validity I.1 Balance Table A17: T1 balance test for subjects included in T1 analyses. OLS regression. All covariates are measured at baseline. Robust standard errors in parentheses. Dependent variable: T1 treatment indicator Female 0.016 (0.025) Age −0.001 (0.001) Muslim 0.017 (0.028) SC/ST 0.001 (0.025) Primary education −0.010 (0.026) Income (INR 000s) 0.002 (0.002) Married −0.038 (0.029) Length of residence in city 0.002∗ (0.001) Owns home in city 0.058∗∗ (0.025) Hadn’t voted previously 0.044 (0.041) How likely to vote in city if registered −0.010 (0.060) Political interest −0.040 (0.041) Sense of political efficacy −0.031 (0.031) Political trust index 0.004 (0.015) Shared meal with non-coethnic −0.012 (0.034) Has hometown voter ID 0.013 (0.037) Returned to vote in hometown 0.047 (0.042) More at home in hometown 0.013 (0.034) Straight-line distance to home district 0.00002 (0.00003) Still receives hometown schemes −0.006 (0.024) Owns hometown property 0.001 (0.026) Pr(>F) of H0: joint orthogonality 0.408 Observations 2,120 Adjusted R2 0.0004 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 23 Table A18: T2 balance test for subjects included in T2 analyses. Weighted least squares regression. Clusters weighted equally. Model includes block fixed effects. All covariates are measured at baseline. Cluster-robust standard errors in parentheses. Dependent variable: T2 treatment indicator Politician visits −0.008 (0.035) Female 0.018 (0.039) Age 0.001 (0.002) Muslim 0.121 (0.078) SC/ST 0.003 (0.056) Primary education 0.001 (0.042) Income (INR 000s) 0.011∗∗ (0.004) Married 0.007 (0.049) Length of residence in city 0.002 (0.002) Owns home in city 0.004 (0.058) Pr(>F) of H0: joint orthogonality 0.406 No. of clusters 87 Observations 1,969 Adjusted R2 0.017 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 24 I.2 Attrition rates by treatment condition Table A19: Comparison of attrition rates across T1 treatment arms using OLS regression. The analysis includes all subjects randomized to T1 treatment or control. Robust standard errors in parentheses. Dependent variable: Attrition Indicator Assigned to T1 treatment −0.007 (0.011) Observations 2,306 Adjusted R2 −0.0003 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Table A20: Comparison of attrition rates across T2 treatment arms using weighted least squares regression. Clusters are weighted equally. Models include block fixed effects. The analysis includes all subjects randomized to T2 treatment or control. (Note that this number is smaller than that for the T1 attrition analysis as baseline geo-coordinates were unavailable for some subjects; hence they were not assigned to clusters or randomized.) Cluster-robust standard errors in parentheses. Dependent variable: Attrition Indicator Assigned to T2 treatment −0.021 (0.021) No. of clusters 87 Observations 2,131 Adjusted R2 0.003 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 25 J Administrative data on T1 application processing Because our field team helped applicants to submit the voter registration forms online, we were able to track the progress of each submitted case. Specifically, the Election Commission’s online portal provides the dates that applications were received and assigned to a Booth Level Officer (BLO). It also gives the final date when the application was either accepted or rejected. (No reasons are given for rejection.) In Figure A12, we present histograms of the time lapse for each of these various stages, as well as the median times. 150 150 100 100 50 50 0 0 0 50 100 150 200 0 50 100 150 200 (a) Time to decision (b) Time to appoint BLO 150 150 100 100 50 50 0 0 0 50 100 150 200 0 50 100 150 200 (c) Time to success (d) Time to reject Figure A12: This figure displays the distributions of the time elapsed between the application submission dates and various stages in the application processing, as reported in the online portal of the Election Commission of India. The sample is for experimental subjects assigned to T1 who submitted applications. Red vertical lines display the median value. Note, cases that took longer than 200 days to process are excluded. 26 K Deviations from pre-analysis plan The study pre-analysis plan (PAP) was filed at the Evidence in Governance and Politics registry before the researchers accessed the endline data and prior to any data analysis being conducted. There were two deviations from the pre-registered plan. First, the observational data analyses described on pp. 6–7 of the PAP were not implemented as there was insufficient variation in the outcome variable of interest (whether eligible subjects wished to acquire a city-based voter ID card) to make these tests feasible. This substantive finding is highlighted in the paper. Second, in the endline data, there were 38 missing values for the variable, e_campaign5_exposure_hardwork. This is one of five variables used in the construction of the index variable, e_campaign_exposure. For the purposes of constructing this index variable, the 38 missing values of e_campaign5_exposure_hardwork were imputed by sampling at random from the non-missing values of that variable within the cluster (polling station) to which the respondent belonged. All remaining pre-registered experimental analyses in this paper are implemented in conformity with the PAP. L Additional ethical considerations In this section, we detail features of our experimental design and implementation that went above and beyond the minimal requirements needed to secure Institutional Review Board (IRB) approval. We note that exceeding IRB requirements has long been advocated by ethicists in the discipline (e.g. Tolleson-Rinehart 2008; Yanow and Schwartz-Shea 2008; Teele 2014; Fujii 2012), and indeed, we heeded the general principles outlined in existing guidelines when designing our research study. Below, we outline additional ways in which we sought to conduct our study to the highest ethical standards, highlighting several pieces of scholarship that influenced our research design and fieldwork. First, there is an emerging consensus among political scientists that experiments fielded in the midst of an election should be designed to minimize the possibility of swaying aggregate electoral outcomes (e.g. which candidate or party ultimately wins a seat; see Desposato 2014). The American Political Science Association’s “Principles and Guidance for Human Subjects Research” notes that, “in general, political science researchers should not compromise the integrity of political processes for research purposes without the consent of indi- viduals that are directly engaged by the research process” (13). In an important recent contribution, Slough (2020) develops a formal model that yields the following recommendations for researchers seeking to abide by this principle. These include: keeping the number of experimental subjects—i.e. voters—comparatively small; avoiding races whose margin of victory is expected to be close; fielding experiments in larger districts where single-member simple plurality rules are in force; and crafting interventions that reduce the chances of spillovers (Slough 2020: 32-3). As noted in the main text, our interventions were situated and implemented to make it highly improbable that any overall electoral results would be affected, in line with these proposals. Second, a number of prominent ethical guidance documents underscore the importance of partnering with local stakeholders. In particular, the American Political Science Association’s “Principles and Guidance for Human Subjects Research” recommends that “researchers should consider the broader social impacts of the research process when deciding whether to engage in the partnership” (13). Humphreys (2015: 89) notes that “the key idea is that if an intervention is ethical for implementing agencies with respect to the ethical standards of their sphere—which may differ from the ethical standards of researchers—then responsibility may be divided between researchers and implementers, with research ethics standards applied to research components and partner standards applied to manipulations.” As described in the paper, we partnered with a local NGO that had been working to promote city-based voter registration among urban migrants for several years. This NGO took the lead role in designing and piloting the interventions, engaging with community leaders, offering contextual input on the design of the study, and providing training assistance and materials to help bring the intervention to scale. Training materials, along with field protocols, were reviewed and approved by the researchers to ensure adherence to ethical and scientific best practices. We further relied on the services of a locally-based research firm that had extensive experience working in the cities where the interventions occurred. It is important to emphasize that our NGO partner, as well as a 27 host of other similar organizations (many of which we consulted in the design of our study), was already conducting similar interventions and would have continued to conduct similar interventions with or without our collaboration. In short, our role as academics in the research partnership was to scientifically evaluate an intervention that was already in existence. Third, Asiedu et al (2021) call attention to what they term “policy equipoise,” based on the clinical principle that “the expert community must not have certainty that any arm in a trial is better therapeutically than any other arm.” They note that in settings where there is not uncertainty about which arm will achieve normatively desirable results compared to another arm, randomization of an intervention may still be ethical on certain grounds, such as scarcity of intervention supply. We approached our experimental design with these principles in mind. There was considerable uncertainty regarding the efficacy of both interventions. Additionally, given the resource-intensive nature of the voter registration intervention as well as the budget constraints facing us and our partners, in no sense were we withholding registration resources that would otherwise have been allocated to subjects in the comparison group. Fourth, we were closely attuned to the treatment of our field teams as well as the imperative to engage in “knowledge transfers” to local populations (cf. Teele 2014). Based on extensive scoping and piloting, we established that the communities where we were working were substantially safe environments in which to carry out research, entailing no appreciable risk to field workers. On knowledge and skills transfers, we were able to enlist the support of research assistance from graduate students at a local university; the students used fieldwork on the project to fulfill one of the requirements for their masters degree program. Field teams were overseen by a professional project manager and compensated at standard local rates for such fieldwork. Finally, our research was designed to be policy-relevant. Highlighting the growing policy impact of much social scientific work, Tolleson-Rinehart (2008: 507-8) notes that “how the knowledge of such improvements can be shared . . . is increasingly urgent.” To boost the overall benefits of experimental research, the broad and proactive dissemination of research findings is critical. For example, Cronin-Furman and Lake (2018: 612) encourage researchers to ask: “Have you made a plan to ensure that your research results are disseminated back to the affected community in ways that are meaningful or valuable to them?” In keeping with this recommendation, we will publicize the results of our study to a wide policy audience— including NGO actors working on migrant advocacy, in addition to bureaucrats at the front lines of electoral administration, for whom our study may hold lessons. Works referenced: • American Political Science Association Council’s “Principles and Guidance for Human Subjects Re- search.” bit.ly/3bId74R. Accessed 3/13/2021. • Asiedu, E., Karlan, D., Lambon-Quayefio, M., and Udry, C. 2021. “A Call for Structured Ethics Appendices in Social Science Papers.” National Bureau of Economic Research Working Paper 28393. bit.ly/38Eh74J. • Cronin-Furman, K. and Lake, M. 2018. “Ethics Abroad: Fieldwork in Fragile and Violent Contexts.” PS: Political Science and Politics 51(3): 607-14. • Desposato, Scott. 2014. “Ethical Challenges and Some Solutions for Field Experiments.” Mimeo, University of California, San Diego. bit.ly/3lbtzOq. • Fujii, L. 2012. “Research Ethics 101: Dilemmas and Responsibilities.” PS: Political Science and Politics 45(4): 717-23. • Humphreys, M. 2015. “Reflections on the Ethics of Social Experimentation.” Journal of Globalization and Development 6(1): 87-112. • Slough, T. 2020. “The Ethics of Electoral Experimentation: Design-Based Recommendations.” Mimeo, Columbia University. bit.ly/3bDfiqy. • Tolleson-Rinehart, Sue. 2008. “A Collision of Noble Goals: Protecting Human Subjects, Improving Health Care, and a Research Agenda for Political Science.” PS: Political Science and Politics 41(3): 507-11. • Teele, D. 2014. “Reflections on the Ethics of Field Experiments.” In D. Teele (ed.), Field experiments and their Critics: Essays on the Uses and Abuses of Experimentation in the Social Sciences, Yale University Press, 115-40. 28 • Wood, E. 2007. “Field Research.” In C. Boix and S. Stokes (eds.), The Oxford Handbook of Compar- ative Politics, Oxford University Press: 123-46. • Yanow, D. and Schwartz-Shea, P. 2008. “Reforming Institutional Review Board Policy: Issues in Implementation and Field Research.” PS: Political Science and Politics 41(3): 483-94. 29 Overcoming the Political Exclusion of Migrants: Theory and Experimental Evidence from India Nikhar Gaikwad, Columbia University Gareth Nellis, University of California, San Diego SUPPLEMENTARY INFORMATION 1 Supplementary Information Contents A Voter registration procedures worldwide 3 B Internal migration and political participation worldwide: Additional case-study evidence 9 C Voter registration field experiments: A systematic review 10 D Description of IHDS-II variables 20 E Additional information on experimental subject characteristics 23 F Summary statistics 24 G Survey instruments for experiment 26 H T1 additional results 30 H.1 T1 interaction with hometown attachment characteristics ...... 30 H.2 T1 effects on political interest decomposed ...... 31 H.3 T1 effects on contacting and non-electoral participation ...... 32 H.4 Heterogeneous effects of T1 by Hindi fluency ...... 33 H.5 T1 main results controlling for Hindi fluency ...... 34 H.6 T1 main results without covariates ...... 35 I Vote choice 36 J T2 additional results 37 J.1 T2 effects on trust and integration ...... 37 J.2 T2 results without covariates ...... 38 K Migrant voting behavior over time 39 L Research locations 40 2 A Voter registration procedures worldwide Table SI1: Description of voter registration rules in the 20 most populous low- and middle-income democracies. We class countries as democratic if their Polity IV score was 6 or greater in 2018. Per the World Bank, low- and middle-income countries are those whose GDP per capita was less than USD 12,375 in 2018. Population and income data are from the World Bank. Information on registration procedures were gathered from official government sources and country experts. Country Population GDP Polity Voter registration procedure Post-migration procedure (2018) per Score capita (2018) (USD, 2018) India 1,352,617,328 2,016 9 Voter-initiated. Citizen must register by De-registering required. Migrant must submitting Form 6 to local election office submit Form 7 (deletion) at voter office and receive at-home verification visit. ID in constituency of origin, and produce is dispatched to home address by post. deletion slip at destination constituency prior to re-registering. 3 Indonesia 267,663,435 3,894 9 State-initiated. Citizens with electronic No formal requirements for ID cards (e-KTP, which records their de-registration. address) are automatically registered to vote and placed on the “temporary” voting roll. Individuals without an e-KTP can register online or in person at their designated voter office. Pakistan 212,215,030 1,473 7 Voter-initiated. Citizen must register by Need to de-register from electoral rolls at filling Form 21, and presenting their constituency of origin before National ID Card as proof of residence re-registering at destination and identity. constituency. Table SI1: (continued) Description of voter registration rules in the 20 most populous low- and middle-income democracies. Country Population GDP Polity Voter registration procedure Post-migration procedure (2018) per Score capita (2018) (USD, 2018) Brazil 209,469,333 8,921 8 Voter-initiated and compulsory. Citizens No requirement to de-register from must register at the local voter office. electoral rolls at place of origin. Need to Eligible voters who do not register by 19 re-register when moving between states years of age, or newly naturalized as well as municipalities. persons who do not register within a year of acquiring citizenship, face a penalty of 3.51 reals (payable at time of voter registration), unless the person was away from their voting district on election day and justifies his/her absence to electoral justice. Registrants receive a bar-coded elector’s card (título eleitoral) 4 proving their registration within at least 15 days of the request. There are other non-pecuniary penalties for unregistered citizens like ineligibility for passports, public sector jobs, and loans from public sector banks. Process is biometric. Nigeria 195,874,740 2,028 7 Voter-initiated. Citizen must report to No need to de-register. The PVC is the local election office to register for a biometric. Citizen must write to the biometric Permanent Voter Card (PVC) Resident Electoral Commissioner (REC) during the Continuous Voter through the Electoral Officer of the Registration period. ID is then current constituency at least 60 days distributed by the election office. prior to elections. If the REC is satisfied that applicant currently resides in the area, they will approve the application and direct that the applicant’s details be transferred to the new location. The transfer is recorded on the centralized computer system and the applicant receives their new voter card. Table SI1: (continued) Description of voter registration rules in the 20 most populous low- and middle-income democracies. Country Population GDP Polity Voter registration procedure Post-migration procedure (2018) per Score capita (2018) (USD, 2018) Mexico 126,190,788 9,698 8 Voter-initiated. Individuals must register No requirement to de-register. in person at their local election office. Re-register at destination constituency Applicants provide a signature, by surrendering the previous voter ID. fingerprint, and photograph to obtain a “Voter’s Mexican Credential.” Photo voting cards are delivered to citizens 20 days after application submissions, or can be collected from the voter office personally. They must be renewed every 10 years. Philippines 106,651,922 3,103 8 Voter-initiated. The applicant personally No need to de-register. At the appears before the Election Officer (EO), destination constituency, EO will issue 5 states his/her name and exact address. migrant an application to re-register, After establishing the identity of the upon approval of which, EO in applicant, the EO verifies the name of constituency of origin is instructed to the applicant from the Local Voter’s delete name from its electoral rolls. Registration Database or in the Printed Lists of Voters. South 57,779,622 6,374 9 Voter-initiated. Registrants must present No requirement to de-register. Africa either a South African bar-coded ID Re-registration in new constituency book or a valid Temporary Identity needed to update status. Certificate in order to register at the local voter office/polling station. No voter registration card is issued; rather, the document provided for registration is marked and becomes proof of registration. Myanmar 53,708,395 1,326 8 Voter-initiated. Citizen must report to Voter must inform election authorities the voter office in designated before moving to new constituency. constituency. Applicants primarily identify themselves with a National Identification Card, though other forms of identification may apply. Table SI1: (continued) Description of voter registration rules in the 20 most populous low- and middle-income democracies. Country Population GDP Polity Voter registration procedure Post-migration procedure (2018) per Score capita (2018) (USD, 2018) Kenya 51,393,010 1,711 9 Voter-initiated. Eligible voters must Citizen must transfer their registration present themselves to the registration (linked to their National ID card) when officer with their original identification moving between constituencies. documents at the designated registration center and complete the registration form (Form A). Registered voters are issued a registration acknowledgement slip bearing the voter’s details. The National ID card is the only document required to prove identity. Colombia 49,648,685 6,651 7 Voter-initiated. To register, eligible De-registering at constituency of origin voters must present their national is not required; migrant only needs to 6 identity card and have their fingerprints re-register at destination constituency. taken by the National Civil Registry. The process is often described as registering the ID card with the Electoral Registrar. Argentina 44,494,502 11,653 9 State-initiated. All Argentine citizens No need to de-register, updating address with an ID book over the age of 18 are on the national ID card leads to automatically enrolled in the electoral automatic re-registration in destination register, known as the Padrón Electoral; constituency. therefore, they do not need to initiate registration. A complex fee scheme applies for new cards, renewals, and data verifications/updates. Iraq 38,433,600 5,878 6 Voter-initiated. Citizen must produce Special polling booths are set up in relevant documents and register for a destination constituency for Internally voter ID at the voter office. Process is Displaced Persons who are not biometric. biometrically registered. Table SI1: (continued) Description of voter registration rules in the 20 most populous low- and middle-income democracies. Country Population GDP Polity Voter registration procedure Post-migration procedure (2018) per Score capita (2018) (USD, 2018) Peru 31,989,256 6,947 9 State-initiated. Voter registration list is No need to de-register, updating address based on the civil registry. All citizens on the national ID card leads to registered in the civil registry are automatic re-registration in destination automatically included in the voter constituency. registry once they turn 18. The National Registry of Identification and Civil Status is responsible for updating the registry. Registration is free of charge. Malaysia 31,528,585 11,239 7 Voter-initiated. Citizen must fill a form The Electoral Center (EC) accepts at the designated local voter office and applications from registered voters who present MyKad (national ID card) to apply to register their new home register as a voter. The officer fills out addresses to determine their new Voting 7 the registration form on behalf of Center. No requirement to de-register. applicant, and the citizen must verify the data. Ghana 29,767,108 2,202 8 Voter-initiated. Citizen must register in Citizen must notify the Electoral the divisional register of the electoral Commission in case of constituency area in which they ordinarily reside. changes. The transfer is then recorded Successful registrants are issued a on the central computer system. No biometric voter card at the time of need to de-register before re-registering. registration. Nepal 28,087,871 1,026 7 Voter-initiated. Citizen must report to No need to de-register before the local voter office with their re-registering at destination constituency citizenship certificate to get registered as as system is biometric. a voter. The process is biometric. Madagascar 26,262,368 461 6 Voter-initiated. Citizen must present No information available. their national ID card at the local voter office to register as a voter. Table SI1: (continued) Description of voter registration rules in the 20 most populous low- and middle-income democracies. Country Population GDP Polity Voter registration procedure Post-migration procedure (2018) per Score capita (2018) (USD, 2018) Sri Lanka 21,670,000 4,102 6 State-initiated, but not compulsory. Since lists are revised annually, there is Registration of electors and revision of no requirement to de-register. electoral registers are done annually on June 1. Enumerator appointed by the Registering Officer of the district provides the Registration form (BC form) to the chief occupant of each house. The filled registration form (BC form) is collected by the enumerator. Registration of a voter is valid for one year only. Burkina 19,751,535 731 6 Voter-initiated. Registrants must present No information available. 8 Faso a passport, national ID card, or military card. Successful registrants are issued a voter registration card. Process is biometric. B Internal migration and political participation worldwide: Ad- ditional case-study evidence Qualitative evidence corroborates the claim that internal migrants tend to participate less, politically, than non-migrants in a variety of settings worldwide. • Nigeria: Akinyemi, A., Olaopa, O., and Oloruntimehi, O. 2005. “Migration Dynamics and Changing Rural-Urban Linkages in Nigeria.” Mimeo: Obafemi Awolowo University. bit.ly/3wTi1Vr. – This study details low rates of political participation among migrants in Oyo, Ondo, and Ogun states: “Migrants’ participation in political activities showed that 30% of males vs 15% of females among migrants participate and belong actively to a political party” (p. 11). • Ukraine: National Democratic Institute. 2019. “Statement of the NDI Election Observation Mission to Ukraine’s April 21, 2019 Second Round Presidential Election.” bit.ly/2WVgPzt. – This report notes: “Labor migrants . . . face particular barriers to voting. Internal migrants and IDPs must apply each election—including both rounds of the presidential election—to change their place of voting. Citizens residing in non-government-controlled areas must cross the ‘line of contact’ multiple times to vote in a government-controlled location, creating both a financial and physical burden. The process is cumbersome and poorly understood. OPORA [Ukrainian civic network] and others have recommended an online system to ease the process for all types of internal migrants. Some 325,604 citizens, just a small fraction of Ukraine’s internal migrants and IDPs, changed their place of voting for the second round” (p. 7). • Colombia: Rozo, S., and Vargas, J. 2018. “Brothers or Invaders? How Crisis-Driven Migrants Shape Voting Behavior.” The Latin American and Caribbean Economic Association 16836. bit.ly/2LIOqea. – This paper notes that “internally displaced populations in Colombia, while entitled to vote, do not do so in practice. This is both because most internal migrants are below the voting age . . . and because many of the adults lack formal identification documents which are required for voting” (p. 16). • Myanmar: Callahan, M., and Oo, M. 2019. “Myanmar’s 2020 Elections and Conflict Dynamics.” United States Institute of Peace. Peaceworks No. 146. bit.ly/2Z11zUI. – This study notes that “for voter list inclusion, much hinges on the possession of an up-to-date household list. If a voter checks the voter list display and finds her name missing, she has to produce her household list or other form of identification to demonstrate residence in that particular constituency. If—as is the case for Myanmar’s many internal migrants—she is on a household list in another constituency, she must return to the local administrative office in that location, apply to remove her name from the household list there, and then return to her present constituency area to apply to have her name placed on the household list there. Once that requirement has been satisfied, the GAD [General Administration Department] at the ward or village level will transmit the information to the election commission to add her name to the voter list. This is an onerous and—for some—expensive process that disincentivizes self-updating of the voter list. Election observers noted that separate from this process, village development committees could confirm the residence and identity of potential voters to add their names to the voter list. As the EU Observation Mission noted, this ‘trust-based approach’ did not serve internally displaced people well, as they are living far from their villages” (pp. 26-27). 9 C Voter registration field experiments: A systematic review We systematically review the body of published field experimental studies on the impacts of voter registration assistance. We note that no published study to date has focused on the specific challenges that migrants encounter in seeking to enroll. Additionally, few studies probe the downstream effects of voter registration on other indicators of political incorporation (yet see Braconnier et al 2017 for an important exception). Studies focusing on general-population voters have documented mixed effects of voter registration assistance, with treatment impacts varying according to the mode of assistance offered. Compare, for example, Bennion and Nickerson 2010, Bennion and Nickerson 2014, and Nickerson 2007, which identify null and even negative effects of some interventions, with Braconnier et al 2017, Gerber et al 2014, and Nickerson 2015 which identify positive effects. Only two studies focus on registration assistance in developing-country contexts. While Harris and van der Windt Forthcoming finds statistically significant positive treatment effects of assistance given in Kenya, Mvukiyehe and Samii 2017 finds no evidence of an impact of town hall meetings or a civic education campaign to promote registration on either registration or turnout in Liberia. Finally, we note that there is suggestive evidence in existing work that non-movers and movers respond differentially to registration assistance. Notably, Braconnier et al 2017 finds an overall treatment effect of +0.052 percentage points, p<0.01, for home registration visits; yet compared to registered voters as a whole, subjects who registered because of the treatment were more likely to be immigrants (+0.202 percentage points, p<0.01) and to have been born in another region of France (+0.215 percentage points, p<0.01). Additionally, Gosnell 1926 finds that the treatment effect on registration was greater among the sub-sample of subjects who had lived in that voting precinct for less than 10 years. 10 Table SI2: A systematic review of published field experimental studies of the effects of voter regis- tration assistance. Study Sample Treatments Outcomes Results Heterogeneity Bennion, E. A., and U.S., 2006. First First experiment: Names of recipients First experiment: Email Nickerson, D. W. experiment: 259,310 Treatment group of matched to voter on downloadable 2010. “The Cost of undergraduate students at students received emails registration records, registration forms Convenience.” Political public universities during with information on identified by name, age, decreased the likelihood of Research Quarterly the 2006 midterm election downloadable voter and address. registering to vote (-0.3 64(4): 858–69. campaign. registration forms. percentage points across treatment groups, p=0.09). bit.ly/36mUdfY Second experiment: 6,372 Second experiment: participants in the US who Treatment group received Second experiment: downloaded voter texts reminding the user to Participants receiving registration forms from an submit their forms. reminder text messages NGO and opted in to were more likely to register receiving text messages (+4.0 percentage points, from the NGO. p<0.01). Bennion, E. A. and U.S., 2010. 7,366 students One treatment group Names of recipients Among the full sample, Nickerson, D. W. at a U.S. public university received an email linking matched to voter there was no significant 2014. “Cheap, But in the months before the to the state’s fully online registration records, effect on voter registrations Still Not Effective: An registration deadline for voter registration system, identified by name, age, in the downloadable form 11 Experiment Showing the 2010 congressional while another group and address. group (-0.009, se=0.01) or that Indiana’s Online elections. received a link to a in the fully online Registration System downloadable, mail-in registration group (0.016, Fails to Make Email registration form. Control se=0.01). Among students an Effective Way to group received no email who were not previously Register New Voters.” communication. registered to vote, the Indiana Journal of downloadable form group Political Science 14: saw a statistically 39-51. significant decrease in registration (-0.049, bit.ly/2UgnZ0h se=0.017) while the online link group saw no significant effect (-0.008, se=0.017). Table SI2: (continued) A systematic review of published field experimental studies of the effects of voter registration assistance. Study Sample Treatments Outcomes Results Heterogeneity Bennion, E. A. and U.S., 2006. 25,256 public Classes in one treatment Names of recipients In-class presentations and Nickerson, D. W. university students group received a short matched to voter form distribution increased 2016. “I Will Register sampled across 1,026 presentation by the registration records, student registration rates and Vote, If You courses. professor on the identified by name, age, (+5.6 percentage points, Teach Me How: A importance of registering and address. p<0.01). Both treatments Field Experiment to vote, followed by the also increased voter Testing Voter professor distributing turnout (2.3 and 2.9 Registration in College registration cards to percentage points Classrooms.” PS: interested students and respectively for professor Political Science & then collecting them. In treatment and student Politics 49(04): the second treatment volunteer treatment; 867–71. group, a student volunteer p<0.01). No statistically conducted the presentation significant difference exists bit.ly/2U9zyGL and distribution of forms. between the effects of the Control classrooms did not professor and student receive any outreach volunteer treatments. efforts. Braconnier, C., France, 2012 elections. Voter registration Number of new Home registration visits Compared to registered 12 Dormagen, J., and 20,500 households across canvassing. Different registrations/voters in 2011 close to the date of the voters as a whole, subjects Pons, V. 2017. “Voter 10 cities, effect then treatment groups received with an address in each election had the greatest who registered because of Registration Costs and studied as effect of information-only visits; apartment building. effect on increasing the treatment were more Disenfranchisement: registrations among those information and home Follow-up survey measured registrations (+0.052, likely to be immigrants Experimental initially unregistered or registration visits; or two interest in and knowledge p<0.01). All but one of the (+0.202, p<0.01); born in Evidence from misregistered. 1,500 separate visits. Treatment about national politics. canvassing treatments had another region of France France.” American households resampled for groups also varied in the a statistically significant (+0.215, p<0.01); or young Political Science post-election survey. timing of visits relative to effect size. 93% of (age coefficient -0.137, Review 111(3): the election. compliers (citizens p<0.01). 584–604. registered from the treatment visits) voted at Voters registered from the bit.ly/35ikWvf least once in 2012, just as treatments had higher likely as new control turnout in the more salient registrants. Treatment presidential election. increased index of "political interest” by 0.06 standard deviations (p<0.05). Table SI2: (continued) A systematic review of published field experimental studies of the effects of voter registration assistance. Study Sample Treatments Outcomes Results Heterogeneity Gerber, A. S., Huber U.S., 2012 in the months Subjects in the treatment Names of recipients The pooled treatments had Among subjects who had G. A., Meredith, M., before the 2012 general groups received a letter matched to voter significant positive effects previously voted in 2008, Biggers, D. R., and election. 6,280 eligible but from the state election registration and turnout on registration (+0.03, the treatment had a Hendry, D. J. 2014. unregistered formerly authority assuring them records, identified by p<.01) and voter turnout greater effect on “Can Incarcerated incarcerated felons. that they were eligible to name, age, and address. (+0.015, p<.05) in 2012. registration (+0.116, Felons Be vote. p<0.01) and voter turnout (Re)Integrated into (+0.106, p<0.01). the Political System? Results from a Field Experiment.” American Journal of Political Science 59(4): 912–26. bit.ly/3kc0no7 Gertzog, I. N. 1970. U.S., 1966 in the months In treatment precincts, the Change from previous Party registration “The Electoral before the midterm local Democratic Party election in the number of canvassing increased the Consequences of a election. Unit of analysis: conducted door-to-door voters in each precinct who number of Democratic 13 Local Party 12 electoral precincts in voter registration were registered to vote as voter registrations per Organizations San Diego, California. assistance in the weeks Democrats on the voting precinct (ATE estimated Registration before the voter rolls. Vote choice as as +12.9). New registrants Campaign: The San registration deadline. self-reported by subjects in treatment precincts Diego Experiment.” Control precincts had no during a post-election self-reported voting in the Polity 3(2): 247–64. local party contact. survey. majority for the Democratic gubernatorial bit.ly/3mZWxAu candidate, but less than new registrants in control precincts (66% Democratic party vote for treatment group, compared to 76% for control). Table SI2: (continued) A systematic review of published field experimental studies of the effects of voter registration assistance. Study Sample Treatments Outcomes Results Heterogeneity Gosnell, H. F. 1926. U.S., 1924. 6,000 Chicago Treatment group received Names of recipients Treatment increased The treatment effect on “An Experiment in residents randomly selected a mailed notice informing matched to voter registration rate by +10 registration was greater the Stimulation of within the same voting them of the need to registration and turnout percentage points (75% to among the sub-sample who Voting.” American precincts. register in order to vote in records. 65% in control). In a had lived in that voting Political Science the presidential election, follow-up experiment precinct for less than 10 Review 20(4): 869–74. with a follow up notice conducted before the 1925 years (+13 percentage sent to those who had not local election, treatment points). bit.ly/38pbwzJ yet registered after a subjects who registered the certain period. previous fall (approximately 2,250) were mailed another notice criticizing voting abstention. Compared to registered voters in the control group, treatment registrants were +10 percentage points more likely to vote in the local 14 elections (57% to 47%). Table SI2: (continued) A systematic review of published field experimental studies of the effects of voter registration assistance. Study Sample Treatments Outcomes Results Heterogeneity Harris, J., and van der Kenya, 2016 at 1,674 One control ("status quo" Number of new Localization treatment has Effect size of localization Windt, P. polling stations. registration policy) and 5 registrations per polling a significant positive effect treatment greater in poorer Forthcoming. Intervention took place treatment groups: station as a proportion of on voter registration (+2 areas (poorest quintile, “Equalizing Access to eight months before the Localization (election 2013 registrations at the percent of the 2013 +4.39 percent, p<0.01; Improve Voter 2017 general election. commission offered local end of the intervention registered voter total, richest quintile, +0.73 Registration: registration at site); period and on election day; p<0.01) by the end of the percent, p<0.05). Poor Experimental canvassing (election change in turnout as a intervention period. SMS areas, rural areas, and Evidence from Kenya.” commission staff visited proportion of 2013 turnout. reminder messages only areas far from the nearest Journal of Politics. households to encourage have an effect when registration office saw the voting and provide combined with localization greatest effect sizes of bit.ly/32ujBj0 information about (combined treatment effect localization. registration); and SMS +2.4 percent, p<0.01). (messages to registered The canvassing program voters urging them to had little to no effect on encourage their registration (+0.1 percent, unregistered contacts to se=0.003). register). The final two treatment groups combined Localization treatment 15 localization with increased the absolute canvassing or SMS. number of votes cast (by 0.04 standard deviations) but decreased the turnout rate (by 0.03 standard deviations). Localization decreased vote margins in some races (-2.8 percent, p<0.01) relative to control. John, P., Macdonald, U.K., 2012. 129,048 Treatment households Households returning their The lottery offer increased Areas above average in an E., and Sanders, M. households in London in received letters reminding registration form to the the proportion of index of poverty indicators 2015. “Targeting the total sample, of which them to register to vote local government by the households submitting had a larger effect (+2.6 Voter Registration 20,000 were each assigned and informing them that deadline. registration forms by the percentage points relative with Incentives: A to the two treatment registrants would be deadline (+1.5 and +1.9 to control, z=-6.5), while Randomized conditions (larger and entered into a lottery for a percentage points, areas below the average Controlled Trial of a smaller prize) and the rest cash prize (5,000 GBP in p<0.001). No statistically (i.e. wealthier areas) had a Lottery in a London to control. one group, 1,000 GBP in significant difference statistically insignificant Borough.” Electoral another). Control between the two prize effect size (+0.3 points Studies 40: 170–75. households received a amounts. relative to control, z=-0.8). reminder letter but no offer bit.ly/2IlBgSy of a prize. Table SI2: (continued) A systematic review of published field experimental studies of the effects of voter registration assistance. Study Sample Treatments Outcomes Results Heterogeneity Kölle, F., Lane, T., U.K., 2015 general Control group received a Recipients registered Informing students about Nosenzo, D., and election, 7,679 eligible but postcard informing between the beginning of the potential fine increased Starmer, C. 2019. unregistered students in recipients they were not treatment and close of rates of registration “Promoting Voter Oxford. yet registered to vote and registration. Student (logistic regression odds Registration: The providing a link to the university data matched to ratio 1.57, p<0.05). Other Effects of Low-Cost online registration website. official voter registration treatments not significant Interventions on Treatment groups had rolls. at p>0.05. Behaviour and messages added to the Norms.” Behavioural control base, including: Public Policy 4(1): Informing students about 26–49. the possible 80 GBP fine for not registering; offering bit.ly/2IkIeaI students entry into a lottery for small prizes (80 GBP); providing a way to sign up for text message reminders to register; and providing a phone number 16 for students to text their intention to register. Treatment randomized at the student residential building level. Krawczyk, K. and U.S., 2010. 505 clients of NGOs offered treatment Subjects were surveyed NGO voter registration Leroux, K. 2014. "Can different human services subjects voter registration after the election and assistance increased Nonprofit NGOs in the months before assistance during subjects’ asked if they voted. self-reported voter turnout Organizations Increase the 2010 midterm election. normal visits to the NGO. (probit marginal Voter Turnout? Control group received no effect=+0.119, p<0.05). Findings From an voting-related contact from Agency-Based Voter the NGO. Mobilization Experiment." Nonprofit and Voluntary Sector Quarterly 43(2): 272–292. bit.ly/3lfmcoi Table SI2: (continued) A systematic review of published field experimental studies of the effects of voter registration assistance. Study Sample Treatments Outcomes Results Heterogeneity Mann, C. B., and U.S., 2012. Citizens who Treatment group received Names of recipients added All treatments effective at Bryant, L. A. 2020. were eligible to vote but postcards from the state to voter registration and increasing registration “If You Ask, They unregistered. 28,867 election agency stating turnout records, identified (ranging from +1.8 to +2.6 Will Come (to households in Delaware they were not registered to by name, and address. percentage points relative Register and Vote): across all 5 experimental vote and providing to control, all effects Field Experiments conditions; 549,748 information on how to do significant at p<0.01). All with State Election households in Oregon. so. Treatment variations: treatments also increased Agencies on Emphasizing the urgency turnout in the following Encouraging Voter of registering by the election (ranging from Registration.” deadline; visual cues +1.6 to +2.4 percentage Electoral Studies 63: (image of a registration points relative to control, 102021. form); and emphasizing all effects significant at voting as a civic duty. p<0.01). No significant bit.ly/3katEzP differences between different treatments’ effect sizes. Mvukiyehe, E., and Liberia, 2011. 142 villages. In one group of treatment In-person surveying asked Neither treatment had a Samii, C. 2017. Took place in the 9 months villages, residents were community members statistically significant 17 “Promoting preceding the 2011 general invited to community town whether they had effect on registration (0.01 Democracy in Fragile elections. hall and civic education registered or voted in the percentage points, States: Field programs—5 or 6 meetings last election. se=0.01) or voter turnout Experimental over the treatment period. (0.02 percentage points, Evidence from In the second group, se=0.01). Liberia.” World facilitators recruited Development 95: treatment village residents 254–67. for community security committees. The security bit.ly/3lfYJDw committees liaised with UN peacekeepers and monitored security incidents. Table SI2: (continued) A systematic review of published field experimental studies of the effects of voter registration assistance. Study Sample Treatments Outcomes Results Heterogeneity Nickerson, D. W. First experiment: 5 U.S. In both experiments, the Names of recipients The email treatments had 2007. “Does Email public universities during samples received a series of matched to voter no positive effect on Boost Turnout?” the 2002 midterm elections, emails encouraging voting registration records, registration (pooled, Quarterly Journal of 58,311 undergraduate and providing information identified by name, age, treatment effect=-0.4 Political Science 2(4): students in sample. on voter registration. and address. percentage points, se=0.2) 369–79. Students did not opt-in to or turnout (pooled, being part of the study. treatment effect=-0.2 bit.ly/3lgkyCK percentage points, se=0.3). Second experiment: 161,633 U.S. residents who signed up on an NGO’s website to receive registration and voting reminder emails. Nickerson, D. W. U.S., 2004 and 2007, In-person canvassing by Rate of new voter In-person canvassing New registration effect was 2015. “Do Voter several months ahead of local NGOs who helped registrations and new voter increased voter registration higher on low-SES streets Registration Drives different elections. residents fill out and turnout per street in voter (+4.4 percentage points, (+6 percentage points in Increase Treatment randomized by submit registration cards. registration records. p<0.01) and voter turnout low-SES streets compared 18 Participation? For street, 620 streets in six (+0.9 percentage points, to +1 percentage point in Whom and When?” cities. Congressional, p<0.01) on treated streets. high-SES streets). But the The Journal of presidential, gubernatorial, Turnout among subjects SES gap in new votes cast Politics 77(1): 88–101. and local elections included registered as a result of the was smaller (+4 percentage in the sample. treatment was 24 percent, points in low-SES streets bit.ly/2U6LN70 lower than the population compared to +1 percentage average. point in high-SES streets). Table SI2: (continued) A systematic review of published field experimental studies of the effects of voter registration assistance. Study Sample Treatments Outcomes Results Heterogeneity Sweeney, M., Service, U.K., 2017. The annual Control addresses received Households submitting Some of the treatments S., Mackinson, L., and electoral register canvass in the city councils’ normal their registration before a increased submissions Northcott, H. 25 the cities of Hackney and form letters, which reminder was sent; before a reminder letter April, 2018. Hull. 226,528 households. provided information on households submitting was sent, particularly the “Increasing Responses how to submit the their registration before eye-catching envelope to the Annual Canvass household response to the the deadline. treatment and the in Hackney and Hull.” electoral register and treatment emphasizing The Behavioural stressed that it was legally hassle to the recipient for Insights Team. required. The treatment non-compliance (Envelope groups received one of 10 treatment: 3.4 percentage bit.ly/3kfNklR treatments that either points, p<0.01. Hassle made the envelope more treatment: 0.9 percentage eye-catching; provided points, p<0.05). more detailed information about how to submit the Treatments had no effect form; or stressed the on the rate of submitting individual hassle or social the enquiry form before costs of not submitting the the end of the canvassing 19 form. Any household period (i.e. but did not (treatment and control) lower the non-response that had not submitted rate). their filing by a certain point received reminder letters. Williamson, S. 2 April, U.S., 2018. 4,353 subjects NGO volunteers offered to New registrations by the Registration assistance Effect size was larger for 2019. “The Filer Voter in two cities filing their help subjects fill out a initially unregistered, during tax filing made subjects under the age of Experiment: How income taxes at an voter registration form measured as names and unregistered subjects 4.9 35 (+10 percentage points, Effective is Voter NGO-run tax preparation during subjects’ tax identifiers of recipients percentage points more p<0.05) and smaller but Registration at Tax assistance site. preparation visit. added to voter registration likely to register (8.8 still significant for subjects Time?” Governance rolls between the start and percent registered in who requested a Spanish Studies program, The end of the treatment treatment group compared language consent form Brookings Institution. period. to 3.9 percent in control, (around +1 percentage p<0.05) by the end of the point, p<0.05). brook.gs/32osLNZ treatment period. D Description of IHDS-II variables Table SI3: Description of IHDS-II variables. Variable code Variable label Question Question language in IHDS-II Note number in IHDS-II rec_migrant_10y Migrant Q1.16 From where did the family come? This question was asked if respondents 1) Same state, Same district answered “less than 90” to question 1.15: 2) Same state, Another district How many years ago did your family 3) Another state first come to this village/town/city? For 4) Another country urban areas, we code as migrant those who stated less than 10 years in answer to this question. Only respondents who answered (2) or (3) to question 1.16 classified as migrants, on our conceptualization. rec_hh_has_shortterm_migrant Migrant Q4.1 Have you or any member of your Sending household left to find seasonal/short 20 Household term work during the last five years and returned to live here? rec_hh_member_of_party Political Q18.11 Does anybody in the household belong Party to a political party? Member rec_hh_panchayat_member Panchayat Q18.14 Is anyone in the household a Member member/official of the village panchayat/nagarpalika/ward committee? rec_hh_attended_meeting Attended Q18.13 Have you or anyone in the household Meeting attended a public meeting called by the village panchayat (gram sabha)/nagarpalika/ward committee in the last year? Table SI3: (continued) Description of IHDS-II variables. Variable code Variable label Question Question language in IHDS-II Note number in IHDS-II rec_confidence_in_panchayats Confidence: Q21.6 As far as the people running these Panchayats institutions are concerned, would you (in rural say you have areas) 1) A great deal of confidence Confidence: 2) Only some confidence Ward 3) Hardly any confidence at all in: Committees (in urban Village Panchayats/Nagarpalika/Nagar areas) Panchayat – to implement public projects rec_confidence_in_pols Confidence: Q21.1 continued from above: Politicians Politicians – to fulfill promises rec_confidence_in_state_gov Confidence: Q21.4 continued from above: State 21 Government State government – to look after the people rec_acquaint_pol_in_community Acquaintance: Q17.2g/SN2g1 Do you or any members of your Politician in household have personal acquaintance Community with someone who works in any of the following occupations: Politicians (beyond gram panchayat) Elected members (such as MP/MLA, Zilla parishad member excluding village panchayat): Among your relatives/caste/community rec_acquaint_pol_out_community Acquaintance: Q17.2g/SN2g2 continued from above: Politician Outside Politicians (beyond gram panchayat) Community Elected members (such as MP/MLA, Zilla parishad member excluding village panchayat): Outside the community/caste Table SI3: (continued) Description of IHDS-II variables. Variable code Variable label Question Question language in IHDS-II Note number in IHDS-II rec_acquaint_party_in_community Acquaintance: Q17.2h/SN2h1 continued from above: Party Worker in Political party officials: Among your Community relatives/caste/community rec_acquaint_party_out_community Acquaintance: Q17.2h/SN2h2 continued from above: Party Worker Outside Political party officials: Outside the Community community/caste rec_benefits_income_winsor Winsorized Q9.5, Q13, This variable is coded as This variable measues the monetary sum Benefits Q1-8 all “INCBENEFITS” in the codebook. of various government benefits Income government respondents receive. benefits INR rec_has_proof_residence Has Proof of Q10.3b Does anyone in the household have the Residence following: 22 Proof of residence such as electricity/phone bill, rent agreement etc.? rec_has_photo_id Has Photo ID Q10.3a continued from above: Photo ID proof such as voter card, ration card, PAN card, etc.? rec_urban_area_census11 Urban URBAN2011 NA Urban residence from 2011 Census of India. rec_muslim Muslim Q1.11 What is the religion of the head of household? rec_sc_st SC/ST Q1.13 Is this (caste/jati and sub caste/sub jati to which you belong) Brahmin, General/Forward, OBC, SC, ST, or Others? rec_assets Asset Index ASSETS I would like to ask you about what This variable is in the codebook, not in things your household owns. Do you own the questionnaire. The variable is made a . . . [various household items listed]. up of multiple questions from section 15 of the questionnaire. E Additional information on experimental subject characteristics Table SI4 presents the reasons given by experimental subjects for why they migrated to Delhi and Lucknow. A clear majority cited employment as the main reason. Table SI5 shows that most migrants migrated with their spouses and children. Table SI6 demonstrates that most migrants in the Lucknow sample were intra-state migrants from other parts of Uttar Pradesh. Most migrants to Delhi came from neighbouring states. (Note that Delhi is effectively a city-state and so all migration is from other states.) Table SI4: Main reasons given for migration (in percentages) by T1 experiment subjects. Sample Employment Marriage Other Both cities (full sample) 84.0 8.7 7.4 Delhi 80.8 9.2 10.0 Lucknow 88.8 7.8 3.3 Table SI5: Family accompanying T1 experiment subjects at the time of migration (in percentages). Sample None Spouse Children Both cities (full sample) 4.0 62.9 57.5 Delhi 4.5 57.7 51.7 Lucknow 3.3 70.9 66.3 Table SI6: States of origins of sampled migrants. Sample Come from same state Come from neighboring state Both cities (full sample) 87.6 41.1 Delhi NA 60.2 Lucknow 87.6 11.5 23 F Summary statistics Table SI7: Summary statistics for variables used in the analysis of the T1 experiment. “(E)” variables were measured at endline and “(B)” variables were measured at baseline. Statistic N Mean St. Dev. Min Max (E) Has city-based voter ID 2,120 0.281 0.450 0 1 (E) Voted in city in 2019 2,120 0.281 0.449 0 1 (E) Likelihood of voting in city in future 2,120 0.872 0.197 0 1 (E) Political interest index 2,120 0.045 0.909 −1.442 1.559 (E) Interest: City politics 2,120 0.472 0.326 0 1 (E) Interest: National/state politics 2,120 0.520 0.341 0 1 (E) Politician accountability perceptions 2,120 0.715 0.336 0 1 (E) Sense of political efficacy 2,120 0.446 0.409 0 1 (E) Political trust index 2,120 0.009 0.652 −1.616 1.170 (E) Contacting city officials index 2,120 0.303 0.610 0 6 (E) Non-electoral participation index 2,120 0.295 0.598 0 5 (B) T1 treatment 2,306 0.493 0.500 0 1 (B) Female 2,306 0.540 0.498 0 1 (B) Age 2,306 28.795 10.063 18 88 (B) Muslim 2,306 0.235 0.424 0 1 (B) SC/ST 2,306 0.379 0.485 0 1 (B) Primary education 2,306 0.650 0.477 0 1 (B) Hindi 2,306 0.908 0.220 0 1 (B) Income (INRs) 2,306 10.252 4.663 1 30 (B) Married 2,306 0.684 0.465 0 1 (B) Length of residence in city 2,306 16.472 9.894 0 78 (B) Owns home in city 2,306 0.672 0.470 0 1 (B) Hadn’t voted previously 2,306 0.747 0.435 0 1 (B) How likely to vote in city if registered 2,306 0.920 0.186 0 1 (B) Political interest 2,306 0.265 0.299 0 1 (B) Sense of political efficacy 2,306 0.570 0.367 0 1 (B) Political trust index 2,306 0.001 0.751 −1.703 1.409 (B) Shared meal with non-coethnic 2,306 0.386 0.344 0 1 (B) Has hometown voter ID 2,306 0.268 0.443 0 1 (B) Returned to vote in hometown 2,306 0.174 0.379 0 1 (B) More at home in hometown 2,306 0.615 0.346 0 1 (B) Straight-line distance in kilometers to home district 2,306 384.075 328.329 2.386 1,761.926 (B) Still receives hometown schemes 2,306 0.468 0.499 0 1 (B) Owns hometown property 2,306 0.276 0.447 0 1 24 Table SI8: Summary statistics for variables used in the analysis of the T2 experiment. “(E)” variables were measured at endline and “(B)” variables were measured at baseline. Statistic N Mean St. Dev. Min Max (E) Campaign exposure index 1,969 0.026 0.590 −0.960 3.648 (E) Basti visits by politicians 1,969 0.586 0.871 0 4 (E) Home visit by politician or party worker 1,969 0.591 0.492 0 1 (E) Number of gifts 1,969 0.020 0.146 0 2 (E) Migrant-focused campaigning 1,969 0.447 0.497 0 1 (E) Perceived campaign intensity 1,931 0.718 0.359 0 1 (E) Trust: National government 1,969 0.730 0.340 0 1 (E) Trust: State government 1,969 0.686 0.349 0 1 (E) Trust: Municipal government 1,969 0.584 0.373 0 1 (E) Trust: Political parties 1,969 0.329 0.358 0 1 (E) Considers city home 1,969 0.956 0.154 0 1 (E) Plans to live in city 1,969 54.117 28.594 0 100 (E) Recommends others live in city 1,969 0.777 0.416 0 1 (B) T2 treatment 1,969 0.530 0.499 0 1 (B) Politician visits 1,969 0.608 0.823 0 3 (B) Female 1,969 0.537 0.499 0 1 (B) Age 1,969 28.617 10.075 18 88 (B) Muslim 1,969 0.241 0.428 0 1 (B) SC/ST 1,969 0.369 0.483 0 1 (B) Primary education 1,969 0.667 0.471 0 1 (B) Income (INRs) 1,969 10.220 4.632 1 30 (B) Married 1,969 0.671 0.470 0 1 (B) Length of residence in city 1,969 17.038 10.036 0 78 (B) Owns home in city 1,969 0.693 0.461 0 1 25 G Survey instruments for experiment Table SI9: Variable definitions in the original survey instrument. Survey / variable name Question text Response options Baseline survey Female What is your gender? 1. Female / 2. Male / 3. Other Age What is your age? 18 - 99 Muslim What is your religion? 1. Hindu / 2. Muslim / 3. Sikh / 4. Christian / 5. Jain / 6. Buddhist / 7. Parsi / 8. No religion / 9. Other SC/ST What is your caste group? 1. SC / 2. ST / 3. OBC / 4. Forward caste / 5. Other-specify Primary education What is the highest level of education you have 1. No formal education (cannot attained? read and write) / 2. No formal education (can read and write) / 3. Primary school / 4. Secondary school / 5. Senior secondary school / 6. Graduate / 7. Postgraduate Hindi How well do you speak Hindi? 1. I do not speak any Hindi / 2. I can speak some Hindi but I am not fluent / 3. I am a fluent speaker of Hindi Income (INR 000s) What is your total monthly household income in 0 - 1,000,000 Rupees? Married Are you currently married? 1. Yes / 2. No Length of residence When did you move to [Delhi/Lucknow] to live or 1920 - 2018 work? Owns home Do you own or rent your home? 1. Rent / 2. Own / 3. Other-specify Politically active in Have you gone back to vote in an election in your 1. Yes / 2. No village home village or town since moving to [Delhi/Lucknow]? More at home in village To what extent do you agree with the following 1. Very much agree / 2. statement? "I feel more at home in my previous Somewhat agree / 3. Somewhat place of residence than I do in [Delhi/Lucknow]." disagree / 4. Very much disagree Family in city Which of the following family members stay with 1. No family members / 2. you here in [Delhi/Lucknow]? Spouse / 3. Children / 4. Parents / 5. Gradparents / 6. Other extended family Number of calls to prior Over the past week, approximately how many 0 - 99 residence phone calls did you make to friends or relatives back in your previous place of residence? Receives schemes in Do you or your immediate family continue to 1. Yes / 2. No prior residence benefit from government schemes in your previous place of residence-for example, PDS, MGNREGA, or cash transfer schemes? Owns property in prior Do you or your spouse personally own land or 1. Yes / 2. No residence property in your previous place of residence? Wants city voter ID card Do you wish to apply for a [Delhi/Lucknow] voter 1. Yes / 2. No ID card that will allow you to vote in national, state, and local elections in [Delhi/Lucknow]? 26 Table SI9: Variable definitions in the original survey instrument. (continued) Survey / variable name Question text Response options Has voter ID for prior Do you currently have a voter ID card allowing 1. Yes / 2. No residence you to vote in your previous place of residence (outside [Delhi/Lucknow])? How likely to vote If an election for the [Delhi/Lucknow] municipal 1. Very likely / 2. Somewhat corporation were going to be held tomorrow, and likely / 3. Somewhat unlikely / you were registered to vote here, how likely do 4. Very unlikely you think it is that you would vote? Political interest In general, how interested are you in politics? 1. Very interested / 2. Somewhat interested / 3. Not very interested Political efficacy To what extent do you agree with the following 1. Strongly agree / 2. statement? "People like me don’t have any Somewhat agree / 3. Somewhat influence on the government in [Delhi/Lucknow]" disagree / 4. Strongly disagree Trust in national How much trust do you have in the national 1. No trust at all / 2. Not government government? much trust / 3. Some trust / 4. A great deal of trust Trust in state How much trust do you have in the 1. No trust at all / 2. Not government [Delhi/Lucknow] state government? much trust / 3. Some trust / 4. A great deal of trust Trust in municipal How much trust do you have in the 1. No trust at all / 2. Not corporation [Delhi/Lucknow] municipal corporation? much trust / 3. Some trust / 4. A great deal of trust Shared meal with ethnic Over the past six months, how many times have 1. Very regularly / 2. out-group you shared a meal with someone from another jati Somewhat regularly / 3. A few or religion? times / 4. Not at all Politician visits to basti Which of the following politicians, if any, have 1. Municipal corporator / 2. visited your basti here in [Delhi/Lucknow] in the MLA / 3. MP past year? Last vote in prior Think about your previous place of residence. 1. I voted in a village as part of residence What best describes the most recent election in a Lok Sabha election / 2. I which you yourself voted there? voted in a village as part of a state assembly election / 3. I voted in a village as part of a gram panchayat election / 4. I voted in a town or city as part of a Lok Sabha election / 5. I voted in a town or city as part of a state assembly election / 6. I voted in a town or city as part of a municipal corporation election / 7. Other-please specify / 8. I have not voted previously 27 Table SI9: Variable definitions in the original survey instrument. (continued) Survey / variable name Question text Response options Officeholders contacted Which of the following [Delhi/Lucknow] 1. [Delhi/Lucknow] corporation officeholders have you contacted at some point official / 2. [Delhi/Lucknow] over the past year? municipal corporator / 3. Local MLA / 4. Local MP / 5. Party worker inside your basti / 6. Party worker outside your basti / 7. Housing society/vikas samiti official in your basti / 8. Ward samiti representative in your basti / 9. Dalal/broker/middleman / 10. Other community worker in your basti Endline survey Has voter ID for city Do you currently have a voter ID card that allows 1. Yes / 2. No you to vote in [Delhi/Lucknow] elections? Voted in city in 2019 Did you vote in [Delhi/Lucknow] during the Lok 1. Yes / 2. No Sabha elections held in May of this year? How likely to vote in city How likely is it that you will vote in the next 1. Very likely / 2. Somewhat state elections held in [Delhi/Lucknow]? likely / 3. Somewhat unlikely / 4. Very unlikely Attention to city politics How much attention do you pay to news about 1. A lot of attention / 2. Some politics in [Delhi/Lucknow]? attention / 3. No attention at all Attention to How much attention do you pay to news about 1. A lot of attention / 2. Some national/state politics national and state politics? attention / 3. No attention at all Political accountability To what extent do you agree or disagree with the 1. Very much agree / 2. following statement? "Elected politicians are Somewhat agree / 3. Somewhat accountable to the citizens of this city." disagree / 4. Very much disagree Political efficacy To what extent do you agree or disagree with the 1. Strongly agree / 2. following statement? "People like me don’t have Somewhat agree / 3. Somewhat any influence on the government." disagree / 4. Strongly disagree Trust in national How much trust do you have in the national 1. A great deal of trust / 2. government government? Some trust / 3. Not very much trust / 4. No trust at all Trust in state How much trust do you have in the 1. A great deal of trust / 2. government [Delhi/Lucknow] state government? Some trust / 3. Not very much trust / 4. No trust at all Trust in municipal How much trust do you have in the municipal 1. A great deal of trust / 2. corporation corporation of [Delhi/Lucknow]? Some trust / 3. Not very much trust / 4. No trust at all Trust in political parties How much trust do you have in political parties? 1. A great deal of trust / 2. Some trust / 3. Not very much trust / 4. No trust at all Considers city home To what extend do you agree or disagree with the 1. Very much agree / 2. following statement? "I consider [Delhi/Lucknow] Somewhat agree / 3. Somewhat to be my home." disagree / 4. Very much disagree Plans to live in city For how many more years do you plan to live in 0-100 [Delhi/Lucknow]? 28 Table SI9: Variable definitions in the original survey instrument. (continued) Survey / variable name Question text Response options Recommends others live Would you recommend to your friends and 1. Yes / 2. No in city relatives in your home village that they come to [Delhi/Lucknow] to live and work? Campaign: basti visits Which of the following politicians, if any, have 1. Sitting municipal corporator visited this basti in the last three months, / 2. Sitting MLA / 3. Sitting including during the Lok Sabha election MP / 4. MP candidate campaign? Campaign: home visit During the recent Lok Sabha campaign in 1. Yes / 2. No [Delhi/Lucknow], did a politician or political party worker come to your door to ask for your vote? Campaign: gifts During the recent Lok Sabha campaign in 1. A gift of money / 2. A gift [Delhi/Lucknow], did a politician or political of clothing / 3. A gift of alcohol party worker offer you any of the following items? / 4. Another kind of gift / 5. If so, which items? Free travel / 6. No, nothing was offered Campaign: pro-migrant During the recent Lok Sabha campaign, did any 1. Yes / 2. No politician or political party try to specifically win the votes of recent migrants to this neighborhood? Campaign: intensity To what extent do you agree or disagree with the 1. Strongly agree / 2. following statement? "During the recent Lok Somewhat agree / 3. Somewhat Sabha campaign, politicians and political party disagree / 4. Strongly disagree workers campaigned hard to win the votes of people in this particular basti." Contacting officials Which of the following officeholders, if any, have 1. [Delhi/Lucknow] corporation you yourself contacted in the past three months? official / 2. Municipal corporator / 3. Local MLA / 4. Local MP / 5. Political party worker inside your basti / 6. Party worker outside your basti / 7. Housing society/vikas samiti official in your basti / 8. Ward samiti representative in your basti / 9. Dalal/broker/middleman / 10. Other community worker in your basti Non-electoral Here is a list of things that people sometimes do 1. Attended a community participation as citizens. Please tell me which of these, if any, meeting / 2. Joined or you have personally done during the past three participated in the meetings of months. a civic association, such as a club, union, or NGO / 3. Gone to a meeting of a political party / 4. Gone to a political rally / 5. Given money to a political party or to a political cause / 6. Handed out leaflets or done door to door campaigning on behalf of a cause or a political party / 7. Voted in an internal political party election 29 H T1 additional results H.1 T1 interaction with hometown attachment characteristics We find that 98 percent of subjects in the omnibus sample expressed a wish to receive assistance to register to vote locally, which we take to be evidence against the claim that hometown attachments are consequential for the take up of registration help. Table SI10 further probes the “voluntaristic detachment” theory by examining whether the T1 intervention was less effective among subjects with stronger attachments to home regions. This may have occurred if, say, subjects more attached to home regions declined to invest in the registration process once it got underway. The null interaction coefficients presented in Table SI10 suggest that this is not the case. Table SI10: [Exploratory] Estimates of heterogeneous effects of T1 treatment by hometown at- tachment characteristics. Models do not include additional covariates. Robust standard errors in parenthesis. Dependent variable: Has City-Based Voter ID Voted in City in 2019 (1) (2) T1 x Length of residence 0.001 −0.00005 (0.002) (0.002) T1 x Kilometers to home 0.00003 0.00001 (0.0001) (0.0001) T1 x More at home at hometown 0.049 −0.054 (0.057) (0.056) T1 x Still gets hometown schemes −0.045 −0.009 (0.041) (0.041) T1 x Owns hometown property 0.0002 0.009 (0.044) (0.044) T1 0.201∗∗∗ 0.239∗∗∗ (0.059) (0.059) Length of residence 0.003∗∗ 0.002 (0.001) (0.001) Kilometers to home −0.00003 −0.0001∗∗∗ (0.00003) (0.00004) More at home in hometown −0.055∗ −0.035 (0.033) (0.033) Still gets hometown schemes 0.012 0.003 (0.024) (0.025) Owns hometown property −0.021 −0.010 (0.025) (0.027) Constant 0.164∗∗∗ 0.213∗∗∗ (0.035) (0.036) Observations 2,120 2,120 Adjusted R2 0.076 0.060 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 30 H.2 T1 effects on political interest decomposed Table SI11: [Exploratory] T1 experimental results for political interest, using the two individual components of the political interest index as outcomes. OLS estimates of intent to treat effects. Models include covariates. Robust standard errors in parentheses. Interest: Interest: City Politics National/State Politics (1) (2) T1 treatment 0.032 0.028 (0.014) (0.015) p-value (upper) 0.011 0.028 Control mean 0.457 0.506 Observations 2,120 2,120 Adjusted R2 0.023 0.020 DV values {0, 0.5, 1}{0, 0.5, 1} 31 H.3 T1 effects on contacting and non-electoral participation Table SI12 considers two outcomes not included in our pre-analysis plan, yet that bear on citizens’ political engagement beyond elections. Column 1 shows the registration campaign somewhat increased the number of urban politicians subjects report having contacted within the past three months (0.039 additional contacts, p=0.069). Non-electoral participation, too, responded to the treatment, with the count of reported activities increasing by 0.043 (p=0.048; column 2). Both behavioral changes are impressive given the comparatively short timespan of the project. They resonate with a growing body of literature stressing the significance of claims-making in developing country contexts: the need for active citizens to make requests of the state if their welfare needs are to be met. The small absolute magnitudes of the effects are in line with the low control group means. The majority of respondents had neither contacted any city official nor engaged in any non-electoral political activity over the past three months, highlighting the depths of this group’s marginalization. Table SI12: [Exploratory] T1 experimental results for contacting and non-electoral participation. OLS estimates of intent to treat effects. Models include covariates. Robust standard errors in parentheses. Contacting City Non-Electoral Officials Index Participation Index (1) (2) T1 treatment 0.039 0.043 (0.026) (0.026) p-value (upper) 0.069 0.048 Control mean 0.286 0.277 Observations 2,120 2,120 Adjusted R2 0.028 0.031 DV values {0,..., 6}{0,..., 5} 32 H.4 Heterogeneous effects of T1 by Hindi fluency Table SI13: [Exploratory] Estimates of heterogeneous effects of T1 treatment by Hindi fluency. Models do not include additional covariates. Robust standard errors in parentheses. Dependent variable: Has City-Based Voter ID Voted in City in 2019 (1) (2) T1 x Hindi fluency 0.032 0.108 (0.092) (0.094) T1 0.214∗∗ 0.110 (0.087) (0.089) Hindi fluency −0.101∗ −0.139∗∗ (0.058) (0.062) Constant 0.253∗∗∗ 0.304∗∗∗ (0.055) (0.059) Observations 2,120 2,120 Adjusted R2 0.074 0.055 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 33 H.5 T1 main results controlling for Hindi fluency Table SI14: [Exploratory] T1 experimental results for primary political outcomes, controlling for Hindi fluency, in addition to the set of pre-registered control variables. OLS estimates of intent to treat effects. Robust standard errors in parentheses. Has City-Based Voted in City Likelihood of Voting Voter ID in 2019 in City in Future (1) (2) (3) T1 treatment 0.236 0.204 0.032 (0.019) (0.019) (0.009) p-value (upper) 0.000 0.000 0.000 Control mean 0.161 0.178 0.856 Observations 2,120 2,120 2,120 Adjusted R2 0.086 0.066 0.011 DV values {0, 1}{0, 1}{0, 0.33, 0.67, 1} Table SI15: [Exploratory] T1 experimental results for additional political outcomes, controlling for Hindi fluency, in addition to the set of pre-registered control variables. OLS estimates of intent to treat effects. Robust standard errors in parentheses. Political Politician Sense of Political Interest Accountability Political Trust Index Perceptions Efficacy Index (1) (2) (3) (4) T1 treatment 0.092 0.039 -0.012 0.027 (0.039) (0.015) (0.018) (0.028) p-value (upper) 0.010 0.004 0.743 0.170 Control mean 0.000 0.697 0.450 0.000 Observations 2,120 2,120 2,120 2,120 Adjusted R2 0.026 0.006 0.002 0.019 DV values [−1.44, 1.56] {0, 0.33, 0.67, 1}{0, 0.33, 0.67, 1} [−1.62, 1.17] 34 H.6 T1 main results without covariates Table SI16: [Exploratory] T1 experimental results for primary political outcomes. OLS estimates of intent to treat effects. Models do not include covariates. Robust standard errors in parentheses. Has City-Based Voted in City Likelihood of Voting Voter ID in 2019 in City in Future (1) (2) (3) T1 treatment 0.243 0.208 0.033 (0.019) (0.019) (0.009) p-value (upper) 0.000 0.000 0.000 Control mean 0.161 0.178 0.856 Observations 2,120 2,120 2,120 Adjusted R2 0.073 0.053 0.007 DV values {0, 1}{0, 1}{0, 0.33, 0.67, 1} Table SI17: [Exploratory] T1 experimental results for additional political outcomes. OLS estimates of intent to treat effects. Models do not include covariates. Robust standard errors in parentheses. Political Politician Sense of Political Interest Accountability Political Trust Index Perceptions Efficacy Index (1) (2) (3) (4) T1 treatment 0.091 0.037 -0.007 0.018 (0.039) (0.015) (0.018) (0.028) p-value (upper) 0.011 0.006 0.660 0.263 Control mean 0.000 0.697 0.450 0.000 Observations 2,120 2,120 2,120 2,120 Adjusted R2 0.002 0.003 0.000 0.000 DV values [−1.44, 1.56] {0, 0.33, 0.67, 1}{0, 0.33, 0.67, 1} [−1.62, 1.17] 35 I Vote choice This section compares the partisan preferences of migrants in our experimental sample with the preferences of immediately surrounding local populations. In the descriptive (non-experimental) analysis presented in Figure SI1, we compare reported vote choice by all migrants in our experimental sample (i.e. both those in the T1 treatment and T1 control groups) to the final 2019 electoral returns (a) in the seven Lok Sabha constituencies in which we worked, and (b) in the specific polling-station localities in which we worked—and therefore where our migrants were entitled to cast a vote. Administrative data was obtained from the Trivedi Centre for Political Data at Ashoka University, and from the “Form 20” polling-station returns published online by the Election Commission of India, which we digitized. We conduct the comparisons separately for Delhi and Lucknow. We do this because, although the two major national parties (the Indian National Congress, INC, and the Bharatiya Janata Party, BJP) received substantial support in both locations, the Aam Aadmi Party (AAAP) was only seriously competitive in Delhi, while the Bahujan Samaj Party (BSP) and Samajwadi Party (SP) were only seriously competitive in Lucknow. The results of this analysis need to be heavily caveated by the fact that only a minority of registered respondents in our endline survey were willing to reveal their vote choice to enumerators. Consequently, our estimates of the sample’s vote choice are noisy and may not be representative of the sample as a whole. Nevertheless, the comparisons are revealing. (a) Delhi 2019 Lok Sabha Results 0.8 Level 0.6 Constituency 0.4 Polling Station Vote Share Vote 0.2 Sample 0.0 BJP INC AAAP Other (b) Lucknow 2019 Lok Sabha Results 0.8 0.6 Level Constituency 0.4 Polling Station Vote Share Vote 0.2 Sample 0.0 BJP INC BSP SP Other Party Figure SI1: [Exploratory] Comparison of the distributions of party-wise support between migrants in the experimental sample and the final vote tallied at the constituency and polling-station levels. 36 J T2 additional results J.1 T2 effects on trust and integration Table SI18: [Exploratory] Estimates of T2 effects on trust in political institutions and social inte- gration. Outcomes are whether respondent has trust in the national (1), state (2), and municipal governments (3), and in political parties (4); whether the respondent considers the city to be “home” (5); for how many more years they plan to live in the city (6); and whether they would recommend to friends and relatives in their hometown to come to the city to live and work (7). Weighted least squares estimates of intent to treat effects. Clusters weighted equally. Models include block fixed effects but no additional covariates. Cluster-robust standard errors in parentheses. Dependent variable: Trust: Trust: Trust: Trust: Considers Plans Recommends National State Municipal Political City to Live Others Live Gov. Gov. Gov. Parties Home in City in City (1) (2) (3) (4) (5) (6) (7) T2 −0.001 0.050∗ 0.043 −0.033 0.017∗∗ 0.677 0.007 (0.023) (0.026) (0.029) (0.026) (0.007) (3.344) (0.030) Observations 1,969 1,969 1,969 1,969 1,969 1,969 1,969 Adjusted R2 0.016 0.034 0.015 0.009 0.004 0.020 0.001 ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 37 J.2 T2 results without covariates Table SI19: [Exploratory] T2 experimental results for exposure to campaigning during the 2019 elections. Weighted least squares estimates of intent to treat effects. Clusters weighted equally. Models do not include covariates. Cluster-robust standard errors in parentheses. Index Components Campaigning Basti Home Visit Number Migrant- Perceived Exposure Visits by by Politician or of Focused Campaign Index Politicians Party Worker Gifts Campaigning Intensity (1) (2) (3) (4) (5) (6) T2 treatment 0.101 0.055 0.039 0.020 0.006 0.073 (0.058) (0.080) (0.039) (0.014) (0.046) (0.032) p-value (upper) 0.043 0.249 0.160 0.078 0.445 0.012 Control mean -0.039 0.559 0.550 0.013 0.425 0.676 Observations 1,969 1,969 1,969 1,969 1,969 1,931 No. of Clusters 87 87 87 87 87 87 Adjusted R2 0.056 0.049 0.033 0.012 0.004 0.012 DV values [−0.96, 3.65] {0,..., 4}{0, 1}{0, 1, 2}{0, 1}{0, 0.33, 0.67, 1} 38 K Migrant voting behavior over time Table SI20: Migrant versus non-migrant voting behavior in Delhi elections over time. Note, in some cases the reported percentages represent weighted averages based on relative population shares described in the given source. Election Vote choice 2020 Vidhan Sabha Elections •INC: •Migrants: 4% Source: Lokniti Pre-Poll Election Survey •Non-migrants: 5% 2020 bit.ly/35FG4ei •BJP: •Migrants: 43% •Non-migrants: 38% •AAP: •Migrants: 51% •Non-migrants: 55% •Others: •Migrants: 3% •Non-migrants: 2% 2015 Vidhan Sabha Elections •INC: •Migrants: 8% Source: Lokniti Post-Poll Election Survey •Non-migrants: 10% 2015 bit.ly/35FG4ei •BJP: •Migrants: 33% •Non-migrants: 32% •AAP: •Migrants: 55% •Non-migrants: 55% •Others: •Migrants: 3% •Non-migrants: 3% 2008 Vidhan Sabha Elections •INC: •Migrant constituencies: 36% Kumar, Sanjay. 2013. Changing Electoral •Rest: 44% Politics in Delhi: From Caste to Class. •BJP: SAGE Publications India. p. 81. •Migrant constituencies: 34% •Rest: 37% Vote choice comparisons between •BSP: constituencies dominated by migrants from •Migrant constituencies: 19% UP/Bihar and constituencies not dominated •Rest: 12% by such migrants (“Rest”) •Others: •Migrant constituencies: 11% •Rest: 6% 39 L Research locations Figure SI2: Approximate locations of the experimental sample: Delhi. Boundaries represent as- sembly constituency segments within the Lok Sabha constituencies in which the experiment was fielded. 40 Figure SI3: Approximate locations of the experimental sample: Lucknow. Boundaries represent assembly constituency segments within the Lok Sabha constituencies in which the experiment was fielded. 41