Does Citizenship Abate Class? Evidence and Reflections from a South Indian City

Abstract

Drawing on data from a large household survey in Bangalore, this paper explores the quality of urban citizenship. Addressing theories that have tied the depth of democracy to the quality and effectiveness of citizenship, we develop an index of citizenship and then explore the extent to which citizenship determines the quality of services and infrastructure that households enjoy. Our findings show that citizenship and access to services in Bangalore are highly differentiated, that much of what drives these differences has to do with class, but we also find clear evidence that the urban poor are somewhat better in terms of the services they receive than they would be without citizenship. Citizenship, in other words, abates the effects of class.

INTRODUCTION

In 1951, India was a mere 17.3 per cent urban, and only five Indian cities had populations greater than 1 million. By 2011, three cities – Mumbai, Delhi, and Kolkata – had more than ten million people each, and 53 cities had populations of more than one million. By 2031, six cities are projected to cross the population threshold of 10 million. Depending on the definition of “urban” and the assumed economic growth rate, India’s population, 32 percent urban in 2011, could well be over 40 percent urban over the next 15–20 years, if not higher (United Nations 2012. The latest Census shows that for the first time, the absolute increase in urban population during 2001–2011 exceeded the increase in rural population in any ten-year period since independence.

Burgeoning urbanization poses critical questions about how growing and expanding cities can ensure the quality of life for all. A spate of recent research and government reports, for example the Ahluwalia committee report (2011), has underscored the problem of governance and, in particular, has pointed to weaknesses in urban citizenship. Though cities, as compared to villages, are often conceptualized as bastions of freedom and opportunity, the prevalence of slums, low levels of civic participation and the inequitable provision of infrastructure point to the poor quality of effective citizenship for many. A truncation of citizenship affects the quality of democracy. Normally, a more active citizenry makes democracy deeper.

As India continues to urbanize, a few critical questions have to be addressed: what is the quality of citizenship in urban India? Are the rising cities witnessing the emergence of citizen consciousness and a rights-based politics, heralding a greater citizen-focused deepening of the polity? Or, do vertical patron-client ties between the political elite and citizens and other forms of dependency remain obdurately strong? Is the exercise of citizenship a function of class, caste or community as much of the literature maintains? Can citizenship lead to substantive improvements in people’s lives? Specifically, can it help improve the extent and quality of public service delivery?

Primarily because of the predominantly rural nature of Indian polity and society, the literature on urban India has remained sparse. Moreover, most of the literature has to date been largely qualitative and there have been few quantitatively informed analyses of the extent and effectiveness of urban citizenship. We need both newer conceptualizations and careful measurement. The Janaagraha-Brown Citizenship Index (JB-CI) project, of which this study is a part, seeks to address this empirical and conceptual gap.

We seek to answer two questions: how is citizenship distributed over various social and economic categories, and to what extent does citizenship, relative to other factors, determine the extent and quality of public service delivery? We go beyond the strictly legal concept of citizenship, which defines citizenship as a bundle of rights (for example, freedom to vote, equal treatment in law, freedom of association, freedom to protest and petition, etc). We concentrate on what we call effective citizenship, conceptualized as the capacity of citizens to use their basic rights (Heller 2000). This is a relational, not legal, conceptualization of citizenship, and we define it as consisting of two critical dimensions: knowledge (what citizens know about their rights) and participation (whether and how much they participate in political and civic life, and what forms such participation takes).

Based on a large survey in Bangalore, we present four key findings. First, Bangalore citizens vote in high numbers but do not participate much in politics beyond voting, or in civic life. In part this no doubt reflects the fact that at the local level formal institutions for engaging in politics (ward councils) are absent or weak.

Second, effective citizenship in Bangalore is highly differentiated. While all Bangaloreans know and cherish their formal rights, their capacity to use those rights is very unevenly distributed. The biggest correlates of high effective citizenship are education and class. On the whole, the higher the class, the greater the effective citizenship. There is one exception though: the highest class exhibits lower effective citizenship. The caste and religious differences are worth noting as well. Dalits and Muslims generally have lower levels of effective citizenship than non-Dalits and Hindus/Christians respectively. This general pattern of differentiated citizenship, however, comes with an important caveat. Inequalities in effective citizenship are largely driven by differences in knowledge of civic and political affairs. In contrast, participation, especially as it relates to voting, is a substantial leveler: lower classes, Dalits and Muslims participate more than higher classes, higher castes and Hindus/Christians.

Third, it is on the translation of rights into outcomes – namely, providing public services (water, power, roads, sanitation) --- that urban governance has failed most conspicuously.¹ Access to basic services and infrastructure in Bangalore is unevenly distributed and is highly correlated with class and caste, though not with religion. Muslims do not fare worse than the Hindus overall.

Fourth, in this overall pattern of unequal effective citizenship, there is however one promising finding. Though the poor have lower effective citizenship, it matters more for them. Specifically, we find that the poor get more in terms of access to basic services and infrastructure from the citizenship they do have than their class position would otherwise predict. To put it simply, if they did not participate in political and civic life, they would receive less from the state. Citizenship, in this sense, is an ally of the poor.

In what follows, we begin with a discussion of theories and concepts deployed in this study. We next describe our research design. We then show how we constructed the two indices -- one for citizenship and another for public service provision. The next two sections describe how citizenship and public services are distributed over the various standard socio-economic categories: class, caste, religion, education, gender, location and migrant status. Finally, we deploy statistical models and engage in an explanatory exercise, asking to what extent citizenship matters relative to other factors, in the provision of public services.

DOES CITIZENSHIP MATTER?

A basic idea runs through the existing literature on citizenship. The literature is marked by “the malodorousness of subjecthood and the fragrance of citizenship” (Jayal, 2013: 3). But what does citizenship entail? To answer this question, we turn to T. H. Marshall, widely regarded as the field’s theoretical pioneer. We address three questions: (a) How did Marshall conceptualize citizen rights? (b) What might be his deficiencies, both generally and especially with respect to India? (c) In what ways do we address these deficiencies and go beyond Marshall in this study?

Marshall’s Formulations

Published originally in 1950 and reprinted many times, Marshall’s Citizenship and Social Class was the first, and highly influential, treatment of the subject. Marshall sought to divide citizenship into three components: civil, political and social. The civil component referred to individual freedoms, such as the freedom of speech, religion and association, and the right to property, contracts and justice. The courts were the main institutions concerned with this aspect of citizenship. The political component of citizenship encompassed franchise as well as the right to run for office. The local governments and parliament were the principal institutional arenas for with these rights. The third, social, element of citizenship, was split by Marshall into two parts: (a) “the right to a modicum of economic welfare and security” and (b) “the right to share to the full in the social heritage and to live the life of a civilized being according to the standards prevailing in the society” (Marshall 1992: 8). The so-called social services, especially, though not only, public provision of health care and education, were the institutions closely associated with the third set of rights. The third aspect of citizenship, also called social citizenship, is inextricably tied up with the rise of a welfare state. Marshall also argued that this conceptual classification was based on the historical evolution of citizenship in Britain. The civil rights were introduced in the 18^th century, political rights in the 19^th, and the social rights in the 20^th.

It is noteworthy that Marshall conceptualized the problem of deprivation entirely in class terms. It was the economically poor, who had “the right to a modicum of economic welfare and security” and “the right to share to the full in the social heritage”. If the state did not guarantee such rights and make allocations for them through state-financed health, housing and education schemes, markets would not be able to provide them. Indeed, left unchecked, markets would deprive the poor of full citizenship. Markets might be consistent with political and civil citizenship, but they were certainly in conflict with social citizenship.

The relative neglect of non-class forms of exclusion comes with some other limitations of the Marshallian model. Most notably, in painting his broad canvas of the history of citizenship in the UK, Marshall had a tendency to privilege rights, and he specifically conflated rights-as-status with rights-as-practice. All citizens are presumed to have the basic rights and the capacity to exercise free will, associate as they choose and vote for who and what they prefer. Following in the relational tradition of analysis, Somers (1993) has argued that the conventional treatment wrongly equates the status of citizenship (a bundle of rights) with the practice of citizenship (a set of relationships). Formal rights matter, but formal rights must also be actionable. Somers goes on to argue that given the highly uneven rates of political participation and influence across social categories that persist in advanced democracies (and especially the United States), the notion of citizenship should always be viewed as contested. But in the context of developing democracies, where inequalities can be very high and access to rights is often circumscribed by social position or compromised by the weaknesses of state institutions, the very notion of citizenship comes into question (Heller, 2000; Mahajan, 1999; Fox, 1994).

Beyond Marshall: Conceptualizing Citizenship in India

Which communities of India, defined in non-class terms, experience truncated citizenship? Given what we know from existing studies, Dalits (Scheduled Castes, or SCs), Adivasis (Scheduled Tribes, or STs), Muslims and women are some of the obvious candidates for investigation. Also relevant here is an Ambedkar idea. He used to call the village a cesspool for Dalits, and viewed the city a site of potential liberation. Is that true? To what extent does caste discrimination exist in urban India, compromising citizenship?² By definition, this question acquires significance in the study of citizenship in urban India.

We thus seek to go beyond Marshall and much of the contemporary literature on citizenship in two ways. First, Marshall’s concentration is on class deprivation; we include non-class forms of deprivation – caste, religion and gender – as well, since in the Indian context these are important sources of social exclusion in their own right. Second, Marshall’s focus is on the legal availability of rights, not on how the legally enshrined rights are experienced on the ground. Our focus is less on the laws or rights in theory, more on the practices on the ground.

Following Somers, we argue that the formal nature of citizenship, rights-as-status or the legal codification of basic rights of citizenship, should be analytically distinguished from its efficacy (rights-as-practice), that is, the degree to which a citizen can effectively use their rights independently of their social position and without compromising their associational autonomy.³ There is no dispute as to the formal character of citizenship in India, at least with respect to basic civic and political rights. These are enshrined in the constitution, have been upheld by the courts and are the bread-and-butter of Indian democratic life⁴. Social rights in the Marshallian sense – right to food and education, if not health -- have only just really come into play as formal rights of citizenship, but the principle of being able to deploy civic and political rights to demand social rights is well established.

The effective dimension of citizenship is in contrast much less clear, and in fact presents the central conceptual and empirical challenge of this study. How effectively Indians make use of their rights to associate, vote, participate and engage remains an open-ended question. There is certainly widespread recognition that citizenship in India is highly differentiated. Chatterjee’s claim that the realm of civil society – the realm in which citizens use their rights - is largely the privileged domain of the middle classes and that the poor have only their electoral clout to work with has even become a dominant trope of the literature (Chatterjee 2006). Is Chatterjee right? Do the poor exercise only political, not civil rights?

We argue that practicing citizenship means essentially two things. First, it requires having sufficient knowledge and understanding to fully engage in public life. This means having, in effect, the basic knowledge of politics and how the state functions. These are necessary for making informed decisions about one’s preferences and about how to make claims on the state, be it by voting or directly interacting with state actors. Second, one must enjoy the freedom to participate in public life. This cannot simply be confined to voting, but means enjoying freedom to engage in activities of public relevance across social boundaries, including gender, religion, caste and class. There is a large literature on the latter. For instance, Rueschemeyer et al. (1992) have systematically linked the participatory dimension of citizenship to substantive outcomes.

Following this reasoning, we take knowledge and participation as the building blocks of the idea of effective citizenship. And we view substantive social outcomes (water, electricity, sanitation and roads), in part, as a function of the exercise of civic and political rights, i.e. effective citizenship.

EMPIRICAL ANALYSIS: DATA, VARIABLES, AND MODELS

Data

The data used in this study was collected from a sample of 4,093 individuals in Bangalore in December 2013-January 2014. We adopted a multi-stage stratified random sampling method to select wards (20) and polling parts (10 from within each selected ward) to ensure geographical representation (central and outer regions), as well as social representation (Dalits, Adivasis, and Muslims) in Bangalore. Polling parts are the smallest political geographic entities in urban India consisting of approximately 7–14 streets and 1500–2000 individuals above the age 18. We selected polling parts because they provide some indication of a neighbourhood due to their small size. Thirty households were randomly selected from each polling part using a systematic sampling method. Individuals were selected from households using randomization of all household members above the age of 18 who had lived in the household for a minimum of one year. We excluded respondents who had not lived in the city for at least a year since such residents might be temporary (and as such not very invested in practicing their citizenship).

The basic demographic characteristics of our sample are presented in Table 1. We also compare these statistics to the Census data for 2011. Our sample over-represents the Muslims and Dalits/Adivasis. The sample Muslim population is 18% compared to 14% in the Census. The proportion of Dalit/Adivasi respondents in the entire sample is approximately 20.4 percent (16.8% Dalit and 3.6% Adivasi), while the comparable proportion for Bangalore reported in the 2011 Census is approximately 14% of the entire population (12% Dalit and 2% Adivasi). The gender break-up of our sample is 55.6% women as opposed to 47.8% as per Census data, 44.3% men as opposed to 52.4% from Census data. Readjusting sample proportions according to census data does not alter our findings.

Table 1 --.

Sample Characteristics: Caste and Religion (percentages)

Religion		Caste (Hindus Only)
Hindu	72.9	Dalit	17.0
Muslim	18.0	Adivasi	24.0
Christian	8.8	OBC	5.0
Jain	0.1	Upper Castes/Other	54.0
Other	0.2

In Table 1, we also present the Dalit and Adivasi numbers as a proportion of the overall Hindu sub-sample. Dalits and Adivasis account for about 29 percent of all Hindus in the sample (24 percent Dalit and 5 percent Adivasi). Within the Hindu sub-sample, it is also noteworthy that upper castes represent 54%, a figure that might strike many as too high. From our analysis of individual respondents, it is clear that many Lingayats and Vokkaligas classified themselves as upper castes, even though legislatively substantial proportions of them have been classified as the “other backward classes” (OBC). The legal and the self-reported categories thus diverge. In line with existing empirical research, our decision is to stick to self-reporting in this case. Politically, Lingayats and Vokkaligas have a dominant status in the state of Karnataka. Empirically driven social science research conducted in Karnataka, therefore, tends to recode them as upper caste in order for the data to make better sense, given these groups’ social standing and access to resources. We take self-classification as a social fact. We would also note that a majority of our respondents who reported having moved to Bangalore in the past year self-identified as upper castes.

Socio-Economic Factors

We present two sets of variables. The first set of variables we describe are the standard socio-economic categories (or respondent background characteristics) meant to capture the hypothesized sources of social exclusion or unequal endowments. The second set of variables are two indices we constructed: the Citizenship Index (CI), which serves as the independent variable in our analysis, and the Basic Service Delivery and Infrastructure Index (BSDII), which is our dependent variable. Appendix One contains a full list of our variables, how we measured them, and how the indices were constructed.

Background variables include caste, religion, education and class. We asked all Hindu respondents their caste (Dalit/Adivasi/OBC/Upper Caste), and all caste data reported here refer to respondents’ self-classification into one of these four categories. Given the relatively small number of Adivasis in our sample, we combine Dalits and Adivasis in the analysis. We also only report findings for Muslims, Hindus and Christians, as the total number of other religions was very small (0.3%). Our education variable was a 5-point classification scheme, running from non-literate to college degree.

All background variables, except class, are easy to define and compute. Measuring class is notoriously difficult. We developed and collected an asset based measure as well as occupational data. Here, however, we report only a third measure of class, based on Housing Type (HT). Both occupational and asset data have serious measurement problems and are also conceptually problematic.⁵ Neither is a good proxy the full array of conditions that capture the class situation. HT on the other hand is a good measure in part because the home itself is the largest asset, but also because homes capture the spatial dynamics of having access to neighborhood assets including locational advantages and social capital. This then comes much closer to the relational views of class increasingly favored in the literature (Portes and Hoffman 2003; Tilly 1998; Massey 2007).

Another significant advantage of our HT variable is that it was not self-reported. Instead, field surveyors, after receiving extensive field training, were asked to classify each household they surveyed into one of five types:

• HT 1: Informal settlement

• HT 2: Designated/Notified slum

• HT 3: Lower middle class housing

• HT 4: Middle class housing

• HT 5: Upper Class housing

Figure 1 summarizes the distribution of our sample over housing types.⁶ Approximately 2 percent in our sample live in informal settlements (HT1) and about 11 percent in one-room notified/designated slum housing (HT2). Taking these two categories together (HT1 and HT2), we find that 13.1% live in slums.⁷ This is higher than the census figure of 8.5%. The census figure has, however, been widely criticized for undercounting slums (Bhan and Jana 2013). About 53 percent of our sample lived in HT 3, that is, lower middle class housing.⁸ HT 4, which we call middle class, is also quite large, accounting for 29.8% of our sample. If housing type, relative to assets and occupation, is a better indicator of class, it becomes very clear that Bangalore has a very sizeable middle class.

Figure 1:

Class Distribution (Using Housing Type)

The Citizenship Index (CI):

As discussed earlier, the effective exercise of citizenship requires having the necessary knowledge and being able to participate in public life. Our citizenship index, thus, has two components: knowledge and participation. To capture each of these we asked a series of questions and developed specific measures.

Knowledge of civil and political affairs was relatively easy to capture: for political/electoral knowledge we asked if the respondent knew which parties and individuals held which positions (i.e. which party or coalition rules at the national and state levels) and for civic knowledge we asked if they knew about different opportunities for participation (e.g. awareness of ward meetings), and if they knew which agencies delivered which services (water, electricity, sewerage etc).

Participation refers to specific forms or instances of direct involvement in political and civic life. The participation index is composed of three dimensions of participation: voting activity, non-voting political participation (political participation henceforth), and civic participation. Voting focuses on whether a respondent voted in the three recent elections (the 2009 parliamentary elections, 2013 state assembly elections, and the 2010 local elections). Political participation refers to a respondent’s political activities outside of voting, i.e. participation in elections and rallies and contributions to political parties. Civic participation measures a respondent’s civic involvement, i.e. participation in neighbourhood redressal of common problems, participation in a variety of associations and participation and frequency of participation in local ward meetings. The overall citizenship index is an aggregation of the average scores for the two components of knowledge and participation.

Before turning to CI and its component parts, we want to make a few descriptive observations. At all levels of elections, Bangaloreans vote in high percentages: 78% at state level, 71 % at the municipal level and 70 % at national level, with the highest rates of voting occurring among the poorest respondents in the sample. Political knowledge is also high. About 83 percent respondents answered the question “ruling party at state-level’ correctly and 85 percent answered “ruling party at national-level” correctly. But only 35 percent respondents knew the name of their municipal corporator. The level at which citizens are most likely be able to use their rights - the local or municipal level - is precisely the level at which they have the least political knowledge. Almost certainly, this reflects how weak local government has historically been at the local level.

But when it comes to participation in politics beyond the voting booth, Bangaloreans are once again not very active. Less than 10 percent contribute time to political campaigns in municipal elections and less than 10 percent participate in politics outside elections. Especially in local politics, the space in which classical democratic theorists from Locke to Gandhi have argued the skills and virtues of citizenship are forged, there clearly continues to be a massive deficit. Thus, 93 percent of our respondents reported that they did not know if there was a ward committee in their community, and only approximately 3 percent reported to having attended a ward meeting, again no doubt reflecting the anemic nature of local institutions of representation. In sum, Bangaloreans vote a lot, know something, but don’t do all that much beyond electoral participation.⁹

The CI consists of both the knowledge and participation measures weighted equally. The index can take on values in a 0–1 range. The mean is 0.34, indicating the typical citizenship of a respondent in Bangalore. To get a substantive sense of what this means, recall that this index is based on 12 questions that focus on knowledge about national, state, and local political actors, institutions and state service provision agencies, and 9 questions on voting, political and civic participation, with each set of questions being equally weighted, as is each component. As such, a perfect score would require answering all questions positively. The mean score indicates that a respondent with mean citizenship tends to vote in two (and sometimes three) elections, participates in one political or civic activity, and has some knowledge about political actors (typically national and state political actors) and state agencies (about 2–3 key ones such as water, electricity, and transportation). Participation, in the form of attending meetings and rallies organized by political parties during or between elections, is typically low. While civic participation in specific caste, religious, or voluntary associations is also low, respondents occasionally participate in neighbourhood meetings that address service problems.

We turn to the CI’s distribution across each of our control variables: education, caste, religion, and housing type. These are reported in Tables 2–5 below. In order to tease out the relationships between the CI and the control variables, we recode the CI into a discrete binary variable. Respondents who score above the mean CI value are coded as having ‘high’ citizenship and those with CI values equal to or less than the mean are coded as having ‘low’ citizenship.¹⁰ This enables us to isolate patterns across the control variables (which are discrete and nominal) clearly and intuitively.¹¹

TABLE 2:

Distribution of CI by level of Education

	Education Level (Respondent)
CI	No Schooling	Primary School	Middle School	Secondary School	College & Above
Low	71	62	60	44	37
High	29	38	40	56	63

TABLE 5:

Distribution of CI by Gender, Location, Migrant Status

	Gender		Location		Migrant
CI	Female	Male	Inner	Outer	Migrant	Non-Migrant
Low	54	39	46	52	56	41
High	46	61	54	48	44	59

Table 2 reveals, as one might expect, that there is a very clear and linear relationship between citizenship and education. The lower one’s educational level the lower the level of effective citizenship. Those with no schooling are most likely to have low citizenship. Those with secondary schooling and college are much more likely to fall into the high categories of citizenship. Caste appears to have some effect on citizenship, with the Dalits and Adivasis having lower citizenship than upper castes. (Table 3). The same is true for religion (Table 3). Muslims are slightly more likely to have lower effective citizenship, and Christians are slightly likely to have higher citizenship, but the differences across the three religious groups are not pronounced.

TABLE 3:

Distribution of CI by Caste and Religion

	Caste			Religion
CI	Dalit/Adivasi	OBC	Upper Caste	Hindu	Muslim	Christian/Others
Low	53	48	43	47	51	44
High	47	52	57	53	49	56

Household type (HT), which is our selected proxy for class, has a very strong association with the distribution of citizenship (Table 4). The majority of those living in HT1 (shacks) have low levels of citizenship and only about 15% score high levels of citizenship. Those living in designated slums also score much lower levels of citizenship than the middle classes (HT3 and HT4) with only 31% having high citizenship. But the overall relationship between class and citizenship is not perfectly linear. Thus, those living in the highest category of housing – the upper class – in fact display lower levels of citizenship than the middle class (HT4) and are really only marginally higher than the lower middle. In sum, the bottom of the class hierarchy has lower than average citizenship, the middle has high citizenship, and the citizenship of the upper classes tapers off somewhat.

TABLE 4:

Distribution of CI by Class

	Household Type
CI	HT1	HT2	HT3	HT4	HT5
Low	85	68	48	36	45
High	15	32	52	64	55

Tables 5 explore the distribution of citizenship across gender, location and migrant status. Though all three of these categories impact the distribution of citizenship in the direction one might have anticipated, the gap between women and men, and between migrants and non-migrants – roughly 16% in both cases - is especially high.

The CI is a highly aggregated measure. When disaggregated into its component parts, we find that knowledge and participation are unevenly distributed across caste, community and housing types, but move in opposite directions. While we do not present the disaggregated tables here, we briefly summarize the observed patterns in the data.¹²

About 44 percent of all respondents fall into the high knowledge category and 67 percent in the high participation category. Hindus, upper castes and higher housing types have more knowledge, but participate less compared to Muslims, Dalits/Adivasis and lower housing types. What is striking is that participation has an equalizing effect. For instance, only 35 percent of the Dalits and Adivasis indicate high knowledge in our sample, but 72 percent exhibit high levels of participation. The numbers are very similar for Muslims as well: only 36 percent have high knowledge, but participation among them is very high, about 71 percent. Finally, while only 18 percent of respondents from designated slum exhibit high knowledge, 68 percent participate in political and civic life. Indeed, the greater propensity of the poor and the relatively marginalized social groups to participate in both political and civic life goes a long way in closing the knowledge gap.

The Basic Service Delivery and Infrastructure Index (BSDII)

Our dependent variable, BSDII, covers water, sanitation, electricity and roads. Each of these carries the same weight in the index. Water provision service is based on five dimensions: source, usability, convenience, gaps in supply, and consistency. The indicators for electricity provision include whether a household has an electricity connection and the number of gaps in power supply experienced by the household. The indicators for sanitation capture whether a household has own toilet, or shared/community toilet, or whether the toilet is an open or shared pit, or open defecation is practiced. Similarly, the measure for infrastructure, i.e. roads, is based on three dimensions: whether the road is unpaved (kuccha) or paved (pucca); in good or poor condition; and if water gets logged during monsoon. The BSDII is, thus, a simple aggregation of these 12 questions.

In this section we provide the overall distribution of BSDII across our socio-economic factors. BSDII is a continuous measure that ranges from 0 to 1.0 and has a mean of 0.648 and standard deviation of 0.189. Households that score above the mean BSDII value are coded as having ‘high’ public services and those with BSDII values equal to or less than the mean are coded as having ‘low’ public services. A perfect BSDII score (1.0) would translate to having quality and convenient water supply with no interruptions, electricity with very infrequent interruptions, excellent roads and drainage, and good sanitation, specifically in-house flush toilets that are connected to sewage systems. A household with an average BSDII score can expect to have a public source of water that is located inside the premises (a tap or hand-pump) with some gaps in provision. Water is typically used for a single purpose (either general use or drinking, mostly the former), and some of these households are likely to have water storage. These households have a metered power connection and typically face power outages between 4 to 6 hours a week, have flush toilets inside the house (as opposed to a community toilet or pit toilet), and are located in areas with roads that tend to be in good (pucca) condition, but with likely poor drainage during monsoon. As the distribution in Figure 2 shows, about 44 percent of households receive low (i.e. below the mean) public services and 56 percent are characterized as with high services.

Figure 2:

Distribution of BSDII

How is this unequal distribution of services related to our basic socioeconomic control variables? Figure 3 shows a strong and linear relationship between education and quality of services. This no doubt reflects the tight relationship between education and class. Figure 4 confirms what one might have predicted, namely that the Dalits/Adivasis receive much lower services than OBCs and upper castes (the difference between OBC and upper caste households is marginal). The fact that about 60 percent of the Dalit/Adivasi households get poor services, compared to 40 percent of the OBC, suggests that a good portion of SCs live in ghettos. In contrast, Figure 5 suggests that religion does not seem to have any statistically significant relationship to services (for instance, 44 percent of Hindu households get poor services compared to 47 percent Muslim households and 46 percent Christian households). Muslim households are as well serviced as any other religion.¹³

Figure 3:

Distribution of BSDII by Level of Education

Figure 4:

Distribution of BSDII by Caste

Figure 5:

Distribution of BSDII by Religion

When we look at BSDII across housing types, we get our strongest finding yet. The relationship here is very linear (Figure 6). It is not surprising that about 90 percent of households in informal settlements and 73 percent in slums receive low level services. In contrast about 77 percent of upper class households and 71 percent of middle class households receive higher level services.

Figure 6:

Distribution of BSDII by Class

MODELS AND RESULTS

We model basic service delivery and infrastructure provision as a function of citizenship and socio-economic controls including class, caste, religion, education of respondent, migrant status, and location of household (in inner or outer wards). The relationship between citizenship and basic service delivery and infrastructure is estimated using Ordinary Least Squares (OLS) regression of the form:

$${BSDII}_i={\beta }_0+{\beta }_1{Citizenship}_i+{\beta }_2{Controls}_i+{\epsilon }_i$$

The results are reported in Tables 6 and 7, (models 1–4).¹⁴ Models 1 and 3 are baseline models for the Hindu sample and the larger sample including all respondents, respectively. The baseline models include only the control variables.¹⁵ These baseline models (1 & 3) confirm what emerged from Figures 3–6. Our class variable is statistically significant and follows expected signs. Informal settlements and households in notified slums (HT 1 and 2) exhibit significantly lower levels of public service provision relative to lower middle housing (the reference category in our models), while middle and upper class housing indicate higher levels. We also find that basic service delivery and infrastructure for the Dalit and Adivasi households are significantly lower than OBC households (the reference category) while there is no difference between OBC and forward caste households. We also find that service provision and infrastructure in Muslim households is not statistically different from non-Muslim households (Model 2).¹⁶

TABLE 6:

The Effect of Citizenship on Basic Service Delivery and Infrastructure

Dependent Variable: Basic Service Delivery and Infrastructure Index
Independent Variable	1 Baseline Model (Hindus Only)	2 Baseline Model + Citizenship (Hindus Only)	3 Baseline Model (All Sample)	4 Baseline Model + Citizenship (All Sample)
Class
HT 1	−0.302 (0.027)***	−0.294 (0.027)***	−0.299 (0.026)***	−0.292 (0.026)***
HT 2	−0.103 (0.012)***	−0.099 (0.012)***	−0.121 (0.010)***	−0.118 (0.010)***
HT 4	0.057 (0.007)***	0.057 (0.007)***	0.065 (0.006)***	0.063 (0.006)***
HT 5	0.071 (0.013)***	0.072 (0.013)***	0.088 (0.011)***	0.089 (0.011)***
Caste
SC/ST	−0.048 (0.009)***	−0.047 (0.009)***	-	-
Upper Caste	0.0006 (0.008)	0.0006 (0.008)
Religion	-	-
Muslim			0.001 (0.007)	0.003 (0.007)
Education Level
Primary	0.022 (0.022)	0.021 (0.022)	0.017 (0.018)	0.016 (0.018)
Middle	0.017 (0.014)	0.016 (0.014)***	0.010 (0.011)	0.010 (0.011)
Secondary	0.058 (0.012)***	0.054 (0.012)***	0.048 (0.010)***	0.046 (0.010)***
College & Above	0.081 (0.012)***	0.078 (0.012)***	0.079 (0.010)***	0.078 (0.010)***
Location (Outer Ward)	−0.046 (0.007)***	−0.046 (0.008)***	−0.050 (0.006)***	−0.049 (0.008)***
Non-Migrant	0.004 (0.006)	0.002 (0.006)	0.002 (0.005)	0.002 (0.005)
Citizenship	-	0.035 (0.019)	-	0.025 (0.017)
Constant	0.614 (0.013)***	0.605 (0.014)***	0.609 (0.010)***	0.602 (0.011)***
Observations	2871	2804	4041	3943
F	70.61	63.40	95.61	85.40
Root MSE	0.164	0.165	0.167	0.167

Robust standard errors in parentheses

^***p<0.01

^**p<0.05

^*p<0.1

TABLE 7:

Conditional Effects of Citizenship on Basic Service Delivery and Infrastructure

Dependent Variable: Basic Service Delivery and Infrastructure Index
Independent Variables	5 (Hindus Only)	6 (All Sample)	7 (Hindus Only)	8 (All Sample)
Citizenship	0.028 (0.019)	0.015 (0.021)	0.049 (0.021)	0.036 (0.019)
Class	−0.212 (0.023)***	−0.218 (0.019)***	−0.223 (0.024)***	−0.228 (0.020)
Citizenship* Class	0.291 (0.081)***	0.244 (0.066)***	0.272 (0.082)***	0.200 (0.067)***
Caste
SC/ST	−0.057 (0.009)***	-	−0.044 (0.018)**	-
Upper Caste	0.003 (0.008)	-	-	-
Citizenship* (SC/ST)	-	-	−0.083 (0.050)	-
Religion
Muslim	-	−0.0006 (0.007)	-	−0.008 (0.017)
Citizenship *Muslim	-	-	-	−0.025 (0.045)
Education
Primary	0.031 (0.023)	0.022 (0.019)	-	-
Middle	0.028 (0.014)	0.017 (0.011)	-	-
Secondary	0.065(0.012)***	0.056 (0.010)***	-	-
College & Above	0.105 (0.012)***	0.109 (0.010)***	-	-
Non-Literate	-	-	−0.129 (0.025)***	−0.092 (0.021)***
Citizenship* (NonLiterate)	-	-	0.233 (0.079)***	0.128 (0.067)*
Location (Outer Ward)	−0.057 (0.007)***	−0.061 (0.006)***	−0.057 (0.007)	−0.059 (0.006)***
Constant	0.621 (0.014)***	0.617 (0.011)***	0.696 (0.008)***	0.682 (0.007)***
Observations	2804	3943	2804	3943
F	63.34	85.65	70.37	75.58
Root MSE	0.167	0.171	0.169	0.173

Robust standard errors in parentheses

^***p<0.01

^**p<0.05

^*p<0.1

Households with respondents having a secondary or higher level of education are associated with higher levels of basic public services relative to households where respondents had no schooling. We don’t observe any statistical differences in public services to households with respondents below the middle school. In sum, education is positively correlated on access to infrastructure and there is also a clear threshold effect. That is, it only makes a difference once one is educated above the middle school.17

We also find that households in wards in the outer areas of Bangalore are characterized by lower levels of basic service provision and infrastructure relative to those in wards that lie in the inner city. However there is no statistical difference between non-migrant (those who have lived their entire lives in Bangalore) and migrant households when it comes to service provision and infrastructure.

Models 2 and 4 include the variable of interest, citizenship.¹⁸ We find that the effect of citizenship, while positive, is not statistically significant in either model. This implies that citizenship has no effect on the levels of public services a household receives. The introduction of citizenship does not change any of the effects of the control variables. The coefficients for the control variables are stable, consistent with expectations, and statistically significant.¹⁹

This general relationship however must be significantly qualified when we introduce the second set of models 5 through 8 presented in Table 7. In models 5 – 8, we test for conditional effects of citizenship on public services. That is, instead of a constant effect on service delivery and infrastructure levels across all housing types, the effect of citizenship is expected to vary across class. We estimate a set of multiplicative interaction models that estimate the effects of citizenship on public services conditional on: (a) class, (b) education (c) caste and (d) religion.²⁰

Specifically, we anticipate citizenship to have a larger (positive) effect on public service delivery for poor households relative to the wealthier households, who can get by without political participation and making demands politically. Similarly, we anticipate citizenship to mitigate the effects of lower levels of education, lower caste, and religious minority (i.e. Muslim) status on basic service delivery and infrastructure.

All models in Table 7 show that citizenship conditional on the lowest housing types has a significant effect on service delivery and infrastructure. That is, an increase in effective citizenship of respondents living in the lowest housing types correlates with an increase in the level of basic service and infrastructure those households receive (relative to the wealthier households i.e. HTs 3, 4 and 5). ²¹ Table 8 presents the marginal effects of citizenship conditional on (a) class (HT1 and HT2) (b) education (non-literate) (c) caste (Dalit/Adivasi households) and (d) religion (Muslim households) along with standard errors, and 95% confidence intervals.

Table 8:

Marginal Effects, Standard Errors, and 95% Confidence Intervals

	Marginal Effect of Citizenship	Standard Error	95% Confidence Interval
Class
HT1 & HT2	0.24	0.07	0.10 to 0.37
Education
Non-Literate	0.16	0.07	0.03 to 0.29
Caste
Dalit/Adivasi	−0.03	0.04	−0.12 to 0.06
Religion
Muslim	0.01	0.04	−0.07 to 0.09

We find that the marginal effect of citizenship on public service delivery conditional on class is positive and statistically significant, and ranges from 0.37 to 0.10 (from model 8). The poor have less of citizenship and less of public service delivery and infrastructure, but they get more services and infrastructure for their citizenship than others. The marginal return to citizenship is higher for the poor. For instance, a standard deviation increase in citizenship is associated with an approximately ten percent increase in infrastructure and services for the poor.²²

Citizenship has similar effects for those without schooling. The marginal effect of citizenship on services for non-literate households is positive, greater than that for literate households, and statistically significant. The magnitude of this relationship is smaller than that for class. However, we find that a conditional effect does not exist for Dalits and Adivasis (from model 7) or Muslim households (from model 8). While citizenship mitigates the effect of class and illiteracy, it does not seem to do the same for caste, particularly Dalits/Adivasis or for religion.

Conclusion

Historically, cities have been associated with greater associational freedom and more social and economic opportunity. In India, constitutional guarantees and political practices have secured basic political and civic rights. But social rights have only recently been made constitutional rights, and this does not include the basic services that most urban residents expect. This then leads to two important questions. First, can all citizens, irrespective of their socio-economic status, use these civic and political rights effectively? Second, to what extent can citizens secure basic services as a matter of rights? Can citizenship, as Marshall so famously argued, abate the effects of class and, more broadly, social exclusion?

We addressed these two core questions on the strength of a survey of over 4,000 households in Bangalore. On the whole, the answer to both questions would appear to be negative. On the one hand we found that citizenship in practice is very unevenly distributed, and that this distribution closely tracks class, caste, religion and gender. On the other hand, we found that basic services and infrastructure are highly unevenly distributed and that class, though not caste and religion, drive much of this effect. Given that Bangalore has not only been the poster child of India’s recent economic success - indeed a global icon of the information technology revolution – and it has also generally been perceived as India’s best governed mega-city, it is alarming that such large swaths of the city are deprived of adequate services.

Taken together, our finding of highly uneven patterns of service delivery and clear evidence of class-based social exclusion might suggest that citizenship doesn’t make a difference, or worse yet, that levels of citizenship reflect and reinforce social inequality. Women, Muslims, Dalits/Adivasis and lower classes enjoy less effective citizenship than men, Hindus/Christians, OBCs/upper castes, and middle and upper classes. This is not entirely surprising, and supports arguments in the literature, most notably by Chatterjee (2004), that citizenship in India is largely the preserve of elites. But lurking behind this aggregate finding are some patterns that suggest a more complicated picture.

Since we disaggregate citizenship into knowledge and participation, we are able to statistically identify that the two components -- knowledge and participation -- work in opposite directions. Socio-economic difference, including gender, drives significant differences in knowledge. The more privileged one is, the more one knows about the system and presumably how to use it. Participation works in quite the opposite direction, with the poor, Dalits, Adivasis and Muslims participating much more than the rich (who in fact participate very little), OBCs/upper castes and Christians and Hindus. Participation is the lifeblood of citizenship for the poor. This supports an existing body of literature that has found that the poor and lower castes are far more active electorally than the rich and upper castes (Yadav and Palshikar 2009).

But our final conclusion is far and away the most important one. While the urban poor have lower effective citizenship than the middle class, the poor get more out of their exercise of citizenship than the middle class, and specifically that if it were not for the citizenship they do have, they would have less access to basic services and infrastructure. In sum, the poor suffer from citizenship deficits as well as public service and infrastructure deficits, but these latter deficits would be greater without the poor exercising their citizenship rights. While citizenship has not closed the gap between the classes, it does make a significant difference for the poor. Citizenship significantly abates class in Bangalore. Only further research will establish whether this and other findings of this study would hold in urban India in general.

Appendix 1: Variable Description and Summary Statistics

Variable	Description	N	Mean	SD	Min	Max
Basic Service Delivery and Infrastructure Index (BSDII)		4041	0.65	0.18	0	1
Water	Provider of primary source of water: Private=0; Public=1• Location of primary water source: Outside=0; Inside=1• Water storage system for primary water source: No=0; Yes=1 • Gaps in supply: No=0;Yes=1• Water use: Only general or drinking (0); Both (1)
Power	Metered electricity connection: No=0; Yes=1•Frequency of power cuts: More than 18 hours =0 to No power cuts=1
Sanitation	Pit (own or shared) or open defecation (=0) •Community septic tank, flush latrine or dry latrine (=1) •Septic tank/flush latrine-own or shared (=2)
Roads	Type of road: Unpaved=0; Paved=1 •Road Condition: Poor=0; Good=1 •Water logging in monsoon: No=0; Yes=1
Citizenship		3994	0.32	0.16	0	1
Political Knowledge	Name of party or coalition of parties is currently ruling at the national level: Incorrect=0; Correct=1•Name of party/coalition of parties is currently ruling at the state level: Incorrect=0; Correct=1•Name of Corporator (of respondent’s ward) Incorrect=0; Correct=1
Civic Knowledge	Name of (respondent’s) ward Incorrect=0; Correct=1•Name of public agency responsible for providing: (a) Water Supply (b) Electricity (c) Public Transport (d) Traffic: Incorrect=0; Correct=1•Purpose of Right to Information Act: Incorrect=0; Correct=1
Electoral Participation (Voting)	Voted in 2010 BBMP Elections: No=0; Yes=1•Voted in 2013 Karnataka State Assembly Elections: No=0; Yes=1•Voted in 2009 Lok Sabha Elections No=0; Yes=1
Political Participation	Respondent (or someone in household) contributes time to campaigns during municipal elections: Never=0; Always/ Sometimes=1•Respondent (or someone in household) participates in meetings or rallies organized by political parties or officials outside of election time: Never =0; Always/ Sometimes=1 •Respondent (or someone in household) talks to friends, neighbors or others in the community about supporting a candidate: No=0; Yes=1.
Civic Participation	Respondent (or someone in household) participates in (a) Non-government organizations (b) Resident Welfare Associations (c) Caste organizations (d) Religious organizations (e) Non-caste, non-religious organizations: No=0; Yes=1•Respondent (or someone in household) attended ward committee meetings: No=0; Yes=1.
Housing Type (HT)		4093
Informal Settlement (HT1)	Self-built dwelling often made from: reclaimed wood, fabric, tarpaulin, corrugated metal, and/or sack-cloth. Not located on street-fronts, often located in vacant lots, behind buildings, on sidewalk, road medians, small green spaces, under overpasses, and construction sites. Sometimes also located in larger vacant or abandoned/under-construction non-self-made structure, but using self-made materials within that building (such as tents). Almost always single floor single room dwellings.		0.01	0.13	0	1
Notified Slum (HT2)	One-room pucca row house with corrugated metal roof and densely packed. Typically located behind buildings, in gullys, and not on main street. Few windows, small windows, with shutters and single entrance.		0.11	0.31	0	1
Lower Middle (HT3)	Single or multi-floored concrete (only) structures, with 2–3 rooms. If housed within an apartment building, they generally have shared balconies, small windows, publically accessible staircases outside, no gate, wall, or security, and may have commercial units on the ground floor.		0.52	0.49	0	1
Middle (HT4)	Independent house or apartment building and often a shared dwelling between independent family units indicated by multiple mailboxes and different entrances. Gate present but usually no high-wall present around house. Apartment buildings often have outdoor staircases, may have a gate entrance to building but generally not part of a complex or gated community. Mostly concrete structures but some have additional materials such as glass, wood, and/or brick. Apartments often have private balconies.		0.29	0.45	0	1
Upper Class (HT5)	Independent house or apartment building, often constructed using concrete, wood, glass with a surrounding wall and gate in front of house, and security guarding entrance. Outdoor staircases are rare, and size of individual apartments is large with multiple balconies for one apartment and large windows		0.04	0.20	0	1
Caste (Hindu Only)	Self-classification into official categories of Scheduled Castes or Scheduled Tribes (ST&ST), Other Backward Castes (OBC), and Forward Castes (FC)	2911
Forward Caste			53.7		0	1
Other Backward Classes			17.6		0	1
Scheduled Caste & Tribe			28.7		0	1
Religion	Self-classification into official categories of: Hindu, Muslim, Christian, Sikh, Jain, and Buddhist	4092
Hindu			72.9		0	1
Muslim			18.0		0	1
Christian			8.6		0	1
Other (Sikh, Jain, Buddhist)			0.3		0	1
Education		4089
No Schooling			11.0		0	1
Primary			3.25		0	1
Middle			14.9		0	1
Secondary			39.2		0	1
College and Above			31.6		0	1
Location	Respondent: lives in outer ward (=0); inner ward (=1)	4093	0.21	0.40	0	1
Migrant	Respondent: has always lived in Bangalore (=0); Migrant (=1)	4093	0.56	0.49	1	1