1

Examining Urban Inequality and Vulnerability to Enhance Resilience: Insights from Mumbai, India

Patricia Romero-Lankao, Daniel M. Gnatz and Joshua Sperling

Climatic Change, 2016 (In press)

1. Introduction

Common to vulnerability and resilience research and policy intervention is the concern that differences in the capacity to respond to floods, social instability and other stresses depend on differences in socioeconomic status, springing from social inequality, i.e., from differential access to the asset base from which response actions and investments can be made (Ribot 2010; Harlan et al. 2007). In this context, an understanding of how households, ranging from poor to wealthy differ in levels of vulnerability and of the mechanisms creating this differential will be fundamental to creating fair and effective interventions to enhance resilience across populations. A complex problem exists, however, in sorting out the relative influences of various indicators of wealth and vulnerability, particularly in urban areas from low and middle income countries. Much of this complexity is rooted in the fact that both wealth and vulnerability are multivariate concepts with differences in definition and scope, often dependent on disciplinary perspective, and a range of different methods to quantitatively characterize the relative importance of indicators (Vyas and Kumaranayake 2006; Romero-Lankao et al., 2012). The development of methods that can constructively engage with these differences may help move diffuse equity and capacity goals towards cohesive and policy relevant narratives. Two main methods have been used to categorize households according to wealth (poverty) and vulnerability. The first aggregates indicators into indices, with many studies assuming, for the sake of simplicity, that all indicators should be weighted equally (Cutter, 2003). This creates uncertainties associated with indicator selection, index construction, and the relative influence of the chosen indicators (Guest 2000; Vincent 2007; Eakin and Bojorquez-Tapia 2008). The second uses principal components analysis (PCA) to derive indices that help cluster households into groups that reflect different socioeconomic or vulnerability levels (Gwatkin et al. 2000; Filmer and Pritchett 2001; McKenzie 2003; (Cutter et al. 2008)). However, by reducing the number of variables under consideration, PCA loses, at least in part, the multidimensional nature of both social inequality and vulnerability. In this paper we combine a fuzzy logic approach with an analytic hierarchic process (ANH), to examine the relative importance of select wealth (poverty), exposure, sensitivity and capacity indicators as they affect differentiated vulnerability to climate hazards in Mumbai, India. We apply a livelihoods framework to characterize urban households by the resources or assets that comprise their livelihoods. While ANH provides us with a robust procedure for analyzing the relative importance of particular indicators of wealth and vulnerability, fuzzy logic helps address the uncertainties involved in assigning any household to a particular class.

2

2. Inequality and vulnerability/resilience

2.1 Concepts

For centuries, concepts and theories of social inequality have been the subject of compelling philosophical, conceptual and ethical discussions. Definitions of what social equality means and what a fair and just situation entails, have provided philosophical and ethical frameworks for analyses of social inequality ((Cowell 1998)). However, disagreements in definition, scope and policy implications have persisted. Of relevance are studies examining the effects of social inequality on a vast array of outcomes, e.g., health, economic performance, risk and vulnerability (Adger 1999; (Blaikie et al. 2014); (Wilkinson and Pickett 2006)), and measuring social inequality with indicators of income, consumption, and assets (Filmer, Scott 2013). Social inequality will be defined here as a condition that ensues when resources in a city are distributed unevenly across households, typically through institutional mechanisms that engender socially defined categories of wealthy or poor persons and communities. We will focus on differences in the assets and options (capacities) that urban households have to pursue the livelihoods and life they value and to respond to the hazards and stresses they encounter. Vulnerability and resilience have also been subject to many definitions and approaches. Of relevance here is that in measuring urban vulnerability the choice of indicators depends on the theoretical approach used. Urban vulnerability is frequently measured within the vulnerability as impact approach, by quantifying the relationship between hazards (e.g., temperature) and outcomes (e.g., illness), while controlling for factors such as age and gender (Makri & Stilianakis, 2008; O’Neill, 2005). This approach does not explicitly look at how the structures of society, such as inequality or determinants of political power influence differences in populations resilience (Ribot 2010; Adger 1999). Rather, it tends to view climate-hazards as a physical phenomenon. Livelihoods and political ecology (i.e., contextual) approaches, on the other hand, examine the structural drivers of differences in populations’ vulnerability to hazards within urban areas, and the interactions among climatic, socioeconomic, and political components of hazard-exposure, sensitivity, and adaptive capacity (Moser, 1998). In this context, household vulnerability is examined in terms of assets, or resources households and communities can draw upon (e.g., self-help housing) to pursue their livelihoods, build their wellbeing and to respond to the hazards they encounter (Collins and Bolin 2009; Manuel-Navarrete, et al. 2011). These indicators are hierarchically structured into multi-criteria models (Baud, et al., 2008; C. Moser 2008; Romero-Lankao et al. 2014). We define the vulnerability of urban populations as the propensity to be adversely affected, and their resilience as the capacity to perceive risk and successfully adapt and recover in the face of stresses (Field C.B. et al., 2014). To be vulnerable, an urban population must not only be exposed and or predisposed (sensitive), it must also have limited capacity to adapt and mitigate hazards (Romero-Lankao and Qin 2011). Hazards, or stresses result from social and environmental mechanisms (e.g., landslides resulting from land use changes induced by urbanization). Hazards interact with patterns of socio-economic segregation, placing some

3

people in hazard-prone areas and endowing different urban populations with access to resources and options differently (Harlan et al. 2007). Our framework takes advantage of an overlap and cross fertilization between capabilities and livelihoods approaches, as noted by Adger (2006), to look at how access to assets affect capacity. The relationship between vulnerability and inequality is not a simple one. Within this relationship, vulnerability and resilience seem to act as inversely related outcomes, the purest states of which, can be seen as extreme ends of a continuum, with a place on that continuum defined by exposure, sensitivity and capacity to cope with or successfully respond to hazards. Among each of these there are sets of conditioning factors that can be related to social inequality. These factors include sociodemographic dynamics; the legacies of past political (governance) decisions and policies around urban land use planning, infrastructures and services; the economic and political mechanisms of social exclusion; and the role of energy, water and other utility networks in determining resource access, environmental problems and the capacities of urban populations. This study examines a select group of these factors, and the uneven distribution of assets and options, to see how they may affect levels of vulnerability differently across household vulnerability classes.

2.2 Measures Indicators are fundamental tools that can be used to support prioritized policies seeking to enhance populations’ capacities, yet scholars are still grappling with how best to determine the most appropriate indicators. Social inequality is often measured with three wealth indicators: income, expenditure and consumption. However, it is not easy to collect income data that captures the diversity of household livelihood strategies. Many urban populations, for instance, have unreported income and at least a portion of their livelihoods supported by barter or trade. Members of these populations may be self or transitorily employed (Vyas and Kumaranayake 2006; McKenzie 2003). While consumption or expenditure measures are easier to collect and relatively more reliable than measures of income (Filmer and Pritchett 2001), they also require extensive and expensive data collection efforts. To overcome this, scholars have increasingly advocated the use of asset indicators, an approach we will use in this paper (Filmer and Pritchett 2001; C. Moser and Felton 2007) Social inequality and vulnerability indices, used by both inequality and vulnerability scholars, can serve as heuristic tools to examine household membership within specific status classes and household structural features, such as assets, options and perceptions associated with different capacities to respond to hazards (Baud, et al. 2008; Moser, 1998). A few caveats need to be kept in mind, however, given the dynamic nature of social inequality and vulnerability. The use of indices to classify urban households may not hold for those households over time; although survey data are better than census data to capture the multidimensionality of social inequality and vulnerability, they cannot measure top percentiles of wealth or vulnerability accurately because of the small sample size of wealthy and very poor households; and robust methods are needed for assigning weights in the aggregation of indicators (Vyas and Kumaranayake 2006). Unfortunately, the most common approach used in index construction is to assign equal weight to each indicator of wealth and vulnerability. While this method has the virtue of simplicity, it

4

often creates overgeneralization. A common practice to overcome these limitations has been the use of principle components analysis (PCA) to aggregate ownership and asset variables into a single dimension (e.g.,(Filmer and Pritchett 2001; Qin et al. 2015)). However, this method runs the danger of reductionism, because the aggregation cannot capture the multidimensionality of social inequality and vulnerability, nor the portfolio of assets and options households draw on to create their livelihoods and respond to hazards and adversities. Vincent (2007) and Eakin and Bojorquez (2008) noted sources of uncertainty derived from the creation of vulnerability indices that can apply to inequality indices as well. Indices of vulnerably and social inequality run the risk of creating uncertainty through: the theoretical assumptions that form the basis for their selection of determinants and framing of relationships between drivers of social inequality and vulnerability; the selection of indicators, which may or may not be sufficient to describe the dynamic nature of social inequality and vulnerability; the weights assigned to each indicator; and the ranking of households into vulnerability (or wealth) categories. In this paper, we apply two procedures to overcome these limitations, Analytic Hierarchic Process (ANH) a multicriteria decision analysis tool to weight inequality and vulnerability indicators; and fuzzy logic to assign households to wealth and vulnerability groups (Rashed and Weeks 2003; Tah and Carr 2000; Eakin and Bojorquez-Tapia 2008; Giordani and Giorgi 2010). While ANH offers clear steps to weight indicators based on expert knowledge, theory and empirical research, fuzzy logic is useful to address the imprecision, uncertainty and complexity associated with both social inequality and vulnerability (Bojorquez et al., 2002; Eaking and Bojorquez, 2008). Our goal is to classify urban households in the city of Mumbai into poverty and vulnerability classes in order to examine the distribution, relative importance and relationships between socioeconomic status and vulnerability variables within and across study populations. Studies of this kind can help decision makers create targeted policies aimed at increasing resilience among their cities’ most vulnerable populations. To group urban households, we use capabilities and livelihood approaches, and define capacity as an array of resources, assets and options households draw on to manage risk and pursue the lives they value. These approaches give consideration to various mechanisms of social exclusion in urban areas - e.g., along lines of class, caste, or formal and informal status (Baud et al. 2010; C. Moser 2008; Romero-Lankao et al. 2014). They also provide insights into the ways in which environmental and socioeconomic hazards affect urban livelihoods and the assets and options households use to respond to risks.

3. Mumbai Mumbai, the capital of Maharashtra state, is located on the western seacoast of India. With an estimated metro area population of 22 million people, an average altitude of 10-15 meters above sea level (with many areas at or below), and a large part of its land reclaimed from the sea, the city is at risk from sea level rise, tropical cyclones, gusty winds, coastal erosion and flooding(Ranger et al. 2011). It has a tropical climate that has changed over the last century ((Field C.B., et al. 2014; Hijioka et al. 2014)et al., 2014), and is projected to have further increases in the number of warm days, and the frequency of heatwaves (IPCC, 2014), rising sea

5

levels and warmer and wetter monsoon seasons (Bhagat, Guha, and Chattopadhyay 2006; Ranger et al. 2011). The city usually receives 50 percent of its July and August rainfall in two or three overwhelming events (Jenannami 2006), with both natural and human factors underling its sensitivity to these hazards (Bhagat 2006; Ranger 2011). As a rapidly developing world mega-city facing multiple interacting climate risks and development challenges, a method for understanding Mumbai might also help understand development and climate concerns in other global cities. Driven by the promise of a “city of dreams” where “no one goes hungry” (MCGM 2010), migration of rural populations has led to rapid demographic growth. While Mumbai has become a center for outsourcing by international business organizations (Kapil Gupta 2007), development patterns have consolidated an urban structure where the western side of the city tends toward a wealthier population than the east, and the percentage of people employed in the informal economy has skyrocketed to about 70 percent since the 1980s (MCGM 2014)). From 41.9 to over 50 percent of Mumbai’s population lives in informal settlements (slums) (MCGM, 2014). These slums have limited or no-access to capacity determinants such as legal status, land, housing, electricity, clean water, and toilets (Patankar et al 2010: 10). Thus, the dynamics of economic growth and governance have shaped inequality and vulnerability. Mumbai has profound deficits in sanitation and health, with the vast majority of slum dwellers depending on public toilet blocks, infant mortality rates of 37 per 1000 and one out of four children stunted (Gupta, et. al. 2009). While its literacy rate is 85.6 percent, higher than India overall (65.4 percent), socioeconomic and gender inequalities persist as do inequalities in access to housing (MCGM, 2014: 199). Such high levels of social inequality, which negatively affect household capacity, result from structural mechanisms of exclusion, often conditioned by unfavorable attitudes towards low-income settlements (Zérah 2008). For instance, while during Colonial times, both real and imagined threats of disease and overcrowding were fundamental drivers behind attempts by Mumbai’s elites to create a modern city, modernity has been elusive, as sanitary cordons and containment have been a primary means used to protect elite groups (McFarlane, 2009). Elites have also recurrently favored the construction of self-serving infrastructures such as roads, railways, and land reclamations over broad-based sanitation, housing and clear water supply infrastructures (McFarlane 2012). Inconsistent and unevenly applied development has left many Mumbai residents exposed and sensitive to high levels of pollution, congestion, and flooding, which often follows lines of formality and informality. In informality, land ownership, access to livelihoods, and protection of rights are not defined by laws or planning instruments (Roy 2009). Informal settlements are frequently viewed as unfit for human habitation and routinely demolished in rehabilitation programs ((Bhide 2009; McFarlane 2012)). While some steps have been taken to ‘‘improve slums”, recent policies such as the Greater Mumbai Development Plan 2014-2034 have favored the powerful builders’ lobby.

4. Study design

6

We selected indicators used in the literature to measure wealth, inequality and vulnerability in cities (Filmer and Pritchett 2001; McKenzie 2005; Romero-Lankao, et al. 2012). In collaboration with the Indian Institute of Tropical Meteorology (IITM), a survey of 1252 households stratified by levels of socioeconomic status and built environment characteristics (e.g., well developed residential areas, commercial areas and slum areas) was implemented between December 2013 and February 2014. The goal was to obtain a varied set of household types across the city. The final selection of households depended upon local partners’ judgements related to responsiveness, availability and personal safety. We focused on households because they are an elemental assets management institution that shapes decisions and behaviors related to livelihoods and capacities to manage risks (Jepson 2014). Survey data were also collected to assess levels of network and governmental support individuals rely on to respond to hazards (Isa Baud et al. 2010; C. O. N. Moser 1998; Romero-Lankao et al. 2014). Additional key-informants were interviewed to gain local expert insight into the robustness of these indicators and to better describe relevant trends in urban development. We used two sets of indicators to measure levels of poverty/wealth (blue boxes, Figure 1). The first is comprised of seven material possessions indicators, six of which are considered solely as indicators of household economic status: fridge, color TV, Internet, motorcycle, oven and car. The seventh, Air conditioning (AC), is also an indicator of capacity. The second, physical capital, comprises housing and infrastructural conditions (Moser and Felton 2007). We selected six indicators based on whether the household has: hand washing facilities; a separate cooking space; uses clean fuels for cooking; access to reliable electricity; piped water; and a flush toilet connected to a sewage system. These indicators do not only define poverty, but are also of relevance for vulnerability and resilience, a point we will touch on later in our discussion. Vulnerability and resilience have the same three components: hazard exposure, sensitivity and capacity (purple and orange boxes, Figure 1). To capture hazard risk, we asked study households to report the major physical hazard events their family had experienced in the last five years (number of events), how much these events disrupted their lives (impact level), and whether they experienced a health episode as a result of the events. We selected three indicators of individual sensitivity: age, gender and pre-existing medical conditions. To define elderly populations, we considered the much younger age population balance of cities such as Mumbai, with 66.7 percent of residents below the age of 34, and therefore defined elderly populations as being more than 50 years old. Climate change is not gender neutral (Dankelman 2010). Because women experience unequal access to assets and decision-making processes and are most often responsible for household needs, we assumed that women are more sensitive than men. Figure 1: Data and methods to analyze the links between inequality and vulnerability. Here we describe three methods and the indicators used to examine the relative influence of wealth/ poverty, exposure/sensitivity and capacity in differences in vulnerability across four household classes in Mumbai. The weights under each indicator (e.g., 0.1272 for car) were calculated using an Analytic Hierarchic Process (ANH).

7

Capacity to respond to urban climatic risk is defined as a function of three groups of indicators. The first, human capital, measures dependency (based on a ratio of non-working to total household members) and household access to education. The second, institutional assets (socio-institutional capital), were probed with questions about the existence or awareness of public measures to enhance the capacity in general (governmental programs, health care) and during hazard emergencies (governmental support emergency, Figure 1). Since household members also access resources embedded in social networks to gain advantages, we included questions about their participation in social networks and personal support from family and community members (Figure 1). While information is a key resource that enables adaptive performance, communication is vital to a two-way conveyance of meaningful and useful information as it creates shared meaning and a chance for community members to express needs and exchange views (Norris et al. 2008). To capture our third set of capacity indicators, information conveyance and communication, we asked respondents about the number and type of warning sources about hazard events they typically had, their awareness of climate risks, and their impressions of the priority given to climate versus urban development policies. Actions to address climate risk issues often get displaced by competing priorities on urban development agendas, and this lower priority position is a key factor undermining capacity to respond to climate change. We decided to include this question under the information rather than the socio-institutional indicators because a perception of prioritization is an indicator of the communication’s emphasis on one issue over another. However, the fact that this question could fall under either category clearly demonstrates how closely linked socio-intuitional strength and informational conveyance are.

8

We applied an analytic hierarchic process (AHP) to develop indices of poverty/wealth, exposure/ sensitivity, and capacity (supplemental material). The AHP was based on judgements by the research team resulting from theory, prior research and key-informant interviews. To produce indicators weights, we organized them into four levels (Figure 1, which builds on Eakin and Bojorquez 2008). The first level includes the concepts of poverty/wealth, exposure/sensitivity, and capacity (first dark-blue, dark-purple, dark-orange boxes). The second level includes two hierarchies for wealth (physical and material possessions), two for exposure/sensitivity (hazard-risk and sensitivity), and three hierarchies for capacity (human, socio-institutional and information). The indicators used in constructing the hierarchies, with weights under each (e.g., 0.1272 for car) make the third level (see Figure 1). Households are the fourth hierarchical level.

Once the hierarchies were established, we organized them in pairwise comparison matrices, which were performed using Saaty’s 1-9 scale comparison scale. In this, 1 represents equal importance, 5 denotes strong importance and 9 indicates extreme importance (Saaty 2008). The elements in the second level were used to add weights to the indicators in the level below through pairwise comparisons that were transformed into [0.1] priority scale numbers. The values of these indicators were based on two assumptions that are supported in the literature: lower access to assets (wealth) reflects higher levels of poverty; and urban household vulnerability is higher as sensitivity and hazard/exposure increase and capacity decreases (Adger 1999; Vyas and Kumaranayake 2006; Romero-Lankao et al. 2012).

Although asset ownership can be used as an indicator of long-term household wealth, it often does not capture the quality of the assets. For instance, TV ownership does not distinguish between better-off households that often own high definition LCD TVs or poorer households that may own older technology. To address this ambiguity prior to performing pairwise comparisons, following Guest (2000), we asked local informants to think about levels of wealth in Mumbai and tell us what separated poor households from wealthier households in terms of assets. While vulnerability shows patterns across cities, it also has a city specific nature; therefore its indicators hold a city specific weight or influence. Through local informants we were able to determine indicators for access to sanitation in Mumbai, where access is generally low and higher levels of sanitation assets revolve around access to toilet facilities that are not shared with other households (see Chatterjee 2010).

Although the households were the lowest level of our hierarchy, it would have been cumbersome to undertake all the necessary pairwise comparisons of households. Therefore we used compromise programing, a method to solve multi-objective optimization problems applied by Eakin and Bojorquez (2008). We obtained a hierarchy of nested indicators and reduced the sample of questionnaires to 755 with full valid responses.

Using fuzzy logic in Mat Lab, we performed the last step of our analysis. Fuzzy logic defines a problem using linguistic variables or classes of objects with boundaries in which membership is a matter of degree (Rashed and Weeks 2003; Tah and Carr 2000; Eakin and Bojorquez-Tapia 2008; Giordani and Giorgi 2010). It offers a broader family of membership functions, also called linguistic or fuzzy sets, than other standardization methods; it can handle qualitative descriptions and, hence, looks like human reasoning in its use of approximate information and uncertainty to

9

generate decisions. We followed three fuzzy logic steps and applied the approach developed by Bojorquez et al. (2002 and supplemental material). In our first step, fuzzyfication, we translated the social inequality and vulnerability concepts into fuzzy sets or linguistic variables for the components of the poverty/wealth index wij, the

exposure/sensitivity index sij and the capacity index cij ., whose delimitation are Low (L),

Moderate (M), High (H) and Very High (VH). The three indices range in a [0.1] interval and U is the universal set such that 𝑈𝑈 = 𝑥𝑥│𝑥𝑥 ∈ 𝑈𝑈 𝑎𝑎𝑎𝑎𝑎𝑎 𝑥𝑥 ∈ [0.1]. The intersections between the fuzzy sets of households’ socioeconomic status and vulnerability represents the tolerable uncertainty in the grouping of households, adjusted to break the data into four categories: L = 0-0.25, M = 0.26-0.50, H = 0.51-0.75 and VH = 0.76-1.0. In the second fuzzy logic step, combination, the logic sentences were aggregated with the fuzzy sets of each index and the degree of membership of the variables into a fuzification space. To do this, we combined the membership functions resulting from our poverty/wealth, exposure/sensitivity and capacity scores into new membership values.

In the third step, defuzzification, we used the centroid method (provided by Mat Lab) to create output values, i.e., poverty and vulnerability “crispy” scores. We finally classified poverty/wealth, exposure/sensitivity and capacity values into vulnerability categories using the following value ranges: low, 0≤ wj ≤ ¼; moderate ¼ ≤ wj ≤2/4; high 2/4 ≤ ¾; and very high ¾ ≤ 1.

5. Findings

While three quarters of the households have high (62.3%) to very high levels of poverty (14.8%), two thirds belong to the high and very high vulnerability class (55.7% and 11.7% respectively). This confirms that Mumbai has high levels of social inequality and vulnerability. Our results shed light on how wealth, exposure/sensitivity and capacity values relate to differences in vulnerability across four household classes. If we look at the average scores across household classes, with the exception of the very high vulnerability class, the average values for the exposure/sensitivity dimension are relatively low (Figure 2). Thus except within the very high vulnerability class, rather than hazard exposure, disparities in capacity and poverty largely account for the differences in vulnerability across the other three classes. This is highly evident in the low vulnerability class, and within the moderate vulnerability class, where capacity and poverty still play a stronger role in determining vulnerability than exposure/sensitivity. However, while average poverty scores are high within both the high and very high vulnerability classes, exposure/sensitivity plays an equally strong role in creating vulnerability in the very high vulnerability class, while capacity is similar between both the very high and high vulnerability classes (Figure 2). Therefore, exposure/sensitivity plays a strong role in pushing households from vulnerable to highly vulnerable

10

Fig. 2: Disparities in capacity and poverty largely account for differences in households’ vulnerability

Note: Y= Index Value; *** indicates a p<0.001 statistical significance (in non-parametric tests) of dimension in characterizing vulnerability classes.

We also found city wide patterns on the relative influence of wealth, exposure/sensitivity and capacity on vulnerability across household classes. Across households, the mean age of the respondents was 36-38, and 59.2 percent of these were women. Between 19% and 32.5% of households had a family member with a disability or medical condition. As for material possessions and physical capital attributes such as infrastructures and services, most households had material possessions such as a TV (91.5%) and a fridge (61.6%); they used clean cooking fuels (83.9%) and had reliable electricity (80.8%, supplemental materials). Interestingly, across household classes, air conditioning penetration was very low (of between 8% and 12.9%), as was access to private transportation. Air conditioning can increase a population’s resilience to heat waves and other temperature-related hazards climate change is expected to aggravate. Also Internet has low penetration in Mumbai, with 33.3% and 16% (respectively) of low to high vulnerability households connected to the network, a point we will come to when discussing our capacity findings. Within our study sample, lack of improved toilet facilities and sewage systems, a city-wide risk and vulnerability concern, varied slightly with socioeconomic status, with 38.9 and 23.9 (respectively) of low to very high vulnerability class households being connected to a sewage system.

While the number of multiple hazards experienced such as heatwaves, air pollution, and rainfall and the impact level were slightly similar between low and high level vulnerability household

.00

.10

.20

.30

.40

.50

.60

.70

.80

.90

1.00

Low vulnerability Moderatevulnerability

High vulnerability Very highvulnerability

Poverty*** Capacity*** Exposure***Linear (Poverty***) Linear (Capacity***)

11

classes (purple bars, Figure 2), the health impacts themselves varied across household classes, and in the very high vulnerability household class, individuals were three times as likely to recall health episodes with greater frequency. The presence of household members with pre-existing medical conditions was slightly higher among the households with very high vulnerability levels. Health issues affect the human capital that people can mobilize to focus on livelihood strategies, particularly during events such as floods, as people with medical conditions need treatments that cannot be interrupted.

We also found that about three quarters of surveyed households in the low to high vulnerability level classes mostly relied on TV (76.2%) as source of information on hazard emergencies, and only 10.3% of respondents in all classes said they had looked for information on the Internet. This is troublesome as the Mumbai decision makers we interviewed often pointed to the Internet as their way of communicating emergency warnings and other critical information to the public.

We identified differences in relative importance of wealth, exposure/sensitivity and capacity indicators within each household class. For instance, of the three sets of capacity indicators used in this study (Figure 1), priority given to climate policies versus urban development policies and awareness of climate risks were somewhat similar across vulnerability level classes (Figure 3). The rest of the variables, consistently combined in specific ways within household classes. For instance, the low vulnerability households were more than twice as likely to have completed college, or a technical training, than the very high vulnerability class. The high vulnerability class was more likely to have more dependent members, hence, to have a lower dependency score than the other three classes. Three other attributes consistently paired with lower capacity in the very high household vulnerability class were participation in social networks, presence of governmental programs in the community and access to health care.

Fig. 3: Select Capacity Indicators by Household Vulnerability Class

.00

.10

.20

.30

.40

.50

.60

.70

Low Moderate High Very High

12

Note: Y= Index Value; ** indicates a p<0.01 statistical significance (in non-parametric tests) of dimension in characterizing vulnerability classes

Also differentiating the very high vulnerability class was a lower level of reported access to governmental support during emergencies and a lower number of warning sources used to respond to climate hazards. A low proportion of households across classes appear to have strong social networks to fall back on. Interestingly, while low, moderate and high vulnerability groups were two to three times more likely to have regularly participated in social networks, respondents from the high vulnerability group were much more likely to know someone who would support them during emergencies than those from the other three classes (Figure 3).

6. Conclusions

While research on urban resilience has grown considerably in recent years, this paper is one of few studies that have examined the relative influence of wealth and vulnerability on patterns of and differences in resilience within and across urban household classes. Under current climate conditions, high household vulnerability levels in Mumbai are accounted for mainly by differences in wealth and capacity. This conclusion is supported by similar findings in other studies that have established the urban development attributes of wealth and capacity to be key in determining vulnerability across global cities, and in select cities in North America, Asia and Latin America (Garschagen and Romero-Lankao 2013; Cutter et al. 2003; Revi 2008; Romero-Lankao et al. 2014). This pattern might change in a future (warmer) world, with hazard-exposure having a higher influence on vulnerability at the upper limits of Mumbai’s capacity to cope with extreme events. However, the data clearly point to the need for capacity enhancing measures to increase the resilience of vulnerable populations. Without a profound transformation towards policies and actions addressing the structural drivers of social marginalization and exclusion of a majority of households, Mumbai could reach these limits in a not far distant future.

Social inequality is a pervasive feature of urban development affecting resilience within and across cities. Yet the mechanisms that create and perpetuate the relative differences in resilience between poor and wealthy household classes can be difficult to parse. The differences we found in access to reliable electricity versus water and sanitation illustrate the equity and vulnerability implications of economic and political mechanisms through which the city has created and perpetuated differences in the development trajectories of its key infrastructures. According to Zérah (2008), three reasons explain why economic compulsions led to a fast expansion of the electricity distribution network and not of the water and sanitation ones. While public policies facilitated investments in electricity, water and sanitation were financed with public investments that suffered from competition with other needs and priorities. Once an electricity grid is constructed, the cost of individual connection is marginal, while the costs of both extending connections to the water distribution and sanitation network and transporting water are high. In contrast with electricity, therefore, the spread of the water and sanitation network is correlated with the spread of formal housing development. However since over half of Mumbai’s population lives in informal settlements, the city’s water and sanitation distribution is also one of its most profound expressions of social inequality.

13

The links between poverty and exposure sensitivity attributes are more nuanced. With the exception of the very high household vulnerability class (which had the highest levels of hazard exposure), the other household classes tended to be similarly exposed to multiple hazards, which are socially constructed and are expected to be aggravated by climate change. For example, floods are likely to be aggravated by sea level rise and intense rainfall; they also result from a particular combination of heavy precipitation events and the historic reclamation of large areas of land that would otherwise act as a water infiltrator, cooler and flood protector. Of equal importance is the fact that access to sewage and water systems remains highly segregated, thus defining a highly vulnerable patchwork of infrastructure. Unless this infrastructure is improved with equity and climate considerations in mind, it will be unable to protect people and places from floods.

Through a combination of interacting hazards and lack of capacity, urban development in Mumbai is becoming a source of hazards that are creating risks of uncertain proportions. As climate change progresses, an increased number of heatwaves is expected that might interact with air pollution to pose additional risks to human health. Notwithstanding this risk, the use of air conditioning was very low across the household classes we surveyed. With an intensification of risks brought by urban growth and interacting hazards, the possibilities for wealthy sectors to escape from and compensate for risks might also diminish. Ironically, the fact that elites could also be put in danger opens the prospect that addressing these challenges will become a higher priority. To effectively limit these hazards across populations, our study suggests that what will be needed are transformative and coherent, city-wide policies that are aware of their equity and mitigation implications.

Vulnerability frameworks and methods such as the one suggested here can help shed light on and serve as tools to inform policies addressing the multidimensional, disproportionate and interconnected nature of attributes of social inequality and vulnerability, thus enhancing resilience in cities. Such methods help determine the levels of influence of dimensions such as wealth, exposure and capacity across household vulnerability classes. By illuminating how specific combinations of attributes at play within and across urban households can create the conditions for vulnerability, this type of knowledge might help move diffuse equity and capacity goals towards the creation of cohesive and relevant policy interventions. Based on the findings of this study, secure land tenure, and improvements to housing, infrastructures, services and information conveyance are fundamental to enhance the resilience of vulnerable households, as are policies that further empower these households to rely on personal sources of support during emergencies.

We are aware that other methods could be used to undertake the ccomplex task of capturing the multidimensionality of urban poverty and vulnerability (e.g., Decil ratios, Division by Max and thresholding; see De Maio 2007; Aceves-Quesada, Díaz-Salgado, and López-Blanco 2007; Cutter, Mitchell, and Scott 2000). Although our study design allowed us to better capture the complexity of social inequality and vulnerability, future studies are needed that accurately capture the temporal dynamics and features of top and bottom quantiles of wealthy or poor and highly vulnerable household classes.

Our combination of methods allowed for a more nuanced understanding of the multilevel socioeconomic, political, environmental and infrastructural factors converging in context specific ways to determine patterns of and differences in resilience within and across household classes in

14

Mumbai. An understanding of the patterns, differences, pervasiveness and intensification of risks brought by urban development will be fundamental to designing interventions that can enhance resilience equitably and effectively. Acknowledgments: This research was supported by funding from the NSF PIRE Award #1243535. Survey support from Dr. Gufran Beig’s team and students at the Indian Institute of Tropical Meteorology was invaluable. The authors would also like to acknowledge the reviewers for their useful suggestions and comments.

7. References

Aceves-Quesada, José Fernando, Jesús Díaz-Salgado, and Jorge López-Blanco. 2007. “Vulnerability Assessment in a Volcanic Risk Evaluation in Central Mexico through a Multi-Criteria-GIS Approach.” Natural Hazards 40 (2): 339–56.

Adger, W Neil. 1999. “Exploring Income Inequality in Rural, Coastal Viet Nam.” The Journal of Development Studies 35 (5): 96–119.

———. 2006. “Vulnerability.” Global Environmental Change 16 (3): 268–81. Baud, Isa, Monika Kuffer, Karin Pfeffer, Richard Sliuzas, and Sadasivam Karuppannan. 2010.

“Understanding Heterogeneity in Metropolitan India: The Added Value of Remote Sensing Data for Analyzing Sub-Standard Residential Areas.” International Journal of Applied Earth Observation and Geoinformation 12 (5): 359–74. doi:10.1016/j.jag.2010.04.008.

Baud, ISA, N Sridharan, and Karin Pfeffer. 2008. “Mapping Urban Poverty for Local Governance in an Indian Mega-City: The Case of Delhi.” Urban Studies 45 (7): 1385–1412.

Bhagat, RB, Mohua Guha, and Aparajita Chattopadhyay. 2006. “Mumbai after 26/7 Deluge: Issues and Concerns in Urban Planning.” Population and Environment 27 (4): 337–49.

Bhide, Amita. 2009. “Shifting Terrains of Communities and Community Organization: Reflections on Organizing for Housing Rights in Mumbai.” Community Development Journal, bsp026.

Blaikie, Piers, Terry Cannon, Ian Davis, and Ben Wisner. 2014. At Risk: Natural Hazards, People’s Vulnerability and Disasters. Routledge.

Chatterjee, Monalisa. 2010. “Slum Dwellers Response to Flooding Events in the Megacities of India.” Mitigation and Adaptation Strategies for Global Change 15 (4): 337–53.

Collins, Timothy W, and Bob Bolin. 2009. “Situating Hazard Vulnerability: People’s Negotiations with Wildfire Environments in the US Southwest.” Environmental Management 44 (3): 441–55.

Cowell, Frank. 1998. “A., 2009, Measuring Inequality.” LSE Perspectives in Economic Analysis. Cutter, Susan L, Lindsey Barnes, Melissa Berry, Christopher Burton, Elijah Evans, Eric Tate, and Jennifer

Webb. 2008. “A Place-Based Model for Understanding Community Resilience to Natural Disasters.” Global Environmental Change 18 (4): 598–606.

Cutter, Susan L., Bryan J. Boruff, and W. Lynn Shirley. 2003. “Social Vulnerability to Environmental Hazards*.” Social Science Quarterly 84 (2): 242–61. doi:10.1111/1540-6237.8402002.

Cutter, Susan L, Jerry T Mitchell, and Michael S Scott. 2000. “Revealing the Vulnerability of People and Places: A Case Study of Georgetown County, South Carolina.” Annals of the Association of American Geographers 90 (4): 713–37.

Dankelman, Irene. 2010. Gender and Climate Change: An Introduction. Routledge. De Maio, Fernando G. 2007. “Income Inequality Measures.” Journal of Epidemiology and Community

Health 61 (10): 849–52. Eakin, Hallie, and Luis A Bojorquez-Tapia. 2008. “Insights into the Composition of Household

Vulnerability from Multicriteria Decision Analysis.” Global Environmental Change 18 (1): 112–27.

15

Field C.B., Maarten van Aalst, Neil Adger, Douglas Arent, Jonathon Barnett, Richard, Betts, Eren Bilir, Joern Birkmann, JoAnn Carmin, Dave Chadee, Andrew Challinor, Monalisa Chatterjee, Wolfgang Cramer, Debra, Davidson, Yuka Estrada, Jean-Pierre Gattuso, Yasuaki Hijioka, Ove Hoegh-Guldberg, He-Qing Huang, Gregory Insarov, Roger Jones, Sari, Kovats, Patricia Romero-Lankao, Joan Nymand Larsen, Iñigo Losada, et al. 2014. “Technical Summary.” In Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, edited by Field, C.B., V.R. Barros, D.J. Dokken, K.J. Mach, M.D. Mastrandrea, T.E. Bilir, M. Chatterjee, K.L. Ebi, Y.O. Estrada, R.C. Genova, B. Girma, E.S. Kissel, A.N. Levy, S. MacCracken, P.R. Mastrandrea, and L.L. White.

Filmer, Deon, and Lant H Pritchett. 2001. “Estimating Wealth Effects without Expenditure Data—Or Tears: An Application to Educational Enrollments in States of India*.” Demography 38 (1): 115–32.

Garschagen, Matthias, and Patricia Romero-Lankao. 2013. “Exploring the Relationships between Urbanization Trends and Climate Change Vulnerability.” Climatic Change, August, 1–16. doi:10.1007/s10584-013-0812-6.

Giordani, Paolo, and GiovanniMaria Giorgi. 2010. “A Fuzzy Logic Approach to Poverty Analysis Based on the Gini and Bonferroni Inequality Indices.” Statistical Methods & Applications 19 (4): 587–607. doi:10.1007/s10260-010-0146-8.

Guest, Greg. 2000. “Using Guttman Scaling to Rank Wealth: Integrating Quantitative and Qualitative Data.” Field Methods 12 (4): 346–57.

Gupta, Kamla, Fred Arnold, and H. Lhungdim. 2009. “Health and Living Conditions in Eight Indian Cities.” Mumbai: Ministry of Health and Family Welfare Government of India. http://pdf.usaid.gov/pdf_docs/PNADQ634.pdf.

Gupta, Kapil. 2007. “Urban Flood Resilience Planning and Management and Lessons for the Future: A Case Study of Mumbai, India.” Urban Water Journal 4 (3): 183–94.

Harlan, Sharon L, Anthony J Brazel, G Darrel Jenerette, Nancy S Jones, Larissa Larsen, Lela Prashad, and William L Stefanov. 2007. “In the Shade of Affluence: The Inequitable Distribution of the Urban Heat Island.” Research in Social Problems and Public Policy 15: 173–202.

Hijioka, Y, E Lin, JJ Pereira, RT Corlett, X Cui, G Insarov, A Surjan, C Field, V Barros, and K Mach. 2014. “Chapter 24: Asia.” Climate Change 2014: Impacts, Adaptation, and Vulnerability, IPCC Working Group II Contribution to AR5.

Jepson, Wendy. 2014. “Measuring ‘no-Win’waterscapes: Experience-Based Scales and Classification Approaches to Assess Household Water Security in Colonias on the US–Mexico Border.” Geoforum 51: 107–20.

Makri, Anna, and Nikolaos I. Stilianakis. 2008. “Vulnerability to Air Pollution Health Effects.” International Journal of Hygiene and Environmental Health 211 (3–4): 326–36. doi:10.1016/j.ijheh.2007.06.005.

Manuel-Navarrete, David, Mark Pelling, and Michael Redclift. 2011. “Critical Adaptation to Hurricanes in the Mexican Caribbean: Development Visions, Governance Structures, and Coping Strategies.” Global Environmental Change 21 (1): 249–58.

McFarlane, Colin. 2012. “From Sanitation Inequality to Malevolent Urbanism: The Normalisation of Suffering in Mumbai.” Geoforum 43 (6): 1287–90.

McKenzie, David J. 2005. “Measuring Inequality with Asset Indicators.” Journal of Population Economics 18 (2): 229–60.

Moser, Caroline, ed. 2008. Assets and Livelihoods: A Framework for Asset-Based Social Policy. Assets, Livelihoods and Social Policy. Washington DC: World Bank.

16

Moser, Caroline, and Andrew Felton. 2007. “The Construction of an Asset Index Measuring Asset Accumulation in Ecuador.” Chronic Poverty Research Centre Working Paper, no. 87.

Moser, C.O.N. 1998. “The Asset Vulnerability Framework: Reassessing Urban Poverty Reduction Strategies.” World Development 26 (1): 1–19.

Municipal Corportation of Greater Mumbai. 2014. “Development for Greater Mumbai 2014-2034 Preparatory Studies.” Mumbai: Municipal Corporation of Greater Mumbai.

Norris, Fran H, Susan P Stevens, Betty Pfefferbaum, Karen F Wyche, and Rose L Pfefferbaum. 2008. “Community Resilience as a Metaphor, Theory, Set of Capacities, and Strategy for Disaster Readiness.” American Journal of Community Psychology 41 (1–2): 127–50.

O’Neill, M. S. 2005. “Disparities by Race in Heat-Related Mortality in Four US Cities: The Role of Air Conditioning Prevalence.” Journal of Urban Health: Bulletin of the New York Academy of Medicine 82 (2): 191–97. doi:10.1093/jurban/jti043.

Qin, Hua, Patricia Romero-Lankao, Jorgelina Hardoy, and Angélica Rosas-Huerta. 2015. “Household Responses to Climate-Related Hazards in Four Latin American Cities: A Conceptual Framework and Exploratory Analysis.” Urban Climate 14 (1): 94–110.

Ranger, Nicola, Stéphane Hallegatte, Sumana Bhattacharya, Murthy Bachu, Satya Priya, K Dhore, Farhat Rafique, P Mathur, Nicolas Naville, and Fanny Henriet. 2011. “An Assessment of the Potential Impact of Climate Change on Flood Risk in Mumbai.” Climatic Change 104 (1): 139–67.

Rashed, Tarek, and John Weeks. 2003. “Assessing Vulnerability to Earthquake Hazards through Spatial Multicriteria Analysis of Urban Areas.” International Journal of Geographical Information Science 17 (6): 547–76.

Revi, Aromar. 2008. “Climate Change Risk: An Adaptation and Mitigation Agenda for Indian Cities.” Environment and Urbanization 20 (1): 207–29.

Ribot, Jesse. 2010. “Vulnerability Does Not Fall from the Sky: Toward Multiscale, pro-Poor Climate Policy.” Social Dimensions of Climate Change: Equity and Vulnerability in a Warming World, 47–74.

Romero-Lankao, Patricia, Sara Hughes, Hua Qin, Jorgelina Hardoy, Angélica Rosas-Huerta, Roxana Borquez, and Andrea Lampis. 2014. “Scale, Urban Risk and Adaptation Capacity in Neighborhoods of Latin American Cities.” Habitat International 42 (0): 224–35. doi:10.1016/j.habitatint.2013.12.008.

Romero-Lankao, Patricia, and Hua Qin. 2011. “Conceptualizing Urban Vulnerability to Global Climate and Environmental Change.” Current Opinion in Environmental Sustainability 3 (3): 142–49. doi:10.1016/j.cosust.2010.12.016.

Romero-Lankao, Patricia, Hua Qin, and Katie Dickinson. 2012. “Vulnerability to Temperature-Related Hazards: A Meta-Analysis and Meta-Knowledge Approach.” Global Environmental Change 22 (3): 670–83.

Roy, Ananya. 2009. “Why India Cannot Plan Its Cities: Informality, Insurgence and the Idiom of Urbanization.” Planning Theory 8 (1): 76–87.

Saaty, Thomas L. 2008. “Decision Making with the Analytic Hierarchy Process.” International Journal of Services Sciences 1 (1): 83–98.

Tah, JHM, and V Carr. 2000. “A Proposal for Construction Project Risk Assessment Using Fuzzy Logic.” Construction Management & Economics 18 (4): 491–500.

Vincent, Katharine. 2007. “Uncertainty in Adaptive Capacity and the Importance of Scale.” Global Environmental Change 17 (1): 12–24.

Vyas, Seema, and Lilani Kumaranayake. 2006. “Constructing Socio-Economic Status Indices: How to Use Principal Components Analysis.” Health Policy and Planning 21 (6): 459–68.

Wilhelmi, Olga V, and Mary H Hayden. 2010. “Connecting People and Place: A New Framework for Reducing Urban Vulnerability to Extreme Heat.” Environmental Research Letters 5 (1): 14021.

17

Wilkinson, Richard G, and Kate E Pickett. 2006. “Income Inequality and Population Health: A Review and Explanation of the Evidence.” Social Science & Medicine 62 (7): 1768–84.

Zérah, Marie-Hélène. 2008. “Splintering Urbanism in Mumbai: Contrasting Trends in a Multilayered Society.” Geoforum 39 (6): 1922–32.