Inequality in India: Life Chances and Caste Matters

Transcription

1 Inequality in India: Life Chances and Caste Matters Omkar Joshi 1 Abstract: The question of inequality, its determinants and the consequences of inequality has been one of the central areas of research in developmental economics and sociology. There are hardly any sociological studies looking at the question of income inequality within the developing country context and those who study this question in a developing country context, study it from the macro perspective of global inequality and world systems approach. This paper is a contribution in the area of inequality in a developing country context. Using a large-scale nationally representative household survey- India Human Development Survey (IHDS)- for and , I look at the extent of inequality with respect to caste in India. I find that over a period, inequality has risen marginally. However, between-caste inequality is going down and within-caste inequality is rising. Keywords: Inequality, Caste, India 1 Doctoral Student, Department of Sociology, University of Maryland, College Park. ojoshi@umd.edu. I thank Prof Andres Villarreal, Prof Wei-hsin Yu for their valuable feedback. This paper has also benefited from the useful discussions that I had with Prof Reeve Vanneman and Prof Sonalde Desai. 1

2 Inequality in India: Life Chances and Caste Matters Omkar Joshi 1. Introduction The question of inequality, its determinants and the consequences of inequality has been one of the central areas of research in developmental economics and sociology. Although a large body of literature theorizes and reviews various type of disparities viz. disparities in health, education, economic inequality as reflected in disparities of income has received a good deal of attention in the literature, early on from economics and lately in sociology as well. Within economics, the question of inequality has been studied in the context of economic growth. After Kuznets s (1955) seminal paper on the relationship between economic growth and inequality, many studies have been carried out to empirically test the so called inverted U- hypothesis, which simply put, states that inequality first increases with the growth and then declines subsequently (Williamson 1985; Lindert 1986; Anand and Kanbur 1993; Deninger and Squire 1996). However, most of these studies have found a mixed evidence in support of Kuznets hypothesis. Despite this inconclusive evidence, the interest in analyzing inequality continues to grow. Partly this is because, as Ravallion (2014) shows, the withincountry inequality has been growing for the developing countries in last two decades. Therefore, understanding the mechanisms driving the inequality has become an important tool to understand and guide the development effort. In the sociological literature, the question of income inequality had received less attention earlier. Morris and Western (1999) remarked that sociologists had been strangely and remarkably silent on the subject of income inequality even though stratification and inequality were among the few undisputed core areas in the field of sociology. In recent decades, though, research interest in income inequality has increased (McCall and Percheski 2010). However, this increased attention towards the question of income inequality is still largely about studying income inequality within the USA and European countries. When one tries to find the sociological analyses of income inequality in the low-income and developing countries, the silence of sociologists becomes almost deafening (Guidetti and Rehbein 2014). There are hardly any sociological studies looking at the question of income inequality within the developing country context and those who study this question in a developing country context, study it from the macro perspective of global inequality and world systems approach. Given the fact that the developing countries face the dual and simultaneous challenge of pursuing high economic growth while trying to make that growth inclusive and equitable, studying income inequality within a developing country context becomes crucial. Additionally, the problem of inequality is compounded for a developing country due to prevailing regional imbalances and strong presence of stratification systems of gender, race, 2

3 and other group characteristics. Such stratification systems may then become the basis for perpetuating already present and growing disparities between groups. This paper is a contribution in the area of inequality in a developing country context. I look at the question about the extent of income inequality India. I then look at what has happened to inequality between and within different caste groups in India using a large nationally representative data set from India Human Development Survey (IHDS). 2 Caste in India remains one of the major axis of stratification and analyzing income inequality for different caste groups can have important theoretical and practical implications. The remaining paper is organized as follows: the second section lays down the theoretical framework for studying income inequality; the third section gives a brief description of caste system in India and drivers of inequality; the subsequent section talk about the research question; I then discuss the data and methodology in the fifth section; sixth section presents the results and discussion; the final section then gives the discussion and conclusion. 2. Theoretical Approaches to Inequality There is a historical legacy in the sociological literature to study inequality. Primarily, three sociological approaches to inequality can be called upon to throw light at the origin and nature of inequality in society. First is the well- known Marxian structural framework analysis based on the conflict theory of classes. Very briefly, the main argument of the Marxist school is that, the different social groups (read classes) have unequal access to resources. This unequal distribution of resources is the reason behind conflict and opposition between those who have control over the capital (bourgeoisie) and those who can only sell their labour (proletariat). Thus, social structure and division of labour became the two pillars of Marxian thought which can be utilized to look at why and how inequality originates. Max Weber critiqued and furthered this structural school of thought by adding the nuances of status. He argued that social structure is a complex multidimensional phenomenon. There is a class of people in the society which are neither capitalists nor proletariat. Weber did not reject Marx s ideas on class analysis, but expanded them. He put forth that the class position is not determined just on the basis of relation to means of production. There are other factors like status and power which affect class relations and positions. Moreover, Weber linked the concept of status with life chances suggesting that status is not entirely achieved, is also influenced by the ascribed characteristics such as race, gender. The opportunities of an individual are pre-conditioned by his ascriptive characteristics which partly, he inherits from his parents, and partly, because due to prevalent external socioeconomic situations. Thus, the Weberian approach hints at the role played by one s ascriptive characteristics in determining his/her life chances (Breen 2001). Functionalism is another dominant school of thought in sociological literature, courtesy Durkheim, and Parsons. This school regards inequality to be necessary for maintaining the 2 I also conduct the analysis using consumption measures of inequality for robustness check. 3

4 social order and integrity. Davis and Moore (1945) explain that, Social inequality is.an unconsciously evolved device by which societies ensure that the most important positions are conscientiously filled by the most qualified persons. So, inequality according to functionalists is not only inevitable but it is beneficial, too. All these above classical schools take more or less a structural view of inequality. Arguing against these structuralist traditions, the Modernization Theory regards that as societies develop socioeconomic achievements become less tied to social background and other ascribed characteristics (Blau and Duncan 1967; DiPrete and Grusky 1990). Industrialization and post-industrialization processes in the society lead to decline of the community and family functions and give more emphasis to individualism. Education becomes an important vehicle to achieve higher rank in the society. Progressively societies become urbanized. Government policies become welfare-oriented trying to correct the inequities in access of resources and promoting investment in human capital (Beller and Hout 2006; Paterson and Iannelli 2007). Thus, modernization theory propounds that the above-mentioned processes act in such a way as to counter the reproduction of socioeconomic inequalities. Thus, mainly one can consider two theoretical propositions regarding inequality. The Weberian line of thought uses ascribed characteristics such as race, gender etc. to explain the persistence and growth of inequalities. In the Weberian world, accident of birth matters more for an individual and it decides his/her life chances. Despite the mitigating factors such as education, the accumulated disadvantage of the gender or race or such other group characteristic trumps. For example, a person who is born in a poor black family in United States will have a very different set of constraints over his life-course as far as education, access to healthcare etc. are concerned than a person who is born in a white and wealthy family. To complicate the matters further, not all black and poor individuals will be affected equally by their background characteristics. Modernization theory doesn t disregard the role played by the circumstances in which an individual is born and the disadvantage ensuing thereof, but it gives more role to the agency and factors like education. It argues that over a period, the hold of ascribed characteristics over individual s life chances becomes weak. Moreover, the normative structure of the society itself undergoes the change due to forces of modernization and the ascribed characteristics matter less and less. Apart from these main sociological theories, one can also think of economic growth as a larger societal process affecting all sections. Often the argument of economic growth is of the one tide lifts all boats nature. That economic growth has a positive impact in terms of growth of opportunities is clear. It can also be accepted that all boats are likely to be lifted due to economic growth, however some boats get lifted higher than others. Not all sections of society are benefitted equally due to growth. Even when economic growth happens, only some people can participate, others cannot. The trickle-down effect of growth is not same for all groups. 4

5 So, what happens to inequality between various groups becomes the matter of empirical investigation. The above-mentioned effects of persistent status disadvantage and modernization along with the economic growth act counter to each other. The following section describes the Indian case and elaborates as to how these two theoretical approaches can be applied to Indian case. 3. Caste System in India: A Primer Caste system in India is one of the longest surviving system of stratification in the world. Its roots are based in the theological model of cosmogony in Hindu scriptures, which talks about the varna 3 system in the society: the Brahmins (priests and teachers), the Kshatriyas (warriors and royalty), the Vaishyas (traders, merchants, and moneylenders) and the Shudras (those engaged in service-based jobs including lowly menial jobs). Besides these four varnas the fifth category of people was also recognized which were called the Ati-Shudra (the lowest of the low). The Ati-shudras were outside the purview of varna system and were considered as untouchables who s even a shadow was considered to be polluting. These varna groups were mutually exclusive, hereditary, endogamous and occupation based. The jati or the caste system as we see it today, is the successor of a more ancient varna system but is much more complex in number and nature 4. After Independence, the Constitution of India abolished the practice of untouchability and classified the erstwhile untouchables and shudras as Scheduled Castes. Traditionally these are the caste groups which were the most depressed and discriminated against. Apart from these scheduled castes, there is a huge group of tribal population in India which is classified according to the constitutional schema as Scheduled Tribes. These two broad groups, SCs and STs, fall behind other groups on the average indicators of income, education, health, and other socio-economic characteristics. The caste groups which traditionally enjoyed the advantage of high status and esteem in Indian society-namely Brahmins and Kshatriyas are referred to as Forward Castes. Thus, the Indian societal structure can be thought of as a hierarchical structure-based on historical practice and contemporary development indicators- as that of a pyramid, where the forward castes are at the top of this pyramid and the SCs and the STs are at the bottom. The middle of the pyramid comprises of many caste groups designated as Other Backward Castes (OBCs) which are also disadvantaged as compared to the forward castes but are vastly heterogenous in nature and have a greater range of backwardness. This hierarchy that I have outlined is of a broad nature and there is a debate surrounding the basis on which all caste-groups should be strictly arranged in this fashion. 5 However, it would not be out of place to say that there is a clear distinction based on degree of backwardness (howsoever constructed) between the forward castes and the SCs/STs. 3 Varna loosely translates to colour, but it is not an exact meaning. 4 About 3000 jatis exist today in India which are spread regionally all over the country and a one-to-one correspondence between Varna and jatis is not always possible (Deshpande 2000). 5 Standard of living is often advocated as the basis on which the caste groups should be arranged but, see Dumont (1980), Gupta (1984) and Chatterjee (1997) for the debate. 5

6 To correct the historical injustice meted out to the depressed castes, the constitution of India provided for a reservation system (a quota system similar to affirmative action measures in the USA), whereby certain percentage of government jobs and seats in educational institutions are reserved for persons from the SC and ST category. There are many welfare programmes launched by the Central and the State governments targeted towards the members of these caste groups. Because of such programmes over a period, there is an improvement in the overall socio-economic conditions of SCs and STs. The link between caste groups and various outcome indicators of development such as income, consumption expenditure can be represented via the following schematic diagram (Figure 1). The pyramid in the left panel of the figure represents the hierarchical nature of Indian society. The forward castes are at the top and the STs are at the bottom. The solid lines between each caste category represent that once a person is born into a caste/caste group, his status is unchangeable. Whereas the dotted lines within each caste group indicate that there is some mobility because of income, education etc. There are two main channels through which the circumstances at birth, in this case, caste, can affect the outcomes. The first is the Direct Channel by which an individual s life chances are affected. If a person belongs to a caste which lies at the lower position in the overall hierarchy of caste structure then by virtue of his or her position at the bottom of the hierarchy of the pyramid, it affects the outcomes in a negative manner than those who are higher in the hierarchy. This is somewhat like in a race the starting lines for runners from different groups are set differently in such a way that some runners have a starting line that is far behind those of the others. The direct channel captures the accumulated caste disadvantage and historical wrongs that people from lower castes are exposed to. Figure 1: Drivers of Inequality: Ascription-Outcomes Relationship The second channel is the Indirect Channel which affects the outcomes such as income and consumption through its effect on the mobility factors such as education and urbanization. The mobility factors are the factors which help in reducing the disparities between caste 6

7 groups, however, access to these factors say education itself may be affected because of the caste. Support mechanisms refer to various government welfare programmes like social security or pension programmes. These mechanisms can be thought of as a cushion, which mitigates the effect of caste. Thus, the outcomes that we see are affected by the total of caste effect, opportunity structure and support mechanisms together and these outcomes differ for different caste-groups. 4. Research Question The concern about the increase in inequality in India can be seen on the background of high rate of economic growth that India has achieved in last decade. Indian economy grew at about 7-8 per cent on average in the last decade. Although there are differences about the extent of reduction in poverty because of this high growth rate, there is a fair amount of consensus that this high rate of economic growth was beneficial to poor and it lifted many people out of poverty 6. However, at the same time concerns about increase in the level of inequality in the country have been expressed in recent times (IMF 2016; India Wealth Report 2016) comparing India with some of the Latin American countries which have high level of income inequality. It is well documented that increasing inequality within the country leads to worsening of growth and mobility opportunities for poor and disadvantaged sections of society (Ravallion 2014). Neckerman and Torche (2007) in their review article document the deleterious effects of growing group disparities and particularly draw attention towards the social reproduction of inequalities. It is here that the role of caste structure in India comes into picture. Several researchers working on this issue have pointed out the stickiness of the caste structure even in the face of economic prosperity (Deshpande 2010; Munshi and Rosenzweig 2009). However, the previous studies that have looked at the issue of inequality and caste groups have either used only income or consumption expenditure to measure inequality (Bros 2005; Borooah et. al 2014) or have only studied inequality for rural areas (Munshi and Rosenzweig 2009). This paper fills this gap in literature by using a large scale nationally representative data set that has information on both income and consumption expenditure for rural as well as urban households. Also, some of previous studies have only looked at the cross-sectional data at one point of time thereby providing just a static picture of inequality or have used too long a time-frame (Munshi and Rosenzweig 2009 use data from 1982) 7. However, inequality is a dynamic phenomenon which does respond to changes in contemporary economic and social policies which have undergone a rapid change in last decade with launch of some ambitious new public works programmes like Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA) (Desai et. al 2015) and increase in trade liberalisation. All these policies and programmes have an impact over inequality. 6 Official estimates show that the rural poverty declined in India from a high of 37% in to 25% in while urban poverty fell from 33% to 12% over the same period. 7 While taking a long time-frame view can be useful on several other occasions, it doesn t necessarily useful in analyzing responses to some more contemporary social and economic policies. 7

8 Therefore, in this paper I look at inequality at two points in time, in and in , for same caste groups. The main research questions that I address in this paper are as follows: a) What is the extent of inequality in India for different caste-groups? Has inequality increased in last few years? b) How much is the between-group and within group caste inequality? The next section describes in detail the data set, variables, and empirical strategy. 5. Data and Methodology The data used in this paper are from a large scale nationally representative household survey, India Human Development Survey (IHDS). IHDS is a collaborative effort between National Council of Applied Economic Research (NCAER), India and University of Maryland, College Park, USA. IHDS is a longitudinal data set with its first round of data collection in and the next round was in The first round of data was gathered for 41,554 households. The same households were traced in the second round. The second round of data is collected on 42,152 households with a re-contact rate of 83% and these households include apart from the original households, the split households and replacement households. Thus, IHDS-I and II cover the entire country with 33 states and union territories (except Andaman and Nicobar Islands and Lakshadweep islands), 971 urban blocks and 1503 villages. It is a multi-topic data set containing extensive information on the household income and consumption, health, education, and gender outcomes. Along with the household information, village information on various facilities is also collected. I shall be using the repeated cross-section data for two rounds for analysis of inequality in this paper. On Choice of the Welfare Aggregate One important question that a researcher analysing inequality faces is that of the selection of a welfare aggregate. Traditionally research on inequality has used two welfare aggregate measures-consumption expenditure and income 8. Each of the measure has its own advantages and disadvantages. For example, income is the direct measure of economic power and hence is more suitable. But income is also subject to measurement errors and underreporting. On the other hand, consumption expenditure is more suitable to capture the standard of living, but is not free from problems such as respondent fatigue in answering detailed questions about consumption or volatility due to marriage, debt, and health crisis (Vanneman and Dubey 2013). Luckily the IHDS captures both income and consumption expenditure data in a reasonably fair manner and therefore, I can provide a comprehensive picture of inequality from both income and consumption measures of welfare. 9 8 To be fair, wealth is another metric that is used to measure inequality, but given the fact that wealth data is hard to measure and collect I have not considered it here. Also, typically wealth and income have a high correlation. 9 See Vanneman and Dubey (2013) for a detailed discussion on income components. 8

9 I have used adjusted household income adjusted for economies of scale. This is following the standard practice of using equivalence income in the literature to adjust for the family size. It is the net, after tax annual income. Though over years, the direct tax net in India has widened, even today very small proportion of individuals pay direct taxes. For consumption expenditure, too, I have used the adjustments for economies of scale. I first present the descriptive statistics to give overall idea of the dataset. Then I present the Kernel Density Plots for income of different caste groups for two rounds and percentile ratios which gives an idea of how concentrated or spread the distribution of income is. Next, I show the extent of inequality present in India for two rounds using Gini Coefficients. Gini coefficients are standard measures to compute inequality that vary between 0 and 1 with the interpretation that higher the Gini, the more unequal is the distribution. Since Gini s give a good picture of levels but cannot be decomposed additively as between caste and within-caste I also present Theil Coefficient. I then use Oaxaca-Blinder decomposition to show how much of the income gaps can be explained due to endowments and how much are a result of the residual factors. 6. Results Descriptive Statistics Table 1: Comparison of IHDS Data with Census on Select Characteristics Characteristics IHDS-1 ( ) Census 2001 IHDS-2 ( ) Census 2011 Percentage urban Percentage literate Age Age Caste Other Backward Classes (OBCs) Schedule Castes/Dalits (SCs) Schedule Tribes/Adivasis (STs) Other castes and non-hindu Religion Hindu Muslim Christian Sikh Other religion Per cent currently in School (ages 5-14)

10 Work participation rate, males* Work participation rate, females* Average family size Percentage of women married (ages 15-49) Percentage of women married (all ages) Percentage with electricity Percentage with piped water Percentage with liquefied petroleum gas Percentage with toilets** Percentage poor Sample Size (Households) 41,554 42,152 Source: IHDS Data, Published Census Report, and Author's calculation for IHDS Data * IHDS: Worked more than 240 hours last year excluding animal care; Census: Main + Marginal workers ** IHDS 1 refers to % with flushed toilets; + % poor as per Tendulkar Poverty Line As one can see from the table above that there is an increase in urbanization and literacy between two rounds. We can also see that there is a rise in work participation rates for males and females although this rise is more for males. Percentage of poor has also gone down by 5% points from 26% in to 21% in Average family size between two rounds has remained constant. Let us now see what has happened to household income and consumption expenditure between two rounds Figure 1a: Annual Household Income (in Rs.) and Caste Forward OBC SC ST Source: Author s calculation from IHDS-I and II data. Total Adjusted Household Income (Mean) 10

11 The above graph shows that the mean household incomes have increased between two rounds for all caste-categories, however the increase in income is the highest for the scheduled caste households almost by 50% and it is the least for the households from scheduled tribes by about 30%. In absolute terms the income for forward castes is still the highest but we can see that the scheduled caste households are catching up fast. However, if we see in terms of the absolute gap that existed between forward castes and other castes in IHDS-I and IHDS-II, then we see that the gap has widened between round I and round II. We see a similar story if we look at the consumption expenditure. Here, though the catch up is faster for scheduled tribes as their consumption expenditure has gone up by almost 40% than the previous level. But in absolute terms, like income, the gap between average consumption expenditure of a SC or ST household and a forward caste household has gone up between two rounds Figure 1b: Annual Household Consumption Expenditure (in Rs.) and Caste Total Adjusted Household Consumption Expenditure (Mean) Forward OBC SC ST Source: Author s calculation from IHDS-I and II data. While looking at the mean income for different caste groups for both rounds gives us some idea of how each caste group is doing relative to other, it doesn t quite tell us about the overall shape and size of the income distribution for each hence, we employ the method of Kernel Density Plots. Symmetry of Income Distributions and Caste When examining the question of income inequality, it becomes important to examine the size and shape of the distribution. The Kernel Density plots, just like histograms, can give us a good idea about the spread and peak of the distribution. In the following diagram, I have plotted Kernel Density Plots for the log of adjusted household income for each caste group 11

12 0 0 Density Density Density Density over two rounds laid over each other. Each panel represents the income distribution of a caste group. For example, the top left panel (a) plots the log adjusted household income of forward castes in and We can see that the peakedness of the distribution has slightly reduced between two rounds, also there is a slight shift towards the right. Figure 2: Kernel Density Plots of Log Adjusted Household Income and Caste (a) (b) Kernel density estimate Kernel density estimate Forward Caste in Foreward Caste in SC in SC in Adjusted Log Household Income kernel = epanechnikov, bandwidth = Adjusted Log Household Income kernel = epanechnikov, bandwidth = (c) (d) Kernel density estimate ST in ST in Kernel density estimate OBC in OBC in Adjusted Log Household Income kernel = epanechnikov, bandwidth = Adjusted Log Household Income kernel = epanechnikov, bandwidth = Source: Author s calculation from IHDS-I and II data. If we look across all panels we can note some common features: a) all the distributions are asymmetric and skewed to the right with heavy tails b) Over two rounds for all caste groups the income distributions have shifted to the right, indicating overall increase in the income. But some distinctive features about the density plots of each caste group should be noted here, unlike the forward castes kernels, kernels for other caste groups are much tightly squeezed, especially the kernels for scheduled tribes. 12

13 Though informative, to have an estimate of inequality, we compute percentile ratios and Ginis for different caste groups for two rounds. Extent of Inequality Percentile shares is a basic measure that gives a good summary of how the income is distributed. If the income distribution is relatively egalitarian, one would expect to see more or less similar concentration of income in each percentile. Caste Groups Forward Caste Table 2a: Percentile Shares of Income Distribution and Caste Adjusted Household Income Percentiles Coefficients Standard Error Coefficients Standard Error OBC SC ST Source: Author s calculation from IHDS-I and II data The table above gives the percentile shares of the income for different caste groups for two rounds. In if we see, the bottom 20% among the forward caste just got 3% share of the total income. For the same year, the percentage share for bottom 20% of the SC and ST 13

14 population was 5% and 4%. After seven years, the percentage shares for bottom 20% for forward caste has worsened to 2.4% and for SCs and STs too it has declined to 4.2% and 2.9% respectively. As against that we can see that most of the income in all caste groups is concentrated in the top 20% of the income distribution and there is a marginal increase in the share at the top quintiles for all caste groups from to If we compare the same results for consumption expenditure instead of income, the following picture emerges. Table 2a: Percentile Shares of Consumption Distribution and Caste Caste Groups Adjusted Household Consumption Expenditure Percentiles Forward Caste Coefficients Standard Error Coefficients Standard Error OBC SC ST Source: Author s calculation from IHDS-I and II data. As we can see, consumption distribution is slightly more equitable than income distribution. Consumption is less concentrated at the top than the income distribution (the top 20% for 14

15 all caste group have about 42-45% of the overall consumption expenditure as compared to income where it was above 50%). But like income, consumption inequalities too seem to be increasing albeit marginally over two rounds. A better way to capture inequality is by computing Gini coefficients for each caste group. Gini Coefficients for Caste Groups The overall income Gini coefficient for all caste-groups is in and it has increased marginally to The consumption Ginis for the same period are and 0.374, almost the same. So, one can infer that on the whole, overall inequality has very increased slightly from to as far as income in concerned. But what is the picture of inequality if we do the sub-group analysis for different castegroups? Table 3: Gini Coefficients and Caste Groups Caste Groups Income Gini Consumption Gini Forward caste OBC SC ST Source: Author s calculation from IHDS-I and II data Looking at the sub-group level, we can say that inequality has increased for all caste groups between two rounds going by the both welfare aggregate measures, income as well as consumption. This increase seems to be relatively more for forward castes and SCs if we consider the income Gini, but if we take consumption Gini as the measure, it has increased relatively more for OBCs and SCs. The Gini coefficients do tell us about the extent of inequality, but they cannot be additively decomposed. We present Theil Coefficients for both rounds and its overall decomposition as between-caste and within-caste components. Table 4: Theil Coefficient and Overall Decomposition of Inequality Income Consumption Expenditure Theil Within-caste Between-caste Within-caste Between-caste Index Source: Author s calculation from IHDS-I and II data. When we decompose the Theil index into within-caste and between-caste inequality, we find that for both income as well as consumption, inequality between groups is reduced from to and with-in caste inequality has gone up. This is an interesting result as 15

16 it indicates that within the same caste group opportunity structure is not same for all members of the caste group. Oaxaca-Blinder Decomposition As we see that caste-specific Ginis have increased, even though slightly, the question that comes in mind is how can we explain this inequality with respect to caste. In the theoretical framework section, we discussed about caste persistent effects and modernization hypothesis. One way to tease out the caste effect against the other mobility factors is to perform a decomposition of income using Oaxaca-Blinder technique. In the following table I present the results for Oaxaca-Blinder decomposition of income for SCs and STs in and For parsimony, I have not presented the results for consumption expenditure here. Often education and urbanization are cited as reasons for mobility and income growth and hence I have used them as predictors to explain income. Log Household Adjusted Income Table 5a: Oaxaca-Blinder Decomposition of Income for SCs Robust Coefficient Standard P>z Robust Standard Error Coefficient Error Overall Non-SC SC difference explained unexplained Explained Highest Education Urban Unexplained Highest Education Urban _cons Source: Author s calculation from IHDS-I and II data As one can see from the above table there is a gap of 29% between income of non-sc and SC households in The predictors that I mentioned before explain about 20% of this gap and remaining 9% gap is unexplained gap, which can be construed as due to discrimination as well as due to unobservable factors. In this income gap between SC and non-sc households has been reduced to 17%. Also, we notice that the explained component of this difference (explained by education and urbanization) is 14% and P>z 16

17 unexplained component is 3%. i.e. education and urbanization could explain about 83% of the income gap. Thus, in case of SC households the decomposition results suggest that there is some catch-up of income happening with non-sc households. Log Household Adjusted Income Table 5b: Oaxaca-Blinder Decomposition of Income for STs Robust Coefficient Standard P>z Robust Standard Error Coefficient Error Overall Non-ST ST difference explained unexplained Explained Highest Education Urban Unexplained Highest Education Urban _cons Source: Author s calculations form IHDS-I and II data. In case of STs the difference with non-st household incomes is 33% in and it grows to be 37% in Between two rounds the explained component s share too has gone down. In , education and urbanization could account for 90% of the explained component, however in this has gone down to about 70%. 7. Discussion and Conclusion In the foregoing sections, I presented the results pertaining to inequality in income and consumption expenditure. The analysis of caste inequality was motivated partly because of the theoretical considerations related to Weberian hypothesis of persistence of ascribed characteristics like caste versus the Modernization theory and the impact of economic growth and partly due to the reports of increased inequality in India. The analyses of inequality using income and consumption expenditure as a welfare aggregate show bunch of interesting results. Firstly, between the two of the welfare aggregates, the inequality picture as shown by income measure is higher than that presented using consumption expenditure. This is in line with P>z 17

18 the established economic rationale, that income being composite measure of consumption and net worth would be generally higher than consumption (of course barring when current income could be zero or negative). Also, given the fact that savings as a percentage of income is generally larger for higher income households, income inequality tends to be higher than the consumption inequality. Secondly, when we consider two time points, overall income inequality has increased slightly. Consumption Ginis don t show any significant change. This too seems to be reasonable as the slight increase in income inequality could be due to higher volatility of income than consumption. Consumption tends to be sticky even in the face of declining income and hence doesn t show substantial increase. It would of course change should there be a significant change in income, as there is a strong positive correlation between these two aggregates. Thirdly, as far as the caste-specific inequality is concerned, there is a rise in inequality for all the caste groups. This rise in inequality is similar for all caste groups (around 2-3% point increase between rounds for each caste group). The decomposition of overall inequality informs us that between-group inequality has gone down between two rounds and withingroup inequality has increased. Kernel density plots showed a clear shift towards right for all caste groups. These three things taken together could be interpreted as a positive sign that caste-groups that are traditionally at the bottom of the hierarchy are catching up with rest of the society. In case of Scheduled Caste household this impact can be seen more visibly as the Oaxaca-Blinder decomposition results indicate. This could be due to expansion of the welfare programmes like MGNREGA as posited earlier. As far as STs are concerned, we see that their incomes have gone up however, not as fast as one would have liked. Unlike SCs which are well-distributed all over India, STs are concentrated in certain geographic pockets and therefore the processes that drive inequality for scheduled tribes could be regionally differentiated, but this is beyond the scope of the present paper. Fourthly, another interesting point to consider here is the increase in within-group inequality. Both Gini coefficients and Oaxaca-Blinder decomposition give the picture of inequality at mean. It is possible that all households within the same caste do not benefit because of the economic growth equally nor are they vulnerable equally. Given that the spread of the overall income distribution has gone up between two rounds 10, it would be interesting to look at inequality at different cut of points of the income distribution in future analysis. Lastly, from the theoretical point of view, this analysis leans in favour of modernization and economic growth hypothesis, though it must be admitted that it doesn t eliminate the caste impact completely. 10 See Appendix II for summary statistics of income and consumption expenditure for IHDS-I and II 18

19 References: Anand S., and R. Kanbur (1993): The Kuznets process and the inequality-development relationship, Journal of Development Studies, 40(1): 25-52; doi: / (93)90103-t Beller E. and M. Hout (1996): Welfare States and Social Mobility: How Educational and Social Policy May Affect Cross-National Differences in Association between Occupational Origins and Destinations. Research in Stratification and Mobility 24(4): Blau P., and O. Duncan (1967): The American Occupational Structure. Wiley Breen R. (2001): A Weberian Approach to Class Analysis, in Wright, E. O. (ed.), Alternative Foundations of Class Analysis, University of Wisconsin, Maddison, pp Chatterjee P. (1997): The Nation and its Outcasts, in S. Kaviraj (ed.) Politics in India, Delhi: Oxford University Press, pp Davis K., and W. Moore (1970 [1945]: "Some principles of stratification." American Sociological Review, 10 (2), Deininger K. and L. Squire (1996): A new dataset measuring income inequality, World Bank Economic Review, 10: Deshpande A (2000): Recasting Economic Inequality, Review of Social Economy 58(3) (2000): Deshpande A (2010): Does Caste Still Define Disparity? A Look at Inequality in Kerala, India The American Economic Review, 90(2) DiPrete, T., and D. Grusky (1990): "Structure and Stratification for American Men and Women." American Journal of Sociology 96(1): Desai S., P. Vashishtha and O. Joshi (2015): Mahatma Gandhi National Employment Guarantee Act: A Catalyst for Rural Transformation, New Delhi: National Council of Applied Economic Research Dumont, L. (1980): Homo Hierarchicus: The Caste System and its Implications, Chicago: University of Chicago Press. Giddetti G. and B. Rehbein (2014): Theoretical Approaches to Inequality in Ecoonomics and Sociology. A Preliminary Assessment, Transcience, 5(1): ISSN Gupta D. (1984): Continuous Hierarchies and Discrete Castes, Economic and Political Weekly XIX

20 Kuznets S. (1955): Economic Growth and Income Inequality, The American Economic Review, XLV:1-28. Lindert P. (1986): Unequal English Wealth Since 1670, Journal of Political Economy, 94: McCall L., and C. Percheski (2010): Income Inequality: New Trends and Research Directions, Annual Review of Sociology, 36: ; DOI: /annurev.soc Morris M., and B. Western (1999): Inequality in Earnings At The Close of The Twentieth Century, Annual Review of Sociology, 25: DOI: /annurev.soc Munshi, K., and M. Rosenzweig (2009): Why is Mobility in India So Low? Social Insurance, Inequality, and Growth, Working Paper No , National Bureau of Economic Research Neckerman K., and F. Torche (2007): Inequality: Causes and Consequences, Annual Review of Sociology, 33: Paterson L. and C. Iannelli (2007): "Social Class and Comparative Study of England, Wales, and Scotland, Sociology of Education, 80(4): Ravallion M. (2014): Income inequality in the developing world, Science, 344 (6186): ; DOI: /science Williamson J. (1985): Did British Capitalism Breed Inequality? Boston, MA: Allen and Unwin Vanneman R., and A. Dubey (2013): Horizontal and Vertical Inequalities in India., in Income Inequality: Economic Disparities and the Middle Class in Affluent Countries, ed. by J.C. Gornick and M. Jantti. Stanford, CA: Stanford University Press 20

21 Appendix I Variables Definitions Highest Education: The highest level of education of an adult member in the household Urban: Urban or Rural area according to Census 2001 for IHDS-I and Census 2011 for IHDS-II Total Adjusted Household Income: Total Household Income Yearly / Number of Persons in the Household Total Adjusted Household Consumption Expenditure: Total Household Consumption Expenditure Yearly / / Number of Persons in the Household Gini Coefficient: The Gini Coefficient is defined as a ratio of the areas on the Lorenz curve diagram. If the area between the line of perfect equality and Lorenz curve is A and the area under the Lorenz curve is B, then the Gini coefficient is A/ (A + B) A B 21

22 Appendix-II Summary Statistics Highest Adult Education in the Household Caste Groups Mean Standard Deviation Mean Standard Deviation Forward Caste OBC SC ST Total Adjusted Household Income Caste Groups Mean Median Standard Deviation Mean Median Standard Deviation Forward Caste OBC SC ST Total Adjusted Consumption Expenditure Caste Groups Mean Median Standard Deviation Mean Median Standard Deviation Forward Caste OBC SC ST

23 0 0 Density Density Density Density Kernel Density Plots for Consumption Expenditure Kernel density estimate Forward Caste in Foreward Caste in Kernel density estimate SC in SC in Adjusted Log Household Consumption Expenditure kernel = epanechnikov, bandwidth = Adjusted Log Household Consumption Expenditure kernel = epanechnikov, bandwidth = Kernel density estimate ST in ST in Kernel density estimate OBC in OBC in Adjusted Log Household Consumption Expenditure kernel = epanechnikov, bandwidth = Adjusted Log Household Consumption Expenditure kernel = epanechnikov, bandwidth =