Poverty Among Elderly in India

Transcription

1 Soc Indic Res DOI /s Poverty Among Elderly in India Akanksha Srivastava Sanjay K. Mohanty Accepted: 24 July 2011 Ó Springer Science+Business Media B.V Abstract Using consumption expenditure data of the National Sample Survey , this paper estimates the size of poor and tests the hypotheses that are not economically better-off compared to non- in India. Poverty estimates are derived under three scenarios by applying the official cut-off point of the to household consumption expenditure (unadjusted), consumption expenditure adjusted to household size and consumption expenditure adjusted to household composition. Results show that an estimated 18 million in India are the. On adjusting the consumption expenditure to household size and composition, there are no significant differences in the incidence of poverty among and non- in India. This is in contrast to the notion that are better off than non- in India. Based on the findings, we suggest that the age dimension should be integrated into social policies for evidence based planning. Keywords Poverty Household size Household composition MPCE Elderly Non- India 1 Introduction Population ageing resulting from an increase in life expectancy and reduction in fertility is gaining momentum in many developing countries including India. The life expectancy at birth in India has increased from 50 years in 1972 to 64 years in 2004 and is projected to reach 70 years by 2020 (Office of the Registrar General and Census Commissioner 1999, 2008). By 2005, nearly 11 of the 29 states in India had reached the replacement level of fertility (IIPS and Macro International 2007). The proportion of population in the age group 60 years and above has increased from 6.8% in 1991 (57 million) to 7.4% in 2001 A. Srivastava (&) S. K. Mohanty International Institute for Population Sciences, Govandi Station Road, Deonar, Mumbai , India akankshaleo@gmail.com S. K. Mohanty sanjayiips@yahoo.co.in

2 A. Srivastava, S. K. Mohanty (77 million) and is projected to be 12.4% by 2026 (174 million) (Office of the Registrar General and Census Commissioner 2004). During , the annual growth rate of the population aged 60 years and above in India was 3.04% compared to 1.8% for the total population. The old age support ratio (ratio of number of persons aged to 60?) is expected to decline from 8.4 in 2001 to 5.2 by Along with age-structural transition, India is experiencing an epidemiological transition, an economic transition, and social and institutional changes. On the epidemiological front, the disease pattern has shifted from communicable to non-communicable diseases and the older adults face an increasing health risk (RGI 2009; WHO 2010). Socioeconomic changes such as increased urbanisation and modernisation, increasing participation of women in economic activities, mobility of the younger generation and the growth of individualism are leading to the breakdown of the joint family structure, which used to be the primary support for the in India (Knodel et al. 1992; Pandey 2009). This has altered the conventional living arrangements of the an increasing number of are living alone or in small and have to endure economic vulnerability and rising poverty (Sherlock 2000; Dreze 1990). During the period, to , the percentage of living alone or with spouse in India has recorded a 20% increase (Table 1). Kumar et al. (2011) have similar findings. Old age poverty is a significant issue because the income of the reduces, while consumption expenditure increases (largely due to an increase in health expenditure) with advancement of age (Mahal and Berman 2001). More specifically, the out of pocket expenditure on health care significantly alters the household budget, reduces consumption of non-health goods and services, reduces accessibility to health care utilization and pushes many families into the medical poverty trap (Whithead et al. 2001). Thus poverty among the tends to be more permanent than that among the non and are Table 1 Percent distribution of by living arrangement and the state of economic independence in India, Source: Computed from 52 (25.0) and 60 (25.0) rounds of NSS Living arrangement/ Economic independence Living arrangement alone/with spouse only with spouse and children without spouse but with children with other relatives with non relatives Total Percent Total (Number) 33,656 34,104 Economic independence Elderly not dependent on others Elderly dependent on spouse Elderly dependent on children Elderly dependent on grandchildren Elderly dependent on others Total percent Total (number) 32,496 34,229

3 Poverty Among Elderly in India unlikely to come out of the poverty trap (Hurd 1990). Empirical evidence suggests a U shaped relationship of age and poverty; population faces a higher incidence of poverty compared to other groups (Barrientos et al. 2003; World Bank 2006; Mujahid et al. 2008). Gasparini et al. (2007) found that in countries with weak social security systems, there is no significant difference between old age poverty and the overall poverty rates, while in countries with a well developed pension system, poverty rates are lower for the than for other age groups. Panda (1998), in his study among the in rural Orissa found that the where women live alone were the poorest and are in the need of additional socio economic support by the government. Mohanty and Sinha (2010) use a simple measure of deprivation and conclude that poverty among the living in nuclear was higher compared to that among the co-residing with children or non-. Studies also documented that poverty among is lower than that of non- and this variation was attributed to the survival bias, that is, the positive correlation of household income and life expectancy (Deaton and Paxson 1995; Pal and Palacios 2006, 2008). The incidence of poverty in general and among the in particular remains a serious issue in India though the economy has been buoyant for the past two decades (Alam and Barrientos 2010). Though old age poverty is a significant issue in developing countries, policy and research mainly focus on the pension program in the organised sector that has little relevance for the poor (Sherlock 2000). Also, the comprehension of old age poverty suffers from methodological problems, international comparability and data limitations (Treas and Logue 1976; Barrientos et al. 2003; Gasparini et al. 2007). Understanding the extent of poverty among in the wake of the breakdown of the joint family structure and other socio-economic changes is essential for evidence based planning. Besides, the existing social security system in the country is inadequate to meet the multifaceted need of the growing. This paper attempts to provide a numerical estimate of the poor and understand the economic deprivation of and non- in India. Consumption expenditure is adjusted to household size and household composition because of the differentials in the demographic structure of the. We hypothesise that there is no significant difference in the incidence of poverty between with the and without the, and the living alone or with other members are the poorest compared to any other type of. The interest for this paper is primarily due to the growing number of living amidst abject poverty in rural India. This may be partly attributed to the fragmentation of land and disintegration of the family soon after the marriage of the son. This leaves the poor with little or no resources and often children do not undertake the responsibility for the care of parents. This study has bearing on the findings of two earlier studies by Dreze and Srinivasan (1997) and Pal and Palacios (2006). Dreze and Srinivasan dealt extensively with poverty among the widowed in India and estimated the incidence of poverty in 20 types of. They found significantly higher incidence of poverty among widowed when consumption expenditure was adjusted for household size and composition. However, their study did not include by their co-residence and this study addresses the issue. Pal and Palacious had examined the incidence of poverty among and non-, but we believe that an analysis by co-residence of the would help to draw valid inferences.

4 A. Srivastava, S. K. Mohanty 2 Data and Methods 2.1 Data In India, the National Sample Survey Organisation (NSSO) collects periodic data on household consumption expenditure in various rounds. From the household consumption expenditure, the monthly per capita consumption expenditure (MPCE) is computed and used to explain the economic differentials and derive the incidence of poverty in the population sub-groups. The is fixed according to the money values of calories intake (2,400 in rural and 2,100 in urban areas) and adjusted for the state level price line. In , the national level was fixed at a monthly cut off of rupees 356 and rupees 539 per person for rural and urban India, respectively (Planning Commission, Government of India 2007). Among other things, these estimates are not adjusted for age composition of household, size of the household and not segregated for and non-. It has been established that the consumption expenditure and the poverty level vary by household size and the demographic composition of the, and the choice of equivalence scale can sometimes systematically affect absolute and relative levels of poverty (Buhmann et al. 1988; Nelson 1988; Meenakshi and Ray 2002; Pal and Palacios 2006). The need to adjust consumption expenditure to household size is governed by the fact that larger tend to have a comparative advantage in their standard of living over smaller due to economies of scale. Several examples of scale economies include house rent, utilities such as electricity and purchase of food items. Similarly, the consumption needs of the children and those of adults are not identical and hence, consumption expenditure is sensitive to the age composition of household members. We have used the 61st round of consumption expenditure data (Schedule 1.0), collected by the NSSO during The 61st round (Schedule 1.0) of NSS was one of the largest ever consumption expenditure surveys conducted in the country covering a total of 124,644. It was specifically designed to provide reliable estimates of poverty for the districts of India (NSS 2006). The data on consumption expenditure were based on a reference period of 30 days, that is, uniform recall period (URP) and with a reference period of 30 days and 365 days that is, mixed recall period (MRP). We have used the consumption expenditure based on URP and the MPCE is the dependent variable in the analyses. We have used the state specific cut-off point of the poverty based on URP separately for rural and urban areas to derive the unadjusted poverty estimates at the state level. The population weight is used to derive the estimates of MPCE and poverty. Population projection by the expert group of the Registrar General of India (Office of the Registrar General and Census Commissioner 2006) has been used to estimate the number of the. The statistical package STATA 10.0 has been used for the analysis. 2.2 Methods For national estimates, we have classified the into three groups, namely, where the lives alone or with other members, where live with non- members, and non-. However, for state level estimates, we have carried out the analysis for two groups, namely, and non due to sample size constraint.

5 Poverty Among Elderly in India We have derived three alternatives estimates of poverty using the same cut off point of poverty to facilitate the comparison. 1. Official cut-off point of poverty applied to household consumption expenditure (unadjusted poverty estimates). 2. Official cut-off point of poverty applied to consumption expenditure adjusted for household size (poverty estimates adjusted to household size). 3. Official cut-off point of poverty applied to consumption expenditure adjusted for household composition (poverty estimates adjusted to household composition). The equation for deriving the poverty adjusted for economies of scale in its simplest form may be given as: X ¼ Y=A h where, X is the total household expenditure, A is the household size, and h is the degree of household economies of scale, 0 B h B 1. The closer the value of h to 0, more weight is assigned for household size and the closer the value of h to 1, less weight is assigned to the value of household size and h may be assigned different intermediate values such as 0.2, 0.4, , 0.8 and 0.9. In the study, we have used the value of h as 0.9 for adjusting MPCE. Equation 1 can be extended to take into account the economies of scale and adult equivalent scale and is given as: ð1þ X ih ¼ Y h = ða 1 C 1 þ a 2 C 2 þ AÞ h ð2þ where, X is the MPCE of an individual i living in the household h, Y is the total expenditure of the household. A is the number of adults, C 1 is the number of children under 6 years of age, and C 2 is the number of children between 6 and 14 years in the household. We have used the values of a as a 1 = 0.5, a 2 = 0.75 and A = 1 as suggested by Deaton and Zaidi (2002) for adjusting MPCE for middle income countries. Coale and Hoover (1958) developed an adult equivalent scale by assigning a weight of 1.0 for male age 10 years and above, 0.9 for an adult female 10 years and above and 0.5 for a child of less than 10 years of age. 3 Results 3.1 Living Arrangement and Economic Dependence We begin the discussion by giving the context of living arrangement and economic dependence of the as these are important indicators to assess their well being in lieu of the changing demography, economy and society (Table 1). To examine the changes in the living arrangements and economic dependence, we have used the 52nd (25.0) and 60th (25.0) rounds of NSS data, conducted in and , respectively. These surveys provide detailed information on socio-economic conditions, living arrangements and health care utilisation of the population. It is observed that about one in five lives alone or with the spouse and this trend is increasing. On the other hand the economic dependence of revealed that about one-third of the do not depend on others, while about half of the depend on their children for old age support.

6 A. Srivastava, S. K. Mohanty Fig. 1 population the by age of the head of the household in India, Source: Computed from 61(1) round data of NSSO data 3.2 Poverty by Age and Household Size To understand the age dimension of consumption poverty, we have plotted the incidence of poverty by age of the head of the household (Fig. 1). The unadjusted poverty is age sensitive; higher in the younger age group and declines with increase in age (32% in the age group of compared to 20% in the age group of 71 years and above). The patterns are similar for rural and urban India, and for major states. We have further examined the variation in poverty estimates by household size. Based on the literature, we have adjusted the age composition of household members to adult equivalent scale by assigning a weight of 1 to adult (15 years and above), 0.75 for children aged 6 14 years and 0.5 for children under 6 years to derive the adult equivalent scale. Similarly, we put a small value of h = 0.9 to adjust for household size. We have used the official poverty cut-off point to compare the incidence of poverty. On adjusting consumption expenditure to household composition, we observed that the poverty differentials across the age group have narrowed down substantially. The incidence of poverty varies in a narrow range in all age groups except in the age group 71 and above. Similarly, by adjusting the consumption expenditure for household size, the poverty differentials among the age group narrowed down substantially. The relatively low poverty among the 71? may be because of survival bias as attributed by Pal and Palacios (2008) (relatively better off survives bit longer). The differential in unadjusted poverty by household size reveals that large tend to have a higher incidence of poverty. For example, among with two members, 10% were the compared to 30% among with five to seven members and 38% among with eight members and more (Fig. 2). The pattern holds true for rural and urban India and for the states. In the following section, we have estimated the number of the and the adjusted poverty estimates among and non in India. 3.3 MPCE Among Elderly and Non Households Table 2 summarises the proportion of, estimated number of and the mean MPCE of and non- by place of residence and selected

7 Poverty Among Elderly in India Fig. 2 the (unadjusted) by household size in India, Source: Computed from 61(1) round data of NSSO states in India. We define an household as one with at least one member of 60 years and above. In 2005, 28% rural and 24% of urban in India had at least one member and the estimated number of was about 62 million in the country (44 million in rural and 17 million in urban India). The size of population varies from 10 million in Uttar Pradesh to 1.2 million in Assam. With respect to household consumption expenditure, there were no significant differentials in the MPCE of and non- in urban areas (rupees 1,052 vs. rupees 1,053) while differentials were small (rupees 579 vs. rupees 550) in rural India. There is a large variation in MPCE in and non- among the states of India. In urban areas where fertility level is close to the replacement level, 14 of the 17 states (except Andhra Pradesh, Haryana, Jharkhand, Tamil Nadu and Uttar Pradesh) had higher MPCE in non- compared to. In the states of Kerala, Maharashtra and Karnataka (rural), the mean MPCE in non- is higher compared to. In general, the state level patterns of MPCE among and non- are mixed. 3.4 MPCE and Poverty Adjusted to Household Size and Composition This section depicts the relative economic conditions of and provides methodological insight on sensitivity of consumption expenditure to economies of scale and adult equivalents. We have carried out this exercise for the country; separately for rural and urban areas and by household type. We begin the discussion by plotting the unadjusted MPCE among non- and in India (Fig. 3a, b). We have truncated 2,469 of 124,642 that had MPCE of more than rupees 3,000. The distribution of MPCE among and non- is similar indicating that the economic conditions are probably similar in and non-. Following this, we had assigned different values of theta (h) and alpha (a) to understand the impact of household size and composition on MPCE for and non. To understand the economic well being of by their co residence, we have categorised into where live alone or with other members, and where live with non members. Table 3 provides the average household size and the adjusted MPCE for different values of h and a for and non- by place of residence. The average

8 A. Srivastava, S. K. Mohanty Table 2 Average household size and mean MPCE (in rupees) by type of household in major states of India, States Rural Urban Average house holds size No. of house holds Percentage of house holds Estimated no of (000 ) Mean MPCE of house holds Mean MPCE of non- house holds Average size No. of house holds Percentage of house holds Estimated no. of household (000 ) Mean MPCE of house holds Mean MPCE of non- house holds Andhra , , , ,059 1,046 1,011 Pradesh Assam 5.2 4, , , ,079 Bihar 5.3 1, , , Chhattisgarh 5.0 3, , ,029 Gujarat 4.7 7, , , ,182 1,034 1,153 Haryana 5.1 3, , ,217 1,111 Jharkhand 5.2 4, , Karnataka 4.7 7, , , ,119 1,023 1,037 Kerala 4.3 5, , , , ,210 1,363 Madhya 5.2 9, , , Pradesh Maharashtra , , , ,627 1,142 1,151 Orissa 4.7 6, , , Punjab 5.1 3, , , ,205 1,386 Rajasthan 5.5 8, , , Tamil Nadu 3.8 8, , , ,163 1,143 1,055 Uttar , , , , Pradesh West , , , ,730 1,193 1,088 Bengal India , , , ,387 1,052 1,053

9 Poverty Among Elderly in India (a) Density 0 5.0e (b) Density 0 5.0e mpce mpce30 Fig. 3 Unadjusted monthly per capita consumption expenditure among and non in India, Source: Computed from 61st round data of NSSO household size varies largely; from 1.6 in where the lives alone or only with other members, 6 in where the live with non- members and 4.5 among non-. As the value of theta tends to 1, there are no economies of scale. On the other hand, when the value of theta tends to 0, the adjusted MPCE increases, indicating that there are full economies of scale. It has been recommended that for transitional economies, a moderate value of theta captures size economies well. In rural areas, the unadjusted mean MPCE was rupees 703 among the where live alone or with other members, rupees 611 among the where live with non members and rupees 597 among the non. On assigning a value of h = 0.8, the mean MPCE for with the living alone or with other members (rural) was rupees 758 compared to rupees 850 for living with other members and rupees 783 for non. The differences are smaller for rural and urban areas. Similarly, the adult equivalent scale under the two scenarios has been derived for both and non-. In the first scenario, a weight of 1 is assigned to an adult (15 years and above), 0.6 for children of 6 14 years of age and of 0.4 for children of 0 5 years of age. In the second scenario, a weight of 1 is assigned to adults (15 years and above), 0.75 to children in the age group 6 14 years, and 0.5 for children in the age group 0 5. By adjusting for household composition, we found that the differences in mean MPCE by household type are smaller. The mean MPCE among the household where live

10 A. Srivastava, S. K. Mohanty Table 3 Unadjusted and adjusted MPCE (in rupees) and percentage of population by type of household in India, Unadjusted and adjusted MPCE for household size and composition MPCE (rural) MPCE (urban) MPCE (combined) alone/with other members with non members Non house holds alone/with other members with non members Non house holds alone/with other members with non members Non house holds Unadjusted MPCE ,889 1,150 1, Average household size Different values of h h = ,414 1,874 2,785 4,129 3,259 1,362 2,848 2,268 h = ,012 1,611 2,640 3,494 2,851 1,297 2,391 1,963 h = ,689 1,388 2,505 2,962 2,502 1,237 2,012 1,705 h = ,418 1,198 2,382 2,517 2,201 1,181 1,696 1,484 h = ,192 1,037 2,267 2,143 1,943 1,129 1,433 1,295 h = , ,161 1,828 1,720 1,081 1,213 1,133 h = ,063 1,563 1,528 1,036 1, h = ,973 1,339 1, Different values of a a1 = 0.4, a2 = 0.6, A = ,889 1,253 1, a1 =0.5, a2 =0.75, A = ,889 1,222 1, a1 =0.5, a2 =0.75, A = 1.0 and ,063 1,645 1,628 1,036 1,097 1,078 h = 0.9 population living below with official cut-off point a population living below by adjusting to household size of 0.9 b

11 Poverty Among Elderly in India Table 3 continued Unadjusted and adjusted MPCE for household size and composition MPCE (rural) MPCE (urban) MPCE (combined) alone/with other members with non members Non house holds alone/with other members with non members Non house holds alone/with other members with non members Non house holds by adjusting for composition c by adjusting for size and composition d a b c d Consumption expenditure is not adjusted for size and composition Household size is adjusted to 0.9 Assigned a weight of 0.5 for children under age 6, 0.75 for children 6 14 and 1 for adult Consumption expenditure adjusted to size and composition

12 A. Srivastava, S. K. Mohanty alone or with other members remained the same (rupees 703), while that of living with non members was rupees 688. On the other hand, the mean MPCE among non was rupees 705. We have adjusted the consumption expenditure to both size (h = 0.9) and composition (0.5, 0.75 and 1 for age groups 0 5, 6 14 and 15?, respectively) to understand the combined effect. We found that living in rural areas are the poorest compared to other groups (MPCE of rupees 757 among living alone or with non- members, rupees 911 among living with non members and rupees 860 among non ). We have estimated the incidence of poverty in three types of by using the official cut off point of poverty. We have used the uniform cut-off point not to estimate the number of poor in each type of household but to understand and compare the sensitivity of poverty estimates to household size and composition. If we apply the official cut-off point to the unadjusted estimates of MPCE, 15% where live alone or with other members are below the, compared to 27% where co-reside with non members and 28% among non-. But the difference in poverty estimates among and non- reduces substantially when adjusted for age composition of the household. On adjusting for both household size and household composition, it was found that the differentials in poverty estimates reduced significantly. From the analyses, it is clear that on adjusting for either household size or household composition or both household size and household composition, the economic condition of and non- are similar. This is in contrast to the notion that the are better off compared to non- in India. 3.5 Estimates of Elderly Poor in India Table 4 presents the estimated number of the in states of India using the state specific for rural and urban areas (Planning Commission, Government of India ). The percentage of the in rural areas varies from 5% in Punjab to 45% in Orissa. It is higher in those states where the overall poverty level is high (Jharkhand, Madhya Pradesh and Chhattisgarh) and lower in those states where the overall poverty level is low. The average number of in rural varies from 1.2 in Assam to 1.4 in Punjab. Based on the number of, the proportion of the and the average number of in, the estimated number of the was 17.7 million in the country of which 12.8 million resides in rural areas and 4.9 million resides in urban areas. The estimated number of (both rural and urban) varies from 3.2 million in Uttar Pradesh to 1.6 million in Bihar and is relatively lower in the state of Punjab. 3.6 Regional Pattern of Elderly Poverty Adjusting for Household Size and Composition We have further examined the sensitivity of poverty estimates to household size and composition for selected states in India. We found that the mean MPCE of the household increases when consumption expenditure is adjusted for either household size or household composition. Thus, applying the official poverty cut-off point to adjusted MPCE yields lower estimates of poverty. The can be inflated, but that will not affect the

13 Poverty Among Elderly in India Table 4 Estimated number of the in major states of India, States Rural Urban Unadjusted poverty cut off point in rupees Estimated number of in (000 0 ) Average number of in Estimated no of (000 0 ) Unadjusted poverty cut off point in rupees Estimated number of in (000 0 ) Average number of in Estimated no of (000 0 ) Andhra , , Pradesh Assam , Bihar , , Chhattisgarh Gujarat , , Haryana Himachal Pradesh Jammu & Kashmir Jharkhand Karnataka , , Kerala , Madhya , , Pradesh Maharashtra , , , Orissa , , Punjab , Rajasthan , Tamil Nadu , ,

14 A. Srivastava, S. K. Mohanty Table 4 continued States Rural Urban Unadjusted poverty cut off point in rupees Estimated number of in (000 0 ) Average number of in Estimated no of (000 0 ) Unadjusted poverty cut off point in rupees Estimated number of in (000 0 ) Average number of in Estimated no of (000 0 ) Uttar , , , Pradesh West , , Bengal India , , , ,978

15 Poverty Among Elderly in India general inferences on the relative economic deprivation of and non-. On adjusting the consumption expenditure for household composition, the differentials in poverty estimates among and non- have narrowed down in many of the states of India. In 9 of the 17 states of India (rural), namely, Assam, Haryana, Jharkhand, Kerala, Madhya Pradesh, Maharashtra, Rajasthan, Tamil Nadu and Uttar Pradesh, poverty among is higher than that of non- where as the differences are small in the states of Gujarat, Punjab and West Bengal (Table 5). The pattern is similar in urban India. This confirms that a small adjustment in household composition affects the poverty estimates in India. We have adjusted the household size by assigning a value of 0.9 and found similar results. The estimated poverty in is high in 12 of the 17 states of India (rural), namely, Bihar, Chhattisgarh, Gujarat, Karnataka, Kerala, Madhya Pradesh, Orissa, Rajasthan, Tamil Nadu, Uttar Pradesh, Haryana and Jharkhand. This brings out the fact that the relative differences in the economic condition of and non are small on adjusting for household size and composition. These findings are similar to those of Dreze and Srinivasan (1997). 3.7 Determinants of Elderly Poverty To understand the significant predictors of poverty, we have carried out two set of logistic regression analysis separately for rural and urban areas. The dependent variable is the economic deprivation of ( being poor or non-poor) and the person file is used in the analysis. Model 1 examines the significant predictor of poverty without adjusting consumption expenditure to household size and composition while Model 2 examines the significant predictor of poverty when consumption expenditure is adjusted to household size and composition. The dependent variable in Model 1 is the being poor or non poor, based on unadjusted poverty estimates. In Model 2 the dependent variable is being poor or non poor by adjusting for adult equivalent and economies of scale. In Model 1, the significant predictors of poverty are age, educational attainment, social group, religion, number of in the household and type of household. For example, in the age group 80 years and above are 32% less likely to be poor compared to in the age group years (Table 6). Similarly, having education up to graduation and above are less likely to be poor. The pattern is similar in urban areas. On adjusting the consumption expenditure to household size and composition and applying the official-cut-off point in defining the poor (Model 2), we found that age is not a significant predictor of poverty both in rural and urban areas. Education, religion and caste are significant predictors of poverty among the. In rural areas, with three or more are more likely to be poor, whereas, in urban areas with three or more are 46% less likely to be poor. 4 Discussion As a consequence of the ongoing demographic transition, India is experiencing a rapid shift in the age structure leading to an increase in the size of population. Ageing diminishes the capacity to participate in the work force, delimits the sources of income generation and increases the likelihood of falling into poverty (Sherlock 2000).

16 A. Srivastava, S. K. Mohanty Table 5 and non by adjusting consumption expenditure to household size and composition in major states of India, States Unadjusted poverty cut off Consumption expenditure adjusting for household composition Consumption expenditure adjusting for household size Rural Urban Rural Urban Rural Urban Poverty cut off point in rupees Poverty cut off point in rupees non Percentage of living below poverty line Percentage of non non living below adjusted non Andhra Pradesh Assam Bihar Chhattisgarh Gujarat Haryana Jharkhand Karnataka Kerala Madhya Pradesh Maharashtra Orissa Punjab Rajasthan Tamil Nadu

17 Poverty Among Elderly in India Table 5 continued States Unadjusted poverty cut off Consumption expenditure adjusting for household composition Consumption expenditure adjusting for household size Rural Urban Rural Urban Rural Urban Poverty cut off point in rupees Poverty cut off point in rupees non Percentage of living below poverty line Percentage of non non living below adjusted non Uttar Pradesh West Bengal India

18 A. Srivastava, S. K. Mohanty Table 6 Determinants of poverty (unadjusted and adjusted) logistic regression results indicating likelihood of being poor by place of residence in India, Background characteristics Model 1 unadjusted poverty Model 2 adjusted poverty Rural Urban Rural Urban Odds ratio Confidence interval Odds ratio Confidence interval Odds ratio Confidence interval Odds ratio Confidence interval Age group Ò ( ) 1.06 ( ) 0.93 ( ) 1.01 ( ) 80? 0.68* ( ) 0.87 ( ) 0.80 ( ) 0.90 ( ) Educational attainment Illiterate Ò Literate without formal schooling 0.65* ( ) 0.31* ( ) 0.77 ( ) 0.15* ( ) Primary 0.47* ( ) 0.42* ( ) 0.41* ( ) 0.44* ( ) Secondary 0.24* ( ) 0.12* ( ) 0.17* ( ) 0.12* ( ) Higher graduate and above 0.07* ( ) 0.05* ( ) 0.11* ( ) 0.05* ( ) Marital status Never married Ò Currently married 0.72 ( ) 1.16 ( ) 0.97 ( ) 1.02 ( ) Separated 0.82 ( ) 0.84 ( ) 1.11 ( ) 1.00 ( ) Divorced/widowed 1.36 ( ) 0.82 ( ) 0.76 ( ) 1.35 ( ) Religion Hindu Ò Muslim 1.54* ( ) 2.53* ( ) 1.03 ( ) 2.26* ( ) Christians 0.53* ( ( ) 0.65 ( ) 1.15 ( ) Others 0.40* ( ) 0.35* ( ) 0.67* ( ) 0.35* ( )

19 Poverty Among Elderly in India Table 6 continued Background characteristics Model 1 unadjusted poverty Model 2 adjusted poverty Rural Urban Rural Urban Odds ratio Confidence interval Odds ratio Confidence interval Odds ratio Confidence interval Odds ratio Confidence interval Social group Scheduled tribe Ò Scheduled caste 0.70* ( ) 1.43 ( ) 0.63* ( ) 0.82 ( ) OBCs 0.39* ( ) 0.86 ( ) 0.34* ( ) 0.45* ( ) Others 0.23* ( ) 0.32* ( ) 0.19* ( ) 0.18* ( ) Number of One Ò Two 1.18* ( ) 0.91 ( ) 0.99 ( ) 0.98 ( ) More than three 1.58* ( ) 0.39* ( ) 1.89* ( ) 0.54* ( ) Type of household alone/other Ò with non 2.14* ( ) 1.57* ( ) 0.59* ( ) 0.58* ( ) Ò is the reference category; *P \ 0.05

20 A. Srivastava, S. K. Mohanty Population ageing in India is taking place in the context of decreased familial support and a weak social security system. There has been research on health and health care utilization, morbidity and living arrangements, but little is known about the extent of poverty and deprivation among the in India. Though India has a long history of providing estimates of consumption poverty by state and residence, we do not have estimates of the poor. We are interested in the relative economic deprivation of the in the context of a changing familial structure that was the primary support of the in India. We found that the poverty estimates vary by age of head of household and by household size. The estimates are robust. However, these inferences does not hold true when we analyse the poverty by type of household ( and non- ) and adjust the consumption expenditure to household size and composition. When consumption expenditure is adjusted for household size (h = 0.9), in rural India where lives alone or with other members are the poorest compared to where the co-resides with non- members and non-. Similarly, the differentials narrowed down when consumption expenditure is adjusted for household composition. This validates our hypotheses and suggests that the economic deprivations are similar among and non-. These findings holds true for many states in India and similar to previous studies on poverty among the widowed in India. Results from the multivariate analyses confirm that poverty among the living alone or with other members is higher, compared to that among the living with non members. Education is a significant predictor of poverty among the. Poverty increases with an increase in the number of in a household in rural areas, but it is relatively less in urban areas. This perhaps can be attributed to the greater economic independence of the in urban areas. We estimated that approximately 18 million in the country were the in , based on only one dimension of poverty, that is, consumption poverty. We have not given any estimates of multidimensional poverty, which will be much higher than the consumption poverty. For instance, age related morbidity is higher at older ages and inclusion of the health dimension will increase the incidence of poverty substantially. Existing literature suggests that the differentiation in health across gradients of wealth and poverty is one of the key determinants of health in later life (Zimmer 2008; Pandey 2009). We also believe that increased health care expenditure is pushing many individuals and into the poverty trap. We suggest that future research be undertaken to understand multidimensional poverty in later life. Based on these findings, the study recommends that incentives be given for co-residence of the to encourage non-nuclear. The universal pension program for the living in nuclear with little or no education should be prioritized. Within the existing program, all the living in nuclear without or with little education should be included in the Old Age Pension Scheme (OASP). Surveys on the should incorporate details about the type of family or household, and analyse the incidence of poverty by type of household. There is an urgent need to undertake a longitudinal study to understand the well being of the. Acknowledgments The authors thank Dr. Rajesh K. Chauhan, Joint Director, Population Research Centre, University of Lucknow, for his help in data decoding.

21 Poverty Among Elderly in India References Alam, M., & Barrientos, A. (2010). Changing demographic landscape of south Asia and emerging issues of employment, ageing, and old age security. In M. Alam & A. Barrientos (Eds.), Demographics, employment and old age security (pp. 1 22). India: Maclillan. Barrientos, A., Gorman, M., & Heslop, A. (2003). Old age poverty in developing countries: Contribution and dependence in later life. World Development, 31(3), Buhmann, B., Rainwater, L., Schmauss, G., & Smeeding, T. (1988). Equivalence scales, well-being, inequality, and poverty: Sensitivity estimates across ten countries using the Luxembourg Income Study (LIS) database. Review of Income and Wealth, 34, Coale, A. J., & Hoover, E. M. (1958). Population growth and economic development in low income countries. Princeton: Princeton University Press. Deaton, A., & Paxson, C. (1995). Measuring poverty among the. NBER working paper no. 5296, Cambridge, Massachusetts. Deaton, A., & Zaidi, S. (2002). Guidelines for constructing consumption aggregates for welfare analysis. LSMS working paper 135. Dreze, J. (1990). Widows in rural India. London School of Economics, Development Economics Research Programme DEP 26. Dreze, J., & Srinivasan, P. (1997). Widowhood and poverty in rural India: Some inferences from household survey data. Journal of Development Economics, 54(2), Gasparini, L., Alejo, J., Haimovich, F., Olivieri, S., & Tornarolli, L. (2007). Poverty among the in Latin America and the Caribbean. Background paper for the World Economic and Social Survey. Hurd, M. D. (1990). Research on the : Economic status, retirement, and consumption and saving. Journal of Economic Literature, XXVIII, International Institute for Population Sciences (IIPS), & ORC MACRO. (2007). National Family Health Survey (NFHS -3), : India (Vol. I). Mumbai: IIPS. Knodel, J., Chayovan, N., & Siriboon, S. (1992). Impact of fertility decline on familial support for the : An illustration from Thailand. Population and Development Review, 18(1), Kumar, S., Sathyanarayana, K. M., & Omer, A. (2011). Living arrangements of in India: Trends and differentials. Unpublished work. Paper presented at international conference on challenges of population aging in Asia, March 14 15, 2011, INSA, New Delhi. Mahal, A., & Berman, P. (2001). Health expenditures and the : A survey of issues in forecasting, methods used, and relevance for developing countries. Research paper Cambridge, MA: Harvard Burden of Disease Unit. Meenakshi, J. V., & Ray, R. (2002). Impact of household size and family composition on poverty in rural India. Accessed through: Mohanty, S. K., & Sinha, R. K. (2010). Deprivation among the in India. In M. Alam & A. Barrientos (Eds.), Demographics, employment and old age security (pp ). India: Maclillan. Mujahid, G., Pannirsalvem, J., & Doge, B. (2008). The impact of social pensions: Perceptions of Asian older persons UNFPA Country technical services team for East and South East Asia. Thailand: Bangkok. National Sample Survey Organisation. (2006). Level and pattern of consumer expenditure, NSS 61st round, Report no 508 (61/1.0/1). Nelson, J. A. (1988). Household economies of scale in consumption: Theory and evidence. Econometrica, 56(6), Office of the Registrar General, Census Commissioner, India. (1999). Sample Registration System. New Delhi: Ministry of Home Affairs. Office of the Registrar General, Census Commissioner, India. (2004). Sample Registration System. New Delhi: Ministry of Home Affairs. Office of the Registrar General, Census Commissioner, India. (2006). Population projection for India and States ( ). New Delhi: Ministry of Home Affairs. Office of the Registrar General, Census Commissioner, India. (2008). Sample Registration System. New Delhi: Ministry of Home Affairs. Office of the Registrar General, India (RGI). (2009). Report on causes of death in India New Delhi: RGI. Pal, S., & Palacios, R. (2006). Old age poverty in the Indian States: What the household data can say? Report no. 16, discussion paper, South Asia: Human Development Sector, World Bank. Pal, S., & Palacios, R. (2008). Understanding poverty among the in India: Implications for social pension policy. Discussion paper no Germany: Brunel University and IZA.

22 A. Srivastava, S. K. Mohanty Panda, P. K. (1998). The in rural Orissa: Alone in distress. Economic and Political Weekly, 33(25), Pandey, M. K. (2009). On ageing health and poverty in Rural India. Delhi, India: Institute of Economic Growth. Planning Commission, Government of India. (2007). Poverty estimates for Accessed through: Sherlock, P. L. (2000). Old age and poverty in developing countries: New policy challenges. World Development, 28(12), Treas, J., & Logue, B. (1976). Economic development and the older population. Population and Development Review, 12, Whithead, M., Dahlgren, G., & Evans, T. (2001). Equity and health sector reforms: Can low income countries escape the medical poverty trap? The Lancet, 358(928), World Bank Report. (2006). Accessed through: Resources/ /AR06_final_LO_RES.pdf. WHO (World Health Organisation). (2010). Accessed through: Program. news/notes/2010/noncommunicable_diseases_ /en/index.html. Zimmer, Z. (2008). Poverty, wealth inequality and health among older adults in rural Cambodia. Social Science and Medicine, 66,