Measurement of Multidimensional Poverty in India: A State-level Analysis

Abstract

This article measures Multidimensional Poverty Index (MPI) in India using National Sample Survey (NSS) data on “Consumption Expenditure” for the period 2004–2005 and 2011–2012, adopting Alkire and Foster’s (2011, Journal of Public Economics, vol. 95, pp. 476–487) methodology. It considers three main indicators, namely standard of living, education and income at the level of households or persons. The results show that multidimensional poverty head count has declined from 62.2 per cent in 2004–2005 to 38.4 per cent in 2011–2012. However, separate rural and urban regional analysis clearly indicates a sharp decline in rural poverty compared to urban poverty reduction. Lack of education of the household members made the highest contribution to poverty, followed by income and standard of living in India. A state-level analysis shows that Jharkhand, Uttar Pradesh, Rajasthan, Orissa, Bihar, Chhattisgarh and Arunachal Pradesh have a higher poverty head count ratio, while Kerala, Mizoram, Nagaland, Punjab, Himachal Pradesh and Haryana have a lower poverty rate.

Keywords

Consumption expenditure multidimensional poverty poverty indices India

Introduction

Impact of economic reforms and economic growth on inequality and poverty are intensely debated issues in India. Due to paucity of data, especially of income in different time periods, there have been limitations in explanations and interpretations in these discussions. Economic growth cannot be the main objective of economic policy; it is also important to ensure that the benefits of growth reach all sections of society. To examine whether growth has been beneficial to all the sections of society, we need to measure poverty in all its manifestations. Also, poverty measurement is essential to assess how an economy is performing in terms of providing a certain minimum standard of living to all its citizens. The measurement of poverty and the perception of poverty certainly vary across countries and states over time. Therefore, the measurement of poverty has very significant policy implications.

C. Rangarajan suggested that poverty is easy to perceive but difficult to be precise about.¹ India has had a long history of studies on the measurement of poverty. The latest poverty measurement by the expert group headed by Rangarajan has taken on a new look at the methodology for the measurement of poverty. The new poverty line not only considers calorie intake but also protein and fat consumption. Also, it has introduced new norms for the measurement of non-food expenditures in the construction of poverty line. However, it considers only consumption expenditure to measure the poverty in India. It is important to note here that due to unavailability of income data in India, consumption is considered as proxy of income.

The unidimensional poverty measurement has been criticised by many economists; for example, Sen (1980) argued that income may not be translated into basic needs. Therefore, deprivations in areas such as education, health, social and political status are very important to measure poverty, as they are also harder to quantify through price. Therefore, measurement of multidimensional aspects of poverty is very important, as it considers two approaches that is, poverty as capability deprivation (Sen, 1999) and poverty as counting measure of deprivation (Atkinson, 2003). Recent studies (e.g., Bourguignon & Chakravarty, 2003; Tsui, 2002) have emphasised on the multidimensional aspect of poverty. Multidimensional Poverty Index (MPI) considers both incidence and intensity of deprivation, which is superior to measuring poverty only on the basis of income and consumption level. The MPI also has advantages over the Human Development Index (HDI), such as HDI measures well-being at country level; on the other hand, MPI uses household-level data.

The Oxford Poverty and Human Development Initiative (OPHI) jointly with the United Nations Development Programme developed the MPI in 2010. The MPI uses different factors to determine poverty beyond income. Alkire and Santos’ (2010) method for calculating MPI has been used by OPHI to analyse poverty status for 104 countries. However, Alkire and Foster’s (2011) methodology is used to measure MPI more widely, as it summarises a plurality of not perfectly overlapping deprivation domains into a consistent parametric class of multidimensional poverty indices (Pacifico & Poege, 2017). The method gained popularity, as it was based on Foster–Greer–Thorbecke (FGT) indices, and it can also be used for decomposition not only for population subgroups but also by deprivation indicators.

Against this backdrop, the main objective of this article is to measure the MPI for India. For this purpose, Alkire and Foster (2011) method has been used. National sample survey household or unit-level data on consumption expenditure for the periods 2004–2005 and 2011–2012 have been used to calculate the MPI. Used for analysis in this article are three major categories of indicators, that is, standard of living, education and income to measure MPI in India. The novelty of article paper is the usage of National Sample Survey (NSS) data to calculate MPI. It is important to note here that government of India makes many important decisions based on NSS data; for example, NSS data on consumption expenditure is used to arrive at poverty lines in the country. Therefore, the usage of NSS data to measure MPI stands can be taken as a new contribution to poverty literature in India. As the Government of India calculates poverty line for rural and urban areas separately, we also calculated MPI separately for rural and urban areas, as it provides completely different pictures and helps to prescribe appropriate policy to reduce multidimensional poverty in India, by formulating rural and urban policies separately. In addition, in this article, statelevel MPIs for rural and urban areas are calculated separately for easy formulation of state-level policies.

The rest of article is organised as follows. The second section provides a brief review of literature. Methodology and data used are explained in third and fourth sections. Results of analysis are presented in fifth section. The main conclusions and discussions are discussed in sixth and seventh sections, respectively. Finally, major limitations of the study are discussed in the eighth section.

Review of Literature

There are very few studies in India which measure multidimensional poverty. These studies could be grouped in terms of different types of data used for measuring MPI.

Using secondary data, collected from different issues of periodic reports produced by OPHI and various other research reports, Kumar et al. (2015) calculated the state-wise MPI for India. The authors used health (measured by child mortality and nutrition), education (years of schooling and child school attendance) and household status (cooking fuel, toilet, water, electricity, floor and assets) to measure the MPI for India. They found that among 28 states in India, Goa, Punjab, Himachal Pradesh and Tamil Nadu are in a vulnerable stage; Kerala is in a very good position in MPI, while the remaining states are in a very bad position. It was also revealed that 81.4 per cent of the scheduled tribes are poor, compared with 33.3 per cent of the general population in India.

Three rounds of National Family and Health Survey (NFHS) data, that is, for the years 1992–1993, 1998–1999 and 2005–2006, were considered in the study by Chaudhuri et al. (2014) to calculate MPI in India. The study considered different variables of standard of living and health and education to measure the state-level MPI. The study also highlighted the existence of intraurban imbalances and female multidimensional deprivation in India. The study found that development has been imbalanced in the country, with poorer states continuing to be poor; for example, Bihar has remained as the most deprived state over the three rounds of NFHS data. Contrary to the results of income poverty that show a systematic decrease in poverty in all states in India, the MPI calculations show an increase in poverty in few states like Arunachal Pradesh, Tripura and Manipur by 4.7 per cent, 5 per cent and 0.7 per cent, respectively, during 1992–1993 and 2005–2006. The study by Alkire and Seth (2008) uses the revised 2002 below the poverty line (BPL) methodology and NFHS to calculate MPI for India. The study found that up to 12 per cent of the poor sample population and 33 per cent of the extreme poor could be misclassified as non-poor by the pseudo-BPL method. Using unit data from the National Family and Health Survey 3, Mohanty (2011) measured poverty in multidimensional space and examined relationship between multidimensional poverty and child survival. The study found that child survival is significantly lower among abject poor compared to moderate poor and non-poor.

Using NSS and National Family Health Survey (NFHS) datasets, Mishra and Ray (2013) considered a wider range of deprivation dimensions and provided a comprehensive and wide-ranging assessment of changes to living standards in India for the periods 1992/1993 and 2004/2005, that is, the period of economic reforms and the immediate post-reforms period. The analysis was carried out both at regionally disaggregated levels and for socio-economic groups.

Using National Sample Survey Organization (NSSO) data, Sarkar (2012) drew up an MPI, taking into consideration eight indicators such as the highest educational attainment of household members, mean per capita expenditure, protein intake, calorie intake, employment, land, electricity and cooking fuel. Considering all the indicators, the author drew up the MPI and analysed the poverty status in rural India, comparing rural NSSO quinquennial rounds. There were two methods to define poverty line, as proposed by the Task Force. One corresponds to minimum calorie requirements and the other was obtained using the Consumer Price Index for agricultural labour of rural India. Sarkar merged these two methods by considering recent (for 61st round) Tendulkar Committee report on poverty line and Consumer Price Index for agricultural labour. These methods of poverty measurement by the Indian government have been criticised by many. It is clear that the Indian government laid more emphasis on growth over poverty removal. The slackening of Tendulkar Committee poverty line norms by the Planning Commission has resulted in presenting accelerated reduction in poverty figures, but there was indeed no reduction in poverty.

Most recently, using unit data from the Indian Human Development Survey (IHDS), 2004–2005, Dehury and Mohanty (2015) estimated and decomposed the multidimensional poverty dynamics in 84 natural regions of India. Multidimensional poverty is measured in terms of indices of health, knowledge, income, employment and household environment. Results indicate that about 50 per cent of India’s population is multidimensionally poor, though with large regional variations. More than 70 per cent of the population is multidimensionally poor in the Mahanadi Basin, the southern region of Chhattisgarh and the Vindhya region of Madhya Pradesh, while it is less than 10 per cent in the coastal regions of Maharashtra, Delhi, Goa, the mountainous region of Jammu and Kashmir, the mountainous region and plain region of Manipur, Puducherry and Sikkim.

Using Karnataka Household Asset Survey (KHAS) data, Vijaya et al. (2014) constructed an individual-level multidimensional poverty measure for Karnataka, India. Results showed that an individual-level measure can identify substantial gender differences in poverty that are masked at the household level. The authors also find large potential for misclassification of poor individuals as non-poor when poverty is not assessed at the individual level.

Review of results clearly shows that the number of studies that measure MPI in India is very limited. Usage of NSS data has also been neglected. Therefore, this study seeks to fill this gap using the latest NSS data.

Methodology

Alkire and Seth’s (2008) method has been used to measure MPI by different countries focusing on different contexts. However, this study uses Alkire and Foster’s (2011) method to measure multidimensional poverty for India.

Before presenting the study results, it is necessary to explain the method in a nutshell. The method is explained by taking data for n individuals with d > 2 dimensions. [y_ij] is the matrix of achievements in this model. Each element y_ij denotes the achievement of individual i = (1(1)n in jth dimension where j = 1(1)d. Let z_j be the cut-off point (or criteria) for each dimension j .

Define an identification function I_yi,z: y_i × z → {0,1}. In particular

I_{y i, z} = 1, i f i \in S_{j}

= 0, i f i \in (S \ S_{j})

where $S = S_{j} \cup (\frac{s}{s_{j}}) \forall j \in d$ and card(S) = n. Hence, we need to first identify the set of individual S_j who are deprived in dimension j.

Three case scenarios can be considered here.

First, all y_ijs are cardinal (categorical). Let dimension j has K_j order (or category) and let k_j ⊂ k_j be the subset which denotes the deprivation set. In that case

I_{y i, z} = 1, i f i \in k_{j}

= 0, i f i \in (K \ k_{j})

Second, all y_ijs are cardinal. We defined z_j is the cut-off point of jth dimension, where j ϵ d. In that case

I_{y i, z} = 1, i f y_{i j} \leq z_{j} \forall j \in d

= 0, i f y_{i j} \geq z_{j} \forall j \in d

Third, most importantly, some y_ijs are cardinal and some ordinal, which is more commonly observed in survey data. We will use both to identify the poverty. Once the dimensional poverty is defined, we will have a matrix ${[I_{i j}]}_{n \times d}$ , such that

\begin{array}{l} I_{i j} = 1, i f y_{i j} \leq z_{j} \forall j \in k_{j} w h e n j i s o r d i n a l a n d \\ \begin{matrix} \begin{matrix} \begin{matrix}  \end{matrix} = 0 o t h e r w i s e . \end{matrix} \end{matrix} \end{array}

The next step is to aggregate that information in order to derive a single value, which distinguishes a person as poor or non-poor. It is worthwhile to mention two very popular processes here: (a) union approach (b) intersection approach. According to union approach, a person is said to be poor if (s)he is deprived at least one dimension, that is,

I_{i ▪} = \sum_{j = 1}^{d} I_{i j} > 1

where I_ij is the row vector of thematrix ${[I_{i j}]}_{n \times d}$ . On the other hand, a person is said to be poor, according to intersection approach if (s)he is deprived in all dimensions, that is,

I_{i ▪} = \sum_{j = 1}^{d} I_{i j} = d

Poor could also be defined as the intermediate situation, that is,

I_{i ▪} = \sum_{j = 1}^{d} I_{i j} = d : m \in (1, d)

In practice, a weighted mean is used where weight w_j is attached for dimension. Then, if

I_{i ▪}^{w} = {\sum^{​}}_{d}^{j = 1} (I_{i j} \times w_{j}) \geq \frac{Σ w_{j}}{j} f o r u n i o n a p p r o a c h

= Σ w_{j} f o r int e r \sec t i o n a p p r o a c h

= \bar{w} \in (\frac{Σ w_{j}}{j}, Σ w_{j}) f o r int e r m e d i a t e c a s e .

When dimensions are equally important, then $w_{j} = \frac{1}{j}$ is a plausible choice. For the purpose of this study, equal importance stance will be adhered to.

Let us define an indicator vector whose element is defined by H_i = 1, if $I_{i ▪}^{W} \geq \sum^{} \frac{w_{j}}{j}$ and 0 otherwise.

Then the column sum $Σ H_{i ▪} = q$ is nothing but the headcount of the poor, and $H = \frac{q}{n}$ is the headcount ratio. Note that each entry of the vector $I_{i ▪}$ denotes the count of deprivation of ith individual. Hence, average deprivation share across the poor is given by

A = \sum_{I = 1}^{n} I_{i ▪} / (q d)

Hence, the adjusted headcount ratio using multidimensional is

M = H \times A

Data Used

To calculate MPI, this study uses two rounds of datasets of NSS data for the years 2004–2005 and 2011–2012. Due to non-availability of income data at the individual level, the urban monthly per capita consumer expenditure (MPCE) data from the 61st and 68th rounds of the NSS are used as proxy for the years 2004–2005 and 2011–2012, respectively. Following the expert group’s (Tendulkar Committee) suggestion, the MRP (Mixed Reference Period)-based consumption data are considered.² Table 1 presents the total sample households used for the estimation of MPI for India.

While computing index, nine indicators are used to measure the index. The nine indicators are regrouped into three major dimensions, that is, education, income and standard of living, and each indicator is given a weighted score following the rule that each dimension is equally weighted 1/3, and each indicator within the same dimension is equally weighted. As education and income have only one indicator each, this study assigns equal weight, that is, 1/3 to them. However, as standard of living has seven indicators, the weighted score for this dimension is equal to 0.047 = (1/3 × 1/7).

The total number of weighted deprivations is aggregated for each household and individual, with the identification of poor based on a poverty cut-off (i.e., k) of 30 per cent as per the methodology of the UNDP-MPI. Therefore, Vijaya et al. (2014) also use 30 per cent poverty cut-off to measure multidimensional poverty for the Indian state of Karnataka. Further, poverty analysis has been carried out for rural and urban areas separately in order to assess the existing rural urban differences in deprivation. The indicators taken are as follows:

Table 2 presents the details of the indicators used to measure the MPI. Based on available NSS household-level data, this study relies on different indicators to measure MPI, and two rounds of NSS data are used to construct the MPI. In the context of employment, NSS 61st round in 2004–2005 provides information for rural areas on self-employed in agriculture, self-employed in non-agriculture, agricultural labour, other labour and others. Further, for urban areas, information on self-employed, regular wage/salary earning, casual labour and others is provided in different NSS rounds. Taking this information into account, this study takes agriculture labour for rural areas and regular wage/salary earner for urban areas as cut-offs to measure state-level MPI. However, to measure all India-level MPI, the study uses regular wage/salary earner as the cut-off.

Table 1.

Sample Size and Poverty Line in Different NSS Years

Year	Rural	Urban	Rural	Urban
Year	Sample Household Size		Poverty Line (Based on Tendulkar Methodology) (₹)
2004–2005	79,298	45,346	447	579
2011–2012	59,695	41,967	816	1,000

Source: The authors.

Table 2.

Details of Indicators Used to Measure

Dimensions	Indicators/Variables	Weights	Cut-offs
Standard of living	Employment	Weight = 0.047 = (1/3) × (1/7)	Labourers
	Cultivated land		1 acre
	Irrigated land		0.5 acre
	Source of lighting		No access of electricity
	Cooking fuels		Firewood and chips, coke and coal, dung cake, charcoal
	Dwelling unit		Not owned
	Ration card		Having ration card
Education	Highest education attainment in household	1/3 = 0.333	Primary schooling
Income	MPCE	1/3 = 0.333	National poverty line

Source: The authors.

On the other hand, NSS 68th round provides information on self-employed workers in agricultural and non-agricultural sectors, regular wage/salary earners, casual labourers in agriculture, non-agriculture, others. Therefore, agriculture labourers from self-employed and casual labour are taken as cut-offs to measure rural MPI at state level. On the other hand, for urban areas, information on self-employed, regular wage/salary earning, casual labour, others is provided; therefore, regular wage/salary earning is taken as cut-off to measure urban MPI at state level. To calculate all India-level MPI, we use wage/salary earner as cut-off.

Two types of land holding information are provided by NSS, that is, cultivated land and irrigated land. In this study, 1 acre cultivated land and 0.5 acre irrigated land, respectively, are taken as thresholds. For household lighting, NSS data provide information on consumption of kerosene, other oils, gas, candle, electricity, others, and also “no lighting arrangement” at the household level. Electricity (availability–consumption) is taken as cut-off to measure state-level MPI in India. In the context of household’s cooking fuels, information on coke, coal, firewood and chips, LPG, biogas (gobar gas), dung cake, charcoal, kerosene, electricity, others, and “no cooking arrangement” is given by NSS. Consumption of firewood and chips, coke and coal, dung cake, charcoal are taken as cut-offs to measure MPI in India. NSS provides three types of information on dwelling units, that is, hired, “no dwelling units,” and others; therefore, “no dwelling unit” has been considered as the cut-off. Ration card is one of the indicators for identifying poor households in India, and households having ration card are treated as deprived. Therefore, households having ration card are treated as cut-offs to measure rural and urban MPI in India. In regard to household education, NSS data provide different types of information on education, that is, literate without formal schooling, below primary, primary, middle, secondary, higher secondary, diploma/certificate course, graduate and postgraduate and above; primary schooling is taken as cut-off in this study. Finally, as NSS data do not provide income data, monthly per-capita expenditure (MPCE) is considered as a proxy. While national poverty lines considered are calculated and recommended by Tendulkar Committee, state-level poverty lines with rural–urban distinction are used to calculate state-level rural and urban MPI, separately.

C o n t r i b u t i o n o f i n d i c a t o r i t o M P I = \frac{w_{i} C H_{i}}{M P I_{s t a t e}} \times 100

CH denotes censored headcount ratio (CH has been calculated by adding up the number of poor households deprived in a particular indicator and then dividing by the total number of households surveyed) and W denotes weights given to each indicators.

Results Analysis

Table 3 presents the calculated values of different MPI. According to poverty figures, 62.2 per cent of people in India were poor in 2004–2005, which has declined to 38.4 per cent in 2011–2012, that is, a total decline of 23.8 per cent in 7 years, and 3.4 per cent decline in each year. The adjusted headcount ratio shows that poverty in India declined from 38.3 per cent in 2004–2005 to 21 per cent in 2011–2012. The average multidimensional poverty intensity also declined from 61.6 per cent in 2004–2005 to 54.7 per cent in 2011–2012. Other measures of MPI such as, adjusted poverty gap, adjusted FGT measure, average poverty gap, and average squared poverty gap also showed a decline in poverty in India in the years 2004–2005 and 2011–2012.³ Rural urban analysis suggests that multidimensional poverty headcount index has declined in the years 2004–2005 and 2011–2012 for both rural and urban areas.⁴ In rural areas, it declined from 60.2 per cent to 16.7 per cent during the above years, which is about 43.5 per cent decline. On the other hand, for urban areas, it declined from 33.4 per cent to 20 per cent in the years from 2004–2005 to 2011–2012. Results show that rural areas have experienced a higher decline in MPI than urban areas. The calculated values of adjusted headcount ratio show that rural (or urban) areas experienced a decline from 33.2 (or 19.2) per cent in 2004–2005 to 8.4 (or 9.8) per cent in 2011–2012. The calculated values of average multidimensional poverty intensity also declined for rural (or urban) areas from 55.1 (or 57.6) per cent in 2004–2005 to 50.4 (or 49.2) per cent in 2011–2012. Other MPI measurements also show similar results, as shown in Table 3.

Table 3.

Estimated Results of the Multidimensional Poverty Index

Various MPI Measurements		2004–2005						2011–2012
Various MPI Measurements		India (in %)	SE	Rural (in %)	SE	Urban (in %)	SE	India (in %)	SE	Rural (in %)	SE	Urban (in %)	SE
H	The multidimensional deprivation headcount	62.2	0.003	60.2	0.003	33.4	0.007	38.4	0.003	16.7	0.003	20	0.003
M(0)	The adjusted headcount ratio	38.3	0.002	33.2	0.002	19.2	0.004	21	0.002	8.4	0.001	9.8	0.002
M(1)	The adjusted poverty gap	15.1	0.001	14.2	0.001	6.4	0.002	9.1	0.001	2.4	0.000	2.6	0.001
M(2)	The adjusted Foster–Greer–Thorbecke (FGT) measure	9.1	0.001	8.9	0.001	3.6	0.001	5.9	0.001	1.3	0.000	1.2	0.000
A	The average multidimensional poverty intensity	61.6	0.001	55.1	0.001	57.6	0.004	54.7	0.001	50.4	0.001	49.2	0.001
G	The average poverty gap	39.3	0.001	42.9	0.001	33.2	0.004	43.3	0.002	28.2	0.003	26.1	0.003
S(2)	Average squared poverty gap	23.8	0.001	26.7	0.001	18.5	0.004	28.1	0.001	15.4	0.003	11.7	0.002

Source: Authors’ calculation using NSS data on consumption expenditure.

Notes:

SE stands for standard error.

H: The share of poor individuals in the population.

M(0): Accounts for both the incidence of poor individuals and the intensity of their multiple deprivations.

M(1): Accounts for the incidence of poverty, the average range of deprivations and the average depth across deprivations. It is computed only with ordinal or real-valued indicators.

M(2): It is computed only with ordinal or real-valued indicators.

A: The average percentage of simultaneous deprivations suffered by the poor individuals.

G: Across all instances where poor individuals are deprived. It is computed only with ordinal or real-valued indicators.

S(2): Average severity across all instances where individuals are deprived. It is computed only with ordinal or real-valued indicators.

Table 4 presents the contribution of each indicator to the overall measure of MPI. The results show that during the study period, the level of education of household members had made the highest contribution to MPI of India, as measured by the adjusted head count ratio (M0), the adjusted poverty gap (M1) and the adjusted FGT measure (M2) in both the time periods. Most interestingly, the contribution from level of education was found increasing in the period from 2004–2005 to 2011–2012. Except M(2), M(0) and M(1) measurements show that income and standard of living made the second and third highest contributions to MPI of India. Sub indicators of standard of living show that ration card distribution made the highest contribution to MPI, as measured by M(0) in 2004–2005 and 2011–2012. The measurement of the adjusted headcount ratio shows that employment status, use of cooking fuels and lighting used by the household occupied second, third and fourth ranks in terms of higher contribution to poverty for the time period from 2004–2005 to 2011–2012. In fact, contribution from employment status and use of cooking fuels slightly increased in the years from 2004–2005 to 2011–2012, whereas the contribution to poverty index from lighting used by the household declined.

Table 5 presents contribution of each of the indicators to the overall measure of MPI at rural urban levels, separately. The results show that education level of the household member contributed the highest (i.e., 53.3%) to the multidimensional rural poverty, as measured by adjusted headcount ratio, adjusted poverty gap and adjusted FGT measure in 2004–2005, followed by standard of living index and income. However in 2011–2012, the highest contribution came from income (i.e., 52.8%) of the household, followed by standard of living (i.e., 30.8%) and education (17.3%) to the adjusted headcount ratio. However, education remains the highest contributor to MPI when it is measured by adjusted poverty gap and adjusted FGT measure. In the context of urban areas, income contributed higher (i.e., 50.4%) than standard of living (31.4%) and education (18.2%) to adjusted headcount ratio in 2004–2005. The highest contributor was again income (63.4%), followed by standard of living (30.6%) and education (6%) to the adjusted headcount ratio in 2011–2012. Among the sub indicators of standard of living, the distribution of ration card contributed the highest to the rural and urban poverty in 2004–2005 and 2011–2012 as well. The contribution of ration card to adjusted headcount ratio increased for rural (or urban) areas from 8.9 (or 7.1) per cent in 2004–2005 to 9.9 (or 9.7) per cent in 2011–2012. Cooking fuels, source of lighting and employability status for both rural and urban areas also contributed heavily to the adjusted headcount ratio in both periods.

Table 4.

Contribution of Different Indicators in MPI Measurement at All India Aggregate Level

Contribution of Main and Sub Indicators (%)		2004–2005			2011–2012
Contribution of Main and Sub Indicators (%)		M(0)	M(1)	M(2)	M(0)	M(1)	M(2)
Standard of living	Employment	6.9	0.8	0.1	7.9	0.9	0.1
	Cultivated land	0.7	1.7	2.8	0.3	0.7	1.1
	Irrigated land	1.2	2.9	4.9	0.5	1.2	1.9
	Cooking fuels	6.8	0.7	0.1	7.1	0.7	0
	Lighting	3.5	1	0.2	2.8	0.7	0.1
	Dwelling unit	0.2	0.1	0	0.2	0.1	0
	Ration card	7.9	6.7	3.7	9.2	7.1	3.6
Contribution of each domain (%)
Standard of living		27	14	11.6	27.9	11.4	6.8
Education		44.8	69.9	79.8	44	75.8	87.3
Income (MRP)		28.2	16.1	8.6	28.1	12.8	5.9

Source: Authors’ calculation based NSS consumption expenditure data in 2004–2005 and 2011–2012.

Table 5.

Contribution of Different Indicators in MPI Measurement at All India Level for Rural Urban Separately

Contribution of Main and Sub Indicators (%)		2004–2005						2011–2012
		Rural			Urban			Rural			Urban
		M(0)	M(1)	M(2)	M(0)	M(1)	M(2)	M(0)	M(1)	M(2)	M(0)	M(1)	M(2)
Contribution of each indicator (%)
Standard of Living	Employment	0.9	0.7	0.4	7	3.5	1	5.2	0.9	0.1	8.4	5.4	2
	Cultivated land	0.5	1.2	1.9	3.4	10.1	18.1	0.3	1.1	2.1	1.2	4.6	10.2
	Irrigated land	1.1	2.5	4	3.8	11.5	20.6	0.6	2.2	4.1	1.5	5.6	12.4
	Cooking fuels	7.7	0.8	0.1	6.3	0.8	0.1	9.1	1.7	0.2	7	1.2	0.1
	Lighting	4.1	1.1	0.2	2.8	0.9	0.2	4.6	1.8	0.4	2.4	1	0.3
	Dwelling unit	0.1	0.1	0	1.1	0.6	0.2	0.1	0.1	0	0.5	0.4	0.2
	Ration card	8.9	6.9	3.7	7.1	7.1	4.2	9.9	11.5	7	9.7	12.3	9.2
Contribution of each domain (%)
Standard of living	23.4	13.3	10.3	31.4	34.7	44.5	30.8	19.2	13.8	30.6	30.4	34.3
Education	53.3	76.1	84.7	18.2	28.6	32.7	17.3	47.3	69	6	17.8	31.5
Income (MRP)	23.3	10.6	5	50.4	36.7	22.8	52.8	33.4	17.1	63.4	51.8	34.3

Source: Authors’ calculation based NSS consumption expenditure data in 2004–2005 and 2011–2012.

The study now moves on to state-level analysis. Figure 1 below shows how poverty as measured by the multidimensional deprivation head count scenario has changed over the years in the different states of India.⁵ The calculated values for rural show that Jharkhand ranked first among 26 states in India with 72 per cent poverty head count in 2004–2005 followed by Uttar Pradesh, Rajasthan, Orissa and Bihar. On the other hand, Kerala ranked the lowest in rural headcount ratio (i.e., 28%) in 2004–2005, followed by Mizoram, Nagaland, Punjab and Maharashtra. The results also show that out of 26 states, 20 states had more than 50 per cent of rural poverty headcount ratio in 2004–2005. As per the results of 2011–2012, Punjab tops the list in terms of lowest rural poverty headcount ratio followed by Kerala, Himachal Pradesh, Haryana and Jammu and Kashmir. This indicates that Mizoram, Nagaland and Maharastra declined in ranks from top five positions in the years from 2004–2005 to 2011–2012. Manipur ranked at the top with the highest level of poverty headcount ratio (i.e., 38%) followed by Arunachal Pradesh, Jharkhand, Orissa and Uttar Pradesh. This indicates that though Manipur and Arunachal Pradesh were not at the top five ranks in terms of higher poverty level in 2004–2005, they could reach higher ranks by 2011–2012. In contrast, Bihar and Rajasthan registered some improvement in reducing rural poverty ratio in the years from 2004–2005 to 2011–2012. As per study results, in 2011–2012, none of the states had more than 50 per cent rural poverty, whereas in 2004–2005, 20 states had more than 50 per cent rural poverty. In 2011–2012, 17 states had more than 10 per cent poverty head count, whereas 9 states had less than of 10 per cent poverty headcount ratio.

Figure 1.

Measurements of multidimensional deprivation headcount ratios

Figure 2.

Measurements of adjusted headcount ratios

Coming to state-wise urban multidimensional deprivation head count ratio, Nagaland had the lowest (i.e., 7%) urban poverty head count ratio followed by Mizoram, Himachal Pradesh, Jammu & Kashmir and Kerala in 2004–2005. In contrast, Chhattisgarh had the highest (i.e., 53%) poverty head count ratio followed by Arunachal Pradesh, Bihar, Manipur and Uttar Pradesh during the same time period. The average state urban head headcount ratio was about 30 per cent in 2004–2005. As per the results of 2011–2012, Meghalaya, Orissa, Bihar, Jharkhand and Uttar Pradesh are at ranks one to five in terms of higher urban headcount poverty ratio. The result indicates that Meghalaya, Orissa and Jharkhand were not listed as top five states in terms of state wise urban poverty measure in 2004–2005, but entered the top-5 list in 2011–2012. This indicates that these states experienced an increase in poverty rate over the period of time. On the other hand, Manipur, Arunachal Pradesh and Chhattisgarh witnessed a decrease in poverty rate over a period of time.

In contrast, Himachal Pradesh had the lowest (i.e., 3%) urban head count ratio in 2011–2012 followed by Haryana, Kerala, Punjab and Tamil Nadu. The results indicate that Jammu & Kashmir, Mizoram and Nagaland lost their top five positions in terms of lowest urban poverty head count ratio in the years from 2004–2005 to 2011–2012. State-wise average urban poverty was about 19 per cent in 2011–2012, which is lower than 30 per cent in 2011–2012.

Most importantly, the results in Figure 1 indicate that an average 32 per cent rural poverty decline in the years from 2004–2005 to 2011–2012. Rajasthan, Uttaranchal, Meghalaya and Jammu and Kashmir experienced a higher decline in rural poverty head count in the period from 2004–2005 to 2011–212 than others states. In contrast, Meghalaya, Nagaland, Orissa and Mizoran witnessed a higher increase in urban poverty head count ratio than other states. In contrast, Chhattisgarh, Arunachal Pradesh, Rajasthan and Gujarat experienced a higher decline in urban poverty. Table A1 presents the complete rankings of the states as per the calculated multidimensional poverty headcount ratio. The ranking is done in the ascending order, that is, from the lowest to the highest poverty head count ratio. In other words, the state which has the lowest poverty headcount ratio gets the first rank.

The study now takes up a discussion on the MPI, which is based on adjusted headcount ratio in Figure 2. The results do not show any change in rural adjusted headcount ratio for the years 2004–2005 and 2011–2012. According to the urban adjusted headcount ratio, Manipur was not listed among top five, but it is listed among top five highest poverty ratios as per multidimensional poverty headcount ratio in 2004–2005. On the other hand, Jammu and Kashmir (or Assam) was listed among top five states with the lowest (or highest) adjusted poverty ratio as of 2004–2005, whereas it does not rank among top five state-list as per poverty headcount ratio of 2011–2012. Overall, the study has not found any differences in the ranks calculated in terms of multidimensional headcount ratio and adjusted headcount ratio.

Now it is important to check how robust the poverty ranking is in terms of value of poverty cut-offs. It may be the case that the choice of cut-off is arbitrary and ranking of states may change drastically due to a change in the cut-off. To deal with this issue, the multidimensional deprivation headcount measures were calculated for all states with rural and urban distinction for different cut-offs for k = 4, 5 and 6, which is based on estimation of Spearman’s rank correlation coefficients as between each pair of rankings. From Table 6, it can be seen that the correlation coefficient is positive and highly statistically significant for the different ranking of the states based on k = 3 with k = 4, 5 and 6. Hence, the conclusion that the rankings for varying poverty cut-offs are highly robust, that is, multiple rankings are not obtained for different values of k. Rankings beyond k = 6 are not attempted in the study because the value of H is very low, and with few observations, the rankings could be biased.

Table 6.

Spearman’s Rank Correlation Matrix for Different Multidimensional Deprivation Headcount (H)

			2004–2005		2011–2012		2004–2005		2011–2012		2004–2005		2011–2012
			K = 4				K = 5				K = 6
			Rural	Urban	Rural	Urban	Rural	Urban	Rural	Urban	Rural	Urban	Rural	Urban
K = 3	2004–2005	Rural	0.95*				0.84*				0.86*
	2004–2005	Urban		0.84*				0.67*				0.54*
	2011–2012	Rural			0.97*				0.80*				0.80*
	2011–2012	Urban				0.97*				0.62*				0.39**

Source: The authors.

Note: * and ** Indicate statistical significant at 1% and 5% levels, respectively.

Main Conclusions

The present article measures the MPI for India by considering two different rounds of NSS on consumption expenditure for the period of 2004–2005 and 2011–2012. Alkire and Foster’s (2011) methodology is used to measure multidimensional poverty by considering three main indicators, that is, standard of living, education and income at the household levels. Standard of living is measured by considering seven sub indicators, that is, employment status, acultivated land possessed, integrated land possessed, source of lighting, cooking fuels, dwelling unit and ration card holding status of the households. Education is measured by the highest education attainment of the household members. Due to paucity of income, monthly per capita consumption expenditure data are used as proxy of income data. A person is identified as poor if the household is deprived in 30 per cent of all weighted indicators. Spearman’s rank correlation coefficients between the ranking of states by considering different poverty cut-offs do not show any significant difference.

Use of latest NSS data to measure MPI and consideration of different household-level characteristics stands as a new contribution of this article. The results show that 62.2 per cent of people in India were poor as measured by multidimensional poverty head count in 2004–2005, which declined to 38.4 per cent in 2011–2012. Also, the adjusted headcount ratio (or average multidimensional poverty) declined from 38.3 (or 61.6) per cent in 2004–2005 to 21 (or 54.7) per cent in 2011–2012. The adjusted poverty gap, adjusted FGT measure, average poverty gap and average squared poverty gap also showed a decline in poverty in India in the years from 2004–2005 to 2011–2012.

Rural (or urban) Multidimensional Deprivation poverty Headcount at all India level declined from 60.2 (or 33.4) per cent in 2004–2005 to 16.7 (or 20) per cent in 2011–2012. This indicates that rural India has experienced a sharper decline in poverty than urban India.

The estimated results indicate that level of education of the household members has made the highest contribution to poverty in India, followed by income and standard of living. Sub-indicators of standard of living (ration card status, employment status, use of cooking fuels and source of lighting of the households) are the major determinants of poverty level in India. At state level, analysis suggests that Jharkhand, Uttar Pradesh, Rajasthan, Orissa and Bihar had the higher multidimensional rural poverty, whereas Kerala, Mizoram, Nagaland, Punjab and Maharashtra had the lower level of poverty as of 2004–2005. As revealed by the study, in 2011–2012, Punjab, Kerala, Himachal Pradesh, Haryana and Jammu & Kashmir had lower level of poverty, whereas Manipur, Arunachal Pradesh, Jharkhand, Orissa and Uttar Pradesh had higher level of poverty. In regard to urban multidimensional poverty Nagaland, Mizoram, Himachal Pradesh, Jammu & Kashmir and Kerala witnessed lower poverty headcount ratio, whereas Chhattisgarh, Arunachal Pradesh, Bihar, Manipur and Uttar Pradesh witnessed higher poverty ratio in 2004–2005. Further, as of 2011–2012, Meghalaya, Orissa, Bihar, Jharkhand and Uttar Pradesh witnessed higher urban headcount poverty ratio, while Himachal Pradesh, Haryana, Kerala, Punjab and Tamil Nadu witnessed lower poverty ratio. Among the other Indian states, Meghalaya, Nagaland, Orissa and Mizoran experienced an increase in urban poverty head count ratio in the years from 2004–2005 to 2011–2012.

Discussion

This article argues that though poverty in India has declined during the period of higher economic growth, degree of poverty reduction varies across states and also across rural and urban regions. Effective distributive policies which benefit the rural population of Jharkhand, Uttar Pradesh, Rajasthan, Orissa, and Bihar have to be formulated and implemented, as these states suffer higher multidimensional rural poverty. A similar consideration is essential for the urban areas of Chhattisgarh, Arunachal Pradesh, Bihar, Manipur and Uttar Pradesh, as they too suffer from higher multidimensional poverty head count ratio. Especially, Bihar and Orissa have to be given special consideration, as both rural and urban areas of these states have higher multidimensional poverty head count ratio. It is to be noted that these states are mainly based on agrarian economy, that is, agriculture, forestry and allied activities. Indian agriculture is characterised by lower productivity and growth rate. Therefore, it is important for the state to switch over from an agriculture-driven to an industry-driven economy through rapid urbanisation. These states also have experienced lower urbanisation than other states in India. Therefore, there is an urgent need to promote urbanisation, so that the state can take advantage of higher productivity and higher economic growth that emanate from urbanisation. This will help agriculture workers not to depend solely on agriculture but also on industry, so that not only their income rises, but they also feel more confident with assured and regular income. These states are also rich in natural resources such as coal, limestone, iron, nickel, cobalt, chromium and marble. There is a paramount need to develop industries based on these local resources in order to ensure that local resources are utilised properly and the people get optimum benefits. Most of the states are also endowed with spots of scenic beauty and therefore have high tourism potential. It is also suggested that tourism industries can be supported for the overall development of these states.

Lack of transport infrastructure is found to be one of the major drawbacks that comes in the way of setting up industries in these states. Other problems haunting these states are shortage of electricity, capital and telecommunication infrastructure, and these issues can be solved only by providing adequate funds. Therefore, the real issue is the lack of funding. Given this scenario, it is imperative to the government to steps to formulate a long-term saving plan, so that the money collected in this manner can be invested in building the required transport, transport infrastructure, communication infrastructure, and improving delivery of basic services for industries to come up.

Apart from this, these states need a boost in education through promoting institutions for basic to higher education. As of 2011, Bihar’s illiteracy rate was whopping 36.2 per cent, much higher than in other Indian states. Gross Enrolment Ratio (GER) in higher education in India is 20.4, which is calculated for 18–23 years age group. Less than 0.5 per cent of the total students are enrolled in doctoral programme (PhD degree). This shows that urgent attention needs to be paid to the education system of India. Also industry-oriented technical and vocational education is required to be provided from school level, so that an industry-complaint labour force is created. Now two ways are open to the government to fund for higher education: initially, government can fund to build higher educational institutions and later can collect income tax from the higher educated employees. Second, the government can offer fellowships for pursuing higher education, and recipients of such fellowships can later pay back their dues after getting employment.

The above-mentioned policies are prescribed as guidelines for the Indian policymakers. Both poor and not-so-poor states will find these suggestions useful in improving their poverty situation. It is hoped that these policies will improve education, income and standard of living of the people, so that they can escape from the clutches of multidimensional poverty and deprivation.

Limitations of the Study

We acknowledge that the analysis in the article has some limitations. The original Global MPI uses indicators of health, education and standard of living for the measurement of poverty. However, as NSS data do not provide information on health, we could not use health indicators to measure the MPI. Instead of a health indicator, we used income indicator which is proxied by MPCE. The results perhaps could vary had we used health indicator. On the other hand, the measurement of MPI can avoid monetary measurement of poverty as income is an outcome indicator and others are input indicators. It is also important to note that we consider MPCE as a proxy of income. Income increases at a higher rate, but consumption which does not increase in the same proportion cannot be a true indicator of income.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

Appendix

Table A1.

Multidimensional Poverty Headcount Rank (k = 30%)

Sr. No.	State	2004–2005		2011–2012
Sr. No.	State	Rural	Urban	Rural	Urban
1	Andhra Pradesh	17	8	14	8
2	Arunachal Pradesh	15	25	25	14
3	Assam	19	15	17	21
4	Bihar	22	24	20	24
5	Chhattisgarh	18	26	18	13
6	Gujarat	11	13	16	10
7	Haryana	8	9	4	2
8	Himachal Pradesh	6	3	3	1
9	Jammu and Kashmir	12	4	5	6
10	Jharkhand	26	20	24	23
11	Karnataka	9	14	11	16
12	Kerala	1	5	2	3
13	Madhya Pradesh	21	21	21	18
14	Maharashtra	5	10	7	7
15	Manipur	20	23	26	19
16	Meghalaya	13	11	8	26
17	Mizoram	2	2	15	9
18	Nagaland	3	1	13	17
19	Orissa	23	17	23	25
20	Punjab	4	6	1	4
21	Rajasthan	24	19	19	12
22	Tamil Nadu	7	7	9	5
23	Tripura	16	16	12	15
24	Uttar Pradesh	25	22	22	22
25	Uttaranchal	14	18	6	20
26	West Bengal	10	12	10	11

Source: The authors.

Notes

References

Alkire

, & Foster

(2011). Counting and multidimensional poverty measurement. Journal of Public Economics, 95(7–8), 476–487.

Alkire

, & Santos

M. E.

(2010, July). Acute multidimensional poverty: A new index for developing countries (Working Paper No. 38). OPHI.

Alkire

, & Seth

(2008, September 1). Measuring multi-dimensional poverty in India: A new proposal (OPHI Working Paper No. 15). http://ssrn.com/abstract=1815355orhttps://dx-doi-org.web.bisu.edu.cn/10.2139/ssrn.1815355

Atkinson

(2003). Multidimensional deprivation: Contrasting social welfare and counting approaches. Journal of Economic Inequality, 1(1), 51–65.

Bourguignon

, & Chakravarty

S. R.

(2003). The measurement of multidimensional poverty. Journal of Economic Inequality, 1(1), 25–49.

Chaudhuri

, Gulati

, Banerjee

, Roy

, Halder

, Karim

, & Vertier

(2014). Multi-dimensional poverty index—A state level analysis of India (Working Paper No. 5). NOPOOR. file:///C:/Users/hp/Downloads/NOPOOR_Working%20Paper%20N%C2%B05_Work%20Package%203%20PovertyDimensions_BANERJEE,CHAUDHURI,MONTIER,ROY_Multidimensional%20poverty%20index%20a%20state%20level%20analysis%20of%20India.pdf

Dehury

, & Mohanty

S. K.

(2015). Regional estimates of multidimensional poverty in India (Economics Discussion Papers, No. 2015-34). Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2015-34

Kumar

, Kumar

, & Sonu (2015). Multi-dimensional poverty index (MPI): A state wise study of India in SAARC countries. International Journal of Enhanced Research in Educational Development (IJERED), 3(1), 14–21.

Mishra

, & Ray

(2013). Multi-dimensional deprivation in India during and after the reforms: Do the household expenditure and the family health surveys present consistent evidence? Social Indicators Research, 110(2), 791–818.

10.

Mohanty

S. K.

(2011, October 27). Multi-dimensional poverty and child survival in India. PLoS ONE, 6(10), e26857.

11.

Pacifico

, & Poege

(2017). Estimating measures of multidimensional poverty in Stata. The Stata Journal, 17(3), 687–703.

12.

Sarkar

(2012). Multi-dimensional poverty in India: Insights from NSSO data (Working Paper). OPHI.

13.

Sen

A. K.

(1980). Equality of what? In McMurrin

(Ed.), The tanner lectures on human values (pp. 197–220). University of Utah Press.

14.

Sen

A. K.

(1999). Development as freedom. Oxford University Press.

15.

Tsui

(2002). Multidimensional poverty indices. Social Choice and Welfare, 19(1), 69–93.

16.

Vijaya

, Lahoti

, & Swaminathan

(2014). Moving from the household to the individual: Multidimensional poverty analysis. World Development, 59, 70–81.