development of methodology towards measurement of...
TRANSCRIPT
1
Final Report
of the project
Development of Methodology
towards
Measurement of Poverty
by
Manoranjan Pal
and
Premananda Bharati
Indian Statistical Institute
203 B. T. Road
Kolkata 700108
2
Foreword and Acknowledgements A national level project like this cannot be completed alone. This fact was realized from the beginning and before we started preparing the project proposal a small team consisting of the following members was formed.
1. Professor Dipankar Coondoo 2. Professor Amita Majumder 3. Professor Manoranjan Pal 4. Dr. Premananda Bharati 5. Dr. Buddhadev Ghosh 6. Dr. Subhendu Chakrabarty 7. Mrs. Saswati Das
I was asked to write the first draft of the proposal. Professor Coondoo gave me some documents to consult with. Other members also helped me in getting other relevant documents. The first draft was circulated to other members in the team and everybody gave his/her own input to make the report as rich and informative as possible. Professor Nikhilesh Bhattacharya inspired us starting from the formulation of the project proposal. The first draft of the project proposal was corrected by him in such a manner that many valuable and necessary information had to be incorporated. We owe him much. The project could not have been completed in such a manner without his active cooperation in the beginning of the project formulation.
Finally the project was given to Indian Statistical Institute (ISI). The Director of ISI made me the Principal Investigator. He did not, however, constitute any team. This does not mean that the other members disassociated themselves from the project. They helped me the same way as they used to help me before. Whenever I convened a meeting to discuss how I should proceed, they would all attend the meeting and give me valuable advice and other help if necessary.
Mrs. Das helped Ms. Lopamudra Choudhury, the Project Assistant of our project, in preparing the NSS 55th round consumption data to SPSS friendly format. Ms. Choudhury prepared many univariate and bivariate tables which helped me in understanding features of the consumption. Ms. Choudhury left the job as she got a better offer. We had to recruit another Project Assistant – Shri Arnab Lahiri. He extracted NSS data of earlier quinquennial rounds and of 61st round to SPSS friendly format. He also left the job as he got a better offer. But within this short span of time he helped me in computing price data and analyzing these data to some extent. We were left with a very little time. I had to apply for extension of the project period. It was granted and I worked hard to complete the project. The draft report was submitted in September 2008. I was then asked to make a presentation on 12th November 2009. Professor Suresh Tendulkar, the then chairman, National Statistics Commission, Shri K. L. Dutta, representative from Planning Commission, Dr. Rajeev Mehta, representative from NSSO Kolkata and many other experts in this field were present in the seminar. The presentation was well acclaimed. It should be mentioned here that there was another presentation of the report on 17 March 2009 at NSSO, Kolkata. Professor Pronab Sen, CSI and Secretary, MoSPI, Professors S.P. Mukherjee and
3
Sharmila Banerjee of University of Calcutta, Professors Bikas K. Sinha, Madhura Swaminathan and Dipankar Coondoo of Indian Statistical Institute and many other academicians and scientists were present in the Seminar. The final report is being submitted after incorporating the remarks and suggestions made by the members present in these two seminars.
The above story looks very simple. But there were ups and downs. Initially when we were thinking what type of data would be needed and how we should proceed with the data, we consulted many scientists in the related fields. We sent the project proposal through email to Professors Angus Deaton of Princeton University and David M. Naiken of Food and Agricultural Organization (FAO) to get their valuable advice. Professor Deaton replied us with a long letter which was full of valuable advices. We could not fully understand the meaning of some of these in the beginning, but understood it more clearly only afterwards, when the results of our analysis started coming out. Professor Naiken sent some of his papers which were relevant for our project. We met Professor Subhashish Gangopadhyay of Institute of Development Studies. He had completed a project on poverty measurement using earlier NSSO data. We got some valuable tips from him.
We went to Nutrition Foundation of India (NFI) and took advices from Professors C. Gopalan and Prema Ramachandran (NFI). We went to National Institute of Nutrition (NIN) to get an idea how calorie norms are calculated. We met Professors GNV Brahmam, Shiva Kumar, K. V. Rao and Drs. K. Venkaiah, N. Balakrishna and Veena Shatrugna and discussed in details various problems that one may face in determining calorie norms at different age‐sex groups and activity patterns. We came to know that the regional food habits may also pose significant problems towards estimation of calorie norms. Finally, there are effects of weight, height and other individual parameters. We had to go to International Institute for Population Sciences (IIPS) to further our knowledge. Professors T. K. Roy, F. Ram, Subrata Lahiri and Dr. Kamla Gupta enlightened us in many aspects. We also met Professors Suresh Tendulkar, Pronab Sen of Ministry of Statistics and Programme Implementation and Professor Santosh Mehrotra of Planning Commission. All of them helped us and extended full cooperation. We gained a lot from inputs received from different sources. We are indebted to all of them and thankful to all of them,
We must have come across many other persons during this period and have forgotten to mention their names. We are thankful to all of them.
Though the report is being submitted by two of us, the contribution of others is no less than that of us.
(Manoranjan Pal)
(Premananda Bharati)
4
Development of Methodology towards Measurement of Poverty
1.1 Introduction: Ministry of Statistics and Programme Implementation (MoS&PI) requested Indian Statistical Institute to take up a research project on development of a statistical methodology towards measurement of poverty (vide letter no. D.O.NO.M‐12012/38/2005‐SSD, dated 19th October, 2005, from Dr. R.C. Panda, Additional Secretary, MoS&PI, Government of India, addressed to the Director, Indian Statistical Institute). This is in view of the fact that the norm of 2400 Kcal for rural India and 2100 Kcal for urban India for calculation of poverty line was prescribed sometime in the beginning of seventies. It is desirable to know whether these norms still hold good as of now as the consumption pattern as well as the quantum of daily energy requirements might have undergone changes during the last 35 years. 1.2 Review of Literature: Most of the existing measures of poverty, viz., Head Count Ratio (HCR), Poverty Gap Index (PGI), Squared Poverty Gap Index (SPGI), Sen’s Index of Poverty (SPI), Foster-Greer-Thorbecke Index (FGTPI) etc., depend on the poverty line. In a country like India, where there are large numbers of poor persons absolutely as well as proportionately, it is a luxury to look into the measures other than Head Count Ratio. In this report we shall only concentrate on the head count ratio. One may find the number of poor persons directly by comparing the actual calorie intake with the corresponding norms. It automatically takes care of the income derived in kind. However, it does not take into consideration of non-food items. If we agree that the calorie intake should be the basis of measurement of poverty then direct measurement is obviously superior to the measures, which use poverty line. On the other hand, once the poverty line is determined, one needs only to know the income/total consumption expenditure of the people in the community. Also there is a major controversy on whether we should consider a person to be poor if he does not consume the required food even when he earns enough income to meet the basic requirements. Of course, the poverty line will depend on the price structures, consumption pattern as well as age-sex-occupation composition of household members in each region. One should bear in mind that the estimates of number of poor are to be supplied each year by GOI and the whole process of collection of data and estimation is to be repeated every year if some other short cut method for updating the figures is not devised. It should be clear from the above discussion that there are two basic problems in measurement of poverty. First problem is the identification of poor and the second is the aggregation of available information so as to arrive at a reasonable estimate of degree of poverty. The complicacies involved in estimating the number of poor persons in India and in the states of India will be more easily understood if we briefly go through the earlier attempts made in this direction. The first attempt to measure absolute poverty in India was made by the working group consisting of eminent economists and other scientists in the social sciences, set up by the Planning Commission in 1962. The recommendations of the Nutrition Advisory Committee of the Indian Council of Medical Research in 1958 on the
5
minimum calorie requirements were taken for granted for this purpose. The second Task Force (henceforth to be referred as Task Force or TF) was constituted by Planning Commission in 1977. The Task Force submitted the report in 1979 (Government of India, 1979). They have taken the recommendation of the Nutrition Expert Group (1968) on the calorie intake norms according to 14 age and sex categories. The Planning Commission again constituted another Expert Group (henceforth to be referred as Expert Group or EG) in 1989. The report was submitted in 1993 (Government of India, 1993). It should also be mentioned here that there was another investigation carried out by a team consisting of, among others, experts from Indian Statistical Institute in 1996 on this subject. They submitted their report to the Department of Statistics, Government of India in 1997 (Gangopadhyay et al., 1997). The poverty line suggested by the expert group in 1962 was Rs. 20 per capita total expenditure (PCTE)1 per month at 1960‐61 all India prices. The basis for calculating the poverty line was the minimum normative food basket. In this approach a fixed set of commodities with specified quantities for consumption is taken as norms. This commodity vector is multiplied by the price vector to get the minimum food cost. The corresponding per capita monthly expenditure is calculated on the basis of observed relation between the food expenditures and the total expenditures of the households whose per capita food costs happen to be around the minimum food cost as calculated by the above procedure. Dandekar and Rath (1971) criticized the poverty line because the study group did not make any rural‐urban distinction. They have recalculated these poverty lines on the basis of minimum calorie requirements and have found poverty lines to be Rs. 170 per capita per annum for rural households and Rs. 271 per capita per annum for urban households which comes down after some rounding off to Rs. 15 per capita per month for rural households and Rs. 22.5 per capita per month for urban households at 1960‐61 prices. They were also aware that the cost of minimum level of living varies not only between rural and urban areas but also among different states. However, they assumed, in their calculations, the same minimum calorie requirements for rural and urban areas of different states, which is 2,250 Kcal per capita per day. Common sense suggests that poverty line should vary over regions mainly because of the variations of the tastes and preferences and the price structures over the regions. The expert committee in 1962 did not consider any regional variation in the estimate of the poverty line. The Task Force in 1979 recommended poverty lines separately for rural and urban areas at national level. They have suggested Rs. 49.09 in rural areas and Rs. 56.64 in urban areas for the base year 1973‐74 as official poverty lines. These correspond to the minimum daily calorie requirements of 2400 Kcal in rural areas and 2100 Kcal in urban areas.2 Though the calorie norms taken by the Task Force were same for rural and urban areas, the variation in the poverty lines and the daily calorie requirements was due to the differences in age‐sex compositions. For updating in subsequent years the consumption basket remained the same, and only the price changes were taken into consideration. The difference between the above two methods should be noted here. In the first case the consumption basket was pre‐specified. Obviously the consumption basket satisfied the specified uniform calorie norm. In the second case for each age‐sex combination, the average per capita
1 Though income is a useful measure of well being of a person, a more direct measure is the consumption expenditure. Per capita total expenditure (PCTE) is taken monthly and may be denoted as MPCTE or MPCE. Consumption expenditure data are more reliable and sTable than income data. 2 To be more precise the daily calorie requirements were worked out as 2435 Kcal for rural and 2095 Kcal for urban areas.
6
food items consumed were first determined for different PCTE classes and then the calorie intakes found. A relation between this per capita calorie intake and the average PCTE was obtained for each combination. This relation was used to find the PCTE against the calorie norm. These PCTEs were then averaged to arrive at the poverty lines. Observe that, in this case also, the consumption baskets had to be constructed, in the process, from the observed data for each age‐sex category. Gangopadhyay et al. (1997) took six alternative poverty lines and found the poverty scenario of different regions of India for each line. The six poverty lines are defined as:
1. OPL: Poverty line based on the official3 norm and updated using disaggregated price adjustment suggested by Minhas et al. (1988).
2. EOPL: Poverty line based on the official norm and updated using price adjustment suggested by Expert Group (1993).
3. APL: Poverty line based on the alternative norm and updated using disaggregated price adjustment suggested by Minhas et al. (1988).
4. AIOPL: All India OPL used for all the states/regions. 5. AIEOPL: All India EOPL used for all the states/regions. 6. AIAPL: All India APL used for all the states/regions.
Ray and Lancaster (2004) derived another set of age‐gender breakdown of daily normative calorie requirement corresponding to the overall per capita calorie norm of 2400 kcal/day for the average rural Indian. These have been obtained from the website www.MedIndia.net. These figures are close to the energy allowances recommended by an expert group of the Indian Council of Medical Research in 2002. The corresponding figures for the urban India can be obtained from the above figures multiplying by 2100/2400. They did not consider the activity patterns of the adult Indians that affected the calorie norms much. Ray and Lancaster (2004) gave four methods to arrive at poverty line from family budget data given the calorie and other nutrient price based expenditure norms.
1. The official poverty line4 2. The Calorie norm, i.e., 2400 Kcal per capita per day for rural India and 2100 Kcal per
capita per day for urban India. 3. Nutrient price based food expenditure norm, i.e., a balanced diet of 2738 Kcal energy
comprising of 467.53 gms. of carbohydrate, 66.6 gms of protein and 66.9 gms of fat (Gopalan, et al., 1999).
4. Nutrient price based total expenditure norm. This is derived from Nutrient price based food expenditure norm dividing by 0.7, because the Engel ratio for food for a poor household was found to be 0.7 by them.
Their proposed procedure incorporates the changes in household size, composition and other characteristics in the calculation of the household specific poverty lines and borrows the idea of Coondoo et al. (2003) in which the unit values of the major nutrients, namely, carbohydrate, protein and fat are estimated using a cross sectional household budget data set on food expenditure, total consumer expenditure, quantities of nutrient consumed and related variables.
3 1973-74 poverty lines were taken as official norm. 4 The official poverty lines for NSS rounds 43 and 50 are the ones reported by Dubey and Gangopadhyay (1998), and for round 55 are the ones used by the Planning Commission [see Government of India (2001), Poverty Estimates for 1999‐2000, Press Information Bureau, Feb. 22.] to provide the official poverty estimates for 1999‐2000.
7
Deaton and Dreze (2002) observed that the official poverty lines are implausible in many cases. The poverty lines were much higher in the urban sectors compared to the corresponding rural sectors in several states. The problem lied, in their opinion, in the use of defective price indices in adjustment of the poverty lines over time. Deaton (2003) preferred superlative indices like Fishers Ideal Index or Tornqvist Index (Weighted geometric mean of the Laspeyres and Paasches indices where the weights are the budget shares). So far as index of urban prices relative to rural prices are concerned, these two formulae give almost the same value. Deaton, however, estimated the poverty lines (state wise rural, urban for 43rd, 50th and 55th rounds) using Tornqvist form starting with the official rural all‐India poverty line for the 43rd round, 1987‐88. Sundaram and Tendulkar (2002) made an entirely independent study and concluded that a large part of the poverty decline between the 50th and 55th rounds of the NSS (36 to 26% in the all India HCR) is ‘real’ and not due to any methodological changes. So far as prices are concerned the following points should be taken into consideration for calculating poverty line.
(a) The prices over consumer goods vary over regions. In particular rural and urban prices are different. Also there are state wise variations of prices.
(b) The prices vary over different expenditure classes. This is due to two reasons: i. The commodity compositions are different for different expenditure groups. ii. Since we are taking commodity groups5 price will change according to its quality
within the group. Thus it is desirable to consider prices for the households having total expenditure less than and around the expected poverty line. In the 55th round data there are 173 separate items on food, beverage, tobacco, fuel items etc. Respondents were asked to report expenditures and quantities purchased of these items for both last 30 days and last 7 days. So there are two sets of quantities and implied prices available.
1.3 Objective of the Present Study: In this study, we propose to review the norms of 2400 Kcal for rural India and 2100 Kcal for urban India for calculation of poverty line. We shall start with the recommended dietary allowances including net energy intake in terms of Kilo Calorie per day (kcal/d or simply denoted as Kcal) for different age‐sex groups, which are already available. For example National Institute of Nutrition published dietary guidelines in 2003 (National Institute of Nutrition, 2003). It was prepared by the Working Group headed by Dr. Kamala Krishnaswamy, Director, National Institute of Nutrition (NIN). This group of NIN was aided by an Expert Advisory Group of eminent scientists of institutions like CSIR, ICMR, AIIMS, NIN etc. The different groups of Indians consist of men and women subdivided by activity patterns and infants, children, boys and girls sub‐classified by age groups. Thus it takes care of age, sex and occupational patterns – the major three determinants of nutritional requirements. These figures are almost same as those referred by Ray and Lancaster obtained from the website www.MedIndia.net. For this we need the family budget data of the latest two quinquennial surveys of NSSO along with occupations and activity patterns6 of all the adult members in the household and the latest nutrition/dietary survey data, if available. NSS employment survey data may help us in finding
5 It is difficult to do away with it. It is not possible to consider all possible types of clothing’s, say. There will be quality differences. Even for food items there are variations with respect to quality. A single commodity ‘rice’ has more than hundred types. 6 It should include information on the number of days worked, the no. of hours worked per day and the intensity of work.
8
activity pattern. It may also be necessary to have data on the percentages of pregnant and lactating women that usually exist at a given point of time. It should be stressed at the outset that we have no intention of comparing our estimates from NSSO data with those derived from National Accounts Statistics (NAS) data. However, we would like to minimize the difference between the direct and indirect estimates found from NSSO data. The National Institute of Nutrition (NIN) gives the “Recommended Dietary Allowances for Indians (Macronutrients and Minerals)” for different age‐sex‐occupation categories (NIN, 2003). “The guidelines promote the concept of nutritionally adequate diets and healthy lifestyles from the time of conception to old age”. These are for maintenance of optimal health. We consider the optimum levels suggested by NIN as minimum requirements. Minimum requirements are those quantities, when administered to one, does not adversely affect one’s health and efficiencies and if administered less than these quantities then health and efficiencies may deteriorate. This can be achieved by administering much lower levels of macronutrients than the optimal level. Moreover, due to other health reasons, many Indians do not consume the quantities prescribed by NIN. Even many of the high‐income‐group persons do not consume these optimal quantities. Should we declare them poor? Of course not. If we go by actual consumption, then we may get an unusually high proportion of poor persons. The minimum requirements are, however, not known. Unless one does a controlled experiment, it is not possible to find the minimum requirements. Only thing one can do is to take different alternative levels below the prescribed optimum level and see the results. The above discussions also lead us to the basic question, i.e., whether we should go by actual calorie consumptions or other requirements or we should go by income method. The decision to some extent has to be subjective. It is rather surprising to note that the calorie norms of 2400 Kcal and 2100 Kcal remained same for a long time. State specific age-sex-occupation distribution should be considered to get a weighted average of the calorie requirements. This should be done for rural and urban India separately. Changes have been incurred in calorie requirements of people. There are labor saving devices in agriculture, in travel/transportations, in household work etc. and perhaps same occupations, which required heavy work 30 years ago require moderate work now. Should we use our judgment to decide which occupation is heavy, which moderate and which sedentary? Moreover, do we have data on number of days worked in the year and number of days without work? Perhaps not. Thus, it appears that there is scope of work for fresh estimate of average calorie norm. The age distribution of population has also changed. NSS estimates of cereals consumption per person per 30 days have declined over time, even though absolute poverty has been declining at the same time. Is it because other food items have become relatively more important in contributing to calorie intake? But it is more likely that calorie needs have become lower. After making a scrutiny of consumption and employment schedules of 50th and 55th rounds of NSS, it appears that the consumption schedule do not give the occupation status of each member in the household. It is, however, given age‐group wise in the employment schedule. But the two schedules need not match. To get a more clear view of the problem let us consider the following formulation of the model: Assume that there are N households in a given sector (rural or urban) of a given state. NSS consumption survey gives us, besides many other information, the consumption vector (q1
(j), q2
(j), …, qk(j)) in quantity terms and (v1
(j), v2(j), …, vk
(j)) in value terms for k commodity groups of food for jth household. NIN on the other hand gives the optimal calorie, fat and protein
9
requirements for each food item. From this we can find the actual calorie consumption vector (C1
(j), C2(j), …, Ck
(j)) of the jth household. Suppose the total calorie intake of the jth household is C0(j).
Again, we have the age‐sex composition of the household. According to the calorie requirements of each category as laid down by NIN, we may define the calorie equivalent scale. In which the category with highest prescribed calorie is given weight 1. Other categories are given weights in comparison with this category. Thus we calculate the total optimal calorie consumption of the household. Suppose this is denoted as C00
(j). If C0(j) < C00
(j) then the household may be declared as poor. The direct calculation refers to the counting of poor by the above method. The indirect calculation of number of poor is through the poverty line method. There is a relation between the per capita food consumption and the per capita total consumption of the households which is known as the Engel relation for food. The Engel relation for food is found using these data for households which are supposed to be poor. These relation is known to be very stable. We use this relation to find the per capita total consumption from the per capita food consumption. The per capita food consumption for this purpose is calculated using the data on households, which just meet the optimal calorie requirements. This gives us a poverty line.
References: 1. Dandekar, V.M. and N. Rath (1971): Poverty in India, Economic and Political
Weekly. Mumbai, Vol. VI, 1 and 2. 2. Deaton, Angus and Dreze, Jean (2002): “Poverty and Inequality in India: A Re-
examination”, Economic and Political Weekly, Sept. 7. 3. Deaton, Angus (2003): “Prices and Poverty in India, 1987-2000”, EPW, Jan. 25. 4. Gangopadhyay, S., L.R. Jain and A. Dubey (1997): “Poverty Measurement and
Socio-economic Characteristics: 1987-88 and 1993-94,” Report submitted to the Dept. of Stat., Govt. of India.
5. Gopalan, C, Sastri B.V. Rama and S.C. Balasubramanian (1999): “Nutritive Value of Indian Foods”, National Institute of Nutrition, ICMR, Hyderabad.
6. Government of India (1979): “Report of the Task Force on Projections of Minimum Needs and Effective Consumption Demand”, Perspective Planning Division, Government of India, Planning Commission.
7. Government of India (1993): “Report of the Expert Group on Estimation of Proportion and Number of Poor”, Planning Commission, Perspective Planning Division.
8. ICMR (2002): Nutrient Requirements and Recommended Dietary Allowances for Indians: A Report of the Expert Group of the Indian Council of Medical Research, New Delhi.
9. Minhas, B.S., L.R. Jain, S.M. Kansal and M.R. Saluja (1988): “Measurement of General Cost of Living for Urban India, All-India and Different States,” Sarvekshana, 12, 1-23.
10. National Institute of Nutrition (2003): “Dietary Guidelines for Indians – A Manual” by National Institute of Nutrition, Indian Council of Medical Research, Hyderabad – 500 007, India, 2003.
11. Ranjan Ray and Geoffrey Lancaster (2004): On Setting the Poverty Line Based on Estimated Nutrient Prices With Application to the Socially Disadvantaged Groups in
10
India During the Reforms Period, Discussion Paper 2004-09, School of Economics, University of Tasmania.
12. Sundaram, K. and Tendulkar, S. (2002): “Recent Debates on Data Base for Measurement of Poverty in India”, Delhi School of Economics. Presented at joint GOI/World Bank poverty workshop, Delhi, Jan. 2002. Available at http://www.worldbank.org/indiapovertyworkshop.
11
Chapter 2
Existing Methods of Finding Poverty Rate
2.1 Introduction The conventional Method of finding the poverty lines is divided into few steps as follows:
(i) Each member in a household is put in the respective group according to the pre‐assigned age‐sex groups and the activity status of adult members prepared for this purpose. In addition to it the status of pregnancy and lactation of female members may also be considered. For each age‐sex‐activity status group there is also a pre‐assigned calorie requirement called calorie norm. Calorie norm of a household is determined by adding the calorie norm of each member of the household.
(ii) For a given region the proportion of population in each age‐sex‐activity status group is found. The average calorie norm of the region (may be termed as Calorie Line) is found by taking the weighted mean of the calorie norms of each category of members where weight is taken as proportional to the total population in that category. It is assumed that the poverty line of the region is a function of the calorie line. The poverty line of the region is thus based on this calorie line. The regions taken in India are the rural and urban sectors of each state. Overall calorie lines of rural and urban India are also found.
(iii) Actual Calorie consumption of the household is calculated by adding the calorie of each food item consumed by the household. This is done by using the calorie conversion factor of each item which is defined as the calorie content of one unit of the item. Naturally to find the calorie intake of the household one should have data on quantities of food items consumed by the household.
(iv) It is assumed that calorie intake or more precisely per capita calorie intake of a household is directly related with the per capita food expenditure and in turn with the per capita total expenditure of the household. In practice the two steps, i.e, finding relation between per capita calorie intake and the per capita food expenditure and then between per capita food expenditure and the per capita total expenditure are merged and only the relation between per capita calorie intake and the per capita total expenditure is found. The relation is established for different expenditure groups to make it as realistic as possible. The Poverty line is the per capita total expenditure which corresponds to the calorie norm of the concerned population. This may be done for each state separately for rural and urban regions. Since per capita Calorie intake is viewed as a function of the per capita total expenditure, the poverty line is found by inverse interpolation method.
It should be noted here that Official Poverty Line in India is not found by the conventional method. Rather it is found by projecting the poverty line from the base year poverty line to the current year poverty line using the relevant price indices. The base year poverty rates were found by a method which is similar to the conventional method. 2.2 Justification of taking Calorie Norm of a Region by the above method Calorie line can also be computed by taking average of calorie norms of all the households. If the two calorie lines coincide then there is no problem. Otherwise questions may arise about the choice of the appropriate calorie line. Fortunately it can be proved that the two calorie lines are
12
same. To prove this, assume that there are N households in the community and the ith household has i1, i2, … ik, …, iK, members in the K categories. The calorie norms of the categories are C1, C2, … Ck, …, CK, respectively. The calorie norm of the ith household can be expressed as
and the average of these calorie norms is = = = ,
which proves the proposition. 2.3 The Proposed Calorie Norms A detailed procedure on the calculation of Human Energy Requirements can be found in the important publication of the final report of the Joint FAO/WHO/UNU Expert Consultation on Human Energy Requirements, convened in October 2001 at FAO headquarters in Rome, Italy7. They estimated the human energy requirements from measures of energy expenditure plus the additional energy needed for growth, pregnancy and lactation. “Energy requirement8 is the amount of food energy needed to balance energy expenditure in order to maintain body size, body composition and a level of necessary and desirable physical activity consistent with long-term good health.” However since there are interpersonal variations, the mean level of dietary energy intake of the healthy, well-nourished individuals who constitute that group has been recommended as the energy requirement for the population group. Average energy requirements of infants from birth to 12 months, children and minors for each age in years, adults and elderly persons of each age group are given in the report. They also supply the daily energy requirements of mothers during pregnancy and lactation. Since NSSO consumption data usually do not cover information of mothers about pregnancy and lactation period, it is not possible for us to incorporate it in this report. It is also necessary to have information on the lifestyles of adults in relation to the intensity of habitual physical activity. All adults are put in one of the three categories (i) sedentary or light activity lifestyle, (ii) active or moderately active lifestyle and (iii) vigorous or vigorously active lifestyle. Total Energy Expenditure (TEE) will be different for different lifestyles. The basis of calculation in each of these groups is the basal metabolic rate (BMR). It is defined as the amount of energy used for a person while at rest during a period of time. BMR mainly depends on the age, gender, body size, and body composition of the
7 http://www.fao.org/docrep/007/y5686e/y5686e01.htm#TopOfPage. Henceforth this report will be referred to as ‘FAO report’ or ‘report of FAO’. 8 The procedure for measuring Total Energy Expenditure (TEE) is through experiments like Doubly Labeled Water technique (DLW), Heart Rate Monitoring (HRM). When experimental data on total energy expenditure are not available, factorial calculations based on the time allocated to activities can be adopted. Factorial calculations combine the energy spent on different components or activities like sleeping, resting, working etc. that are performed habitually.
13
individual. The energy spent during sedentarily, moderately or vigorously active life style will thus be more than BMR. The TEE of a person can be expressed as constant time of BMR, known as Physically Active Level (PAL). It is the TEE for 24 hours expressed as a multiple of BMR, and is calculated as TEE/BMR for 24 hours. The energy requirements of infants and children are the average values of the groups of infants and children and may be considered to represent the moderately active individuals. For adults also it may seem to be appropriate to take persons with moderately active lifestyle because they are the representative groups being in the middle of the other two groups namely with sedentary and vigorously active lifestyle. However sedentary lifestyles are increasing in most societies owing to the access to effort saving technology. We have automobiles and buses for transportation. Many time- and effort- saving devices are used in the day to day household works. In the work places there are electronic and mechanical devices to save our time and hard work. More and more efficient machines for plowing the land, building houses, construct roads etc. have been invented. Thus the proportions of persons with sedentary lifestyles are much more than what it was about 10 or 20 years back. Also, the information on the lifestyles of the members in the families are not available for any of the NSS rounds. The process of finding the activity level from the occupation group in most cases is somewhat fuzzy. Thus it is best to take all adults as sedentary. The energy requirements of infants during the first year of life are given in the report of FAO (Table 2.1). As mentioned earlier, infants need extra calorie for growth. The amount of energy deposition in the tissues during growth (Eg) is shown in column (4) and (9) respectively for boys and girls in Table 2.1. The average weight of boys and girls given in the report will not be same as in India. Indian Council of Medical Research (ICMR, 2004) has carried out a similar study in India. They have found the average weight of infants of age 0-5 months and 6-11 months (i.e., first and second six months of the first year of life) as 5.4 Kg. and 8.6 Kg. respectively. Since there were not much difference in the average weights between boys and girls, the average figures were given after combining boys and girls. A simple calculation taking average over six months of daily energy requirements per kg. of body weight (columns (6) and (11) in the Table 2.1) separately for boys and girls leads to the daily energy requirements as 93 and 92 kcal/d/kg respectively during the first six months of their lives. The corresponding figures in the next six months are 80 and 79 kcal/d/kg respectively. The difference in the daily energy requirements between boys and girls is not much and can be safely taken to be as 93 and 82 respectively. The corresponding figures given by ICMR are 108 and 98 which are much higher than these figures.
Table 2.1. Energy Requirements of Infants During the First Year of Life Boys Girls
TEE Eg Daily energy requirements
TEE Eg Daily energy requirements
Age in months Weight
Kg. Kcal/d Kcal/d Kcal/d Kcal/d/kg
WeightKg.
Kcal/d Kcal/d Kcal/d Kcal/d/kg(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) 0-1 4.58 306 211 518 113 4.35 286 178 464 107 1-2 5.50 388 183 570 104 5.14 356 161 517 101 2-3 6.28 457 139 596 95 5.82 416 134 550 94 3-4 6.94 515 53 569 82 6.41 469 68 537 84 4-5 7.48 563 45 608 81 6.92 514 57 571 83 5-6 7.93 603 36 639 81 7.35 552 47 599 82
14
6-7 8.30 636 17 653 79 7.71 584 20 604 78 7-8 8.62 664 16 680 79 8.03 612 17 629 78 8-9 8.89 688 14 702 79 8.31 637 15 652 78
9-10 9.13 710 21 731 80 8.55 658 18 676 79 10-11 9.37 731 21 752 80 8.78 679 15 694 79 11-12 9.62 753 22 775 81 9.00 698 14 712 79
* Eg: energy deposition in tissues during growth. We now turn to the requirements of children and minors of ages 1 to 17 years lbd9. In this case also we take sum of the TEE and energy deposition in tissues during growth. Observe that the amount of energy deposition is very less compared to the TEE and gradually diminishes to zero as the age of boys and girls increase.
Table 2.2. Energy Requirements of Boys and Girls at Different Age Groups: FAO
Boys Girls Weight TEE Eg* Daily energy
requirements Weight
kg TEE Eg* Daily energy
requirements
Age years
Kg Kcal/d Kcal/d Kcal/d Kcal/d/kg Kg Kcal/d Kcal/d Kcal/d Kcal/d/kg(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) 1-2 11.5 934 14 948 82.4 10.8 851 14 865 80.1 2-3 13.5 1117 11 1129 83.6 13.0 1035 12 1047 80.6 3-4 15.7 1240 12 1252 79.7 15.1 1145 11 1156 76.5 4-5 17.7 1349 11 1360 76.8 16.8 1231 10 1241 73.9 5-6 19.7 1456 11 1467 74.5 18.6 1320 10 1330 71.5 6-7 21.7 1561 12 1573 72.5 20.6 1415 13 1428 69.3 7-8 24.0 1679 14 1692 70.5 23.3 1537 17 1554 66.7 8-9 26.7 1814 16 1830 68.5 26.6 1678 21 1698 63.8
9-10 29.7 1959 19 1978 66.6 30.5 1831 23 1854 60.8 10-11 33.3 2128 22 2150 64.6 34.7 1981 25 2006 57.8 11-12 37.5 2316 25 2341 62.4 39.2 2123 25 2149 54.8 12-13 42.3 2519 29 2548 60.2 43.8 2250 26 2276 52.0 13-14 47.8 2737 33 2770 57.9 48.3 2355 24 2379 49.3 14-15 53.8 2957 33 2990 55.6 52.1 2430 19 2449 47.0 15-16 59.5 3148 30 3178 53.4 55.0 2478 12 2491 45.3 16-17 64.4 3299 24 3322 51.6 56.4 2499 5 2503 44.4 17-18 67.8 3396 15 3410 50.3 56.7 2503 0 2503 44.1
*. Eg: energy deposition in tissues during growth.
It is again a simple exercise to find the daily energy requirements for Indian children for different age groups as given by ICMR. The results are presented in Table 2.3.
Table 2.3. Energy Requirements of Boys and Girls at Different Age Groups: A Comparison Between FAO and ICMR Estimates
Boys Girls
Daily energy requirements Daily energy requirements Age groups Body weight FAO ICMR
Body weight FAO ICMR
Kg. Kcal/d/kg Kcal/d Kcal/d Kg. Kcal/d/kg Kcal/d Kcal/d (1) (2) (3) (4) (5) (6) (7) (8) (9)
9 Last birth day.
15
0-5 months 5.4 93 502 583 5.4 92 497 583 6-11 months 8.6 80 688 843 8.6 79 679 843
1-3 years 12.2 82 1000 1240 12.2 79 964 1240 4-6 years 19.0 75 1425 1690 19.0 72 1368 1690 7-9 years 26.9 69 1856 1950 26.9 64 1722 1950
10-12 years 35.4 62 2195 2190 31.5 55 1733 1970 13-15 years 47.8 56 2697 2450 46.7 47 2195 2060 16-17 years 57.1 51 2912 2640 49.9 44 2160 2060
One can compare our values with that given by ICMR (Columns (5) and (9) of the Table 2.3. ICMR values are higher for boys up to 12 years and for girls up to 15 years. After that ICMR recommendations are less than those obtained from FAO recommendations. We shall see the position of different states in India for both the cases to see whether there is any difference between the two estimates. In order to calculate daily energy requirements for adults it is necessary to consider the three groups of persons with respect to lifestyles. The Intensity of Habitual Physical Activity, or PAL values for each of these groups are given in the Table 2.4.
Table 2.4. Classification of Lifestyles in Relation to the Intensity of Habitual Physical Activity, or PAL
Category PAL value Sedentary or light activity lifestyle 1.53 Active or moderately active lifestyle 1.76 Vigorous or vigorously active lifestyle 2.25
Though the BMR values are given in the FAO report for different mean weights such as 50 kg., 55 kg., 60 kg., 65kg., etc. separately for men and women, we shall consider only the average weight of 60 kg. for men and 50 kg. for women as has been considered by ICMR. We multiply the BMR by the PAL values to get the TEE per kg. weight. This is then multiplied by the average weight to get the TEE of an individual. Physical growth, more or less, stops as soon as a person reaches adulthood. This TEE is same as the daily energy requirement since no extra energy is necessary for growth.
Table 2.5. Daily Energy Requirements for Men and Women in India Men Women
Daily energy requirements
Daily energy requirements
Mean weight
BMR
FAO ICMR
Mean weight BMR
FAO ICMR
Age group
Lifestyle
Kg. Kcal/d/kg
Kcal/d
Kcal/d/kg
Kcal/d Kg. Kcal/d/kg
Kcal/d
Kcal/d/kg
Kcal/d
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Sedentary 60 2479 41.3 - 50 1912 38.2 - Moderate 60 2851 47.5 - 50 2200 44.0 -
18-29.9
Heavy 60
27
3645 60.7 - 50
25
2812 56.2 - Sedentary 60 2387 39.8 - 50 1912 38.2 - Moderate 60 2746 45.8 - 50 2200 44.0 -
30-59.9
Heavy 60
26
3510 58.5 - 50
25
2812 56.2 - 60 or more Sedentary 60 22 2020 33.7 - 50 22 1683 33.7 -
16
Moderate 60 2323 38.7 - 50 1912 38.2 - Heavy 60
2970 49.5 - 50
2475 49.5 -
Sedentary 60 2367 39.4 2425 50 1882 37.6 1875 Moderate 60 2722 45.4 2875 50 2165 43.3 2225
18 or more
Heavy 60
25.8
3480 58.0 3800 50
24.6
2768 55.4 2925 The mean BMR values are different for different age groups as well as for different mean weights. Since the mean weights have been fixed, we need only to consider different age groups i.e., 18 to 29.9 years, 30 to 59.9 years and 60 years or more. The weighted average of the BMR values, where weights are proportional to the population sizes of the respective age groups, is computed separately for males and females and for different lifestyles. The Table 2.5 presents the daily energy requirements for 60 kg. and 50 kg. respectively for males and females along with the recommended values of ICMR. It can be seen from the table that the average daily energy requirements for men and women are very close to the values corresponding to the age groups 30 to 59.9 years. The ICMR recommendations are slightly higher in all cases except for women with sedentary lifestyle. Table 2.6 gives the summary of daily energy requirements for males and females at all age groups. In order to find the average daily calorie requirement different committees and groups were formed from time to time. The Task Force on Projection of Minimum Needs and Effective Consumption Demand set up by the Planning Commission in 1979 has used the Calorie Norms proposed by the Indian Council of Medical Research (ICMR) prior to 1979. The Expert Group, also set up by the Planning Commission, has used a different calorie norm. However all the proposed calorie norms are very close to one another. The calorie norms and the weighing diagrams are summarized in the Table 2.7. It should be mentioned here that the calorie norms given under the column heading of ICMR (2002) are the latest norms found so far. The differences in the calculated calorie lines calculated from the values given in the Table 2.7 are thus due to two reasons, (i) the differences in the calorie norms and (ii) the differences in the weighing diagram. We shall fix the weighing diagram as found from Census 2001 and the weighing of the activity status proportional to those taken by the Expert Group and calculate the Calorie lines to see whether there are differences in these values. This will give us an idea about the effect of the calorie norms on the calorie lines.
Table 2.6. Energy Requirements at Different Age Groups Separately for Males and Females: A Comparison between FAO and ICMR Estimates
Males Females Females/Males
Daily energy requirements
Daily energy requirements
Body weight
FAO ICMR Age groups
Body weight
FAO ICMR
Body weight
FAO ICMR Kg. Kcal/d Kcal/d Kg. Kcal/d Kcal/d - - -
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) 0-3y 10.5 865 1064 10.5 839 1064 1.00 0.97 1.00 4-6y 19.0 1425 1690 19.0 1368 1690 1.00 .96 1.00
7-12y 31.1 2025 2070 29.2 1727 1960 0.95 0.86 0.95 13-18y 57.1 2912 2640 49.9 2160 2060 0.87 0.74 0.78
19 or more 60.0 2367 2425 50.0 1882 1875 0.83 0.89 0.77 2.4 Calorie Poverty Rates
17
Table 2.7 gives the energy requirements at different age‐sex combination groups. We can compare between FAO, ICMR, the Task Force and the Expert Group estimates presented in the table. Using this estimates one can calculate the calorie lines (Table 2.8). In addition to the calorie lines Table 2.8 also gives us the proportion of persons below the calorie line (may be termed as Calorie Poverty Rate or simply the Calorie Poverty) by direct comparison of households’ total calorie intakes with the calorie line (Columns 9, 10 and 11) assuming that all adults are at sedentary level. Column (12) is found by combining the rural and urban proportions given in columns (9) and (10), i.e., weighted mean of columns (9) and (10) weights being proportional to the rural and urban population of 2001. Since there is not much difference between the values in columns (11) and (12) one can take any of these columns as the calorie poverty of India. There are considerable differences in the calorie lines found by different methods which is reflected in the calorie poverty. Thus the calorie norms matter much. One should use FAO and ICMR calorie norms because these are the latest calorie norms available. Calorie lines for urban India have been found to be more than those of rural India. Hence the same is true for the Calorie Poverty. When activity status is considered the calorie lines are inflated and so are the calorie poverties (Table 2.9). However, in this case the urban poverties are found to be less than rural poverties. Thus one of the reasons for differences between rural and urban poverties is the activity status and it should be considered in any analysis while calculating poverty ratios.
18
Table 2.7. Energy Requirements of age-sex combination at Different Age Groups: A Comparison between FAO, ICMR, the Task Force and the Expert Group Estimates
Daily energy requirements Daily energy requirements
Age groups FAO ICMR Task
Force Expert Group FAO ICMR Task
ForceExpert Group
Age groups taken by the Task Force
Weights taken by the Task Force of the population for 1982-83 (projected) and 27th round NSS Employment data
Age groups taken by the
Expert Group
Weights taken by the Expert Group of ICMR in 1988
Weights taken according to Census 2001
Males Females Rural Urban Rural Urban Rural Urban Year Kcal/d Kcal/d Kcal/d Kcal/d Kcal/d Kcal/d Kcal/d Kcal/d Year Male Female Male Female Year Male Female Male Female Male Female Male Female (1) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (14) (13) (15) (16) (17) (19) (18) (20) (21) (23) (22) (24)
0 year 596 713 650 700 588 713 650 700 0 1.360 1.360 1.245 1.245 0 0.92 0.92 0.80 0.80 1.76 1.71 1.29 1.28 1-3 years 1000 1240 1200 1240 964 1240 1200 1240 1‐3 4.105 4.105 3.830 3.830 1‐3 3.68 3.68 2.84 2.84 7.16 7.18 5.55 5.65 4-6 years 1425 1690 1500 1690 1368 1690 1500 1690 4‐6 3.995 3.995 3.765 3.765 4‐6 4.18 4.18 3.28 3.28 8.22 8.03 6.18 6.16 7-9 years 1856 1950 1800 1950 1722 1950 1800 1950 7‐9 3.735 3.735 3.830 3.830 7‐9 3.74 3.74 3.17 3.17 7.74 7.63 6.18 6.3
10-12 years 2195 2190 2100 2190 1733 1970 2100 1970 10‐12 3.730 3.730 3.830 3.830 10‐12 4.44 3.79 3.91 3.36 8.56 8.05 7.28 7.16 13-14 2.370 2.220 2.360 2.230 13-15 years 2697 2450 2500 2450 2195 2060 2200 2060 13-15 6.52 6.13 6.22 6.36
16-17 years 2912 2640 2400 2640 2160 2060 1900 2060 3.57 3.26 3.98 3.93 18 year 2367 2425 2800 2640 1882 1875 2200 2060
16-18 6.62 5.69 7.00 6.24
2.81 2.47 2.95 2.67 19 or more sedentary 2367 2425 2400 2425 1882 1875 1900 1875
15 or more sedentary 6.58* 18.41* 22.72* 25.27*
19 or more sedentary 7.05* 18.24* 20.71* 26.47* 13.97
† 37.05†
40.69†
56.02†
19 or more moderate 2722 2875 2800 2875 2165 2225 2200 2225 15 or more
moderate 2.13 0.93 7.72 1.40 19 or more moderate 3.32 1.26 7.05 1.24 6.58† 2.56† 13.85† 2.63†
19 or more heavy 3480 3800 3900 3800 2768 2925 3000 2925 15 or more
heavy 22.63 10.88 3.83 1.470 19 or more heavy 16.71 7.84 2.97 0.87 33.11
† 15.93† 5.83† 1.84†
19 or more total - - - - - - - - 15 or more
total 31.34 30.22 34.27 28.14 19 or more
total 27.08 27.34 30.73 28.58 53.66 55.54 60.37 60.49
* Weights of non‐workers have been added to the weights of sedentary because they have the same calorie norm †Weights are proportional to those of Expert Group
19
Table 2.8: Calorie Lines Assuming All Sedentary Using Census 2001 Populations as Weights and the Corresponding Calorie Poverty Rates by Direct Method Using Individual Multiplier: NSS61
Calorie Lines Calorie Poverty Rural India Urban India All India All India
Method Male Female Male Female Rural Urban All Rural Urban All
Rural‐Urban Combined*
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) FAO 2147 1757 2208 1792 1957 2011 1972 0.5084 0.5784 0.5153 0.5162 ICMR 2202 1825 2255 1843 2019 2060 2030 0.5596 0.6162 0.5643 0.5644
Task Force 2370 1997 2456 2036 2189 2257 2208 0.6831 0.7280 0.6956 0.6956 Expert Group 2208 1830 2262 1848 2024 2066 2036 0.5592 0.5932 0.5693 0.5687 Weight2001 0.5139 0.4861 0.5262 0.4738 0.7218 0.2782 1.0000 0.7218 0.2782 1.0000 1.0000
* Column (12) is found after combining columns (9) and (10) with weights 0.7218 and 0.2782. Table 2.9. Calorie Lines Assuming Census 2001 Populations as Weights and Activity Status Weight Proportional to Those of Expert Group and the Corresponding Poverty Ratios by Conventional Method Using Individual Multiplier: NSS61
Rural India Urban India All India All India
Male Female Male Female Rural Urban All Rural Urban All Rural‐Urban Combined*
Method
Calorie poverty lines Calorie poverty rates (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) FAO 2539 1905 2322 1816 2231 2082 2190 .7066 .6255 .6838 .6840 ICMR 2687 2002 2398 1872 2354 2149 2297 .7758 .6721 .7507 .7470
Task Force 2678 2013 2357 1883 2355 2132 2293 .7762 .6614 .7482 .7443 Expert Group 2693 2006 2404 1877 2360 2154 2302 .7792 .6747 .7535 .7501 Weight2001 0.5139 0.4861 0.5262 0.4738 0.7218 0.2782 1.0000 0.7218 0.2782 1.0000 1.0000
* Column (12) is found after combining columns (9) and (10) with weights 0.7218 and 0.2782. 2.5 Direct Calculation of Calorie Poverty Rates The direct calculation of calorie poverty rates refers to aggregating the calorie status of each household by comparing the actual calorie consumption with the specific calorie norm of the household10. It is first seen whether the household calorie consumption is below the specific household calorie norm implied by the age‐sex composition of the household. If it is so then all the members in the household are declared as poor. The ratio of weighted aggregate of number of poor persons to the weighted aggregate of total number of persons gives the calorie poverty rate. In practice we define a dummy variable 1 if a person is poor and 0 if the person is non‐poor and compute the weighted mean of the dummy variable. This is given in Table 2.10.
10 The specific calorie norm of the household is found by taking the following sum m0003*1064+m0406*1690+m0712*2070+m1318*2640+m19ps*2425+m19pm*2875+m19ph*3800+ f0003*1064+f0406*1690+f0712*1960+f1318*2060+f19ps*1875+f19pm*2225+f19ph*2925 for ICMR and m0003*865+m0406*1425+m0712*2025+m1318*2912+m19ps*2367+m19pm*2722+m19ph*3480+ f0003*839+f0406*1368+f0712*1727+f1318*2160+f19ps*1882+f19pm*2165+f19ph*2768 for FAO, where m and f values are the corresponding numbers of members in the specific age groups.
20
The estimated poverty ratios can be compared with the direct calorie poverty ratios to get an idea about the discrepancy of the household behavior so far as the consumption of commodities are concerned. Table 2.10. Calorie Poverty Rates by the Direct Calculation and the Fixed Calorie Line Method (Conventional Method) Using Individual Multiplier and 2001 Census Weighing Diagram
Without using activity status Using activity status
Direct Method Fixed Calorie Line Method SECT Group
NSS 43rd Round11
NSS 50th Round
NSS 55th Round
NSS 61st Round
NSS 61st Round
NSS 61st Round
Year 1987‐88 1993‐94 1999‐2000
2004‐05 2004‐05 2004‐05
(1) (2) (3) (4) (5) (6) (7) (8) ICMR 0.40 0.47 0.50 0.56 0.82 0.78 FAO 0.36 0.42 0.45 0.51 0.76 0.71
Task Force ‐ ‐ ‐ ‐ ‐ 0.78 Rural
Expert Group ‐ ‐ ‐ ‐ ‐ 0.78 ICMR 0.50 0.57 0.55 0.62 0.76 0.67 FAO 0.46 0.53 0.51 0.58 0.71 0.63
Task Force ‐ ‐ ‐ ‐ ‐ 0.66 Urban
Expert Group ‐ ‐ ‐ ‐ ‐ 0.67 The Table 2.10 clearly shows that the calorie poverty rates by direct method are always higher than those by fixed calorie line method (compare columns (7) and (8)). The variation of calorie consumption much below the calorie line will not affect the calorie poverty rate. So is the variation of calorie consumption much above the calorie line. Only marginal households, i.e., the households with actual calorie consumption close to the calorie line, will affect the calorie poverty rate. The number of marginal households is more in the direct methods then in the fixed calorie line methods. Thus the poverty rate by the direct method is more sensitive to changes in the consumption. Moreover, it is less likely for the marginal households to increase the consumption, whereas those who are supposed to lie above the calorie line may have reasons to consume less. The net effect is the increase of calorie poverty rate by the direct method compared to the fixed calorie line method. Also, calculation of poverty rates by the fixed calorie line method is easier because in this method the per capita calorie consumption of a household is compared with a fixed calorie line and one does not have to compute the poverty line of each household. On the other hand, the calorie line of the household may be very much different from the fixed calorie line because the age‐sex‐activity status of the household may be much different from the average age‐sex‐activity pattern of all the households. The direct method thus seems to be superior to the fixed calorie line method in this respect. Urban poverties are found to be more when activity levels of adults are not considered. This does not seem to be possible. Activity status should be considered. There are mainly two reasons for differences in the poverty rates between rural and urban sectors. The first is the differences of consumptions due to the differences of incomes. The MPCE of urban households is certainly more than the MPCE of rural households and it is expected that the households with 11 In the NSSO 43rd round data multiplier was missing for 1054 households and multiplier was ‘0’ for 4 households.
21
more income will consume more food. But our findings nullify it. The second reason is the differences of consumptions due to differences in the activity status. Our findings support it. One can further investigate whether less calorie consumption in the urban areas are due to price differences. This is discussed in the latter chapters. One can also observe from Table 2.10 that the direct calorie poverty has systematically increased over the time. This is incomprehensible. One would expect just the opposite to happen. One of the reasons is that the activity status is not considered in calculating the poverty line and also a fixed weighing diagram corresponding to the 2001 population census has been used throughout. To take care of this deficiency, we recalculate the poverty norms for each age‐sex‐activity group by taking the same set of groups for all the periods and using the available weighing diagram nearest to the time point. Task Force calorie norm is weighted by the population (projected) of 1982‐83 and 27th round NSS Employment data, The Expert Group norms are weighted by the same weights taken by the Expert Group. The FAO and the ICMR norms have weights proportional to the population of 2001 Census for age groups up to 18 years and proportional to the weights as taken by the Expert Group for age groups above 18 years. The norms have been approximated to the nearest multiples of 5. The final calorie norms, thus obtained, are given in the Table 2.11. Table 2.11. Recalculated Age‐Sex‐Activity Status Wise Calorie Norms for the Same Set of Age‐Sex‐Activity Status Groups
Norm → Task Force (TF) Expert Group (EG) FAO ICMR Age-Sex-Activity Status
Groups Male Female Male Female Male Female Male Female
0-3 years 1060 1060 1120 1120 865 840 1065 1065 4-6 years 1500 1500 1690 1690 1425 1370 1690 1690
7-12 years 1950 1950 2080 1960 2025 1725 2070 1960 13-18 years 2580 2100 2550 2060 2910 2165 2640 2060
19 or more sedentary 2400 1900 2425 1875 2365 1880 2425 1875 19 or more moderate 2800 2200 2875 2225 2720 2165 2875 2225
19 or more heavy 3900 3000 3800 2925 3480 2770 3800 2925 Values are approximated to the nearest multiples of 5. Observe that calculation of calorie poverty rates by direct method does not need any weighing diagram of the population. This is automatically taken care of by the multiplier of each member (may be termed as individual multiplier), which is the product of the household multiplier and the household size. We now calculate the Calorie Poverty Rates of different NSS rounds taking these norms by direct method12. 12 The specific calorie norm of the household is found by taking the following sum for calculation of Calorie Poverty Rates by the direct method: Task Force: m0003*1060+m0406*1500+m0712*1950+m1318*2580+m19ps*2400+m19pm*2800+m19ph*3900+ f0003*1060+f0406*1500+f0712*1950+f1318*2100+f19ps*1900+f19pm*2200+f19ph*3000
Expert Group: _m0003*1120+m0406*1690+m0712*2080+m1318*2550+m19ps*2425+m19pm*2875+m19ph* 3800+f0003*1120+f0406*1690+f0712*1960+f1318*2060+f19ps*1875+f19pm*2225+f19ph*2925
FAO:
22
Table 2.12. Calorie Poverty Rates of Different Rounds of NSS by the Direct Method using Activity Status and Individual Multiplier with Calorie Norms as in Table 2.12
43th Round 50th Round 55th Round 61st Round Sector TF EG ICMR FAO TF EG ICMR FAO TF EG ICMR FAO TF EG ICMR FAORural 0.67 0.68 0.68 0.60 0.74 0.75 0.75 0.68 0.76 0.77 0.77 0.70 0.81 0.82 0.82 0.76Urban 0.57 0.60 0.60 0.54 0.63 0.65 0.65 0.60 0.61 0.63 0.63 0.59 0.74 0.76 0.76 0.71All 0.65 0.66 0.66 0.59 0.72 0.73 0,73 0.66 0.72 0.73 0.73 0.67 0.79 0.80 0.80 0.75
The problem that the calculated calorie poverty rates increase over time, is not solved yet. To answer this question, one has to look into the details of the activity status found from National Classification of Occupation made in 1968 (NCO 1968) standards. The correspondence between NCO 1968 and the activity status has undergone a sea change. The life styles have changed much due to the introduction of many work and time saving devices. Many new commodities have come into the market. The tests and preferences on the commodities by the people have changed. The workers who were designated as hard workers have ceased to be so. So are the moderate workers. And this is reflected in the trend of Calorie Poverty Rates shown in Table 2.12. Calorie poverty rates show an increasing trend whichever method is used except for urban sector during 50th and 55th rounds of NSS. Table 2.12 also clearly shows that there is not much difference among the three Calorie Poverty Rates found by direct methods, namely the Task Force, Expert Group and the ICMR Calorie norms. This is because all these norms were based on the then ICMR calorie Norms. Only FAO estimates differ much from the rest. It thus suffices to compare only ICMR and FAO estimates. 2.5 The Poverty Lines and the Poverty Rates To get the poverty line we now see the correspondence between the Daily Per Capita Calorie Intake (DPCI) and the (Monthly) Per Capita Consumption Expenditure (MPCE) of Households. For fixed Calorie Line method this process is rather simple. One can make different groups of households according to the MPCE and find the relation between DPCI and MPCE and get the appropriate value of MPCE corresponding to the average calorie norm derived from the individual calorie norms and the weighing diagrams as set from time to time. Using NSS 61st round data we regress DPCI on MPCE for each expenditure class separately for rural and urban sector and the regression coefficients are given in the Table 2.13. Table 2.13. Results of Linear Regression of DPCI on MPCE by Expenditure Class Separately for Rural Urban sector Using Individual Multiplier: All India, NSS 61st Round Data
SECT EXPGRP Intercept(a)
Coefficient of MPCE (b)
0–235 256.9 5.610 235–270 729.2 3.331 270–320 947.0 2.468
Rural
320–365 1283.8 1.506
m0003*865+m0406*1425+m0712*2025+m1318*2910+m19ps*2365+m19pm*2720+m19ph*3480+ f0003* 840+f0406*1370+f0712*1725+f1318*2165+f19ps*1880+f19pm*2165+f19ph*2770
ICMR: m0003*1065+m0406*1690+m0712*2070+m1318*2640+m19ps*2425+m19pm*2875+m19ph*3800+ f0003 *1065+f0406*1690+f0712*1960+f1318*2060+f19ps*1875+f19pm*2225+f19ph*2925
23
365–410 907.8 2.521 410–455 1314.9 1.499 455–510 1213.0 1.721 510–580 1280.0 1.617 580–690 2480.7 ‐0.303 690–890 1773.3 0.784 890–1155 1573.1 0.995
1155+ 2518.3 0.255 0–335 565.5 3.030
335–395 1496.9 0.300 395–485 1422.0 0.603 485–580 1038.8 1.488 580–675 1300.6 0.888 675–790 1643.0 0.411 790–930 2007.5 0.0183 930–1100 1554.5 0.549
1100–1380 1150.2 0.863 1380–1880 1703.8 0.399 1880–2540 2141.4 0.188
Urban
2540+ 2760.5 0.0186
It can be seen from Table 2.13 that the coefficients are very much erratic. It calls for scrutiny of data. We deleted all the households with DPCI ≤ 100 Kcal or ≥ 10000 Kcal and recalculated the regression coefficients which are given in Table 2.14.
24
Table 2.14. Results of Linear Regression of DPCI on MPCE by Expenditure Class Separately for Rural Urban sector Using Individual Multiplier: All India, NSS 61st Round Truncated Data With 100 Kcal < DPCI < 10000 Kcal)
Sector
Expenditure Group
Intercept (a)
Coefficient of MPCE (b)
0–235 473.2 4.578 235–270 747.0 3.264 270–320 947.7 2.468 320–365 1099.8 2.008 365–410 923.2 2.480 410–455 1499.4 1.058 455–510 1187.7 1.766 510–580 1496.7 1.176 580–690 1510.1 1.161 690–890 1810.0 0.721 890–1155 1633.8 0.926
Rural
1155+ 2707.4 0.09544 0–335 780.9 2.304
335–395 1638.0 ‐0.07299 395–485 1433.5 0.575 485–580 1114.5 1.270 580–675 1351.1 0.806 675–790 1601.8 0.465 790–930 1487.6 0.606 930–1100 1563.0 0.535
1100–1380 1216.8 0.798 1380–1880 1723.3 0.376 1880–2540 1528.6 0.437
Urban
2540+ 2625.7 0.01448 The predicted values of DPCI for lower and upper boundary points of each expenditure class separately for rural and urban sector for the truncated data are given in the Table 2.15. This will help us in identifying the exact class interval containing the poverty line. In the ideal case the predicted values of DPCI at the upper boundary values of each expenditure class should be equal to that at the lower boundary values of the next expenditure class. Not surprisingly, in our case it is not so. This will not however create much problem in the estimation, because the corresponding predicted values do not differ much for the concerned intervals.
25
Table 2.15. Predicted Values of DPCI at the Boundary points and the Mean Values of Each Expenditure Class Interval Using the Results of Linear Regression of DPCI on MPCE Separately for Rural Urban sector Using Individual Multiplier: All India, NSS 61st Round Truncated Data With 100 Kcal < DPCI < 10000 Kcal)
Coefficients Predicted DPCI Sector Expenditure
Groups Mean MPCE
Intercept
(a)
Coefficient of MPCE
(b)
At Lower Boundary
At Upper Boundary
At Mean
0–235 199.5 473.226 4.578 473.2 1549.1 1386.5 235–270 253.8 746.961 3.264 1514.0 1628.2 1575.4 270–320 296.6 947.699 2.468 1614.1 1737.5 1679.7 320–365 342.4 1099.837 2.008 1742.4 1832.8 1787.4 365–410 387.7 923.156 2.480 1828.4 1940.0 1884.7 410–455 432.1 1499.399 1.058 1933.2 1980.8 1956.6 455–510 481.6 1187.717 1.766 1991.2 2088.4 2038.2 510–580 543.3 1496.662 1.176 2096.4 2178.7 2135.6 580–690 630.4 1510.056 1.161 2183.4 2311.1 2242.0 690–890 775.0 1810.039 0.721 2307.5 2451.7 2368.8 890–1155 999.9 1633.836 0.926 2458.0 2703.4 2559.7
Rural
1155+ 1956.6 2707.436 0.09544 2817.6 NA 2894.2 0–335 279.7 780.930 2.304 780.9 1552.8 1425.4
335–395 368.2 1637.980 ‐0.07299 1613.5 1609.1 1611.1 395–485 441.3 1433.481 0.575 1660.6 1712.4 1687.2 485–580 533.2 1114.463 1.270 1730.4 1851.1 1791.6 580–675 625.8 1351.107 0.806 1818.6 1895.2 1855.5 675–790 730.2 1601.801 0.465 1915.7 1969.2 1941.3 790–930 858.0 1487.592 0.606 1966.3 2051.2 2007.5 930–1100 1014.3 1562.988 0.535 2060.5 2151.5 2105.6
1100–1380 1226.4 1216.833 0.798 2094.6 2318.1 2195.5 1380–1880 1594.4 1723.298 0.376 2242.2 2430.2 2322.8 1880–2540 2157.2 1528.592 0.437 2350.2 2638.6 2471.3
Urban
2540+ 4235.6 2625.684 0.01448 2662.5 NA 2687.0 We have also found the quadratic regression of DPCI on MPCE hoping that it would give a better estimate. Even with truncated data the quadratic relation gives somewhat erratic values of the coefficients. But so far as prediction is concerned it will not pose much problem. This time all the households with DPCI ≤ 500 Kcal or ≥ 5000 Kcal are deleted for a better result. The regression coefficients and the Predicted MPCEs are given in Table 2.16.
26
Table 2.16. Predicted Values of DPCI at the Boundary points and the Mean Values of Each Expenditure Class Interval Using the Results of Quadratic Regression of DPCI on MPCE Separately for Rural Urban sector Using Individual Multiplier: All India, NSS 61st Round Truncated Data With 500 Kcal < DPCI < 5000 Kcal)
Coefficients Predicted MPCE Sector
Expend. Group
Mean MPCE
Intercept
(a)
Coefficient of MPCE
(b)
Coefficient of MPCE2
(c)
At Lower Boundary
At Upper Boundary
At Mean
0–235 199.5 331.0 6.933 ‐0.007936 331.0 1522.0 1398.3 235–270 253.8 4182.8 ‐23.759 0.05310 1531.9 1638.8 1573.2 270–320 296.6 ‐5073.3 43.109 ‐0.06841 1579.0 1716.4 1694.7 320–365 342.4 ‐112.0 9.416 ‐0.01129 1745.0 1820.7 1788.4 365–410 387.7 ‐11952.8 69.074 ‐0.08603 1797.8 1905.9 1895.9 410–455 432.1 7117.4 ‐25.011 0.03021 1941.2 1991.6 1950.7 455–510 481.6 ‐5668.7 30.183 ‐0.02942 1973.9 2072.5 2043.8 510–580 543.3 4776.7 ‐10.956 0.01119 2099.6 2186.5 2127.3 580–690 630.4 1197.7 2.098 ‐0.0007131 2174.7 2305.8 2236.9 690–890 775.0 2780.7 ‐1.811 0.001630 2307.1 2460.0 2356.2 890–1155 999.9 590.0 3.008 ‐0.001052 2433.9 2660.9 2545.9
Rural
1155+ 1956.6 2695.3 0.03978 ‐0.000000571 2740.5 NA 2771.0 0–335 279.7 576.2 4.007 ‐0.003331 576.2 1544.7 1436.3
335–395 368.2 2133.8 ‐2.960 0.004234 1617.4 1625.2 1618.0 395–485 441.3 ‐1932.7 15.981 ‐0.01756 1640.0 1687.5 1699.9 485–580 533.2 ‐429.4 7.180 ‐0.005646 1724.8 1835.7 1793.8 580–675 625.8 6058.2 ‐14.251 0.01202 1836.1 1915.4 1847.2 675–790 730.2 16116.2 ‐39.319 0.02720 1968.8 2029.7 1908.3 790–930 858.0 6957.6 ‐12.267 0.007551 1979.2 2080.1 1991.3 930–1100 1014.3 ‐437.4 4.060 ‐0.001545 2002.1 2159.2 2091.2 1100–1380 1226.4 3778.8 ‐3.363 0.001682 2114.7 2341.0 2184.2 1380–1880 1594.4 416.3 2.095 ‐0.0005641 2233.1 2361.1 2322.5 1880–2540 2157.2 3735.8 ‐1.607 0.0004674 2366.6 2669.5 2444.2
Urban
2540+ 4235.6 2559.3 0.01657 ‐0.000000102 2600.7 NA 2627.6
From the given Calorie Lines we have found the Poverty Rates using both linear and quadratic methods. The linear and quadratic methods almost give same results. The two results would have been more close if we take the length of each class interval less than what has been taken here or equivalently increase the number of expenditure classes.
27
Table 2.17. Poverty Lines and Poverty Rates by Fixed Calorie Line Method using Linear as well as Quadratic Regression for Interpolation of Poverty Lines Separately for Rural Urban sectors Using Individual Multiplier: All India, NSS 61st Round Truncated Data with 100 Kcal < DPCI < 10000 Kcal for Linear Interpolation and with 500 Kcal < DPCI < 5000 Kcal for Quadratic Interpolation)
Sector Method Used Calorie and Poverty Lines and Rates
FAO ICMR TF EG
Calorie Line 2231 2354 2355 2360 Conventional Calorie Poverty Rate
(Fixed Calorie Line) 0.71 0.78 0.78 0.78
Direct Calorie Poverty Rate 0.76 0.82 0.81 0.82 Poverty Line 621.0 754.5 755.8 762.8 Linear Poverty Rate 0.75 0.85 0.85 0.85 Poverty Line 625.5 772.0 773.3 780.2
Rural
Quadratic Poverty Rate 0.75 0.85 0.85 0.86 Calorie Line 2082 2149 2132 2154
Conventional Calorie Poverty Rate (Fixed Calorie Line)
0.63 0.67 0.66 0.67
Direct Calorie Poverty Rate 0.71 0.76 0.74 0.76 Poverty Line 980.9 1095.3 1063.6 1104.7 Linear Poverty Rate 0.63 0.70 0.68 0.70 Poverty Line 1004.6 1085.1 1062.2 1092.3
Urban
Quadratic Poverty Rate 0.64 0.69 0.68 0.69
Poverty Rates are still too high to be accepted. The Poverty Rates found by the Fixed Poverty Line Method using linear interpolation are higher than the Calorie Poverty Rates found by Calorie Line Methods found earlier in this report for rural India. Almost the opposite is the case for urban India. This anyway is not a solace to us given the fact that all the rates are too high to be acceptable.
28
Chapter 3
New Methods of Finding Poverty Rate: Decomposition of Total Calorie Consumption over Members in the Households
3.1 Introduction In this chapter we shall see whether the calorie norms prescribed by FAO differ from the calorie norms prescribed by ICMR with special emphasis to average consumption of the male and female members in the households and hence we shall try to find the Poverty Rates. For this we shall first introduce a novel methodology of finding within house distribution of calorie intakes in terms of averages. Information on the mean calorie intake for each age-sex composition of members in the households of a given community is often needed in many of the health and nutrition studies. Most of the existing measures of poverty, viz., Head Count Ratio (HCR), Poverty Gap Index (PGI), Squared Poverty Gap Index (SPGI), Sen’s Index of Poverty (SPI), Foster-Greer-Thorbecke Index (FGTPI) etc., use poverty line. The calculation of poverty line in turn needs data on the calorie intakes of members of households considering the age-sex composition of households. In particular we may be interested in knowing whether there is any difference in the consumption of food, which is usually summarized through the mean calorie consumption, of male and female members in the households in a given community. Calorie consumption should be different for different types of members in the household. Average energy requirements13 of infants from birth to 12 months, children and minors of each age in years, adults and elderly persons of each age group separately for males and female members are given in the final report of the Joint FAO/WHO/UNU Expert Consultation on Human Energy Requirements, convened in October 2001 at FAO headquarters in Rome, Italy14. They estimated the human energy requirements from measures of energy expenditure plus the additional energy needed for growth, pregnancy and lactation. It is also necessary to have information on the lifestyles of adults in relation to the intensity of habitual physical activity. All adults are put in one of the three categories (i) sedentary or light activity lifestyle, (ii) active or moderately active lifestyle and (iii) vigorous or vigorously active lifestyle. Total Energy Expenditure (TEE)15 will be different for different lifestyles. National Sample Survey Organization (NSSO) collects data on consumption of commodities at household level at regular time intervals called ‘rounds’. A round corresponds to a year and
13 “Energy requirement is the amount of food energy needed to balance energy expenditure in order to maintain body size, body composition and a level of necessary and desirable physical activity consistent with long-term good health.” However since there are interpersonal variations, the mean level of dietary energy intake of the healthy, well-nourished individuals who constitute that group has been recommended as the energy requirement for the population group. 14 http://www.fao.org/docrep/007/y5686e/y5686e01.htm#TopOfPage. Henceforth this report will be referred to as ‘FAO report’ or ‘report of FAO’. 15 The procedure for measuring Total Energy Expenditure (TEE) is through experiments like Doubly Labeled Water technique (DLW), Heart Rate Monitoring (HRM). When experimental data on total energy expenditure are not available, factorial calculations based on the time allocated to activities can be adopted. Factorial calculations combine the energy spent on different components or factors like sleeping, resting, working etc. that are performed habitually.
29
hence collection of data is uniformly spread over the year to eliminate seasonal variation. Moreover, the sample size is increased substantially at about every five year interval of time so that the estimates can be obtained relatively more reliably and at more disaggregate levels. These are known as ‘quinquennial rounds’. The latest quinquennial round data on consumption is the 61st round data, collected during July 2004 – June 2005. However, it is not apparently clear how one can estimate consumption pattern among the members of the households when the data on consumption are available only at household level. Within house distribution of resources especially to see the gender bias within the household has been addressed by many authors like Deaton (1989), Pitt et al. (1990), Ahmad and Morduch (1993), Haddad and Reardon (1993), Haddad et al (1996), Udry (1997), Bhalotra and Attfield (1998)) and Duflo (2005). The issue of gender bias in the intra household allocation of resources could explain many phenomena including the phenomenon of missing women in South Asia (Dreze and Sen, 1989). Deaton (1989, 1997) argues that for a given level of income, families with
children will spend less on adult goods in order to purchase children's goods. If household purchase favors boys over girls, smaller expenditures on adult goods would be made by families
with boys as compared with those with girls. This procedure for detecting gender bias was applied by him to data from Côte d'lvoire and Thailand. The data show no evidence of inequality between boys and girls in Côte d'lvoire, and a small and statistically insignificant bias in favor of boys in Thailand. Deaton (1997) also observed the changes in the household expenditure on a particular good due to changes with the gender composition of the household. In particular change can be seen due to addition of a boy or a girl member in the household of a given composition. The difference in the average changes will give the gender bias. Lancaster et al. used and extended the collective household model (Bourguignon, et. al., 1993, Browning and Chiappori, 1998) to find the gender bias within household. It was derived from an economic model in which utilities of male and female members were maximized under some constraints. Song (2008) took wife's education relative to her husband's as a measure of bargaining power. This was calculated taking the ratio of wife's education in years to the sum of the years both husband and wife received. The paper then examines three hypotheses. One of the hypotheses was “whether households allocate fewer resources to daughters than to sons”. The hypothesis was not rejected for health expenditures. Chesher (1997), Bidani and Ravallion (1997), and Deaton and Paxson (2000) and others have proposed schemes for breaking up an aggregate expenditure into its household members. Mason et al. (1999) offered a straightforward way of disaggregating household expenditure into age‐sex composition of the household by taking linear regression with no constant. But there is a danger of not taking the intercept term. The intercept term absorbs expenditures not reported or not taken into account because of various reasons. Ray and Lancaster (2004) derived a set of age‐gender breakdown of daily normative calorie requirement corresponding to the overall per capita calorie norm of 2400 kcal/day for the average rural Indian. These have been obtained from the website www.MedIndia.net. These figures are close to the energy allowances recommended by an expert group of the Indian Council of Medical Research (ICMR, 2002). Their proposed procedure incorporates the changes in household size, composition and other characteristics in the calculation of the household specific poverty lines and borrows the idea of Coondoo et al. (2003) in which the unit values of the major nutrients, namely, carbohydrate, protein and fat are estimated using a cross sectional household budget data set on food expenditure, total consumer expenditure, quantities of nutrient consumed and related variables. The National Institute of Nutrition (NIN) gives the “Recommended Dietary Allowances for Indians (Macronutrients and Minerals)” for different
30
age‐sex‐occupation categories (NIN, 2003). “The guidelines promote the concept of nutritionally adequate diets and healthy lifestyles from the time of conception to old age”. Following the path of Mason et al. (1999) and Coondoo et al. (2003) decomposition of total calorie consumption among the members of the households is possible to some extent if we modify their model to suite the requirements in case of calorie consumption. In particular, if we are interested in the estimation of a certain aspects of consumption at the aggregate level, say mean calorie consumption of each of the different groups of members in the households, taking all households into consideration, then it is possible to estimate the same after some modifications which leads to Generalized Linear Regression Model (GLRM). In the next section we shall describe how the problem can be visualized through a model which leads to a special case of GLRM and discuss the estimation procedure of the associated parameters of the model for the specific case. The technique will then be applied to the 61st round NSSO data on consumption to see whether mean calorie consumptions vary among male and female members of the households.
3.2 The Model and the Methodology Suppose there are altogether K possible categories of members in a household h with number of members xh1, xh2, …, xhK, respectively. The total daily calorie consumption of the household is yh. Since the total calorie consumption is the sum of individual calorie consumption we have the following identity.
yh = ch1 xh1 + ch2 xh2 + … + chK xhK, … (3.01) where ch1, ch2, …, chK are the actual per head calorie consumptions of the respective categories. In general ch1, ch2, …, chK will vary from household to household. If the mean consumptions are β1, β2, …, βK, respectively, then we can write
chk = βk + uhk, for k = 1, 2, 3, …, K, … (3.02) where uh1, uh2, …, uhK are the deviations of the actual calorie consumption from the respective mean values. The deviations uh1, uh2, …, uhK, have zero means. We also assume that this deviation for a single person in that category have same variance for all the households and is denoted by νk
2, k = 1, 2, …, K. We can thus rewrite the above identity as y = (β1 + uh1) xh1 + (β2 + uh2) xh2 + … + (βk + uhk) xhK
= β1 xh1 + β2 xh2 + … + βk xhK + (uh1 xh1 + uh2 xh2 + … + uhK xhK) = β1 xh1 + β2 xh2 + … + βk xhK + ∑k uhk xhk = β1 xh1 + β2 xh2 + … + βk xhK + εh, … (3.03) where εh = ∑k uhk xhk. Observe that uhk and εh are random variables. Since the expectation of uhk, i.e., E(uhk) = 0 for all h and k, we have E(εh) = 0 for all h. We assume that the errors are independent over households. I.e., the errors in one household are independent of the errors in the other household. We assume that the dispersion matrix of uh1, uh2, …, uhK is Φh. The variance of εh, V(εh), is thus V(εh) = σh
2 (say) = xh´ Φh xh, … (3.04) where xh is the household composition vector and Φh is
Φh = , … (3.05)
where , i≠j, is the covariance between uhi and uhj and is the variance of uhi also denoted by
νi2.
The equation (2.03) can now be written as y = xh´β + εh. … (3.06)
31
The above equation is valid so far as the assumption that the sum of individual calorie consumption is same as that of total calorie consumption which is implied by the total food consumption of the household. But in real life the total food expenditure on food includes expenditure on food offered to guests, servants, cattles and other visitors of the house. Total calorie consumption of the household implied by the food consumption will then be greater than the sum of calorie consumption of the individual members. Also some members of the household may avail food outside the house. Thus the inequality may be other way round also. It is necessary to introduce one more component in the equation to accommodate it. So far we didn’t include any intercept term in the above equation. This deviation can be accommodated in the intercept column of the regression, if introduced. We thus redefine equation (2.06) by redefining xh as xh´ = (xh0, xh1, xh2, …, xhK), … (3.07) where xh0 = 1 for all h. The coefficient of this is β0, which is the expected value of the deviation. The new β vector is β´ = (β0, β1, …, βK). εh, Φh, σh
2 etc. are suitably reformulated as σh
2 = xh´ Φh xh, … (3.08) where Φh is
Φh = , … (3.09)
where , i≠j, is the covariance between uhi and uhj and is the variance of uhi also denoted by
νi2. Specifically ν0
2 can be regarded as the variance of the equation error uh0. With this formulation we can now write all the equations in a compact form as y = X´β + ε, … (3.10) where y = (y1, y2, …, yH)´, ε = (ε 1, ε 2, …, ε H)´, and
X = = (x.0, x.1, x.K) = , … (3.11)
where H is the total number of units (households) taken for regression. The dispersion matrix of ε is Ω = diag(σ1
2, σ22, …, σH
2). … (3.12) We may now discuss how one should interpret the estimates of the regression coefficients. Each element in β gives the expected amount of calorie consumption for a member in the respective category. This may also be interpreted as the increase in the average amount of calorie consumption due to increase by one person in the respective category. Usually in regression analysis one can give both the interpretations of the estimated coefficient if the intercept term is not significant. Interestingly, in this case, both the interpretations are plausible even if the intercept term is significant i.e., the sum of individual average calories is not the total calorie on the average. If the intercept term is positive, it means that there is extra consumption, possibly by guests or servants for many of the households which outweighed the consumption of food by the members of the households outside the house. If it is negative the interpretation is the other way round. The variable associated with the intercept term always takes value 1. Thus it may be interpreted as a ‘ghost’ member in the household which may consume or produce extra calories for consumption of other members.16
16 One of the reasons behind the discrepancy between poverty ratios calculated through calorie‐intake and through per capita income/total expenditure is the existence of the intercept term in the regression.
32
The Generalized Least Squares (GLS) estimate of β is = (X´ Ω‐1X)‐1(X´ Ω‐1y). … (3.13)
It is easy to verify that expectation of , E( ), is β.
E( ) = β. … (3.14)
However Ω is not known. It is a diagonal matrix with σh
2 (= xh´ Φ xh) as its hth element.
σh2 = (xh0, xh1, …, xhK) . … (3.15)
To estimate the values we first get the usual regression (weighted least squares using weights
as multipliers) estimate of β as . Use this to get the residuals . Regress 2 on (1, 2xh1, 2xh2, …,
xh12, 2xh1xh2, …). The regression coefficients will give us the estimates of the distinct elements of ’s, namely ( , , …, , , …, , …, ). This can be used as GLS estimate of β. The
process can be repeated until convergence up to desired level of precision. The non‐negativity of the estimated value of σh
2 for each h is not guaranteed because the estimated value of Φ may not be nonnegative definite. It was in fact found to be so with our consumption data of 61st round of NSSO. Deleting the data associated with negative Φ did not help. There were further some negative Φs and the process did not end quickly. As a first approximation we have taken Φ to be diagonal with only φ11, φ22, …, φKK. Thus we regressed
2 on xh1
2, xh22, …, xhK
2. All the coefficients of xh12, xh2
2, …, xhK2, this time, were positive for most of
the subsets of data considered for our analysis. For the few cases where all the coefficients were not found to be positive, we deleted a very few observations to achieve the desired result. The subsequent tables present the results of the analysis applied to our proposed model using NSS 61st round data. NSS 61st round data has 124644 households. After scrutiny and elimination of outlying observations it reduced to 124362 households. The following gives the summary of the data before and after scrutiny.
Table 3.1: Sample Sizes: Before and After Scrutiny Before Scrutiny After Scrutiny Rural 79298 79170 Urban 45346 45192 Total 124644 124362
The number of observations deleted is not much compared to the total sample size.
The NSS data provides multiplier to each household. The multiplier is calculated from the sampling scheme adopted by NSSO. This can be used as weights to get more accurate estimates. We have used SPSS 11 and SPLUS 2000 for our calculations. 3.4 The Results The coefficients of the regression model, where the regressors are the numbers of members in each of the age‐sex groups belonging to the households, along with the intercepts terms are presented in Table 3.2. In most of the cases coefficients associated with the male members have higher values than those of the female members. Three types of methods of estimation have been followed. The first set of estimated values refers to the estimation method where no If the income based poverty index could be adjusted by this intercept term the two indices would become closer.
33
weights or multipliers have been used. The second method used only the weights, known as multipliers, which arise due to given sampling scheme adopted by the NSSO. The third method used weights arising not only due to the specific sampling scheme but also due to the specific model. It can be seen that the estimates varied to some extent. The coefficients, i.e., mean consumption of calories, decrease for most especially among members in the lower age groups. However, the mean calorie consumptions for female members are less than those of male members. Table 3.2. All India Average Calorie Intake of Members of Households by Age Group and Sex:
Results of the Regression Method Using NSS 61st Round Data Rural Urban Rural + Urban
Coefficients
W/o usingany Wt.
Using sampling
Wt.
Using both sampling & model Wt.
F/MW/o usingany Wt.
Using sampling
Wt.
Using both sampling & model Wt.
F/M W/o usingany Wt.
Using sampling
Wt.
Using both sampling & model Wt.
F/M
Intercept 289.3 374.3 435.4 ‐ 1096.8 989.2 699.3 ‐ 539.2 540.3 520.2 ‐ Males 1099.0 1089.1 1006.5 901.2 857.9 831.6 1089.0 1071.2 978.8 0‐3
yrs. Females 1003.0 983.7 937.0 .93
850.7 877.1 847.9 1.02
996.7 993.3 925.2 .95
Males 1571.7 1536.0 1483.4 1344.8 1294.9 1299.0 1524.3 1502.1 1452.3 4‐6 yrs. Females 1391.0 1418.1 1349.2
.911112.6 1285.3 1108.9
.85 1334.3 1411.6 1318.8
.91
Males 1766.1 1750.0 1748.7 1521.1 1634.0 1571.5 1714.6 1743.3 1728.9 7‐12 yrs. Females 1698.2 1657.6 1656.2
.951388.2 1373.0 1514.3
.96 1616.1 1610.7 1624.7
.94
Males 2141.5 2117.6 2115.6 1862.9 1849.8 1915.8 2067.0 2064.1 2067.9 13‐18 yrs. Females 2060.8 2039.6 1965.7
.931675.8 1637.4 1660.8
.87 1930.2 1937.6 1886.9
.91
Males 2379.0 2352.6 2328.7 2047.9 2071.7 2093.6 2277.8 2272.6 2269.2 19 or more yrs. Females 2248.8 2068.0 2097.0
.901896.3 1893.1 2118.0
1.01 2106.9 1999.1 2074.2
.91
It should be remembered here that the calorie norms of female members given by FAO and ICMR (Table 2.6) are also less than or equal to those of corresponding male members. So it is difficult to say whether the differences in the calorie consumption between male and female members are as expected or due to gender inequality without comparing these ratios. The comparison is given in the Table 3.3. The gender ratios of the proposed model are seen to be less than the gender ratios found from the norms given by FAO and ICMR among members in the lower age groups and higher among the members in the higher age groups and adults.
Table 3.3. Age Group wise Comparison of F/M Ratios with Corresponding FAO and ICMR Norms: NSS 61st Round Data
Rural Urban Rural+Urban Age group
FAO
ICMR Our F/G Our/FAO Our/ICMR Our F/G Our/FAO Our/ICMR Our F/G Our/FAO Our/ICMR
0‐3 yrs. 0.97 1.00 .93 .96 .93 1.02 1.05 1.02 .95 .98 .95 4‐6 yrs. 0.96 1.00 .91 .95 .91 .85 .89 .85 .91 .95 .91 7‐12 yrs. 0.86 0.95 .95 1.10 1.00 .96 1.12 1.01 .94 1.09 .99 13‐18 yrs. 0.74 0.78 .93 1.26 1.19 .87 1.18 1.12 .91 1.23 1.17
19 or more yrs. 0.89 0.77 .90 1.01 1.17 1.01 1.13 1.31 .91 1.02 1.18 Average 0.88 0.90 0.92 ‐ ‐ 0.94 ‐ ‐ 0.92 ‐ ‐
It is felt that the treatment on the members would be different for different income/expenditure levels. We grouped the households into 12 expenditure groups. Group 1 has the lowest and the Group 12 has the highest per capita expenditures. Rural and urban expenditure groups are different. The groups are same as the ones taken by NSSO. The results of the regression analysis are much different now. This time all the coefficients have increased substantially (Tables 3.4‐3.6). This is seen more among the members in the lower Age Groups and in the households with low per capita expenditures. The intercept terms are found to be very small or negative for the lower expenditure groups. This signifies that some consumptions were not taken into account for the lower expenditure group households. The members
34
consumed food outside the house or received food in kind which have not been reported. Similarly, by the same logic, because of high positive values of the intercept terms, it can be concluded that some expenditure on food have been incurred by higher expenditure group households and possibly consumed by members from outside the households which have not been reported. The consumption ratios between female and male members are found to be more or less same for all age groups for each expenditure class. Most of the ratios are less than 1. Number of expenditure groups with ratio more than 1 was more among the lower Age Group members especially in Urban India. The Expenditure Group wise regression results and ratios relative to those of FAO and ICMR are given in Tables 3.7 through 3.9. The Gender ratios are found to be more than the gender ratios obtained from the FAO and ICMR norms for most cases. Among the adults and the members in the Age Group 13‐18 years it is found to be true for almost cent percent. Even among the lower Age Group members our estimates of the ratios are seen to be more than the corresponding ratios of FAO and ICMR in almost half of the cases. There are some exceptions in the higher expenditure groups especially the top three expenditure groups. We removed the top three expenditure groups and found the averages of the calorie ratios. The results are given in Table 3.10. We have also presented in the table the mean values of the ratios (Arithmetic Mean) taking all expenditure groups just to show the erratic behavior if the top three expenditure groups are included in the analysis. The result is a quite surprise. The mean ratios excluding the top three expenditure groups give almost same value for rural and urban sectors at each age group. Taking arithmetic mean may not be proper in case of ratios. So we took geometric mean also. The result is similar. The geometric means of female‐male ratios of average calorie consumptions of bottom nine expenditure groups for rural and urban sectors in India are also as close to each other as that of the arithmetic means. So it establishes the fact that, except for higher income groups, there are no differences in the calorie consumptions of female members relative to that of male members between rural and urban sectors in India. We can thus deflate male consumptions or inflate female consumptions accordingly and forget male female distinctions. Assuming that each male member has unit ‘1’, we can convert the number of females using this ratio so that the female members can be taken as equivalent to male members. The only thing we have to do is to multiply the number of female members by 0.96. From Table 3.10 it also becomes clear that these means are more than those found from calorie norms of FAO and ICMR (except for the age group 4‐6 years) signifying that women are not deprived of their due food compared to men. There are however certain limitations in our analysis. We have not considered the activity patterns of the adult members in the households. Though for calculations of calorie norms the adults are usually put in one of the three groups according to the activity pattern or life style – sedentary life style, moderately active life style and vigorously active life style, we have taken sedentary life style for all adults.
35
Table 3.4. Average Calorie Intake of Members of Households by Age Group, Sex and Expenditure Group in Rural India
Age Groups 0 – 3 years 4 – 6 years 7 – 12 years 13 – 18 years 19 years or more
Exp. Group Intercept
Males Females F/M Males Females F/M Males Females F/M Males Females F/M Males Females F/M
Gr1 ‐657.1 1072.1 1089.8 1.02 1357.1 1306.5 0.96 1556.4 1528.6 0.98 1546.0 1533.5 0.99 1553.6 1746.9 1.12Gr2 273.6 1404.7 1227.6 0.87 1301.7 1382.4 1.06 1656.1 1425.3 0.86 1697.6 1767.4 1.04 1699.1 1461.9 0.86Gr3 10.1 1199.0 1175.2 0.98 1633.1 1562.4 0.96 1700.7 1732.9 1.02 1947.3 1739.0 0.89 1783.7 1724.9 0.97Gr4 156.9 1328.7 1482.9 1.12 1617.7 1617.1 1.00 1865.5 1772.8 0.95 1936.9 1767.4 0.91 1854.8 1786.5 0.96Gr5 30.3 1609.0 1328.7 0.83 1677.8 1717.5 1.02 1970.0 1854.7 0.94 2139.7 1888.5 0.88 2037.2 1838.2 0.90Gr6 131.4 1411.4 1582.8 1.12 1830.5 1788.6 0.98 2034.9 1899.2 0.93 2148.6 2121.4 0.99 2001.7 1927.8 0.96Gr7 198.5 1593.3 1577.1 0.99 2034.0 1744.3 0.86 1985.5 1926.9 0.97 2129.2 2087.8 0.98 2104.6 2000.2 0.95Gr8 15.42 1477.6 1479.4 1.00 2075.1 1795.1 0.87 2085.8 1929.0 0.92 2418.1 2165.1 0.90 2267.2 2199.9 0.97Gr9 145.5 1456.7 1635.3 1.12 1940.0 1655.3 0.85 2117.0 2036.1 0.96 2343.8 2381.4 1.02 2331.4 2316.6 0.99Gr10 266.3 1614.7 1347.8 0.83 1975.1 1867.2 0.95 2238.1 2184.3 0.98 2458.5 2349.9 0.96 2423.0 2395.0 0.99Gr11 257.8 1433.9 1392.6 0.97 2446.0 1955.2 0.80 2276.3 2102.9 0.92 2640.7 2315.7 0.88 2693.1 2634.4 0.98Gr12 306.3 1267.9 1055.3 0.83 1767.7 1584.1 0.90 2250.9 2747.5 1.22 2947.1 2677.5 0.91 2975.9 3287.6 1.10
Table 3.5. Average Calorie Intake of Members of Households by Age Group, Sex and Expenditure Group in Urban India
Age Groups 0 – 3 years 4 – 6 years 7 – 12 years 13 – 18 years 19 years or more Exp. Group Intercept Males Females F/M Males Females F/M Males Females F/M Males Females F/M Males Females F/M
Gr1 ‐346.2 852.0 934.5 1.10 1303.6 1366.4 1.05 1503.8 1533.7 1.02 1638.2 1454.0 0.89 1571.4 1687.4 1.07Gr2 ‐77.8 1300.4 1204.3 0.93 1569.0 1276.5 0.81 1513.2 1562.1 1.03 1778.3 1930.6 1.09 1858.1 1552.5 0.84Gr3 ‐39.1 1129.3 1144.2 1.01 1601.5 1565.7 0.98 1654.1 1555.5 0.94 1949.7 1732.1 0.89 1818.2 1771.5 0.97Gr4 17.9 1142.1 1154.6 1.01 1482.1 1677.6 1.13 2145.5 1667.3 0.78 1908.8 1879.9 0.98 2019.2 1797.4 0.89Gr5 200.7 1458.6 1372.2 0.94 1436.5 1226.0 0.85 1834.3 1770.5 0.97 1970.8 1932.4 0.98 1902.0 1872.8 0.98Gr6 402.8 1314.0 1263.4 0.96 1676.2 1630.4 0.97 1581.8 1660.8 1.05 2087.1 2060.4 0.99 1995.7 1879.9 0.94Gr7 344.8 1298.5 1401.2 1.08 1829.2 1605.8 0.88 1921.5 1727.3 0.90 2139.9 1871.9 0.87 2058.1 2000.5 0.97Gr8 280.5 1380.6 1475.8 1.07 1796.6 1408.1 0.78 1836.1 1997.5 1.09 2057.5 1862.4 0.91 2205.9 2134.1 0.97Gr9 448.2 1318.0 1283.9 0.97 1498.1 1455.0 0.97 1882.0 2015.8 1.07 2111.5 2003.8 0.95 2180.7 2236.2 1.03Gr10 581.4 1274.4 1502.3 1.18 1887.2 1439.6 0.76 2042.9 1875.5 0.92 2434.3 2048.0 0.84 2250.1 2228.9 0.99Gr11 786.6 1164.7 1391.5 1.19 1869.8 1269.4 0.68 2174.9 1877.7 0.86 2514.9 2096.8 0.83 2269.8 2408.4 1.06Gr12 948.1 1082.7 1300.0 1.20 960.1 993.0 1.03 2281.3 1862.7 0.82 2425.4 2236.1 0.92 2423.5 2653.1 1.09
36
Table 3.6. All India Average Calorie Intake of Members of Households by Age Group, Sex and Expenditure Group Age Groups 0 – 3 years 4 – 6 years 7 – 12 years 13 – 18 years 19 years or more
Exp. Group Intercept
Males Females F/M Males Females F/M Males Females F/M Males Females F/M Males Females F/M
Gr1 ‐552.3 1013.4 1048.3 1.03 1338.0 1311.5 0.98
1533.3 1511.1 0.99 1559.7 1555.7 1.00 1562.4 1718.9 1.10
Gr2 113.9 1356.0 1232.4 0.91 1353.7 1375.4 1.02
1628.7 1468.8 0.90 1728.5 1817.9 1.05 1744.7 1527.3 0.88
Gr3 ‐4.5 1190.1 1181.6 0.99 1631.5 1560.6 0.96
1691.7 1695.5 1.00 1949.5 1739.3 0.89 1795.0 1730.2 0.96
Gr4 339.3 1165.6 1360.0 1.17 1584.2 1686.5 1.06
1972.9 1677.6 0.85 1849.6 1741.3 0.94 1911.6 1679.5 0.88
Gr5 76.1 1588.7 1358.7 0.86 1626.5 1642.3 1.01
1946.6 1851.3 0.95 2088.8 1910.5 0.91 2003.7 1827.8 0.91
Gr6 237.9 1412.3 1515.0 1.07 1792.6 1748.5 0.98
1935.0 1817.6 0.94 2149.5 2117.1 0.98 1993.7 1891.4 0.95
Gr7 271.4 1488.2 1489.3 1.00 2037.2 1757.5 0.86
1988.7 1856.9 0.93 2115.7 2011.7 0.95 2098.1 1975.1 0.94
Gr8 120.6 1446.8 1478.8 1.02 2070.5 1723.0 0.83
2047.0 1957.1 0.96 2335.9 2077.9 0.89 2264.7 2131.1 0.94
Gr9 248.5 1444.0 1595.5 1.10 1881.3 1621.6 0.86
2079.6 2058.9 0.99 2298.6 2303.3 1.00 2300.8 2248.4 0.98
Gr10 358.1 1565.4 1407.1 0.90 1976.6 1785.1 0.90
2186.3 2116.4 0.97 2465.7 2286.0 0.93 2403.0 2310.2 0.96
Gr11 422.6 1417.4 1387.7 0.98 2228.2 1850.4 0.83
2297.3 2085.0 0.91 2604.0 2283.3 0.88 2580.6 2539.8 0.98
Gr12 460.1 1346.1 1133.1 0.84 1698.3 1542.4 0.91
2318.8 2555.6 1.10 2887.9 2612.2 0.90 2842.7 3075.5 1.08
Table 3.7. Age and Expenditure Group wise Comparison of F/M Ratios with Corresponding FAO and ICMR Norms: Rural India
Age Group 0 – 3 years 4 – 6 years 7 – 12 years 13 – 18 years 19 years or more
Exp. Group Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAOOur/ ICMR
Our F/G
Our/ FAOOur/ ICMR
Our F/G
Our/ FAOOur/ ICMR
Our F/G
Our/ FAOOur/ ICMR
Gr1 1.02 1.05 1.02 0.96 1.00 0.96 0.98 1.14 1.03 0.99 1.34 1.27 1.12 1.26 1.45 Gr2 0.87 0.90 0.87 1.06 1.10 1.06 0.86 1.00 0 .91 1.04 1.41 1.33 0.86 0.97 1.12 Gr3 0.98 1.01 0.98 0.96 1.00 0.96 1.02 1.19 1.07 0.89 1.20 1.14 0.97 1.09 1.26 Gr4 1.12 1.15 1.12 1.00 1.04 1.00 0.95 1.10 1.00 0.91 1.23 1.17 0.96 1.08 1.25
37
Gr5 0.83 0.86 0.83 1.02 1.06 1.02 0.94 1.09 0 .99 0.88 1.19 1.13 0.90 1.01 1.17 Gr6 1.12 1.15 1.12 1.02 1.02 1.02 0.93 1.08 0 .98 0.99 1.34 1.27 0.96 1.08 1.25 Gr7 0.99 1.02 0.99 0.86 0.90 0.86 0.97 1.13 1.02 0.98 1.32 1.26 0.95 1.07 1.23 Gr8 1.00 1.03 1.00 0.86 0.91 0.86 0.92 1.07 0 .97 0.90 1.22 1.15 0.97 1.09 1.26 Gr9 1.12 1.15 1.12 0.85 0.89 0.85 0.96 1.12 1.01 1.02 1.38 1.31 0.99 1.11 1.29 Gr10 0.83 0.86 0.83 0.95 0.99 0.95 0.98 1.14 1.03 0.96 1.30 1.23 0.99 1.11 1.29 Gr11 0.97 1.00 0.97 0.80 0.83 0.80 0.92 1.07 0 .97 0.88 1.19 1.13 0.98 1.10 1.27 Gr12 0.83 0.86 0.83 0.90 0.94 0.90 1.22 1.42 1.28 0.91 1.23 1.17 1.10 1.24 1.43
Average 0.97 1.00 0.97 0.94 0.97 0.93 0.97 1.13 1.02 0.95 1.28 1.21 0.98 1.10 1.27
38
Table 3.8. Age and Expenditure Group wise Comparison of F/M Ratios with Corresponding FAO and ICMR Norms: Urban India Age Group 0 – 3 years 4 – 6 years 7 – 12 years 13 – 18 years 19 years or more
Exp. Group Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Gr1 1.10
1.13 1.10 1.05
1.09 1.05 1.02
1.18 1.07 .89 1.20 1.14 1.07
1.20 1.38
Gr2 .93 .95 .93 .81 .84 .81 1.0
3 1.19 1.08 1.0
9 1.47 1.39 .84 .94 1.09
Gr3 1.01
1.04 1.01 .98 1.02 .98 .94 1.09 .98 .89 1.20 1.14 .97 1.08 1.25
Gr4 1.01
1.04 1.01 1.13
1.17 1.13 .78 .90 .82 .98 1.32 1.25 .89 1.00 1.15
Gr5 .94 .96 .94 .85 .88 .85 .97 1.12 1.02 .98 1.32 1.25 .98 1.10 1.27
Gr6 .96 .98 .96 .97 1.01 .97 1.0
5 1.22 1.10 .99 1.33 1.26 .94 1.05 1.22
Gr7 1.08
1.11 1.08 .88 .91 .88 .90 1.04 .94 .87 1.17 1.11 .97 1.08 1.25
Gr8 1.07
1.10 1.07 .78 .81 .78 1.09
1.26 1.14 .91 1.22 1.16 .97 1.08 1.25
Gr9 .97 1.00 .97 .97 1.01 .97 1.0
7 1.24 1.12 .95 1.28 1.21 1.0
3 1.15 1.33
Gr10 1.18
1.21 1.18 .76 .79 .76 .92 1.06 .96 .84 1.13 1.07 .99 1.11 1.28
Gr11 1.19
1.22 1.19 .68 .70 .68 .86 1.00 .90 .83 1.12 1.06 1.06
1.19 1.37
Gr12 1.20
1.23 1.20 1.03
1.07 1.03 .82 .95 .86 .92 1.24 1.17 1.09
1.22 1.41
Average 1.05 1.08 1.05
0.91 0.94 0.91
0.95 1.10 1.00
0.92 1.25 1.18
0.98 1.10 1.27
Table 3.9. Age and Expenditure Group wise Comparison of Estimated F/M Ratios with Corresponding FAO and ICMR Norms: All India Age Group 0 – 3 years 4 – 6 years 7 – 12 years 13 – 18 years 19 years or more
Exp. Group Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Gr1 1.03 1.06 1.03 .98 1.02 .98 .99 1.15 1.04 1.00 1.35 1.28 1.10 1.23 1.42Gr2 .91 .93 .91 1.02 1.06 1.02 .90 1.04 .94 1.05 1.41 1.34 .88 .98 1.14Gr3 .99 1.02 .99 .96 1.00 .96 1.00 1.16 1.05 .89 1.20 1.14 .96 1.07 1.24
39
Gr4 1.17 1.20 1.17 1.06 1.10 1.06 .85 .98 .89 .94 1.27 1.20 .88 .98 1.14Gr5 .86 .88 .86 1.01 1.05 1.01 .95 1.10 1.00 .91 1.22 1.16 .91 1.02 1.18Gr6 1.07 1.10 1.07 .98 1.02 .98 .94 1.09 .98 .98 1.32 1.25 .95 1.06 1.23Gr7 1.00 1.03 1.00 .86 .89 .86 .93 1.08 .97 .95 1.28 1.21 .94 1.05 1.22Gr8 1.02 1.05 1.02 .83 .86 .83 .96 1.11 1.01 .89 1.20 1.14 .94 1.05 1.22Gr9 1.10 1.13 1.10 .86 .89 .86 .99 1.15 1.04 1.00 1.35 1.28 .98 1.10 1.27Gr10 .90 .92 .90 .90 .93 .90 .97 1.12 1.02 .93 1.25 1.19 .96 1.07 1.24Gr11 .98 1.01 .98 .83 .86 .83 .91 1.05 .95 .88 1.18 1.12 .98 1.10 1.27Gr12 .84 .86 .84 .91 .94 .91 1.10 1.27 1.15 .90 1.21 1.15 1.08 1.21 1.40
Average 0.99 1.02 0.99
0.93 0.97 0.93
0.96 1.11 1.00
0.94 1.27 1.20
0.96 1.08 1.25
Table 3.10. Age and Expenditure Group wise Comparison of Estimated F/M Ratios with Corresponding FAO and ICMR Norms After Truncating Top Three Expenditure Groups: All India
Age Group
0 – 3 years 4 – 6 years 7 – 12 years 13 – 18 years 19 years or more
Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Our F/G
Our/ FAO
Our/ ICMR
Rural 0.97 1.00 0.97 0.94 0.97 0.93 0.97 1.13 1.02 0.95 1.28 1.21 0.98 1.10 1.27
Urban 1.05
1.08 1.05 0.91
0.94 0.91 0.95
1.10 1.00 0.92
1.25 1.18 0.98
1.10 1.27
AM All 12 Exp.
Groups
All India 0.99
1.02 0.99 0.93
0.97 0.93 0.96
1.11 1.00 0.94
1.27 1.20 0.96
1.08 1.25
Rural 1.01
1.04 1.01 0.95
0.99 0.95 0.95
1.10 1.00 0.96
1.29 1.23 0.96
1.08 1.25
Urban 1.01
1.03 1.01 0.94
0.97 0.94 0.98
1.14 1.03 0.95
1.28 1.21 0.96
1.08 1.24
AM Lowest 9 Exp.
Group
All India 1.02
1.04 1.02 0.95
0.99 0.95 0.95
1.10 0.99 0.96
1.29 1.22 0.96
1.08 1.24
Rural 1.00
1.03 1.00 0.95
0.99 0.95 0.95
1.10 1.00 0.95
1.29 1.22 0.96
1.08 1.25
Urban 1.00
1.03 1.01 0.93
0.96 0.93 0.98
1.13 1.03 0.95
1.28 1.21 0.96
1.07 1.24
GM Lowest 9 Exp.
Group
All India 1.00
1.03 1.01 0.95
0.98 0.95 0.95
1.10 1.00 0.95
1.29 1.22 0.96
1.08 1.24
40
41
3.5 The Poverty Rates by Calorie Decomposition Method The results of the previous section are not only useful in determining the female male ratio of the calorie intake, but also has other uses. In this section we use the member wise expected calorie consumption of the households to arrive at the poverty rates. Let us take the calorie consumption table for rural and urban India as obtained in the earlier chapter using 61st round NSS data. The following tables give the extracts of the earlier tables separately for male and female members. We have taken only the first nine expenditure groups, because the top three expenditure groups are not considered to be poor and may be excluded from analysis. Only at the last step the number of excluded households with proper weight will be considered to revise our poverty rates. Table 3.11 Calorie Intakes of Rural Males by Age and Expenditure Groups: NSS 61st Round Data Truncated by Top Three Expenditure Groups Exp. Grp. Lower
Limit Intercept Age
00‐03 Age 04‐06
Age 07‐12
Age 13‐18
Age 19+
Av. Exp.*
Group Multiplier
Gr1 0 ‐657.1 1072.1 1357.1 1556.4 1546.0 1553.6 200.94 1435 Gr2 235 273.6 1404.7 1301.7 1656.1 1697.6 1699.1 253.75 1635 Gr3 270 10.1 1199.0 1633.1 1700.7 1947.3 1783.7 296.52 3073 Gr4 320 156.9 1328.7 1617.7 1865.5 1936.9 1854.8 342.29 3182 Gr5 365 30.3 1609.0 1677.8 1970.0 2139.7 2037.2 387.62 3017 Gr6 410 131.4 1411.4 1830.5 2034.9 2148.6 2001.7 431.91 2679 Gr7 455 198.5 1593.3 2034.0 1985.5 2129.2 2104.6 481.40 2782 Gr8 510 15.4 1477.6 2075.1 2085.8 2418.1 2267.2 543.12 2736 Gr9 580 145.5 1456.7 1940.0 2117.0 2343.8 2331.4 630.19 2627
Table 3.12 Calorie Intakes of Rural Females by Age and Expenditure Groups: NSS 61st Round Data Truncated by Top Three Expenditure Groups
Exp. Grp. Lower Limit
Intercept Age 00‐03
Age 04‐06
Age 07‐12
Age 13‐18
Age 19+
Av. Exp.
Group Multiplier
Gr1 0 ‐657.1 1089.8 1306.5 1528.6 1533.5 1746.9 200.94 1435 Gr2 235 273.6 1227.6 1382.4 1425.3 1767.4 1461.9 253.75 1635 Gr3 270 10.1 1175.2 1562.4 1732.9 1739.0 1724.9 296.52 3073 Gr4 320 156.9 1482.9 1617.1 1772.8 1767.4 1786.5 342.29 3182 Gr5 365 30.3 1328.7 1717.5 1854.7 1888.5 1838.2 387.62 3017 Gr6 410 131.4 1582.8 1788.6 1899.2 2121.4 1927.8 431.91 2679 Gr7 455 198.5 1577.1 1744.3 1926.9 2087.8 2000.2 481.40 2782 Gr8 510 15.4 1479.4 1795.1 1929.0 2165.1 2199.9 543.12 2736 Gr9 580 145.5 1635.3 1655.3 2036.1 2381.4 2316.6 630.19 2627
42
Table 3.13 Calorie Intakes of Urban Males by Age and Expenditure Groups: NSS 61st Round Data Truncated by Top Three Expenditure Groups
Exp. Grp.
Lower Limit
Intercept Age 00‐03
Age 04‐06
Age 07‐12
Age 13‐18
Age 19+
Av. Exp.
Group Multiplier
Gr1 0 ‐346.2 852.0 1303.6 1503.8 1638.2 1571.4 280.89 202 Gr2 335 ‐77.8 1300.4 1569.0 1513.2 1778.3 1858.1 368.09 216 Gr3 395 ‐39.1 1129.3 1601.5 1654.1 1949.7 1818.2 441.18 438 Gr4 485 17.9 1142.1 1482.1 2145.5 1908.8 2019.2 533.09 496 Gr5 580 200.7 1458.6 1436.5 1834.3 1970.8 1902.0 625.54 483 Gr6 675 402.8 1314.0 1676.2 1581.8 2087.1 1995.7 729.78 526 Gr7 790 344.8 1298.5 1829.2 1921.5 2139.9 2058.1 857.80 571 Gr8 930 280.5 1380.6 1796.6 1836.1 2057.5 2205.9 1014.12 580 Gr9 1100 448.2 1318.0 1498.1 1882.0 2111.5 2180.7 1226.20 659
Table 3.14 Calorie Intakes of Urban Females by Age and Expenditure Groups: NSS 61st Round Data Truncated by Top Three Expenditure Groups
Exp. Grp.
Lower Limit
Intercept Age 00‐03
Age 04‐06
Age 07‐12
Age 13‐18
Age 19+
Av. Exp.
Group Multiplier
Gr1 0 ‐346.2 934.5 1366.4 1533.7 1454.0 1687.4 280.89 202 Gr2 335 ‐77.8 1204.3 1276.5 1562.1 1930.6 1552.5 368.09 216 Gr3 395 ‐39.1 1144.2 1565.7 1555.5 1732.1 1771.5 441.18 438 Gr4 485 17.9 1154.6 1677.6 1667.3 1879.9 1797.4 533.09 496 Gr5 580 200.7 1372.2 1226.0 1770.5 1932.4 1872.8 625.54 483 Gr6 675 402.8 1263.4 1630.4 1660.8 2060.4 1879.9 729.78 526 Gr7 790 344.8 1401.2 1605.8 1727.3 1871.9 2000.5 857.80 571 Gr8 930 280.5 1475.8 1408.1 1997.5 1862.4 2134.1 1014.12 580 Gr9 1100 448.2 1283.9 1455.0 2015.8 2003.8 2236.2 1226.20 659
For further calculations it is necessary to smooth the data in Tables 3.11‐3.14. It is a two way smoothing. First we assume that the mean calorie consumption is a function of mean expenditure. We have tried linear, quadratic and cubic relations for each of the four tables. In each case linear fit was very good. The quadratic fit was slightly better. The cubic relation did not improve the fit significantly. Suppose the quadratic equation is
CalConsni = a + b*AvExpi + c*AvExpi2 + ei, … (3.16)
where CalConsni is the average calorie consumption of the ith expenditure group, AvExpi is the Average per capita expenditure of the ith expenditure group, a, b and c are the regression coefficients and ei is the equation error. The regression may be run for all expenditure groups. Since the regression coefficients depend on g, the age group, we can think of finding a, b and c in terms of g. So we write
ag = a1 + b1*g + c1*g2 + e1g, … (3.17)
bg = a2 + b2*g + c2*g2 + e2g, … (3.18)
and cg = a3 + b3*g + c3*g2 + e3g, … (3.19)
with usual assumptions. We substitute (3.17), (3.18) and (3.19) in (3.16) to get the following equation:
43
CalConsn = α+β10*AvExp+β01*Age+β20*AvExp
2+β02*Age2+β11*Age*AvExp
+β21*Age*AvExp2 +β12*AvExp*Age
2 + β22*AvExp2*Age2 … (3.20)
where α and βs are the regression coefficients and the subscripts and the equation error have been omitted for convenience. We discard β22 because it is the coefficient of the variables with order 4 when AvExp and Age are taken together.17 Thus we take the following equation:
CalConsn = α+β10*AvExp+β01*Age+β20*AvExp2+β02*Age
2+β11*Age*AvExp +β21*Age*AvExp
2 +β12*AvExp*Age2 … (3.21)
Instead of regressing step by step, we run a single equation for each of the four sets of data, namely rural‐male, rural‐female, urban‐male and urban‐female. We get the following four regressions:
Male Rural: ‐97.179 + 5.044*AvExp + 366.361*Age – 0.005525*AvExp*AvExp – 32.320*Age*Age + 0.05593*Age*AvExp + 0.0006737*Age*AvExp*AvExp – 0.06959*AvExp*Age*Age
Female Rural: ‐ 321.484 + 5.716*AvExp + 694.802*Age – 0.005352*AvExp*AvExp – 65.739*Age*Age – 1.387*Age*AvExp + 0.001465*Age*AvExp*AvExp + 0.05036*AvExp*Age*Age
Male Urban: ‐ 62.175 + 2.186*AvExp + 669.643*Age – 0.001132*AvExp*AvExp – 77.725*Age*Age – 0.368*Age*AvExp + 0.0001195*Age*AvExp*AvExp + 0.04047*AvExp*Age*Age
Female Urban: 133.356 + 1.904*AvExp + 585.404*Age – 0.000988*AvExp*AvExp – 63.599*Age*Age – .453*Age*AvExp + 0.0001948*Age*AvExp*AvExp + 0.04026*AvExp*Age*Age
Table 3.15. Bivariate Quadratic Estimates of Calorie Intakes of Rural Males by Age and Expenditure Groups: NSS 61st Round Data Truncated by Top Three Expenditure Groups.
Exp. Grp.
Lower Limit
Intercept Age 00‐03
Age 04‐06
Age 07‐12
Age 13‐18
Age 19+
Av. Exp.
Group Multiplier
Gr1 0 ‐657.1 1051.6 1317.5 1490.8 1571.5 1559.6 200.94 1435 Gr2 235 273.6 1200.8 1474.8 1648.8 1722.9 1697.0 253.75 1635 Gr3 270 10.1 1301.7 1585.1 1762.5 1834.0 1799.6 296.52 3073 Gr4 320 156.9 1390.1 1686.1 1869.9 1941.3 1900.5 342.29 3182 Gr5 365 30.3 1457.6 1769.0 1961.8 2036.0 1991.6 387.62 3017 Gr6 410 131.4 1504.3 1833.4 2037.7 2117.2 2072.0 431.91 2679 Gr7 455 198.5 1534.0 1885.9 2106.2 2194.9 2151.9 481.40 2782 Gr8 510 15.4 1537.7 1922.8 2167.7 2272.3 2236.7 543.12 2736 Gr9 580 145.5 1480.0 1920.7 2208.9 2344.9 2328.5 630.19 2627
Average Expenditures are found using individual multiplier. Table 3.16. Bivariate Quadratic Estimates of Calorie Intakes of Rural Females by Age and Expenditure Groups: NSS 61st Round Data Truncated by Top Three Expenditure Groups.
Exp. Lower Intercept Age Age Age Age Age Av. Group 17 If we start with the linear equation CalConsni = a + b*AvExpi + ei, which has been found to give good fit then we get CalConsn = α+β10*AvExp+β01*Age+β02*Age
2+β11*Age*AvExp +β12*AvExp*Age
2.
44
Grp. Limit 00‐03 04‐06 07‐12 13‐18 19+ Exp. MultiplierGr1 0 ‐657.1 1030.6 1339.0 1536.2 1622.2 1596.9 200.94 1435 Gr2 235 273.6 1168.5 1446.9 1619.3 1685.9 1646.5 253.75 1635 Gr3 270 10.1 1264.3 1524.3 1682.8 1739.5 1694.7 296.52 3073 Gr4 320 156.9 1351.1 1597.4 1746.7 1799.0 1754.3 342.29 3182 Gr5 365 30.3 1421.0 1659.8 1806.1 1860.0 1821.5 387.62 3017 Gr6 410 30.3 1473.9 1711.1 1860.4 1921.7 1895.0 431.91 2679 Gr7 455 198.5 1514.9 1757.3 1916.6 1992.9 1986.3 481.40 2782 Gr8 510 198.5 1539.4 1798.2 1980.1 2085.3 2113.6 543.12 2736 Gr9 580 145.5 1523.7 1824.5 2057.3 2222.1 2318.9 630.19 2627
Table 3.17. Bivariate Quadratic Estimates of Calorie Intakes of Urban Males by Age and Expenditure Groups: NSS 61st Round Data Truncated by Top Three Expenditure Groups.
Exp. Grp.
Lower Limit
Intercept Age 00‐03
Age 04‐06
Age 07‐12
Age 13‐18
Age 19+
Av. Exp.
Group Multiplier
Gr1 0 ‐346.2 971.7 1348.3 1592.2 1703.3 1681.8 280.89 202 Gr2 335 ‐77.8 1076.4 1438.2 1674.4 1785.0 1769.8 368.09 216 Gr3 395 ‐39.1 1152.3 1503.1 1734.3 1845.7 1837.3 441.18 438 Gr4 485 17.9 1232.3 1571.2 1797.8 1912.1 1914.1 533.09 496 Gr5 580 200.7 1295.6 1624.5 1848.6 1967.8 1982.3 625.54 483 Gr6 675 402.8 1346.1 1666.2 1889.9 2017.2 2048.1 729.78 526 Gr7 790 344.8 1378.1 1690.9 1917.6 2058.4 2113.1 857.80 571 Gr8 930 280.5 1372.1 1681.3 1917.1 2079.6 2168.7 1014.12 580 Gr9 1100 448.2 1284.9 1598.5 1856.0 2057.2 2202.2 1226.20 659
Table 3.18. Bivariate Quadratic Estimates of Calorie Intakes of Urban Females by Age and Expenditure Groups: NSS 61st Round Data Truncated by Top Three Expenditure Groups.
Exp. Grp.
Lower Limit
Intercept Age 00‐03
Age 04‐06
Age 07‐12
Age 13‐18
Age 19+
Av. Exp.
Group Multiplier
Gr1 0 ‐346.2 1011.4 1328.0 1540.1 1647.5 1650.4 280.89 202 Gr2 335 ‐77.8 1096.6 1395.2 1596.3 1699.8 1705.8 368.09 216 Gr3 395 ‐39.1 1158.6 1444.5 1638.7 1741.2 1752.1 441.18 438 Gr4 485 17.9 1224.7 1497.4 1685.9 1790.1 1810.1 533.09 496 Gr5 580 200.7 1277.5 1540.4 1726.5 1835.8 1868.2 625.54 483 Gr6 675 402.8 1320.9 1576.7 1764.0 1882.9 1933.4 729.78 526 Gr7 790 344.8 1350.6 1603.4 1798.1 1934.6 2013.0 857.80 571 Gr8 930 280.5 1351.6 1609.4 1821.7 1988.4 2109.6 1014.12 580 Gr9 1100 448.2 1290.9 1570.8 1822.2 2045.2 2239.6 1226.20 659
Observe that if we start with the linear equation “CalConsni = a + b*AvExpi + ei
”, which has been found to give good fit then we get
CalConsn=α+β10*AvExp+β01*Age+β02*Age2+β11*Age*AvExp
+β12*AvExp*Age2. … (3.22)
In practice we take a slightly different equation. We omit the term AvExp*Age2 and include the term AvExp2 so that the equation becomes quadratic with AvExp and Age and get the following equation. CalConsn=α+β10*AvExp+β01*Age+β20*AvExp
2+β02*Age2+β11*AvExp*Age. … (3.23)
In this case we get the following eight regression equations: Table 3.19. Coefficients of Quadratic Regression of Calorie Intakes on Average Expenditure and Age: NSS 61st Round Data.
45
9 Exp Group 7 Exp Group Rural Urban Rural Urban
Regressors Male Female Male Female Male Female Male Female (Constant) 37.399 457.657 317.147 626.417 ‐49.521 239.848 118.243 344.175AvExp 3.812 2.365 1.378 .748 4.005 2.626 1.820 1.456Age 426.123 290.345 432.339 311.537 461.137 409.167 499.573 383.725AvExp2 ‐.003503 ‐0.00196 ‐.0007938 ‐.0004224 ‐.003033 ‐.001127 ‐.001085 ‐.0008609Age2 ‐60.850 ‐42.029 ‐48.485 ‐34.671 ‐56.356 ‐47.483 ‐55.933 ‐40.184AvExp*Age .211 .223 0.05991 0.08968 0.02183 ‐.03996 0.01898 0.01850
It has been found that the cross product term is not significant in all the cases. If we omit this term and use the equation CalConsn=α+β10*AvExp+β01*Age+β20*AvExp
2+β02*Age2. … (3.24)
We get the following eight regression equations in that case. Table 3.20. Coefficients of Quadratic Regression of Calorie Intakes on Average Expenditure and Age (Without Interaction Term): NSS 61st Round Data.
9 Exp Group 7 Exp Group Rural Urban Rural Urban
Regressors Male Female Male Female Male Female Male Female (Constant) ‐222.467 183.401 187.716 432.668 ‐72.948 282.737 84.863 311.628AvExp 4.445 3.033 1.558 1.017 4.070 2.506 1.877 1.512Age 512.745 381.763 475.483 376.120 468.946 394.871 510.700 394.574AvExp2 ‐.003503 ‐0.00196 ‐.0007938 ‐.0004.224 ‐.003.033 ‐.001127 ‐.001085 ‐.0008609Age2 ‐60.850 ‐42.029 ‐48.485 ‐34.671 ‐56.356 ‐47.483 ‐55.933 ‐40.184
We shall however take the equation (3.23) for all subsequent calculations. Moreover it is not necessary to take 9 expenditure groups for urban India so far as calculations of poverty rates are concerned. We shall thus take 9 expenditure groups for rural India and 7 expenditure groups for urban India. Thus the four equations are: Rural Male 9 Expenditure Groups: CalConsn=37.399+3.812*AvExp+426.123*Age‐.003503*AvExp2‐60.850*Age2+.211*AvExp*Age Rural Female 9 Expenditure Groups: CalConsn=457.657+2.365*AvExp+290.345*Age‐0.00196*AvExp2‐42.029*Age2+.223*AvExp*Age Urban Male 7 Expenditure Groups: CalConsn=118.243+1.820*AvExp+499.573*Age‐.001085*AvExp2‐55.933*Age2+0.01898*AvExp*Age Urban Female 7 Expenditure Groups: CalConsn=344.175+1.456*AvExp+383.725*Age‐.0008609*AvExp2‐40.184*Age2+0.01850*AvExp*Age These four equations are very important in our poverty calculations. For a given household, we compute the per capita average expenditure of each member of the household depending on whether the household belongs to rural or urban sector. The weighted sums of the expected amount of calories consumed by the members are then found separately for male and female members of the household, where the weight is the number of members in each category. It may be thought that the total of these two sums (i.e., the sum of expected amount of calories consumed by male members and the sum of expected amount of calories consumed by female members) is the estimated calorie consumption of the household. This is not true. We must identify the expenditure group containing the household. The intercept term of the expenditure group should be added to the weighted sum to get the estimated calorie consumption of the household. In a similar manner we get the sum of calorie norms of members in the household.
46
Here the question of intercept term does not arise. The calorie norm of the household is compared with the estimated calorie consumption to determine whether the household is poor. If a household is poor then it is given a dummy value ‘1’, otherwise it is given the value ‘0’. Weighted means of these dummy values give us the poverty ratios. Weights should be taken as individual multiplier and not the multiplier of the household. The individual multiplier is just the product of the household multiplier and the total number of members in the household. This calculation can be carried out separately for rural and urban India (Table 3.21). Table 3.21. Estimates of Poverty Rates Assuming that Activity Status of All Adults are in the Sedentary Level Using Quadratic Regression of Calorie Intakes on Average Expenditure and Age: NSS 61st Round Data.
Rural Urban All India Method of Calculation Norm W/o
Weight With Weight
W/o Weight
With Weight
W/o Weight
With Weight
7 Exp Groups 0.33 0.48 0.52 0.56 0.40 0.50 9 Exp Groups
ICMR 0.36 0.51 0.68 0.74 0.48 0.57
7 Exp Groups 0.27 0.40 0.46 0.49 0.34 0.42 Our
9 Exp Groups FAO
0.31 0.43 0.59 0.63 0.41 0.48 ICMR ‐ 0.56 ‐ 0.62 ‐ 0.57
Direct* FAO ‐ 0.51 ‐ 0.58 ‐ 0.52
*. See Table 2.10. In case of All India estimates we should not take any of the estimates as shown in the above Table, It should be weighted average of 0.508 and 0.556 for ICMR norm and weighted average of 0.434 and 0.488 for FAO norm, weights being 0.7468 and 0.2532 respectively. Applying these weights we get the all India poverty ratios as 0.520 for ICMR norm and 0.447 for FAO norm. In any case these estimates are not plausible. The urban poverty ratios are found to be higher than the corresponding rural poverty ratios. What has gone wrong? Clearly the activity status! And if we take activity status then it will further inflate the poverty rates.
47
References 1. Ahmad, A. and J. Morduch (1993): "Identifying Sex Bias in the Allocation of Household
Resources: Evidence from Linked Household Surveys from Bangladesh", Mimeo, Harvard University.
2. Bhalotra, S. and C. Attfield (1998): "Intrahousehold Resource Allocation in Rural Pakistan: A Semiparametric Analysis", Journal of Applied Econometrics, 13, 463 ‐ 480.
3. Bidani, Benu and Martin Ravallion, 1997. “Decomposing social indicators using distributional data,” Journal of Econometrics, 77, 125‐139.
1. Bourguignon, F., M. Browning, P. A. Chiappori and V. Lechene (1993): "Intrahousehold Allocation of Consumption: Some Evidence on French Data", Annales d'Economie et de Statistique, 29, 137 – 156.
2. Browning, M. and P. A. Chiappori (1998): "Efficient Intra Household Allocations: A General Characterisation and Empirical Tests", Econometrica, 66(6), 1241‐1278.
3. Chesher, Andrew (1997): “Diet revealed?: semiparamatric estimation of nutrient intake‐age relationships,” Journal of the Royal Statistical Society A, 160(3):389‐428.
4. Coondoo, D., Majumder, A. and R. Ray (2003): “Estimation of a set of Multi Lateral Consumer Price Indices for Nutrients”, Working Paper No. ERU/2003‐12, Indian Statistical Institute, Kolkata, India.
5. Coondoo, D., Majumder, A. and R. Ray (2004), “A Method of Calculating Regional Consumer Price Differentials with Illustrative Evidence from India”, Review of Income and Wealth, 50(1), 51‐68.
6. Deaton, A. (1989): "Looking for Boy‐Girl Discrimination in Household Expenditure Data", World Bank Economic Review, 3(1), 1 ‐ 15.
7. Deaton, A. (1997): The Analysis of Household Surveys: A Microeconometric Approach to Development Policy, The World Bank, Washington DC.
8. Deaton, Angus and Christina Paxson, 2000, “Growth and saving among individuals and households,” Review of Economics and Statistics, 82(2), 212–25.
9. Dreze, J. and A. Sen (1989): Hunger and public action, Oxford, Oxford University Press. 10. Duflo, E. (2005): "Gender Equality in Development", Mimeo, MIT. 11. FAO (2001): Report of the Joint FAO/WHO/UNU Expert Consultation, Rome, 17‐24
October, 2001, http://www.fao. org/docrep/007/y5686e/y5686e01.htm# TopOfPage. 12. Gibson, J. and S. Rozelle (2004): "Is It Better to be a Boy? A Disaggregated Outlay
Equivalent Analysis of Gender Bias in Papua New Guinea", Journal of Development Studies, 40 (4), 115‐136.
13. Gong, X., A. Van Soest and P. Zhang (2005): "Sexual Bias and Household Consumption, a semiparametric analysis of Engel curves in rural China", Journal of Applied Econometrics, 20, 509 ‐ 527.
14. Haddad, L. and T. Reardon (1993): "Gender Bias in the Allocation of Resource within Households in Burkina Faso: A Disaggregated Outlay Equivalent Analysis", Journal of Development Studies, 29(2), 260 ‐ 276.
15. Haddad, L., C. Peña, C. Nishida, A. Quisumbing and S. Alison (1996): Food Security and Nutrition Implications of Intrahousehold Bias: A Review of Literature, FCND Discussion Paper No. 19, Food Consumption and Nutrition Division, International Food Policy Research Institute, Washington, D.C.
16. ICMR (2002): Nutrient Requirements and Recommended Dietary Allowances for Indians: A Report of the Expert Group of the Indian Council of Medical Research, New Delhi.
48
17. James WPT & Schofield EC (1990): Human Energy Requirements: a Manual for Planners and Nutritionists. New York: Food and Agriculture Organization of the United Nations/Oxford University Press.
18. Kingdon, G. G. (2005): Where has all the bias gone? Detecting gender bias in the Intrahousehold Allocation of Educational Expenditure" Economic Development and Cultural Change, 53 (2), 409‐452.
19. Lancaster, G., Maitra, P. and Ray, R (2007): Household Expenditure Patterns and Gender Bias: Evidence from Selected Indian States, Paper presented at the Indian Statistical Institute Platinum Jubilee Conference on “Gender Issues and Empowerment of Women”, held at Kolkata during 1‐2 February 2007.
20. Mason, Andrew D., Claudia E. Montenegro, and Shahidur R. Khandker (1999): “Can we say anything more about gender and poverty using household consumption data,” Poverty Reduction and Economic Management Network, The World Bank.
21. NIN (2003): “Dietary Guidelines for Indians – A Manual”, p. 1. 22. Pitt, M., M.R. Rosenzweig and M.N. Hassan (1990): "Productivity, health and inequality
in the intrahousehold distribution of food in low‐income countries", American Economic Review, 80, 1139‐1156.
23. Ray, R. and Lancaster, G. (2004): On Setting the Poverty Line Based on Estimated Nutrient Prices With Application to the Socially Disadvantaged Groups in India During the Reforms Period, Discussion Paper 2004‐09, School of Economics, University of Tasmania.
24. Song Lina (2008): “In Search of Gender Bias in Household Resource Allocation in Rural China” University of Nottingham and IZA, Discussion Paper No. 3464 April 2008.
25. Subramanian, S. and A. Deaton (1991): "Gender Effects in Indian Consumption Patterns", Sarvekshana, 14, 1 ‐ 12.
26. Udry, C. (1997): "Recent Advances in Empirical Microeconomic Research in Poor Countries: An Annotated Bibliography", Journal of Economic Education, 28(1), 58 – 75.
27. WHO (1985): Energy and protein requirements: Report of a joint FAO/WHO/UNU expert consultation. WHO Technical Report Series No. 724. Geneva.
28. WHO (1995): Physical status: The use and interpretation of anthropometry. Report of a WHO expert committee. WHO Technical Report Series No. 854. Geneva.
49
Appendix-3.1 GLOSSARY AND ABBREVIATIONS The following terms and abbreviations are relevant in this paper. These are consistent with the definitions used in other related WHO and FAO documents (FAO, 2001, 2002; James and Schofield 1990; WHO, 1985, 1995). Basal metabolic rate (BMR): The minimal rate of energy expenditure compatible with life. It is measured in the supine position under standard conditions of rest, fasting, immobility, thermoneutrality and mental relaxation. Depending on its use, the rate is usually expressed per minute, per hour or per 24 hours. Body mass index (BMI): The indicator of weight adequacy in relation to height of older children, adolescents and adults. It is calculated as weight (in kilograms) divided by height (in meters), squared. The acceptable range for adults is 18.5 to 24.9, and for children it varies with age. Doubly labelled water (DLW) technique: A method used to measure the average total energy expenditure of free-living individuals over several days (usually 10 to 14), based on the disappearance of a dose of water enriched with the stable isotopes 2H and 18O. Energy requirement (ER): The amount of food energy needed to balance energy expenditure in order to maintain body size, body composition and a level of necessary and desirable physical activity, and to allow optimal growth and development of children, deposition of tissues during pregnancy, and secretion of milk during lactation, consistent with long-term good health. For healthy, well-nourished adults, it is equivalent to total energy expenditure. There are additional energy needs to support growth in children and in women during pregnancy, and for milk production during lactation. Heart rate monitoring (HRM): A method to measure the daily energy expenditure of free-living individuals, based on the relationship of heart rate and oxygen consumption and on minute-by-minute monitoring of heart rate. Total energy expenditure (TEE): The energy spent, on average, in a 24-hour period by an individual or a group of individuals. By definition, it reflects the average amount of energy spent in a typical day, but it is not the exact amount of energy spent each and every day. Physical activity level (PAL): TEE for 24 hours expressed as a multiple of BMR, and calculated as TEE/BMR for 24 hours. In adult men and non-pregnant, non-lactating women, BMR times PAL is equal to TEE or the daily energy requirement. Physical activity ratio (PAR): The energy cost of an activity per unit of time (usually a minute or an hour) expressed as a multiple of BMR. It is calculated as energy spent in an activity/BMR, for the selected time unit. Conversion Factors: 1 joule (J) is the amount of mechanical energy required to displace a mass of 1 kg through a distance of 1 m with an acceleration of 1 m per second (1 J = 1 kg × 1 m2 × 1 sec‐2). Multiples of 1 000 (kilojoules, kJ) or 1 million (megajoules, MJ) are used in human nutrition. The conversion factors between joules and calories are: 1 kcal = 4.184 kJ, or conversely, 1 kJ = 0.239 kcal. Energy equivalents: 1 g protein = 5.65 kcal; 1 g fat = 9.25 kcal.
50
Appendix-3.2
Average Calorie Consumption of Male and Female Members of the Households: A Caution
What we have established in this chapter is that there is no difference between urban and rural sectors so far as calorie consumption of female members relative to calorie consumption of male members is concerned for each age group regardless which expenditure group is considered. In other words we can multiply calorie consumption of a female member by a constant factor to make it equivalent to a male member of the same age group and carry out our analysis without being bothered about the expenditure groups or the rural‐urban sectors. Here we further consider the average calorie consumption of the members in the household, but concentrate on the male female distribution of the same and try to show that the results of statistical analysis if not done properly may be misleading. Ratio of average calorie requirements of female members to that of male members are given in the following two tables. This ratio is around 0.92 for age groups upto age 12 years and then becomes less in the higher age groups. The overall average of Ratios is 0.92.
Table A3.2.1 Energy requirements of infants during the first year of life: FAO Boys Girls Girls/Boys Daily energy requirements
Daily energy requirements
Weight Daily energy requirements
Age in months Weight
Kg. Kcal/d Kcal/d/kg
WeightKg.
Kcal/d Kcal/d/kg - /d /d/kg (1) (2) (5) (6) (7) (10) (11) (12) (13) (14)
Age WtB CalB CalKgB WtG CalG CalKgG RatWt RatCal RatCalKg0-1 4.58 518 113 4.35 464 107 0.95 0.90 0.95 1-2 5.50 570 104 5.14 517 101 0.93 0.91 0.97 2-3 6.28 596 95 5.82 550 94 0.93 0.92 0.99 3-4 6.94 569 82 6.41 537 84 0.92 0.94 1.02 4-5 7.48 608 81 6.92 571 83 0.93 0.94 1.02 5-6 7.93 639 81 7.35 599 82 0.93 0.94 1.01 6-7 8.30 653 79 7.71 604 78 0.93 0.92 0.99 7-8 8.62 680 79 8.03 629 78 0.93 0.93 0.99 8-9 8.89 702 79 8.31 652 78 0.93 0.93 0.99
9-10 9.13 731 80 8.55 676 79 0.94 0.92 0.99 10-11 9.37 752 80 8.78 694 79 0.94 0.92 0.99 11-12 9.62 775 81 9.00 712 79 0.94 0.92 0.98
Mean* 7.72 649 86 7.20 600 85 0.93 0.92 0.99 *. Unweighted
51
Table A3.2.2. Energy requirements of boys and girls at different age groups: FAO
Boys Girls Girls/Boys
Weight Daily energy requirements
Weightkg
Daily energy requirements
Weight Daily Energy Requirements
Age years
Kg Kcal/d Kcal/d/kg Kg Kcal/d Kcal/d/kg - /d /d/kg (1) (2) (5) (6) (7) (10) (11) (12) (13) (14)
Age WtB CalB CalKgB WtG CalG CalKgG RatWt RatCal RatCalKg1-2 11.5 948 82.4 10.8 865 80.1 0.94 0.91 0.97 2-3 13.5 1129 83.6 13.0 1047 80.6 0.96 0.93 0.96 3-4 15.7 1252 79.7 15.1 1156 76.5 0.96 0.92 0.96 4-5 17.7 1360 76.8 16.8 1241 73.9 0.95 0.91 0.96 5-6 19.7 1467 74.5 18.6 1330 71.5 0.94 0.91 0.96 6-7 21.7 1573 72.5 20.6 1428 69.3 0.95 0.91 0.96 7-8 24.0 1692 70.5 23.3 1554 66.7 0.97 0.92 0.95 8-9 26.7 1830 68.5 26.6 1698 63.8 1.00 0.93 0.93
9-10 29.7 1978 66.6 30.5 1854 60.8 1.03 0.94 0.91 10-11 33.3 2150 64.6 34.7 2006 57.8 1.04 0.93 0.89 11-12 37.5 2341 62.4 39.2 2149 54.8 1.05 0.92 0.88
Mean* 22.8 1611 72.9 22.7 1484 68.7 0.98 0.92 0.94 12-13 42.3 2548 60.2 43.8 2276 52.0 1.04 0.89 0.86 13-14 47.8 2770 57.9 48.3 2379 49.3 1.01 0.86 0.85 14-15 53.8 2990 55.6 52.1 2449 47.0 0.97 0.82 0.85 15-16 59.5 3178 53.4 55.0 2491 45.3 0.92 0.78 0.85 16-17 64.4 3322 51.6 56.4 2503 44.4 0.88 0.75 0.86 17-18 67.8 3410 50.3 56.7 2503 44.1 0.84 0.73 0.88
Mean* 55.9 3036 54.8 52.1 2434 47.0 0.94 0.81 0.86 Overall Mean* 34.5 2114 66.5 33.0 1819 61.1 0.97 0.88 0.91
*. Unweighted
We can also compare FAO estimates with the corresponding ICMR estimates of the requirements. This is given in Table A3.2.3. Table A3.2.3. Energy requirements of boys and girls at different age groups: A comparison between FAO and ICMR estimates taking body weights of ICMR Boys Girls Girls/Boys
Daily energy requirements
Daily energy requirements
Daily energy requirements
Age groups
Body weight
FAO ICMR
Body weight
FAO ICMR
Body weight
FAO ICMR Kg. Kcal/d/
kg Kcal/d Kcal/d Kg. Kcal/d/
kg Kcal/d Kcal/d Kg. Kcal/d/
kg Kcal/d Kcal/d
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) 0-5m 5.4 93 502 583 5.4 92 497 583 1.00 0.99 0.99 1.00 6-11m 8.6 80 688 843 8.6 79 679 843 1.00 0.99 0.99 1.00 1-3y 12.2 82 1000 1240 12.2 79 964 1240 1.00 0.96 0.96 1.00 4-6y 19.0 75 1425 1690 19.0 72 1368 1690 1.00 0.96 0.96 1.00 7-9y 26.9 69 1856 1950 26.9 64 1722 1950 1.00 0.93 0.93 1.00
10-12y 35.4 62 2195 2190 31.5 55 1733 1970 0.89 0.89 0.79 0.90 13-15y 47.8 56 2697 2450 46.7 47 2195 2060 0.98 0.84 0.81 0.84 16-17y 57.1 51 2912 2640 49.9 44 2160 2060 0.87 0.86 0.74 0.78
52
Since ICMR takes the same weight for boys and girls up to 9 years we get the ratio as 1. Even in the higher age groups the female male ratio of calorie requirements proposed by ICMR is more than that of FAO. Since the requirements are based on actual calculations from a sample of apparently normal individuals drawn from Indian population, ICMR estimates of the ratios seem to be more reliable especially in the higher age groups. The actual ratio of consumptions as given in the Table A3.2.4 shows that the female male ratio of calorie consumption should be taken as 0.92. Table A3.2.4: Age Group wise Comparison of Estimated F/M Ratios for all Expenditure Groups: Rural, Urban and All India, 61st Round NSS data
Age Groups
Rural (F/M)
Urban (F/M)
Rural+Urban (F/M)
0‐3 yrs. 0.93 1.02 0.95
4‐6 yrs. 0.91 0.85 0.91
7‐12 yrs. 0.95 0.96 0.94
13‐18 yrs. 0.93 0.87 0.91
19 or more yrs. 0.90 1.01 0.91
Average 0.92 0.95 0.92 Source of Data: NSSO 61st Round This ratio can also be arrived at if we find the mean consumption separately for the male and female population by regressing the total calorie intake on number of male and female members in the household (Tables A3.2.5‐A3.2.8). In fact, we have found the average calorie consumption of male and female members of the households after removing the effects of rural/urban sectors and the expenditure groups. For this we had to take appropriate dummy variables for sector (rural‐urban) and the expenditure groups. The coefficients of the dummy variables are not important to us. We have taken the coefficients of numbers of male and female members as the corresponding average calorie consumptions. The four regressions correspond to the case of without and with using multipliers along with whether number of observations truncated from above. The truncation points are such that the top three expenditure groups are deleted. The results of the four tables are summarized in Table A3.2.9. It can be clearly seen from Table A3.2.9 that the ratio is 0.92.
53
54
Table A3.2.5. Average calorie consumption of male and female members of the households after removing the effects of rural/urban sectors and the expenditure groups: Without using multiplier
Unstandardized Coefficients Model
Regressors Coefficients T Sig. F,Sig.,d.f., R2
and F/M All India CONSTANT
SECTOR MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 MPCEGR10 MPCEGR11 MPCEGR12 ALLMALES ALLFEMAL
‐3462.259‐263.9921098.7921763.9572518.7162913.3793336.2813718.9034146.9614559.7445013.0675618.3796691.5972081.3391908.581
‐2788.330‐545.265718.3851330.3071929.0592228.3302543.5962865.6813214.0343565.0093915.4213983.9154813.875
12857.16311496.243
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
Sig. = .000, R‐Sq.=0.691
F/M=0.92
Rural India CONSTANT MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 MPCEGR10 MPCEGR11 MPCEGR12 ALLMALES ALLFEMAL
‐3739.6741043.0991761.2032491.9832979.9163394.0503788.2274253.4174700.5305208.2565906.9907205.4152112.1071944.923
‐2561.429572.0441115.2921601.7851915.2872162.8032441.7612764.5953077.4883400.8783481.5854322.099
11072.7389829.872
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
Sig. = .000, R‐Sq.=0.694
F/M=0.92
Urban India CONSTANT MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 MPCEGR10 MPCEGR11 MPCEGR12 ALLMALES ALLFEMAL
‐2763.5971214.2801720.6982521.1702643.7493059.6953393.4083711.9484019.5184321.2704703.2965218.4671962.2101790.352
‐1230.197441.759719.8321072.3751118.7271308.5411462.9981598.5961751.5591886.0871889.6142111.7716491.0025983.125
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
Sig. = .000, R‐Sq.=0.676
F/M=0.91
55
Table A3.2.6. Average calorie consumption of male and female members of the households after removing the effects of rural/urban sectors and the expenditure groups: Using multiplier
Model
Regressors Coefficients T Sig. F (Sig.)
All India CONSTANT SECTOR
MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 MPCEGR10 MPCEGR11 MPCEGR12 ALLMALES ALLFEMAL
‐3589.74923.888
1071.1611708.1752305.1972753.3983220.1803599.7384018.5654530.6945032.5415671.5866904.2432099.0951939.409
‐62.1071.15615.01127.73437.85145.02152.61259.49267.24077.38986.89691.512111.994293.737267.192
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
Sig. = .000, R‐Sq.=0.676 F/M = 0.924
Rural India CONSTANT MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 MPCEGR10 MPCEGR11 MPCEGR12 ALLMALES ALLFEMAL
‐4036.074917.7611621.1332319.0952811.2713283.8093684.4824106.5284656.9595227.6895989.1637464.6332153.8122003.510
‐49.2838.56017.67525.84831.65636.98942.31347.94155.21962.49568.11285.958236.077213.161
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
Sig. = .000, R‐Sq.=0.683 F/M=0.930
Urban India CONSTANT MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 MPCEGR10 MPCEGR11 MPCEGR12 ALLMALES ALLFEMAL
‐2596.8921151.9941726.7102194.6892595.2303045.5473388.8413788.6414197.8164517.3354860.6595367.1101955.6881796.315
‐34.59212.84922.10228.00332.18837.66341.53046.23652.79857.66556.02560.400173.031161.212
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
Sig. = .000, R‐Sq.=0.650 F/M=0.919
56
Table A3.2.7. Average calorie consumption of male and female members of the households after removing the effects of rural/urban sectors and the expenditure groups: Without using multiplier and last three groups deleted
Unstandardized Coefficients Model Regreesors Coefficients
t Sig. F,Sig.,d.f., R2 and F/M
All India CONSTANT SECTOR MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 ALLMALES ALLFEMAL
‐2835.229147.3311045.4361662.9142238.1852678.3103128.9813496.9943899.6574377.9201947.3211818.936
‐54.6947.03416.91331.16442.40950.51658.96166.62075.17286.101
275.627253.532
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
F=20331, Sig. = .000, D.f. = (11,87375) R‐Sq.=0.719, F/M=0.934
Rural India CONSTANT MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 ALLMALES ALLFEMAL
‐3150.974927.1431608.3772289.1832762.0483218.3193600.3284005.4524524.7441979.2851876.901
‐45.20910.43521.15930.78437.52343.73149.87356.39464.684
223.016206.387
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
F=15176, Sig. = .000, D.f. = (10,53292) R‐Sq.=0.740 F/M=0.948
Urban India CONSTANT MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 ALLMALES ALLFEMAL
‐2106.0281116.5681674.4762115.8772503.5832942.9153268.9373644.5954029.0461879.0951713.343
‐28.58913.03322.42428.23332.46538.04541.86046.44852.870
161.155146.982
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
F=7025, Sig. = .000, D.f. = (10,34073) R‐Sq.=0.673 F/M=0.912
57
Table A3.2.8. Average calorie consumption of male and female members of the households after removing the effects of rural/urban sectors and the expenditure groups: Using multiplier and last three groups deleted
Unstandardized Coefficients Model Regreesors Coefficients
t Sig. F,Sig.,d.f., R2 and F/M
All India CONSTANT SECTOR MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 ALLMALES ALLFEMAL
‐2832.488‐109.7271096.7191730.0992454.3532830.3363224.6073592.3483995.2844376.7221957.2851807.138
‐2532.002‐223.221818.0451488.4592143.8762468.5592802.6303154.8763527.3853895.694
12454.06811188.184
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
R‐Sq.=0.730, F/M=0.923
Rural India CONSTANT MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 ALLMALES ALLFEMAL
‐2991.2831053.8841736.1452436.1032902.8453294.0253675.6334112.6904529.9831971.7841830.049
‐2383.172690.7851313.9871871.1502229.2152507.4272829.7113191.4023539.090
11100.6579953.426
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
R‐Sq.=0.745 F/M=0.928
Urban India CONSTANT MPCEGR02 MPCEGR03 MPCEGR04 MPCEGR05 MPCEGR06 MPCEGR07 MPCEGR08 MPCEGR09 ALLMALES ALLFEMAL
‐2425.5401197.1761686.7922468.0012579.3602980.2143301.0333602.8343893.6071907.3801735.196
‐1046.474432.030699.8251040.5871081.6171262.6391408.9621535.0801677.2685746.0365192.029
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
R‐Sq.=0.683 F/M=0.910
58
Table A3.2.9 Ratio of calorie consumptions of female and male members of the households after removing the effects of rural/urban sectors and the expenditure groups: A Summary
Expenditure groups taken
Whether multiplier used Rural Urban All India
Without Using Multiplier 0.92 0.91 0.92 Taking all expenditure groups Using Multiplier 0.93 0.92 0.92
Without Using Multiplier 0.95 0.91 0.93 Taking only first nine expenditure groups Using Multiplier 0.93 0.91 0.92
We have seen in this chapter that the ratio of actual calorie consumption is almost same as those of FAO and ICMR for lower age group members but is prominently greater than 1 in the higher age groups. This means that the trend of the actual consumption ratio in India is not as decreasing as those of FAO and ICMR ratios of norms. The value that female male ratio of calorie consumption is 0.92 can be very much misleading though we have considered the effect of rural‐urban sectors and the expenditure groups to some extent. The effect of interactions has not been considered at all. In other words the regression should have been carried out for each expenditure group separately for rural‐urban sectors as it was done in this chapter. Lower values of the ratios are due to pooling of expenditure groups, because the higher expenditure groups have very low values of these ratios and also the relation may be nonlinear if the groups are pooled. From Table A3.2.10 (also from Table 3.10), it is clear that the ratio of calorie intakes of female to male members is 0.96, which is a much higher value than found by ignoring or not giving due consideration to the expenditure groups. So one should always take 0.96 and not 0.92 regardless whether it is rural India, urban India or all India. Table A3.2.10: Age and Expenditure Group wise Comparison of Estimated F/M Ratios for all Expenditure Groups and for bottom 9 Expenditure Groups: Rural, Urban and All India, 61st Round NSS data
Method
Age Group
0 – 3 years
4 – 6 years
7 – 12 years
13 – 18 years
19 years or more
Average
Rural 0.97 0.94 0.97 0.95 0.98 0.96 Urban 1.05 0.91 0.95 0.92 0.98 0.96 AM
All 12 Exp.
Groups All India 0.99 0.93 0.96 0.94 0.96 0.96 Rural 1.01 0.95 0.95 0.96 0.96 0.97 Urban 1.01 0.94 0.98 0.95 0.96 0.97 AM
Lowest 9 Exp.
Group All India 1.02 0.95 0.95 0.96 0.96 0.97 Rural 1.00 0.95 0.95 0.95 0.96 0.96 Urban 1.00 0.93 0.98 0.95 0.96 0.96 GM
Lowest 9 Exp.
Group All India 1.00 0.95 0.95 0.95 0.96 0.96 We can thus deflate male consumptions or inflate female consumptions accordingly and forget male female distinctions. But this is not possible. We have only the numbers of males and females in our data. Assuming that each male member has unit ‘1’, we can convert the number of females using this ratio so that the female members can be taken as equivalent to male members. The only thing we have to do is to multiply the number of female members by 0.96.
59
Chapter 4
New Methods of Finding Poverty Rate: Error Distribution Method 4.1 Introduction In this chapter we assume that the variables like calorie consumption, expenditure and the calorie norm follow certain distribution which is specific for specific variable. This assumption is quite plausible, because the values of these variables vary from one household to the other household. Let us first try to visualize the direct method in the distribution set up. Each household has a fixed age‐sex‐activity status configuration and thus Daily Per Capita Calorie Norm (DPCN) for the household is fixed. Consider all the households with this same configuration. The monthly per capita expenditure (MPCE) of the households will vary even for a fixed DPCN. Let us denote this random variable (MPCE) by X. Y is similarly a random variable representing the Daily Per Capita Calorie Intake (DPCI) of the household. This also varies from one household to another household. DPCN is fixed. However if we allow the configuration to change, i.e., take all households regardless of the configurations then this can even be treated as a random variable. Let us assume that this random variable is Z. What we do in the direct calorie method is compute Prob(Y<Z). In actual practice, for a given household, i.e., for Z=z, say, we check whether the corresponding value y of Y is less than z. If it is so then we put the value ‘1’ for the household otherwise we put the value ‘0’. Count all such 1’s with appropriate weights. This gives us the Calorie Poverty Rate by the direct method. The similar expression in the population set up is Ez[Prob(Y<Z|Z=z)]. This is in fact Prob(Y<Z). It is not clear how one should set the Poverty Line by the Direct Method. We assume that a tri‐variate distribution exist between DPCI, Daily Per Capita Calorie Norm (DPCN, i.e., Calorie norm of the household) and MPCE. We shall have to find the Probability of DPCI to be less than DPCN given MPCE and get the weighted sum of these probabilities. In other words, we should find E[Pr.(DPCI<DPCN|MPCE)]. We assume tri‐variate normal distribution for each expenditure class. Find means, variances and covariances of the three random variables. This will give us the relation among the three variables. This relation can be used to get the desired Poverty Rate. The problem of getting Poverty Line by Direct method still remains unsolved. Symbolically, we assume a tri‐variate distribution among Y, Z and X and find Prob(Y<Z) given X=x and get the weighted sum of these probabilities, i.e., Ex[Prob(Y<Z|X=x)]. If the joint density function is known, then this can be found from the following integral.
.
We assume a tri‐variate normal distribution for each expenditure class. Find means, variances and covariances of the three random variables. This will give us the relation among the three variables. This relation can be used to get the desired Poverty Rate. There are many practical problems in the above mentioned assumptions. First, the trivariate distribution is different for different expenditure class. Moreover we are assuming X to take values within a given interval. Thus the assumption of normal distribution is also questionable. It may be controversial to assume normal distribution for calorie norm too, because this is fixed for a fixed family composition and there are only a limited number of such combinations.
60
4.2 The Formulation of the Model To overcome the above mentioned problems we assume that DPCI is a linear function of MPCE except for a random error, which is assumed to be normally distributed. This is in fact found to be so for each interval of MPCE. Symbolically, we can write
yh = a + bxh + εh, for all h such that xh є (A,B),
where A and B are the boundaries of the given interval. Here, X need not be normal. What we are assuming is that yh – (a + bxh) is normal for a given xh. Similarly, we assume
yh = a' + b'zh + ε'h, for all h such that xh є (A,B).
Prob(Y<Z) for a given interval of X is the expected value of Prob(Y<Z) where the expectation is taken over all x values in the interval (A,B). Prob(Y<Z) can be approximated to Prob(a' + b'Z + ε' – Z < 0) or Prob(ε' < ‐ (a' + (b'‐1)Z)). Even if we assume that the length of the class interval of x is very small, we cannot assume that Z can be approximated by E[Z]. We can combine the above two assumptions by taking a multiple linear regression model as given below.
yh = a + bxh + czh + εh, for all h such that xh є (A,B), We can find the weighted least squares estimates , of a, b and c and then Prob(y<z|x & z) or Prob(y‐z<0|x & z), or Prob(a+bx+cz+ε‐z<0|x & z), or Prob(ε<‐(a+bx+cz‐z) |x & z), or
Φ[‐{a+bx+(c‐1)z}/ є], assuming that ε follows a normal distribution with mean ‘0’. We face the same problem of fixing x and z in practice. We can use the lower and upper boundary values of x and see if the two values differ much for a fixed value of z. We can take the mid value of the two. To fix the z value we can take the weighted mean value of z values in the given interval. All these are however approximations. Table 4.1 shows the calculations to find the poverty rates by trivariate regression method and Table 4.2 gives the poverty rates assuming MPCE values same as the (i) lower boundary, (ii) upper boundary and (iii) mean value and the calorie norm as the actual mean per capita norm of all households in the given interval taking FAO norms (PCNFAO). The rural and urban poverty rates are found as 0.77 and 0.68 taking mean MPCE and PCNFAO values of each interval. These values are at the higher ends of the values found earlier (0.71‐0.76 for rural and 0.63‐0.71 for urban India). Since FAO has the lowest calorie norms, the poverty rates by trivariate regression method using other calorie norms would be higher than these values and thus other calorie norms are not explored in this section. The poverty rates need to be adjusted, because the rates have been found taking truncated data, i.e., for data with DPCI > 100 Kcal and DPCI < 10000 Kcal. We assume that all households with DPCI < 100 Kcal are poor and all households with DPCI > 10000 Kcal are non‐poor and then
61
re‐estimate the poverty rates giving appropriate weights. The adjusted poverty rates however has been found to be same as the existing poverty rates when rounded off to two decimal places, firstly because the weights of excluded households are too small compared to the weights of households included in the analysis and secondly because the proportion of poor persons of the excluded households does not differ much.
62
Table 4.1. Results of Trivariate Regression of DPCI on MPCE and PCNFAO Separately for Rural Urban Sectors using Household Truncated Data With 100 Kcal < DPCI < 10000 Kcal and Individual Multiplier: All India, NSS 61st Round
Sector Lower Bound
Upper Bound
Mean MPCE
Mean DPCIMean
PCNFAOSD
PCNFAOMean
PCNICMRSD
PCNICMRMean PCNTF
SD PCNTF
Mean PCNEG
SD PCNEG a b c SD of
residuals (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17)
0 235 199.5 1376 2213 352.5 2368 350.8 2359 379.6 2371 345.4 153.4 4.555 .147
358.830
235 270 253.8 1574 2261 353.5 2417 356.5 2410 386.1 2420 351.6 308.1 3.425 .176
305.527
270 320 296.6 1679 2295 370.5 2445 376.9 2441 408.8 2447 372.0 690.7 2.304 .133
325.575
320 365 342.4 1799 2326 377.5 2468 385.2 2465 417.7 2469 380.0 748.3 2.035 .147
350.990
365 410 387.7 1885 2365 387.0 2511 400.1 2512 434.5 2511 395.5 596.7 2.266 .173
379.505
410 455 432.1 1962 2376 406.5 2515 421.1 2517 457.0 2514 416.1 1073.6 1.100 .172
366.726
455 510 481.6 2041 2410 392.9 2548 410.9 2551 447.2 2547 406.6 703.0 1.765 .201
401.726
510 580 543.3 2158 2437 404.1 2573 426.6 2578 463.6 2571 422.3 841.9 1.179 .268
417.731
580 690 630.4 2290 2475 411.6 2608 439.0 2615 477.8 2606 435.1 854.0 0.987 .310
499.042
690 890 775.0 2380 2469 410.4 2596 442.3 2600 482.2 2593 438.9 937.7 0.691 .363
534.773
890 1155 999.9 2568 2474 415.9 2594 452.2 2596 492.5 2591 448.9 516.4 0.962 .437
611.513
Rural
1155 NA 1956.6 3018 2444 402.3 2555 441.0 2552 477.5 2552 438.0 1209.7 0.099 .610
1012.313
0 335 279.7 1413 2097 317.3 2205 306.7 2172 327.1 2206 300.5 301.3 2.090 .258
321.756
335 395 368.2 1607 2152 344.0 2253 340.9 2226 364.7 2253 334.8 1227.0 ‐0.187 .211
322.789
Urban
395 485 441.3 1688 2151 333.4 2246 328.3 2215 349.4 2246 322.0 931.6 0.529 .243
359.787
63
485 580 533.2 1832 2194 340.6 2286 335.3 2254 355.9 2285 328.8 645.1 1.094 .257
404.367
580 675 625.8 1856 2217 327.0 2306 327.9 2274 348.5 2304 322.5 733.6 0.845 .267
381.713
675 790 730.2 1943 2213 323.8 2300 323.0 2266 340.7 2298 317.5 738.8 0.615 .341
410.383
790 930 858.0 2023 2245 309.8 2324 308.4 2289 323.4 2321 302.9 665.8 0.662 .345
411.790
930 1100 1014.3 2111 2252 298.2 2330 295.1 2293 306.8 2327 289.7 980.5 0.455 .295
566.200
1100 1380 1226.4 2209 2260 295.6 2341 295.2 2303 305.6 2338 290.3 467.7 0.757 .354
471.319
1380 1880 1594.4 2340 2295 270.7 2372 268.7 2332 274.0 2369 264.1 613.6 0.349 .503
565.151
1880 2540 2157.2 2546 2301 273.1 2381 268.7 2341 272.4 2378 263.7 509.4 0.348 .526
584.564
2540 NA 4235.6 2839 2312 260.6 2387 257.1 2346 252.5 2384 252.6 1080.3 0.015 .668
778.707
a, b and c are the regression coefficients of DPCI=a+b×MPCE+c×PCNFAO using household Truncated Data With 100 Kcal < DPCI < 10000 Kcal
64
Table 4.2: Poverty Rates by Trivariate Regression of DPCI on MPCE and PCNFAO Separately for Rural and Urban Sectors using Household Truncated Data With 100 Kcal < DPCI < 10000 Kcal and Individual Multiplier: Rural and Urban India, NSS 61st Round
Sector Lower
Boundary Upper
Boundary Mean MPCE
Weight Pov. Rate1* Pov. Rate2** Pov. Rate
(1) (2) (3) (4) (5) (6) (7) 0 235 199.5 0.048 1.00 0.97 0.99
235 270 253.8 0.051 0.99 0.98 0.99270 320 296.6 0.099 0.98 0.96 0.97320 365 342.4 0.105 0.95 0.92 0.94365 410 387.7 0.102 0.92 0.87 0.90410 455 432.1 0.093 0.89 0.86 0.87455 510 481.6 0.099 0.85 0.79 0.82510 580 543.3 0.102 0.79 0.73 0.76580 690 630.4 0.103 0.71 0.64 0.68690 890 775.0 0.098 0.62 0.51 0.57890 1155 999.9 0.050 0.51 0.35 0.44
1155 ∞ 1956.6 0.050 0.36 0.00 0.33
Rural
Total ‐ 1.000 0.81 0.74 0.790 335 279.7 0.050 1.00 0.96 0.98
335 395 368.2 0.051 0.95 0.95 0.95395 485 441.3 0.097 0.91 0.89 0.90485 580 533.2 0.104 0.87 0.81 0.84580 675 625.8 0.097 0.85 0.80 0.83675 790 730.2 0.100 0.77 0.72 0.75790 930 858.0 0.103 0.75 0.68 0.72930 1100 1014.3 0.097 0.63 0.57 0.60
1100 1380 1226.4 0.102 0.63 0.46 0.551380 1880 1594.4 0.099 0.53 0.41 0.481880 2540 2157.2 0.051 0.45 0.30 0.39
Urban
2540 ∞ 4235.6 0.049 0.32 0.00 0.31 Total ‐ 1.000 0.73 0.64 0.70*. Assuming Lower Bound of Exp. Group and Mean PCNFAO, **. Assuming Upper Bound of Exp. Group and Mean PCNFAO. If there is a calorie norm below which all households are assumed to be poor then gradual decrease of proportions of poor persons does not make any sense. But where do we put the cut‐off point. The best point should be the point where we have 50% below and also 50% above the point. This can be found by quadratic interpolation method. All the households above this cut of point should be taken as non‐poor. The logic behind this is the following. Suppose more than 50 percent of population with a given per capita income can consume food having calorie intake more than the calorie norm then the rest of the households with the same per capita income should be able to consume food as the same level as this group. By a similar logic a portion of households taken as poor should be non‐poor. This will be clear if we take the following figure.
65
Figure 4-1: A Diagrammatic Representation of Poverty Regions
In the Figure 4.1 horizontal axis represents Per Capita Expenditure and the vertical axis represents the degree of poverty. Area under the curve ACD is the poverty rate. From ACD we take the mirror image of AC as OC. So we also remove the portion under the curve OC. The actual poverty region is the portion AOC. There are methods which are independent of any assumption of distribution. This is found by subtracting 1 from twice the poverty rates of each interval. Table 4.3: Improvements of Poverty Rates by Trivariate Linear Regression of DPCI on MPCE and PCNFAO for Rural Sector using Household Truncated Data With 100 Kcal < DPCI < 10000 Kcal and Individual Multiplier: NSS 61st Round
Rural Urban Mean MPCE x
P(Y<Z|X=x)
P(Y<Z|X=x) – (1‐P(Y<Z|X=x))
(Z=897)
Wt Mean MPCE x
P(Y<Z|X=x) P(Y<Z|X=x)‐(1‐P(Y<Z|X=x)) (Z=897)
Wt
(1) (2) (3) (4) (5) (6) (7) (8) 199.5 0.99 0.98 0.048 279.7 0.98 0.96 0.050 253.8 0.99 0.98 0.051 368.2 0.95 0.90 0.051 296.6 0.97 0.94 0.099 441.3 0.90 0.80 0.097 342.4 0.94 0.88 0.105 533.2 0.84 0.68 0.104 387.7 0.90 0.79 0.102 625.8 0.83 0.66 0.097 432.1 0.87 0.74 0.093 730.2 0.75 0.49 0.100 481.6 0.82 0.65 0.099 858.0 0.72 0.43 0.103 543.3 0.76 0.53 0.102 1014.3 0.60 0.20 0.097 630.4 0.68 0.36 0.103 1226.4 0.55 0.11 0.102 775.0 0.57 0.15 0.098 1594.4 0.48 0.00 0.099 999.9 0.44 0.00 0.050 2157.2 0.39 0.00 0.051 1956.6 0.33 0.00 0.050 4235.6 0.31 0.00 0.049
Pov. Rate 0.79 0.60 1.000 Pov. Rate 0.70 0.43 1.000
Degree
of
Per Capita Expenditure
A
C
D
BO
66
Table 4.4: Improvements of Poverty Rates by Trivariate Loglinear of DPCI on MPCE and PCNFAO for Rural Sector using Household Truncated Data With 100 Kcal < DPCI < 10000 Kcal and Individual Multiplier: NSS 61st Round
Rural Urban Mean MPCE x
Poverty Rate under Log‐Normal
2×Col.(2) ‐ 1
Wt Mean MPCE X
Poverty Rate under Log‐Normal
2×Col.(6)‐1
Wt
(1) (2) (3) (4) (5) (6) (7) (8) 199.5 0.97 0.94 0.048 279.7 0.94 0.88 0.050 253.8 0.96 0.93 0.051 368.2 0.91 0.82 0.051 296.6 0.94 0.89 0.099 441.3 0.87 0.75 0.097 342.4 0.91 0.83 0.105 533.2 0.84 0.68 0.104 387.7 0.88 0.76 0.102 625.8 0.82 0.64 0.097 432.1 0.85 0.70 0.093 730.2 0.75 0.50 0.100 481.6 0.80 0.61 0.099 858.0 0.73 0.47 0.103 543.3 0.77 0.54 0.102 1014.3 0.65 0.30 0.097 630.4 0.70 0.39 0.103 1226.4 0.58 0.17 0.102 775.0 0.60 0.20 0.098 1594.4 0.51 0.03 0.099 999.9 0.47 0.00 0.050 2157.2 0.41 0.00 0.051 1956.6 0.34 0.00 0.050 4235.6 0.32 0.00 0.049
Pov. Rate 0.78 0.58 1.000 Pov. Rate 0.71 0.44 1.000 It was already noted that the calculations of poverty rates in each interval needs fixing up values of x and z. While no satisfactory solution exists on how we should fix the x values other than taking the weighted mean value of x in the given interval or taking the boundary points, there exists an alternative satisfactory solution for fixing z value. This is done by transforming calorie norms of all members in the given household into adult equivalent scale. In this case all households will have the same calorie norm which is the calorie norm of an adult member. Adult equivalent calorie intake of a household is found from the following relation.
AECI = CNA × TotCal/SumCalNorm, where, AECI stands for Adult Equivalent Calorie Intake of a Household, CNA is the calorie norm of a sedentarily active adult member in the household, TotCal is the total calorie intake of all the members in the household and SumCalNorm is the sum of calorie norms of all the members in the Household. It should be noted here that calorie norm of each member in the household depends on the age‐sex specification of the member in the household. In the subsequent calculations in this chapter we shall take the norms specified by FAO with modifications considering the average weight of all the members in the given category vis‐a‐vis the average weight of the members in the same category as found by FAO. Hence the CNA value is taken as 2367 Kcal per day for both rural and urban sectors. In this case we should take the following linear regression model.
yh = a + bxh + εh, for all h such that xh є (A,B), The weighted least squares estimates of a and b are used to find Prob(y‐z<0|x & z), or Prob( + x+ε‐z<0|x & z), or Prob(ε<‐( + x‐z)|x & z), or Φ[‐{ + x‐z}/ є], assuming that ε follows a normal distribution with mean ‘0’. We face the same problem of fixing x in this case also. We can use the lower and upper boundary values of x. We can take the mid value of the two boundary points. The best way to fix it is at the weighted mean values of x’s in the given
67
interval. The z value is already fixed at 2367 Kcal per day. The poverty rates thus found along with the improvements suggested for the tri‐variate case is given in Table 4.5. The poverty rates are now less than the corresponding poverty rates found from tri‐variate regression method. The most interesting part of this method is that we get almost same poverty rates for both rural and urban sectors.
68
Table 4.5: Regression of DPCI on MPCE Separately for Rural and Urban Sectors using Household Truncated Data With 100 Kcal < DPCI < 10000 Kcal and Individual Multiplier: Rural and Urban India, NSS 61st Round
Log‐linear Linear Exp. Group
Mean MPCE
Weight Mean
Ln(MPCE) Mean DPCI
Mean Ln(DPCI)
Mean PCN
Mean Ln(PCN) a b
SD of Res.
Pov. Rate (LN)
2×Col.(12) ‐ 1
a b SD of Res.
Pov. Rate (Normal)
2×Col(12)‐ 1
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) 01 199.5 0.048 5.291 1376 7.203 2210.2 7.688 3.073 .781 .2578 0.969 0.938 473.2 4.578 362.48 0.998 0.997 02 253.8 0.051 5.535 1574 7.341 2261.1 7.711 4.315 .547 .2095 0.960 0.921 746.9 3.264 311.80 0.986 0.973 03 296.6 0.099 5.690 1679 7.406 2295.0 7.725 4.820 .455 .2007 0.942 0.884 947.6 2.468 329.29 0.975 0.950 04 342.4 0.105 5.834 1799 7.468 2326.6 7.738 5.048 .415 .2020 0.908 0.816 1099.8 2.008 355.35 0.946 0.893 05 387.7 0.102 5.959 1885 7.521 2365.3 7.755 4.416 .521 .2027 0.876 0.752 923.1 2.480 385.36 0.915 0.830 06 432.1 0.093 6.067 1962 7.560 2375.8 7.758 6.159 .231 .1962 0.842 0.685 1499.3 1.058 373.30 0.874 0.748 07 481.6 0.099 6.176 2041 7.599 2409.8 7.773 5.034 .415 .2060 0.803 0.606 1187.7 1.766 409.44 0.826 0.653 08 543.3 0.102 6.296 2158 7.646 2436.5 7.784 5.857 .284 .2011 0.755 0.510 1496.6 1.176 431.55 0.770 0.541 09 630.4 0.103 6.444 2290 7.690 2474.9 7.799 5.725 .305 .2222 0.687 0.374 1510.0 1.161 515.02 0.696 0.393 10 775.0 0.098 6.649 2380 7.743 2468.5 7.797 6.225 .228 .2308 0.595 0.191 1810.0 .721 555.11 0.582 0.164 11 999.9 0.050 6.904 2568 7.818 2473.7 7.799 5.497 .336 .2408 0.470 0.000 1633.8 .926 637.96 0.454 0.000 12 1956.6 0.050 7.431 3018 7.914 2441.8 7.786 6.936 .132 .3183 0.340 0.000 2707.4 .0954 1041.66 0.336 0.000
All Rural ‐ 0.778 0.575 0.797 0.616 01 279.7 0.050 5.626 1413 7.231 2094.1 7.635 4.543 .478 .2648 0.935 0.871 780.9 2.304 331.72 0.984 0.968 02 368.2 0.051 5.907 1607 7.360 2151.3 7.661 7.545 ‐.0313 .2308 0.903 0.807 1637.9 ‐.073 330.78 0.953 0.906 03 441.3 0.097 6.087 1688 7.406 2151.1 7.661 6.375 .169 .2258 0.872 0.745 1433.4 .575 368.77 0.904 0.809 04 533.2 0.104 6.277 1832 7.465 2194.2 7.681 5.431 .324 .2260 0.830 0.661 1114.4 1.270 413.72 0.859 0.719 05 625.8 0.097 6.437 1856 7.503 2217.0 7.693 5.994 .234 .2141 0.816 0.632 1350.8 .806 391.60 0.832 0.665 06 730.2 0.100 6.591 1943 7.548 2212.3 7.691 6.384 .177 .2144 0.743 0.487 1601.5 .466 424.90 0.750 0.500 07 858.0 0.103 6.753 2023 7.583 2245.4 7.707 5.624 .290 .2045 0.728 0.457 1487.2 .607 425.47 0.730 0.460 08 1014.3 0.097 6.920 2111 7.624 2251.6 7.710 4.744 .416 .2343 0.645 0.290 1562.6 .535 572.98 0.648 0.297 09 1226.4 0.102 7.109 2209 7.670 2259.7 7.714 4.784 .406 .2176 0.579 0.159 1217.6 .798 482.91 0.554 0.109 10 1594.4 0.099 7.370 2340 7.723 2294.9 7.731 5.937 .242 .2306 0.518 0.036 1724.2 .376 581.28 0.487 0.000 11 2157.2 0.051 7.672 2546 7.785 2300.9 7.733 4.986 .365 .2303 0.408 0.000 1526.2 .438 602.01 0.389 0.000 12 4235.6 0.049 8.236 2839 7.857 2311.7 7.739 7.270 .0714 .2692 0.328 0.000 2623.9 .0147 796.68 0.313 0.000
All Urban ‐ 0.702 0.431 0.708 0.450
69
Chapter 5
Introducing New Activity Status 5.1 Introduction It was already pointed out that the activity patterns for many of the occupational groups have changed a lot after 1968 when the first NCO codes were introduced. It may not be proper to take the same activity status now as was taken at that time. Even in the same NCO code, activity status may be different for different persons. It is necessary to investigate it thoroughly before one fixes a new activity status to each code. Any improvement in the estimation of poverty rates through reclassification of activity status cannot make the poverty rates below the rates found by assuming all adults to be at the sedentary level. The minimum poverty rate found by assuming all adults to be sedentary is 0.50, which is also very high. We have however introduced a new activity status and applied the same to get the poverty lines and the rates. It should be taken only as an illustration. The new activity statuses are as follows.
1. Sedentary Workers: NCO Codes: 000-519, 600-619 and all others (not classified in 2 and 3). 2. Moderate Workers: NCO Codes: 520-599, 610-689, 790-809, 840-870, 880-950, 710,720,730,740,750 760, 770, 775, 776, 778, 780, 781, 789, 810, 820, 830, 960,970 and 980. 3. Heavy Workers: NCO Codes: 711-719, 721-729, 731-739, 741-749, 751-759, 761-769, 771-774, 777, 779, 782-788, 811-819, 821-829, 831-839, 872-879, 951-959, 961-969, 971-979, 981-989.
With this definition of activity status we can now compute the poverty rates. The direct method gives poverty rates as given in Table 5.1.
Table 5.1. Poverty Rates by Direct Method with New Activity Status Separately for Rural and Urban sectors Using Individual Multiplier: All India, NSS 61st Round Sector Method Used Poverty Rates FAO ICMR TF EG
Direct Calorie Poverty Rate 0.76 0.82 0.81 0.82 Direct (New Activity Status) Calorie Poverty Rate 0.64 0.71 0.68 0.71 Rural Linear Poverty Rate 0.75 0.85 0.85 0.85 Direct Calorie Poverty Rate 0.71 0.76 0.74 0.76 Direct (New Activity Status) Calorie Poverty Rate 0.67 0.71 0.69 0.71 Urban Linear Poverty Rate 0.63 0.70 0.68 0.70
These poverty rates are less in rural India and more in Urban India compared to those of Direct Method with the existing activity status. The salient feature in this new activity status is that it almost equalizes the poverty rates between rural and urban India. Manna (2007) also proposed a new grouping of NCO-1968 codes and got similar results. Observe that the Per Capita Total Expenditure in urban India is more than the Per Capita Total Expenditure in rural India. It means that the calorie intakes of members of urban households are less than the calorie intake of members of rural households given the same per capita expenditure level. 5.2 Calorie consumption by Expenditure Levels and Activity Status
70
Let us find out the mean calorie consumption of household members in the same expenditure groups separately for rural and urban sectors to verify whether it is actually so. This is given in Table 5.2. The calorie intake of urban people is found to be less than 90% of the calorie consumption of rural people in each expenditure group. Moreover, the ratio more or less decreases as per capita expenditure increases, i,e., the contrast is more in higher expenditure groups. One may raise doubt about the validity of this result by questioning whether the proportions of people in the different activity status of adult members are same in rural and urban sectors. This problem can be solved if we divide the whole population into three groups according to the activity status of heads of households and compute the mean calorie intakes as in Table 5.2. This is shown separately in the Tables 5.3, 5.4 and 5.5 for sedentarily, moderately and heavily active groups respectively. The ratios of mean calorie intakes are given in the last column of each table. The tables show the same trend of ratios in all the cases if we ignore the first few expenditure groups. Table 5.2. Comparison of Calorie Intake of members of All Rural and Urban Households by Expenditure Groups Using Individual Multiplier: All India, NSS 61st Round
Rural Urban MPCEGR
Mult PCCal MPCE Mult PCCal MPCE Urban Calorie by Rural Calorie
0 – 50 0.00019 2.09 32.97 0.00009 12.07 20.3028 5.7751 50 – 100 0.00044 175.10 73.38 0.00013 146.96 77.4093 0.8392 100 – 150 0.00327 1043.07 133.42 0.00014 479.29 127.76 0.4594 150 – 200 0.01477 1321.95 180.31 0.00237 1081.77 181.32 0.8183 200 – 250 0.04784 1496.11 228.47 0.00897 1341.70 228.83 0.8967 250 – 300 0.08755 1630.36 276.28 0.01752 1420.23 278.07 0.8711 300 – 350 0.11461 1760.50 325.49 0.03082 1524.04 325.27 0.8656 350 – 400 0.11396 1860.76 375.27 0.04580 1618.13 376.44 0.8696 400 – 450 0.10614 1947.69 424.49 0.05267 1676.19 424.96 0.8606 450 – 500 0.09200 2030.43 474.19 0.05515 1715.19 474.43 0.8447 500 – 550 0.07816 2116.54 524.14 0.05465 1870.49 524.24 0.8837 550 – 600 0.06237 2242.95 573.31 0.05497 1824.94 574.84 0.8136 600 – 650 0.05031 2279.31 623.41 0.05303 1837.71 624.17 0.8062 650 – 700 0.03977 2302.44 675.48 0.04454 1920.61 674.31 0.8341 700 – 750 0.03093 2330.39 724.03 0.04772 1931.71 724.74 0.8289 750 – 800 0.02432 2361.11 774.05 0.03767 2014.27 774.27 0.8531 800 – 850 0.02072 2452.45 824.93 0.03888 1992.30 825.23 0.8123 850 – 900 0.01638 2446.66 874.85 0.03563 2000.09 874.38 0.8174 900 – 950 0.01322 2501.32 923.74 0.03395 2062.60 924.10 0.8246 950 – 1000 0.01016 2538.27 973.92 0.02676 2060.64 975.87 0.8118
1000 & above 0.07278 2897.44 1676.73 0.35847 2368.80 1844.65 0.8175
71
Table 5.3. Comparison of Calorie Intake of members of Only Sedentary Rural and Urban Households by Expenditure Groups Using Individual Multiplier: All India, NSS 61st Round
Rural Urban MPCEGR
Mult PCCal MPCE Mult PCCal MPCE Urban Calorie by Rural Calorie
0 – 50 5274 3.10 21.54 344 .00 41.00 NA 50 – 100 31919 422.47 70.89 2390 5.92 86.58 0.0140 100 – 150 248812 993.71 135.03 15172 766.83 137.13 0.7716 150 – 200 1886121 1313.42 181.18 263475 1092.64 181.24 0.8319 200 – 250 4748720 1466.82 226.85 693099 1352.64 228.35 0.9221 250 – 300 10272087 1601.79 276.09 1469694 1351.64 277.34 0.8438 300 – 350 12726924 1787.93 325.64 2238175 1536.62 325.62 0.8594 350 – 400 13171869 1800.24 374.24 3749664 1612.31 375.91 0.8956 400 – 450 13461762 1869.46 425.92 4559378 1663.77 425.66 0.8899 450 – 500 10876397 1975.67 475.07 4562176 1717.41 475.89 0.8692 500 – 550 8358677 2066.51 523.16 4054039 1789.11 523.37 0.8657 550 – 600 7697638 2313.75 573.27 4039202 1833.14 574.78 0.7922 600 – 650 5362723 2100.65 623.25 4054581 1842.38 623.70 0.8770 650 – 700 4426018 2190.05 674.60 3374927 1878.53 674.68 0.8577 700 – 750 3933445 2197.00 724.64 3578504 1937.15 724.07 0.8817 750 – 800 3180778 2213.22 773.55 2521279 2076.19 775.01 0.9380 800 – 850 2268154 2232.66 827.05 2801845 1982.02 824.38 0.8877 850 – 900 2006746 2350.40 875.85 2314089 1980.61 872.82 0.8426 900 – 950 1602795 2424.71 920.09 2190851 2211.62 923.87 0.9121 950 – 1000 1033990 2395.21 973.18 1812464 2056.09 974.18 0.8584
1000 & above 10988367 2787.99 1763.95 20265869 2360.06 1801.43 0.8465
Table 5.4. Comparison of Calorie Intake of members of Only Moderate Rural and Urban Households by Expenditure Groups Using Individual Multiplier: All India, NSS 61st Round
Rural Urban MPCEGR
Mult PCCal MPCE Mult PCCal MPCE Urban Calorie by Rural Calorie
0 – 50 NA NA NA NA NA NA NA 50 – 100 28746 345.80 56.15 839 41.60 94.00 0.1202 100 – 150 143394 1093.48 135.67 5726 772.10 122.89 0.7060 150 – 200 990340 1226.06 180.25 230329 1160.43 181.04 0.9464 200 – 250 2674543 1405.48 227.39 794272 1281.37 227.71 0.9117 250 – 300 6040954 1575.17 277.19 1481182 1390.95 279.79 0.8830 300 – 350 9685002 1748.27 326.17 3428389 1483.58 324.34 0.8485 350 – 400 10429393 1819.63 376.01 4638857 1572.82 376.27 0.8643 400 – 450 10702725 1897.72 425.23 5788141 1635.95 424.55 0.8620 450 – 500 10112426 1966.20 475.06 6989414 1691.37 473.74 0.8602 500 – 550 9852757 1992.36 524.46 6787037 1941.03 525.36 0.9742 550 – 600 7859418 2073.03 573.76 7330188 1808.63 575.15 0.8724 600 – 650 7266314 2106.58 623.26 6810504 1795.36 624.80 0.8522 650 – 700 5888115 2140.31 676.16 5928266 1876.25 673.86 0.8766 700 – 750 5452055 2217.56 723.70 6900124 1899.13 725.06 0.8564 750 – 800 4614891 2223.22 774.61 5754126 1940.43 773.65 0.8728 800 – 850 4016196 2237.33 825.11 5749549 1965.41 825.32 0.8784 850 – 900 3166114 2317.08 875.61 5605676 1978.59 875.38 0.8539 900 – 950 2750864 2377.93 923.91 5720362 1985.41 924.22 0.8349 950 – 1000 2520364 2432.35 973.24 4123559 2042.60 976.67 0.8397
1000 & above 18187848 2731.86 1794.56 64070983 2358.52 1875.70 0.8633
72
Table 5.5. Comparison of Calorie Intake of members of Only Hard Working Rural and Urban Households by Expenditure Groups Using Individual Multiplier: All India, NSS 61st Round
Rural Urban MPCEGR
Mult PCCal MPCE Mult PCCal MPCE Urban Calorie by Rural Calorie
0 – 50 NA NA NA NA NA NA NA 50 – 100 52248 535.36 82.25 5664 713.60 99.20 1.3329 100 – 150 1923271 1077.69 133.87 NA NA NA NA 150 – 200 7943897 1337.90 180.10 85710 971.59 182.28 0.7262 200 – 250 27623008 1511.42 228.85 730160 1418.19 230.40 0.9383 250 – 300 47712107 1642.06 276.22 1376405 1530.20 276.96 0.9318 300 – 350 61464919 1756.46 325.35 1937499 1606.74 326.24 0.9147 350 – 400 59818805 1881.43 375.38 2980638 1695.45 377.35 0.9011 400 – 450 53545163 1975.81 423.96 2724925 1781.21 424.71 0.9015 450 – 500 46434919 2057.25 473.80 2137247 1789.36 473.51 0.8697 500 – 550 38965359 2158.03 524.26 2716589 1823.28 522.67 0.8448 550 – 600 30140881 2269.05 573.19 2229189 1871.68 574.09 0.8248 600 – 650 24201254 2370.14 623.48 2274942 1952.34 623.01 0.8237 650 – 700 18827219 2379.39 675.48 1672529 2118.07 674.96 0.8901 700 – 750 13248068 2419.69 724.04 1362251 2088.45 724.92 0.8631 750 – 800 10016080 2472.25 773.94 1059931 2257.62 775.72 0.9131 800 – 850 8896301 2605.60 824.32 1063577 2151.66 827.11 0.8257 850 – 900 6755915 2536.14 874.26 895739 2154.06 872.38 0.8493 900 – 950 5329777 2588.05 924.73 491746 2267.93 923.78 0.8763 950 – 1000 3865856 2646.21 974.58 682881 2164.22 975.78 0.8178
1000 or more
24016579 3075.40 1549.09 4394456 2529.31 1576.82 0.8224
If we ignore the first few expenditure groups and the last expenditure group and pool the four adjacent expenditure groups together, we get four pooled expenditure groups. The Geometric Means of these ratios for the four pooled expenditure groups and for all expenditure groups are given in Table 5.6. Table 5.6. Geometric Mean of Rural‐Urban Ratios of Calorie Intake for Four Pooled Expenditure Groups Separately for Sedentarily, Moderately and Strongly Active Households Using Individual Multiplier: All India, NSS 61st Round
MPCEGR Sedentarily Active
Households Moderately Active
Households Strongly Active Households
All Households (GM of Columns 2, 3
& 4) (1) (2) (3) (4) (5)
200 – 400 0.8796 0.8765 0.9213 0.8922 400 – 600 0.8534 0.8909 0.8597 0.8678 600 – 800 0.8881 0.8644 0.8718 0.8747 800 – 1000 0.8747 0.8515 0.8419 0.8559 200 – 1000 0.8722 0.8703 0.8701 0.8708
Though there are some variations about the movement of these ratios over expenditure groups, the overall geometric means do not differ over the activity patterns and this ratio is 0.87. It thus settles the issue that less calorie intake of urban people is not due to the variation of number of members in the different activity status and further it proves that the calorie intake of
73
urban people is about 0.87 times the calorie intake of rural people given the same expenditure group regardless the activity status.18 The following reasons can be thought of behind the differences in calorie consumptions of urban and rural people.
(i) Life styles of members of urban households in the same activity status are different from those of rural households. There are more work saving devices like transport, electronic gadgets and semi‐automatic devices easily accessible or more frequently availed by urban people.
(ii) Urban people take lower level of calorie than what is needed because of the higher preferences given to the other goods and amenities.
(iii) There may be variation of body weights between urban and rural areas. The average body weight which has been considered may need corrections downwards.
We cannot address any of these problems because of lack of data. We do not have detailed information on the lifestyles of rural and urban people. We do not have any idea about the preferences on the different goods and services by the rural and urban people. Finally we do not have any data on weights of people. Consideration of body weights may be more crucial when we calculate poverty line for each state. We can however verify some of the questions indirectly. We can verify whether there are price variations of goods between rural and urban sectors. If there are price differences, it will have impact on the choices of goods and ultimately on the intake of calories. 5.3 Price Variations by Commodity and Expenditure Groups To find the price differences between rural and urban sectors we took the following steps.
1. Found quantities consumed and values (expenditures incurred) for each item by each of the households.
2. Aggregated the quantities consumed and values over commodities in each Commodity Group. The list of commodity groups is given in Table 5.7.19
3. Aggregated these amounts over households in each of the 21 expenditure groups and 13 commodity groups separately for rural and urban sectors taking individual multipliers as weight.
18 The ratios of calorie intakes have been found to be less than 0.85 for almost all the expenditure groups. But activity status wise ratios were higher than 0.85 regardless the expenditure groups. This finding may look to be perplexing at first sight. However, this is quite plausible. This is mainly because, the proportion of people in different activity status groups are different for rural and urban sectors. To verify it one can take a hypothetical case where average calorie intake of the rural people is just 0.90 times of the average calorie intake of urban people for each activity level and the proportion of sedentarily, moderately and strongly active people as 0.5, 0.3 and 0.2 for rural and 0.2, 0.3, and 0.5 for urban sectors respectively. One can also see what happens if the proportions of people at different activity levels are just the reverse of these values. 19 It was observed that the units for quantities are different for different commodities. So aggregation of quantities needed careful scrutiny and judgment for inclusion or exclusion. It is not necessary to include all the items in a commodity groups for aggregation. Some items can safely be excluded if the units sharply differ from other units. But then the corresponding values also are to be excluded.
74
4. Took ratios of aggregated values (expenditures) to aggregated quantities to get the prices.
5. Took median prices over commodity groups. 6. Took 3 point moving average of prices over expenditure groups.
We have altogether 12 commodity groups as given in Table 5.7.
Table 5.7: Descriptions of Commodity Groups: NSS 61st Round†
Item Group Item Serials Item Descriptions Comments
Med. Price (U/R)*
01 101 – 108, 110 – 118,
120 – 122 & 139 Cereals All units for quantities are in Kg. 1.11
02 140 – 148, 150 – 153 Pulses All units for quantities are in Kg. 1.08
03 160 – 165 & 167 Milk and milk products Only values are given for item 166 (ice cream). Mixed units.
1.21
04 170 – 174 Oil (edible) All units for quantities are in Kg. 1.02
05 181 – 186 Non‐vegetables All units for quantities are in Kg. except 180 (egg).
0.98
06 190 – 198, 200 – 208, 210 – 217, 220 & 224
Vegetables All units for quantities are in Kg. except 221 (lemon)
1.01
07 230 – 238, 240 – 246 Fruits Mixed units 1.04 08 250 – 257 Dry fruits All units for quantities are in Kg. 0.78
09 260 – 264 Sugar and sugar products
All units for quantities are in Kg. 1.03
10 280 – 286 Spices All units for quantities are in gm. 1.13 11 290 – 298, 300 – 308 Beverages Mixed units 0.95
12 310 – 315, 320 – 327,
330 – 335 Pan, liquor and other intoxicants
Mixed units 1.29
*. From Table 5.10. †. There are some minor variations in the item numbers taken for 55th round of NSS data. In the 6th group item nos. 220 & 224 have been replaced by item nos. 218 & 222. In the 10th group two new item nos. namely 287 & 288 have been added.
The rural and urban prices and their corresponding ratios by commodity as well as expenditure groups are given in the following three tables.
75
Table 5.8: Rural Prices by expenditure and commodity groups: NSS 61st Round, Truncated Data ExpGr Multiplier G01 G02 G03 G04 G05 G06 G07 G08 G09 G10 G11 G12 200 – 250 35075474 0.7338 2.182 1.0812 5.237 4.350 0.5750 0.2181 1.014 1.761 0.00555 0.0382 0.01723250 – 300 64186552 0.7553 2.368 1.0616 5.211 4.756 0.5863 0.2023 3.407 1.781 0.00562 0.0382 0.01699300 – 350 84021199 0.7805 2.389 1.1190 5.266 4.816 0.5881 0.2030 0.614 1.780 0.00583 0.0377 0.01831350 – 400 83559369 0.7909 2.462 1.1234 5.289 4.795 0.6138 0.2049 1.936 1.779 0.00591 0.0372 0.01940400 – 450 77812665 0.8093 2.502 1.1541 5.258 5.186 0.6213 0.2155 2.034 1.783 0.00601 0.0372 0.02103450 – 500 67443298 0.8156 2.523 1.1680 5.309 5.337 0.6271 0.2198 2.646 1.784 0.00610 0.0375 0.02136500 – 550 57303255 0.8307 2.553 1.1900 5.285 5.387 0.6293 0.2259 2.237 1.787 0.00621 0.0383 0.02149550 – 600 45727165 0.8502 2.572 1.1280 5.248 5.447 0.6415 0.2241 2.363 1.782 0.00624 0.0378 0.02054600 – 650 36882886 0.8605 2.566 1.1970 5.161 5.453 0.6849 0.2368 0.986 1.788 0.00634 0.0351 0.02273650 – 700 29161497 0.8643 2.626 1.2036 5.336 5.832 0.6717 0.2440 2.684 1.799 0.00637 0.0403 0.02181700 – 750 22679420 0.8757 2.611 1.2072 5.304 3.981 0.6801 0.2461 1.771 1.773 0.00632 0.0414 0.02435750 – 800 17830967 0.8773 2.599 1.2189 5.343 5.804 0.7003 0.2574 1.657 1.786 0.00631 0.0439 0.02480
Table 5.9: Urban Prices by expenditure and commodity groups: NSS 61st Round, Truncated Data
ExpGr Multiplier G01 G02 G03 G04 G05 G06 G07 G08 G09 G10 G11 G12 200 – 250 2230616 0.7921 2.524 1.1864 5.246 4.295 0.6094 0.2013 3.936 1.844 0.00590 0.0410 0.02587250 – 300 4355382 0.8220 2.501 1.3358 5.298 3.903 0.5668 0.1866 0.286 1.791 0.00624 0.0397 0.02186300 – 350 7659659 0.8788 2.549 1.3475 5.312 4.181 0.5877 0.1913 3.973 1.859 0.00661 0.0355 0.02146350 – 400 11383593 0.8874 2.652 1.3649 5.367 4.640 0.6276 0.2075 3.743 1.868 0.00652 0.0354 0.02136400 – 450 13089387 0.8945 2.689 1.3490 5.166 4.752 0.6499 0.2091 1.770 1.832 0.00682 0.0346 0.02153450 – 500 13706524 0.9093 2.715 1.4930 5.457 4.847 0.6529 0.2060 1.900 1.807 0.00689 0.0351 0.02373500 – 550 13582091 0.9174 2.766 1.3470 5.266 5.405 0.5907 0.2430 0.162 1.828 0.00700 0.0346 0.02744550 – 600 13660986 0.9230 2.796 1.3597 5.354 4.892 0.6441 0.2431 0.797 1.832 0.00712 0.0347 0.02914600 – 650 13178241 0.9616 2.790 1.4597 5.411 5.942 0.6844 0.3061 5.144 1.855 0.00735 0.0399 0.03111650 – 700 11069244 0.9637 2.763 1.5100 5.506 6.124 0.7653 0.2595 2.303 1.869 0.00743 0.0393 0.03692700 – 750 11859303 0.9788 2.751 1.5648 5.393 5.607 0.6950 0.2700 0.379 1.859 0.00727 0.0436 0.03132750 – 800 9361904 0.9677 2.830 1.4802 5.291 5.894 0.5900 0.2782 0.518 1.840 0.00763 0.0378 0.03965
76
Table 5.10: Ratios of Rural Prices to Urban Prices by expenditure and commodity groups: NSS 61st Round, Truncated Data
ExpGr Multiplier G01 G02 G03 G04 G05 G06 G07 G08 G09 G10 G11 G12 Median (R/U)
Median (U/R)
200 – 250 2230616 0.9263 0.8645 0.9113 0.9982 1.0128 0.9435 1.0834 0.2576 0.9549 0.9406 0.9317 0.6660 0.94 1.07
250 – 300 4355382 0.9188 0.9468 0.7947 0.9835 1.2185 1.0344 1.0841 11.9125 0.9944 0.9006 0.9622 0.7772 0.97 1.03
300 – 350 7659659 0.8881 0.9372 0.8304 0.9913 1.1518 1.0006 1.0611 0.1545 0.9575 0.8819 1.0619 0.8532 0.95 1.06
350 – 400 11383593 0.8912 0.9283 0.8230 0.9854 1.0334 0.9780 0.9874 0.5172 0.9523 0.9064 1.0508 0.9082 0.94 1.06
400 – 450 13089387 0.9047 0.9304 0.8555 1.0178 1.0913 0.9559 1.0306 1.1491 0.9732 0.8812 1.0751 0.9767 0.97 1.03
450 – 500 13706524 0.8969 0.9292 0.7823 0.9728 1.1010 0.9604 1.0669 1.3926 0.9872 0.8853 1.0683 0.9001 0.97 1.03
500 – 550 13582091 0.9054 0.9229 0.8834 1.0036 0.9966 1.0653 0.9296 13.8086 0.9775 0.8871 1.1069 0.7831 0.95 1.05
550 – 600 13660986 0.9211 0.9198 0.8295 0.9802 1.1134 0.9959 0.9218 2.9648 0.9727 0.8764 1.0893 0.7048 0.95 1.05
600 – 650 13178241 0.8948 0.9197 0.8200 0.9537 0.9177 1.0007 0.7736 0.1916 0.9638 0.8625 0.8796 0.7306 0.89 1.13
650 – 700 11069244 0.8968 0.9504 0.7970 0.9691 0.9523 0.8776 0.9402 1.1654 0.9625 0.8573 1.0254 0.5907 0.94 1.06
700 – 750 11859303 0.8946 0.9491 0.7714 0.9834 0.7100 0.9785 0.9114 4.6728 0.9537 0.8693 0.9495 0.7774 0.93 1.07
750 – 800 9361904 0.9065 0.9183 0.8234 1.0098 0.9847 1.1869 0.9252 3.1988 0.9706 0.8269 1.1613 0.6254 0.95 1.05
Median (R/U) 0.90 0.93 0.82 0.98 1.02 0.99 0.96 1.28 0.97 0.88 1.06 0.78 0.94 ‐
Median (U/R) 1.11 1.08 1.21 1.02 0.98 1.01 1.04 0.78 1.03 1.13 0.95 1.29 ‐ 1.06
77
Apart from some random fluctuations there is an increasing trend of prices over total per capita expenditure in both rural and urban areas. To summarize, we have taken the median values of the ratios to minimize the error due to outlying observations if any. Neglecting the bottom and top few expenditure classes which show erratic behavior, the rural urban ratio of prices have been found to be around 0.94 when median is taken. The table shows that the median price ratios do not vary over expenditure groups, whereas these values vary over the commodity groups. The urban prices compared to rural prices are high in cereals, pulses, milk and milk products, and spices. The urban prices are low for dry fruits and beverages. These are comparatively low calorie goods. There are thus two features which emerge from this analysis. As total per capita expenditure increases people move to higher quality goods and thus price increases for both rural and urban sectors. Rural prices compared to urban prices do not have any trend. Moreover the rural urban ratios of prices are close to 0.94. We have also found the ratios of prices of the 61st round to the prices of 55th round data of NSSO by expenditure and item wise groups. These are given in the Tables 5.11 and 5.12. There is no trend of these price ratios over the expenditure groups. It seems the prices increased by the same ratio for all the expenditure groups. But there are variations in the increases of the prices if seen for the item groups. The average of the price ratios are approximately 1.19 and 1.17 for rural and urban sectors respectively, whereas the medians of these ratios are 1.17 and 1.14 respectively for rural and urban sectors. Since the mean values are very much sensitive to the extreme observations, medians are preferable especially when the data are in terms of ratios. Thus the rural prices have on the average increased by 1.17 times and the urban prices have increased by 1.14 times. Urban prices have increased less compared to rural prices. Here it should be noted that the comparison has not been made with similar categories of people in the strict sense. If we assume that there has been a price rise during the period between the two quinquennial rounds, then the group of persons must have reached a higher category of expenditure groups or at least a portion of people have changed the expenditure group moving to a higher category of expenditure groups. So the actual price ratios are expected to be higher than these values, but by almost same proportion, because the price ratios do not vary over expenditure groups. The price ratios NSS 61st round compared to NSS 55th round is very high for edible oil and sugar and sugar products. These are the high calorie items. Thus these price ratios explain to some extent why the calorie consumptions are decreasing over time and also why calorie consumption of urban people is lower than the calorie consumption of rural people.
78
Table 5.11: Ratios of Rural Prices of 61st Round to Rural Prices of 55th Round by Expenditure and Commodity Groups: Truncated Data ExpGr G01 G02 G03 G04 G05 G06 G07 G08 G09 G10 G11 G12 Median Mean
200 – 250 0.940 1.044 1.109 1.396 1.203 1.309 1.270 0.841 1.298 1.151 1.486 0.951 1.177 1.167 250 – 300 0.944 1.151 1.043 1.476 1.605 1.319 1.193 1.179 1.287 1.156 1.264 0.911 1.186 1.211 300 – 350 0.962 1.073 1.125 1.431 1.246 1.230 1.215 0.487 1.280 1.180 1.375 1.017 1.197 1.135 350 – 400 0.951 1.086 1.097 1.401 1.402 1.240 1.178 0.821 1.275 1.154 1.323 1.088 1.166 1.168 400 – 450 0.963 1.079 1.096 1.469 1.219 1.232 1.207 1.802 1.309 1.180 1.362 1.150 1.213 1.256 450 – 500 0.969 1.085 1.103 1.459 1.267 1.181 1.130 1.206 1.301 1.182 1.311 1.092 1.182 1.190 500 – 550 0.967 1.071 1.150 1.484 1.306 1.147 1.171 1.645 1.278 1.187 1.377 1.102 1.179 1.240 550 – 600 0.969 1.116 1.023 1.508 1.254 1.126 1.064 3.240 1.354 1.134 1.285 0.942 1.130 1.335 600 – 650 0.979 1.326 1.087 1.374 1.290 1.187 1.062 0.305 1.341 1.174 1.193 1.046 1.180 1.114 650 – 700 0.982 1.127 1.069 1.408 1.301 1.130 NA 1.565 1.274 1.153 1.279 0.932 1.153 1.202 700 – 750 1.001 1.110 1.079 1.435 0.905 1.135 1.050 0.703 1.244 1.130 1.361 0.957 1.094 1.092 750 – 800 0.995 1.080 1.078 1.524 1.285 1.147 1.134 0.750 1.329 1.130 1.298 0.948 1.132 1.142 Median 0.968 1.086 1.091 1.447 1.276 1.184 1.171 1.010 1.293 1.155 1.317 0.987 1.172 ‐ Mean 0.969 1.112 1.088 1.447 1.274 1.199 1.152 1.212 1.297 1.159 1.326 1.011 ‐ 1.188
Table 5.12: Ratios of Urban Prices of 61st Round to Urban Prices of 55th Round by Expenditure and Commodity Groups: Truncated Data ExpGr G01 G02 G03 G04 G05 G06 G07 G08 G09 G10 G11 G12 Median Mean
200 – 250 0.967 1.057 1.111 1.407 1.374 1.366 1.259 1.372 1.341 1.163 1.614 1.243 1.300 1.273 250 – 300 0.982 1.050 1.494 1.498 1.050 1.100 1.176 0.099 1.277 1.206 1.407 1.109 1.143 1.121 300 – 350 NA 1.038 1.120 1.423 1.118 1.145 1.052 4.641 1.319 1.285 1.163 0.865 1.145 1.470 350 – 400 1.016 1.042 1.123 1.419 1.135 1.098 1.128 1.358 1.552 1.258 1.120 0.936 1.126 1.182 400 – 450 1.001 1.131 1.110 1.695 1.217 1.121 1.075 0.646 1.422 1.289 1.074 0.928 1.115 1.142 450 – 500 0.977 1.194 1.209 1.615 1.138 1.047 1.067 0.921 1.325 1.273 1.080 0.895 1.109 1.145 500 – 550 0.980 1.201 1.061 1.584 1.286 0.924 1.243 0.072 1.328 1.294 1.039 0.902 1.131 1.076 550 – 600 0.966 1.207 1.063 1.643 1.079 1.004 1.189 0.252 1.332 1.285 1.042 0.977 1.071 1.086 600 – 650 0.984 1.094 1.107 1.559 1.211 0.983 1.377 1.444 1.438 1.283 1.170 1.047 1.190 1.224 650 – 700 0.963 1.141 1.132 1.691 1.246 1.102 1.278 0.787 1.371 1.292 1.135 1.070 1.138 1.184 700 – 750 0.970 1.278 1.238 1.595 1.236 0.980 1.143 0.108 1.356 1.278 1.178 0.910 1.207 1.105 750 – 800 0.949 1.161 1.042 1.599 1.211 0.795 1.237 0.168 1.330 1.291 1.018 1.048 1.104 1.071 Median 0.977 1.136 1.115 1.589 1.211 1.073 1.182 0.716 1.336 1.280 1.128 0.957 1.141 ‐ Mean 0.978 1.133 1.151 1.561 1.192 1.056 1.185 0.989 1.366 1.264 1.170 0.994 ‐ 1.171
79
Chapter 6
State Wise Poverty Rates 6.1 Introduction We now turn to the problem of estimating the poverty rates of each state in India. The methods which have been used for estimating the poverty rates at all India level, should be applicable for estimating the same in each state. For example, the calorie poverty rates found by direct method using the FAO norm is given in the Table 6.1. Table 6.1: State Wise Calorie Poverty Rates by Direct Method Using the FAO Norms and Introducing the New Activity Status: NSS 61st Round
Rural Urban State Code
State Pov Weight Food/Tot Pov Weight Food/Tot
01 Jammu & Kashmir 0.47 5064954 0.553 0.46 1705217 0.494 02 Himachal Pradesh 0.46 5557780 0.507 0.33 580729 0.423 03 Punjab 0.55 15707309 0.494 0.57 7449622 0.398 04 Chandigarh 0.58 90308 0.546 0.49 793606 0.375 05 Uttaranchal 0.52 6372994 0.534 0.50 1943805 0.492 06 Haryana 0.54 15821343 0.503 0.61 5742439 0.429 07 Delhi 0.64 839487 0.483 0.64 11578574 0.429 08 Rajasthan 0.51 42977141 0.548 0.60 12318850 0.434 09 Uttar Pradesh 0.52 132536421 0.538 0.61 32414301 0.466 10 Bihar 0.55 66754159 0.648 0.53 6810931 0.548 11 Sikkim 0.77 446464 0.555 0.74 56803 0.435 12 Arunachal Pradesh 0.56 771324 0.502 0.58 99823 0.513 13 Nagaland 0.69 572126 0.542 0.40 237935 0.455 14 Manipur 0.63 1451655 0.550 0.49 469116 0.494 15 Mizoram 0.64 427979 0.574 0.60 278872 0.468 16 Tripura 0.88 2751130 0.637 0.64 448806 0.504 17 Meghalaya 0.79 1805288 0.560 0.69 277007 0.410 18 Assam 0.63 22897503 0.662 0.58 2336499 0.515 19 West Bengal 0.65 59616906 0.597 0.69 19319991 0.451 20 Jharkhand 0.64 20342724 0.617 0.47 3910099 0.479 21 Orissa 0.65 32108075 0.617 0.63 5082848 0.520 22 Chhattisgarh 0.72 18192303 0.560 0.62 3290988 0.414 23 Madhya Pradesh 0.72 46018428 0.526 0.69 14069201 0.411 24 Gujrat 0.75 30935588 0.587 0.70 16283678 0.457 25 Daman & Diu 0.90 107005 0.432 0.83 57952 0.460 26 Dadra & Nagar Haveli 0.89 181421 0.549 0.65 24245 0.404 27 Maharastra 0.77 55114230 0.520 0.78 37218600 0.419 28 Andhra Pradesh 0.73 54227197 0.556 0.72 18642349 0.428 29 Karnataka 0.83 34112160 0.568 0.72 15167630 0.453 30 Goa 0.86 670764 0.504 0.81 402822 0.331 31 Lakshadweep 0.33 29279 0.467 0.29 28769 0.478 32 Kerala 0.72 23560213 0.458 0.68 7230318 0.414 33 Tamilnadu 0.82 34508294 0.531 0.72 21563536 0.434 34 Pondicherri 0.76 310565 0.482 0.72 568093 0.457 35 Andaman & Nicober 0.72 196654 0.520 ‐ ‐ ‐
80
All India 0.64 0.71 These values are difficult to believe in the relative as well as in the absolute sense. Most of the south Indian states show very high values of the poverty rates. Also these values are relatively higher than those of the states known to be among the poorest states in India. For example rural Bihar has the rural poverty rate as 0.55, whereas the states like Karnataka and Tamilnadu have poverty rates higher than 0.80. Urban poverty rate is seen to be higher than rural poverty rates for a few states namely, Punjab, Haryana, Rajsthan, Uttar Pradesh, Maharastra, West Bengal and Arunachal Pradesh. Since these states have higher weights the overall poverty for urban India is found to be higher than rural India, though the difference is not much. There is a negative correlation between proportion of expenditure on total food to the poverty rates (Table 6.2). We have also found the correlation of poverty rates with proportion of expenditures on different item groups in Table 6.2. We would expect positive correlations with necessary food items. But this feature is not reflected in the correlation table. It seems there is also an effect of relative prices, i.e., price of one commodity relative to price of other commodity. The prediction of poverty rates through linear regression using state wise proportions of expenditures on food item groups does not lead to encouraging result (Table 6.3). Because of lack of sufficient data some of the states have been omitted for urban sector. Table 6.2: Correlation of the state wise poverty indices with the state wise proportions of expenditures of each group of items: NSS 61st Round
Food Items With proportion of expenditure of each item to the total expenditure
With proportion of expenditure of each item to the total food expenditure
Rural Urban Rural Urban Total Food ‐0.126 ‐0.506 ‐‐‐ ‐‐‐ Item Group 01 ‐0.100 ‐0.543 ‐0.044 ‐0.204 Item Group 02 0.111 ‐0.518 0.193 ‐0.104 Item Group 03 ‐0.531 ‐0.425 ‐0.513 ‐0.223 Item Group 04 0.207 ‐0.220 0.357 0.195 Item Group 05 0.285 0.036 0.348 0.217 Item Group 06 ‐0.167 ‐0.427 ‐0.169 ‐0.215 Item Group 07 0.663 ‐0.360 0.470 0.496 Item Group 08 0.327 ‐0.273 0.329 ‐0.045 Item Group 09 ‐0.233 ‐0.428 ‐0.184 ‐0.229 Item Group 10 0.357 ‐0.484 0.491 0.067 Item Group 11 0.596 ‐0.325 0.612 0.605 Item Group 12 0.280 ‐0.333 0.360 0.439
81
Table 6.3: Prediction of Poverty Rates Through Linear Regression Using Item Wise Food Expenditure Relative to Total Expenditure and Relative to Total Food Expenditure: NSS 61st Round
Rural Urban
State Code
State Name Pov.
Prediction: Relative to
Total Expenditure
Prediction: Relative to Total Food Expenditure
Pov.
Prediction: Relative to
Total Expenditure
Prediction: Relative to Total Food Expenditure
01 Jammu & Kashmir 0.47 0.59 0.53 0.46 0.55 0.56 02 Himachal Pradesh 0.46 0.53 0.47 0.33 0.51 0.51 03 Punjab 0.55 0.56 0.57 0.57 0.62 0.61 04 Chandigarh 0.58 0.50 0.50 0.49 0.55 0.54 05 Uttaranchal 0.52 0.58 0.58 0.50 ‐ ‐ 06 Haryana 0.54 0.44 0.43 0.61 0.60 0.61 07 Delhi 0.64 0.55 0.46 0.64 0.60 0.61 08 Rajasthan 0.51 0.51 0.53 0.60 0.59 0.59 09 Uttar Pradesh 0.52 0.57 0.56 0.61 0.60 0.60 10 Bihar 0.55 0.52 0.53 0.53 0.52 0.52 11 Sikkim 0.77 0.56 0.52 0.74 0.73 0.74 12 Arunachal Pradesh 0.56 0.63 0.60 0.58 0.62 0.61 13 Nagaland 0.69 0.44 0.42 0.40 0.62 0.60 14 Manipur 0.63 0.58 0.54 0.49 0.49 0.50 15 Mizoram 0.64 0.62 0.65 0.60 0.60 0.61 16 Tripura 0.88 0.65 0.65 0.64 0.61 0.58 17 Meghalaya 0.79 0.83 0.86 0.69 0.58 0.61 18 Assam 0.63 0.61 0.65 0.58 0.59 0.59 19 West Bengal 0.65 0.67 0.65 0.69 0.69 0.69 20 Jharkhand 0.64 0.63 0.65 0.47 0.62 0.62 21 Orissa 0.65 0.66 0.67 0.63 0.58 0.58 22 Chhattisgarh 0.72 0.69 0.67 0.62 ‐ ‐ 23 Madhya Pradesh 0.72 0.65 0.65 0.69 ‐ ‐ 24 Gujrat 0.75 0.74 0.75 0.70 0.70 0.70 25 Daman & Diu 0.90 0.84 0.86 0.83 0.43 0.46 26 Dadra & Nagar Haveli 0.89 0.73 0.71 0.65 0.67 0.66 27 Maharastra 0.77 0.74 0.75 0.78 ‐ ‐ 28 Andhra Pradesh 0.73 0.76 0.75 0.72 ‐ ‐ 29 Karnataka 0.83 0.84 0.85 0.72 0.72 0.72 30 Goa 0.86 0.98 0.91 0.81 ‐ ‐ 31 Lakshadweep 0.33 1.01 0.97 0.29 ‐ ‐ 32 Kerala 0.72 0.70 0.72 0.68 ‐ ‐ 33 Tamilnadu 0.82 0.78 0.79 0.72 ‐ ‐ 34 Pondicherri 0.76 0.79 0.80 0.72 ‐ ‐ 35 Andaman & Nicober 0.72 ‐ ‐ ‐ ‐ ‐ R‐Sq 0.847 0.866 0.694 0.700
82
Chapter 7
Findings of the Study and Discussions
7.1 Introduction The entire study is based on NSS data as supplied by the Ministry. NSS data need rigorous scrutiny before being applied in a fruitful manner. For example, we had to delete all the households with DPCI ≤ 100 Kcal or ≥ 10000 Kcal to carry out regression analysis of DPCI on MPCE with meaningful and stable result. We have used the calorie norms supplied by the Task Force, Expert Group, ICMR and FAO. It suffices to compare only ICMR and FAO estimates. The calorie poverty rates by direct method are always higher than the fixed calorie line method. The variation of calorie consumption much below the calorie line will not affect the calorie poverty rate. So is the variation of calorie consumption much above the calorie line. Only marginal households, i.e., the households with actual calorie consumption close to the calorie line, will affect the calorie poverty rate. The number of marginal households is more in the direct methods than in the fixed calorie line methods. Thus the poverty rate by the direct method is more sensitive to changes in the consumption. Moreover, it is less likely for the marginal households to increase the consumption, whereas those who are supposed to lie above the calorie line may have reasons to consume less. The net effect is the increase of calorie poverty rate by the direct method compared to the fixed calorie line method. The calorie line of the household may be very much different from the fixed calorie line because the age‐sex‐activity status of the household may be much different from the average age‐sex‐activity pattern of all the households. The direct method thus seems to be superior to the fixed calorie line method in this respect. Observe that calculation of calorie poverty rates by direct method does not need any weighing diagram of the population. This is automatically taken care of by the multiplier of each member (may be termed as individual multiplier), which is the product of the household multiplier and the household size. Urban poverties are found to be more than corresponding rural poverties when activity levels of adults are not considered. This does not seem to be probable. There are mainly two reasons for differences in the poverty rates between rural and urban sectors. The first is the differences of consumptions due to the differences of incomes. The MPCE of urban households is certainly more than the MPCE of rural households and it is expected that the households with more income will consume more food. But our findings nullify it. The second reason is that the differences of consumptions are due to differences in the activity status. Our findings support it. Calorie poverty rates show an increasing trend whichever method is used except for urban sector during 50th and 55th rounds of NSS. One of the reasons is due to the change in the activity status over time, which is not considered at all. The correspondence between National Classification of Occupation made in 1968 (NCO‐1968) and the activity status has undergone a sea change. The life styles have changed very much due to the introduction of many work‐ and time‐saving devices. Many new commodities have come into the market. The tests and preferences on the commodities by the people have changed. The workers who were designated as hard workers have possibly ceased to be so. So are the moderate workers. And this is reflected in the trend of Calorie Poverty Rates. We also apprehend that many of the members, who were designated as sedentary workers by NCO‐1968, are now leading a sub‐sedentary life. We have found the Poverty Rates using both linear and quadratic methods of regression of DPCI on MPCE for each class interval. The linear and quadratic methods almost give same result. The two results would have been more close if we take the length of each class interval less than what has been taken here or equivalently increase the number of expenditure classes.
83
The Poverty Rates found by the Fixed Poverty Line Method using linear interpolation are higher than the Calorie Poverty Rates found by Calorie Line Methods for rural India. Almost the opposite is the case for urban India. This anyway is not a solace to us given the fact that all the rates are too high to be acceptable. Except for higher income groups, there are no differences in the calorie consumptions of female members relative to that of male members between rural and urban sectors in India. We can thus deflate male consumptions or inflate female consumptions accordingly and forget male female distinctions. Assuming that each male member has unit ‘1’, we can convert the number of females using this ratio so that the female members can be taken as equivalent to male members. The only thing we have to do is to multiply the number of female members by 0.96. Two entirely new methods have been proposed in this report – Calorie Decomposition Method and the Error Distribution Method. Poverty Rates found by both the methods are higher than expected. Some modifications of the error decomposition method have also been proposed. The modifications lead to better result in the sense that the poverty rates are considerably lower than the other methods using calorie intakes. It may be possible by this method to compensate for the decreasing trends in the calorie consumption by choosing the appropriate cut off points. The logic behind choosing a cut‐off point is simple. One may argue that the best point should be the point, which divides the population of the concerned interval equally. Suppose more than 50 percent of population with a given per capita income can consume food having calorie intake more than the calorie norm then the rest of the households with the same per capita income should be able to consume food at the same level as this group. By a similar logic, for other lower MPCE intervals, a portion of households taken as poor should be non‐poor. But this portion should be estimated by assuming a suitable distribution. The cut‐off point need not be based on 50:50 criteria. It may be, for example, based on 40:60 or 60:40 criteria. We have regrouped the NCO‐1968 codes according to the activity status and calculated the poverty rates by direct method. It improved the estimates to some extent. The noteworthy feature of this regrouping is that the rural and urban poverties become almost equal.20 We have also seen whether the there is a variation of calorie intakes between rural and urban sectors separately for each activity status. It is seen that less calorie intake of urban people is not due to the variation of number of members in the different activity status and further it proves that the calorie intake of urban people is about 0.87 times the calorie intake of rural people given the same expenditure group regardless the activity status. Apart from some random fluctuations there is an increasing trend of prices over total per capita expenditure in both rural and urban areas. Ignoring the bottom and top few expenditure classes which show erratic behaviour, the rural urban ratio of prices have been found to be around 0.94 when median is taken. As total per capita expenditure increases people move to higher quality goods and thus price increases for both rural and urban sectors. Rural prices compared to urban prices do not have any trend. Though the price ratios remain more or less same over expenditure groups, but these values vary over the commodity groups. The urban prices compared to rural prices are high in cereals, pulses, milk and milk products, and spices. The urban prices are low for dry fruits and beverages. These are comparatively low calorie goods. We have also found the ratios of prices of the 61st round to the prices of 55th round data of NSSO by expenditure and item wise groups. There is no trend of these price ratios over the expenditure groups. But there are variations in the increases of the prices if seen for the item groups. The average of the price ratios are approximately 1.19 and 1.17 for rural and urban sectors respectively, whereas the medians of these ratios are 1.17 and 1.14 respectively for rural and urban sectors. The price ratios NSS 61st round compared to NSS 55th round is very high for edible oil and sugar and sugar products. These are the high calorie items. Thus these price ratios may explain to some extent why the calorie consumptions are decreasing over time and also why calorie consumption of urban people is lower than the calorie consumption of rural people. 20 Manna (2007) also proposed a new grouping of NCO-1968 codes and found similar results.
84
Most of the south Indian states show very high values of the poverty rates. Also these values are relatively higher than those states known to be the poorest states in India. For example Bihar has the rural poverty rate as 0.55, whereas the states like Karnataka and Tamilnadu have poverty rates higher than 0.80. Urban poverty rate is seen to be higher than rural poverty rates for a few states namely, Punjab, Haryana, Rajsthan, Uttar Pradesh, Maharastra, West Begal and Arunachal Pradesh. Since these states have higher weights the overall poverty for urban India is found to be higher than rural India, though the difference is not much. The consumer expenditure of a household on food items in NSS data relates to the actual consumption by the members of the household and also by the guests during ceremonies or otherwise. Thus, to avoid double counting, cooked meals received (and not purchased from market) are not recorded in the recipient household. This procedure only leads to bias-free estimates of average per capita consumption as well as total consumer expenditure. However, it does not give unbiased estimates of the nutritional status of households and consequently any estimate based on the nutritional status may be erroneous. An adjustment to nutrient intakes on the basis of the meal accounting suggested by Minhas (1991) is called for. The suggested adjustment by Minhas is given by
Ca = C × (Mh + Mf)/(Mh + Mg + Me), where Mh = number of meals consumed by the household member in the household, Mg = number of meals consumed by guests, Me = number of meals consumed by employees and Mf = number of meals received by members of households free as guest or as employees. The poverty rates using the adjusted calorie as suggested by Minhas along with the corresponding poverty rates with unadjusted calories are given in Table 7.1. Table 7.1 Poverty Rates with and without adjusted Calorie: A Comparison
Without Using Adjusted Calorie Using Adjusted Calorie All Sedentary With Activity Status All Sedentary With Activity Status
Method Rural Urban Comb Rural Urban Comb Rural Urban Comb Rural Urban CombDirect Method (DM) 0.51 0.58 0.53 0.76 0.71 0.75 0.48 0.56 0.50 0.74 0.70 0.73 Fixed Cal. Line Method (FCLM) 0.50 0.56 0.52 0.71 0.63 0.69 0.47 0.55 0.49 0.69 0.61 0.67 Fixed Pov. Line Method (FPLM) 0.45 0.55 0.48 0.75 0.63 0.72 0.39 0.53 0.43 0.72 0.62 0.69 DM after CD (DMCD) 0.46 0.60 0.50 0.81 0.79 0.80 0.40 0.55 0.44 0.81 0.78 0.80 FCLM after CD (FCLMCD) 0.46 0.59 0.50 0.76 0.65 0.73 0.40 0.54 0.44 0.75 0.67 0.73 FPLM after CD (FPLMCD) 0.43 0.61 0.48 0.76 0.67 0.73 0.39 0.52 0.43 0.76 0.69 0.74
Poverty rates using adjusted calories are uniformly found to be less than the corresponding poverty rates using unadjusted calories though the differences between these two figures are not much for most of the cases. This is certainly an improvement and should always be followed whenever poverty rates are calculated through calorie intakes. Since many measures of poverty or rather many measures of Head Count Ratio (HCR) have been discussed in this report, it is necessary to take a comprehensive picture of all the poverty measures together in a tabular form. In doing so, some more poverty measures come up into the picture and thus are reported in the table. We can get the estimated total calorie intakes of households by Calorie Decomposition Method (CD). These estimates have been used to get the poverty rates. We have also found the poverty rates using Direct Method (DM), Fixed Calorie Line Method and Fixed Poverty Line Method. These methods have been explained in the report in great details. We can combine these three methods with that of Calorie Decomposition Method, i.e., getting poverty rates by each of these three methods using the estimated calorie intakes of the households that have been found by CD. Estimated values of calorie intakes eliminates the household wise variations of the calorie intakes that are due to the factors other than the age‐sex composition and the total/per capita expenditures of the households and thus should give more stable and reliable estimates. In Table 7.2, these estimates have been reported. We have given the estimates assuming all people are in sedentary level as well as the estimates assuming the activity status implied by the occupation codes. In the same Table the results of Error Distribution Methods have been presented using both “all sedentary” and “with usual activity” status. It should be mentioned here that we have
85
sometimes neglected the top expenditure group to get a better approximation of quadratic fit of the mean values over different expenditure groups. The coefficients of the regressions in the different expenditure groups are not important. Only the regression residuals are used to get estimates of the error variances and hence the probabilities. The results of error distribution methods are found to be similar to the earlier methods. The poverty rates with and without “Adult Equivalent Scales” give almost the same results. Also, the poverty rates assuming bivariate/trivariate normal and bivariate/trivariate lognormal distribution of the associated variables give almost the same results. Thus the decision on whether we should assume a particular joint distribution of the associated variables or whether we should introduce “Adult Equivalent Scales” in the Error Distribution Method do not matter much. Table 7.2 Poverty Rates without adjusted Calorie with and without Activity Status
All Sedentary With Activity Status Method used
Rural Urban Comb Rural Urban Comb(1) (2) (3) (4) (5) (6) (7) (8) (9)
Direct Method (DM) 0.51 0.58 0.53 0.76 0.71 0.75 Calorie Line
Fixed Calorie Line Method (FCLM) 0.50 0.56 0.52 0.71 0.63 0.69 Existing Expend. Line Fixed Poverty Line Method (FPLM) 0.45 0.55 0.48 0.75 0.63 0.72
DM of CD (DMCD) 0.46 0.60 0.50 0.81 0.79 0.80 FCLM of CD (FCLMCD) 0.46 0.59 0.50 0.76 0.65 0.73
Calorie Decomposition
(CD) FPLM of CD (FPLMCD) 0.43 0.61 0.48 0.76 0.67 0.73 EDM of Linear Regression (EDMLR) 0.50 0.56 0.52 0.79 0.70 0.76 EDMLR Truncation (EDMLRT) 0.19 0.24 0.20 0.60 0.43 0.55 EDM of Log‐linear Regression (EDMLLR) 0.51 0.57 0.53 0.78 0.71 0.76 EDMLGR Truncation (EDMLLRT) 0.22 0.26 0.23 0.58 0.44 0.54 EDMLR Adult Equiv. Scale (EDMLRAES) 0.49 0.54 0.51 0.80 0.71 0.77 EDMLRAES Truncation (EDMLRAEST) 0.20 0.23 0.20 0.62 0.45 0.57 EDMLGRAES 0.51 0.56 0.52 0.78 0.70 0.76
New Error
Distribution (ED)
EDMLGRAES Truncation 0.21 0.26 0.22 0.58 0.43 0.54
There are eight different ways of finding poverty rates using Error Distribution Method. Firstly there are linear as well as log‐linear regressions which can be applied assuming Adult Equivalent Scale or without assuming Adult Equivalent Scale. And then one can decide whether truncation should be used or not. In each of these eight ways we can again find the poverty rates by Direct or Fixed Calorie Line Methods. Thus altogether we have 16 ways of finding poverty rates. Table 7.3 shows the poverty rates for all these 16 ways separately for rural, urban sectors and for rural and urban sectors combined. Table 7.3 Comparison of Poverty Rates with Activity Status: Direct Method Vs. Fixed Calorie Line
Method Method Direct Fixed Calorie
Sector Rural Urban Comb Rural Urban Comb Per capita Calorie Data 0.76 0.71 0.75 0.71 0.63 0.69 Calorie Decomposition 0.81 0.79 0.80 0.76 0.65 0.73 Error Distribution Linear (EDL) 0.79 0.70 0.76 0.69 0.60 0.66 EDL Truncated (EDLT) 0.60 0.43 0.55 0.44 0.30 0.40 Error Distribution Log‐linear (EDLgL) 0.78 0.71 0.76 0.70 0.62 0.68 EDLgL Truncated (EDLgLT) 0.58 0.44 0.54 0.46 0.33 0.42 EDL Adult Equivalent Scale (EDLAES) 0.80 0.71 0.77 0.691 0.601 0.66 EDLAES Truncated 0.62 0.45 0.57 0.451 0.311 0.41 EDLgL Adult Equivalent Scale (EDLgLAES) 0.78 0.70 0.76 0.701 0.621 0.68 EDLgLAES Truncated 0.58 0.43 0.54 0.451 0.321 0.41
1: fixed norm poverty, i.e., Φ[‐{ + x‐2230.917}/ є] for rural and Φ[‐{ + x‐2082.259}/ є] for urban.
86
Each one of the truncated method gives almost the same result. So is for the non‐truncated methods. But there is much difference between a truncated method and the corresponding non‐truncated method. Truncated method seems to be a more logical way of finding the poverty rates. This will be more clear once we take different hypothetical limiting values for the error variances. If the error variances tend to zero and the number of expenditure groups become sufficiently large then the error distribution method gives the same value as that of fixed calorie method and also the same value as the fixed expenditure method. The difference between the fixed calorie method and the fixed expenditure method increases as the error variance increases. 7.2 Reviews and Discussions It has long been observed that official poverty lines and the poverty lines based on calorie consumption are very much different, the poverty lines based on calorie norms being much higher than those obtained by official poverty lines. Consequently, the head count ratios based on calorie norms have been found to be much higher than those based on official poverty lines. Meenakshi and Viswanathan (2003) have found the head count ratios taking different calorie norms and found that these poverty rates are much higher for most of the cases (Table 7.4). Even the rank Correlations of these poverty rates with those of official poverty lines are not high and not statistically significant. It means that the factors which are responsible for such differences are not uniformly related with the poverty rates over the states.
87
Table 7.4: Headcount Ratios Based on Official Method and Direct Calorie Norms – Rural Areas
Headcount Ratios: 1999‐2000
States Based on Official Poverty Line
Based on 2400 Calorie Norm
Based on 2200 Calorie Norm
Based on 1800 Calorie Norm
1 2 3 4 5 Andhra Pradesh 11.1 80.7 69.7 36.9 Bihar 44.0 74.9 62.4 32.5 Gujarat 13.2 80.5 70.4 41.0 Haryana 8.3 55.1 43.5 18.4 Himachal Pradesh 8.0 56.5 42.7 12.1 Jammu and Kashmir 4.0 39.7 28.9 7.3 Karnataka 17.4 78.9 69.9 41.8 Kerala 9.4 81.2 70.3 42.8 Madhya Pradesh 37.1 78.4 68.0 38.5 Maharashtra 23.7 83.3 70.5 39.2 Orissa 48.0 74.6 61.7 29.1 Punjab 6.4 62.8 48.1 20.6 Rajasthan 13.7 56.7 43.0 15.5 Tamil Nadu 20.6 86.5 78.7 55.4 Uttar Pradesh 31.2 64.5 52.0 23.0 West Bengal 31.9 75.6 63.3 34.1 Source: Meenakshi and Viswanathan (2003) They have also observed that there is calorie deprivation even for the rich people. The state wise head count ratios among the richest quintile are found to be as high as 56.6% in Maharashtra. The problem of the persistent discrepancy between the two methods has also been pointed out by many authors including Saith (2005), Sen (2005), Bhalla (2003), Palmer‐Jones and Sen (2001), and it has unequivocally been agreed that the poverty in India has declined (E.g., Sundaram and Tendulkar, 2003) Sen (2005) has made some important and crucial observations.
1. He has found the intake of proteins to be higher than what would be necessary for a balanced diet.
2. The southern states exhibit lower average calorie intake especially among the poor than the rest of the country, But they are in a better position when health indicators are considered.
3. The observed reduction in the per capita consumption of calories has arisen mainly from lower consumption of cereals, which has decreased in absolute terms, especially in rural areas, in recent years.
The above three observations, made by Sen, lead to the question whether we should take all the three main nutrients, namely, calorie, protein and fat into the considerations. In that a case one should find a device by which one can combine the three indicators. Sen himself raised the same question by saying “… Indian dietary habits are steadily moving away from an excessively calorie focused diet to a more balanced one. Does this then imply that it should now be possible to evolve a more multi‐dimensional measure of nutritional adequacy which could form the basis of a new poverty line?” Ray and Lancaster (2005) have found the state wise poverty lines based on estimated nutrient prices to be higher than official poverty lines for rural India and almost same for urban India using 55th round NSS data. Himangshu (2007) has observed that food prices compared to non‐food prices have increased over time except for rural Kerala. Deaton (2003) pointed out that 7 day reference period produced higher average calorie consumption (See also Mahendra Dev and Ravi, 2007). Should we
88
then adopt 7 day reference period for food items instead of 30 days? The fact that tastes and preferences have changed significantly and scopes for choices have increased over time, have been observed by many authors including Radhakrishna and Ravi (1992), Meenakshi (1996), Meenakshi and Ray 1999, Meenakshi and Viswanathan (2003) and so on. In short, from above mentioned findings and discussions we observe that:
1. It is necessary to take different sets of consumption baskets and hence different calorie norms for different states?
2. There may be an under-reporting of consumptions, especially the food consumption? 3. We should take population structure of each state into consideration while calculating
the state wise poverty rates? Besides age-sex composition, activity structure may be different for different states. Will there be an aggregation problem in that case? I.e., Shall we get all India level calorie line same as the weighted average of state wise calorie norms?
4. The calorie requirements depend on climatic and topological situations? 5. Consumption pattern/food habit must have changed a lot. 6. Population structure has changed over time 7. The grouping of activity status through NCO-1968 does not seem to be valid still
now. 8. NSS consumption schedule does not through any light towards the activity status
other than NCO code. In the same occupation status the employees do not do so much physical work that they used to do. We have work saving devices at work place as well as in the house and other places. Also the health status of the members are not considered at all.
7.3 Recommendations: (A) The NSS consumption schedule should have a separate block consisting of items necessary for finding calorie requirements of each member in the household. More specifically the following data are needed for each member:
(1) Age, sex, weight, height and occupation status of each member in the household. (2) Average number of hours spent on each activity in a day along with detailed
description of the activities: These data are to be collected for each member and not only for adult members. It is apprehended that the physical work exerted by non-adult members are also decreasing over time. Weight is an important factor necessary for measuring calorie requirements. Along with this the status of pregnancy and lactation of mothers are also required.
(3) Information on any chronic or acute illness of each member: It should be taken along with data on decrease in the amount of food (calorie/fat/protein) intake of the concerned member due to this illness.
(B) The data on food consumption should have a reference period of 7 days. Thirty days reference period for food items should not be taken for the same household. (C) There should be a system of verification of the data collected for at least a portion of the sampled households by a different agent altogether. (D) It is also possible to find the poverty rates using protein or fat intake of members. Should we get a poverty rate combining all the three factors namely, Calorie, Protein and Fat? If so, how should we combine these factors? There are different measures of poverty discussed in the report. Which one should we take as an official index? A committee should be set up to look into all these matters. The
89
committee may also look into other dimensions of poverty, such as property, possession of selected items like mobiles, cars, cattle etc. Minimum energy requirements of calorie, protein and fat as recommended by ICMR or FAO are based on scientific experiments and thus cannot be questioned. What we can question is whether these experiments are done in an environment suitable for Indian situation. One can question whether ICMR has taken representative people from most of the states in India. It should be noted that FAO recommendations not only take age, sex and activity status into considerations while calculating the average norms but also give functional relations so that one can calculate the norms if the exact values of age, sex, activities throughout the day and weight are known. NCO codes are not sufficient to capture the activity status.
90
References Bhalla, S. (2003): Recounting the Poor: Poverty in India, 1983‐99, Economic and Political Weekly, January 25, 2003, pp. 338‐349. Deaton, Angus and Jean Dreze (2002): ‘Poverty and Inequality in India: A Re‐examination’, Economic and Political Weekly, Vol 37, No 36. Deaton, A. (2003): Adjusted Indian Poverty Estimates for 1999‐2000, Economic and Political Weekly, January 25, 2003, pp. 322‐326. Himanshu (2007): Recent Trends in Poverty and Inequality: Some Preliminary Results, Economic and Political Weekly, February 10, 2007, pp. 497‐08. Mahendra Dev, S. (2005): “Calorie Norms and Poverty”, Economic and Political Weekly, pp. 789‐792 February 19, 2005. Mahendra Dev, S. and C. Ravi (2007): Poverty and Inequality: All‐India and States, 1983‐2005, Economic and Political Weekly, February 10, 2007, pp. 509‐521. Manna, G. C. (2007): On Calibrating the Poverty Line for Poverty Estimation in India, Economic and Political Weekly, July 28, 2007, pp. 3108‐3115. Meenakshi, J V (1996): ‘How Important are Changes in Taste? A State‐Level Analysis of Food Demand’, Economic and Political Weekly, December 14. Meenakshi, J V and Ranjan Ray (1999): ‘Regional Differences in India’s Food Expenditure Pattern: A Complete Demand Systems Approach’, Journal of International Development, vol 11. Meenakshi, J V and B Viswanathan (2003): ‘Calorie Deprivation in Rural India’, Economic and Political Weekly, Vol 38, No 4. Minhas, B.S. (1991) : “On Estimating the Inadequacy of Energy Intake : Revealed Food Consumption Behaviour versus Nutritional Norms (Nutrition Status of Indian People in 1983)”, the Journal of Development Studies, Vol 28, no. 1(October), pp 1‐38. NSSO Report No. 513(61/1.0/6): Nutritional Intake In India, 2004‐2005, NSS 61st Round, July 2004‐ June 2005, National Sample Survey Organisation, Ministry of Statistics & Programme Implementation, Government of India, May 2007 Palmer‐Jones, R., and K. Sen (2001): On India’s Poverty Puzzles and Statistics of Poverty, Economic and Political Weekly, January 20, 2001, pp. 211‐217 Radhakrishna, R and C Ravi (1992): ‘Effects of Growth, Relative Prices and Preferences on Food and Nutrition’, Indian Economic Review, Special Number, Vol 27, pp 303‐23. Rao, C H Hanumantha (2000): ‘Declining Demand for Foodgrains in Rural India: Causes and Implications’, Economic and Political Weekly, Vol 35, No 4. Ravallion, M (2003): ‘Fanciful Numbers and Fictitious Intrigues’, Economic and Poltical Weekly, November 1. Ray, R and G Lancaster (2005): ‘On Setting the Poverty Line Based on Estimated Nutrient Prices: Condition of Socially Disadvantaged Groups during the Reform Period’, Economic and Political Weekly, January 1, Vol XL, No 1. Saith, A (2005): Poverty Lines versus the Poor: Method versus Meaning, Economic and Political Weekly, October 22, 2005, pp. 4601‐4610. Sen, P. (2005): Of Calories and Things Reflections on Nutritional Norms, Poverty Lines and Consumption Behaviour in India, Economic and Political Weekly, October 22, 2005, pp. 4611‐4618. Subramanian, S (2005): ‘Unravelling a Conceptual Muddle: India’s Poverty Statistics in the Light of Basic Demand Theory’, Economic and Political Weekly, January 1, Vol XL, No 1. Sundaram, K and S D Tendulkar (2003): ‘Poverty in India in the 1990s: Revised Results for All India and 15 Major States for 1993‐94’, Economic and Political Weekly, November 15. – (2003): ‘Poverty among Social and Economic Groups in India in 1990s’, Economic and Political Weekly, December 13. – (2003): Poverty Has Declined in the 1990s: A Resolution of Comparability Problems in NSS Consumer Expenditure Data, Economic and Political Weekly, January 25, 2003, pp. 327‐337.