On estimating inadequacy of energy intakes : revealed food consumption behavior versus nutritional norms (nutritional status of Indian people in 1983)

THE WORLD BANK Internal Discussion Paper ASIA REGIONAL SERIES Report No. IDP 85 On Estimating Inadequacy of Energy Intakes: Revealed Food Consumption Behavior versus Nutritional Norms (Nutritional Status of Indian People in 1983) B. S. Minhas September 1990 The views presented here are those of the author, and they should not be interpreted as reflecting those of the World Bank. ASIA REGION D". 7USSION PAPER SERIES Title Author Date Originator IDP2 The Labor Force Participation of Women in the Republic of Korea: Evolution and Policy Issues C. Grootacrt May 1988 F. Iqbal IDPIS The Role of Exchange Rate Policy in Four East Asian Countries Sang-Woo Nam May 1988 D. Lcipzigcr (81388) IDP28 The Small-Scalc Enterprise Credit Program (S.S.E.P.) Under the Second and Third Calcutta Urban Development Projects F. Kahaert March 1988 F. Kahnert (81413) (CUDP II and CUDP III) - An Assessment IDP3S Improving Tax Policy Advice: Lessons and Unresolved Issues from Asia Experience H. Fleisig Juno 1989 H. Flcisig (81413) IDP36 Direct Taxes and Fiscal Policy Issues: An Illustration for East Asia A. Virmani June 1989 H. Fleisig (81413) IDP37 Commodity Taxation in Selected Countries in Soth East and East Asia Z. S'ializi June 1989 H. Fleisig (81413) IDP38 Tax Analysis in Developing Country Settings R. Musgrave June 1989 H. Fleisig (81413) IDP39 Indonesia: External Shocks, Policy Response and Adjustment Performance S. Ahmed June 1989 Sadiq Ahmed (82467) IDP42 An Analysis of the Nature of W.T. Dickens July 1989 R. Zagha (80433) Unemployment in Sri Lanka and K. Lang IDP44 Assisting Poor Rural Areas Through Groundwater Irrigation F. Kahnert August 1989 C. Chamberlin (81409) IDPS1 Educational Development in Asia: A Comparative Study Focussing on Cost Jee-Peng Tan October 1989 Jet-Peng Tan (81408) and Financing Issues Alain Mingat IDP52 Chinese Reforms, Inflation and the Allocation of Investment in a Socialist Economy Okiay Yenal October 1989 Oktay Yenal (81415) IDP63 Public Policy to Promote Industrialization: The Experience of the East Asia* NICs and Lessons for Thailand D. Dollar May 1990 D. Dollar (80518) IDP65 A Study of the Poor in Sri Lanka C. Rouse June 1990 Y. Huang (80434) IDP68 Health Sector Financing in Asia C. Griffin August 1990 Jce-Peng Tan (81408) Note: Extra copies may be obtained from the Asia Information Service Center. ON ESTIMATING INADEQUACY OF ENERGY INTAKES: REVEALED FOOD CONSUMPTION BEHAVIOR VERSUS NUTRITIONAL NORMS (NUTRITIONAL STATUS OF INDIAN PEOPLE IN 1983) By B.S. Minhas Indian Statistical Institute 7 Sansanwal Marg New Delhi, India 110016 and Visiting Research Fellow The World Bank Washington, D.C. 20433 September 1990 ACKNOWLEDGEMENTS Some of the spade work on this paper was done between March and July 1990 with the financial assistance from the Indian Council of Social Science Research linder the terms of an award as its National Fellow. Most of the writing, however, was done during the past six weeks as a World Bank Visiting Research Fellow attached to the Office of the Chief Economist, Asia Region. This provisional version is being circulated with the purpose of receiving comments and suggestions, which would be welcomed. I am grateful to Warren Sanderson for many insightful conversations with him: He was also generous with his time in cranking out the regressions reported in Table 11. Thanks are also due to Ms. Allison Tsatsakis for her competent help and patience in stitching together and composing this manuscript from many small bits which were handed over to her intermittently. ABSTRACT Combining some notions from the Darwinian theory of evolution with the laws of physics, the nutrition experts often derive normative recommendations on the meAn level of energy intake required by the members of a large population. That such norms are fraught with serious conceptual difficulties, should be evident from the lively but inconclusive debates which have been raging among the nutrition theorists since the early 1980s. This paper undertakes an empirical investigation into the food intake behavior of about 123,000 households covered during the 1983 consumer expenditure survey of India. A procedure is suggested for deriving the average caloric intake levels in different regions at which people begin to experience "food abundance" levels inferred only from the self-expressed behavior of consumers. The observed inter-state differences in these behavior-based food abundance thresholds are then examined to explore the question of adequacy/inadequacy of energy intakes in different states of the Indian Union. Estimates of the incidence of hunger emerging from the analysis of food consumption behavior of households (revealed in the 1983 survey) are compared with the corresponding estimates obtained with the use of nutritional norms. In addition to the detailed analysis of food consumption (during the reference period) reported by the households, a direct question relating to their self-perceived (subjective) adequacy/inadequacy of food availability throughout the year was also asked from a very large subset of households in the sample. Some preliminary and essentially exploratory results of this subjective inquiry are also presented. On Estimating Inadequacy of Energy Intakes: Revealed Food Consumption Behavior Versus Nutritional Norms (Nutritional Status of Indian People in 1983) Table of Contents Page No. I. Introduction . . . . . . . . . . . . . . . . . . . 1 II. Consumer Expenditure Surveys in India . . . . . . . 1 III. Procedure for Collection of Data on Food Consumption . . . . . . . . . . . . . . . . . 2 IV. Adjustment of Reported Food Consumption of Households by Netting of Free Meals . . . . . . . . . . 5 V. Reported Consumption and Simple Meal Adjusted Consumption Distributions . . . . . . . . . . 9 VI. Cross-Validation of Caloric Intakes Derived From Household Expenditure Surveys . . . . . . . . 13 VII. Energy Balance and Nutrition Theory . . . . . . . . 14 VIII. Adequacy of Energy Intakes: Food Consumption Behavior Versus Norm-Driven Requirements . . 15 IX. Can We Directly Ask People Whether They Are Hungry? 19 X. Different Estimates of Hunger and its Correlates: (An Exercise Towards Partial Validation of the Estimates) . . . . . . . . . . . . . . . . . XI. Some Concluding Remarks . . . . . . . . . . . . . . 27 Appendix A: Tables and Figures . . . . . . . . . . 29 References . . . . . . . . . . . . . . . . . . . . 34 List of Text Tables Table No. Page No. 1. Percentage Distribution of Households Over per Consumer Unit per Diem Calorie Intake Level by Sub Samples (SS) (All India: Rural and Urban) 4 2. Average NumbeA: of Meals Consumed per Household for Entertaining Guests, for Distribution to Employees (Non-Member) and the Average Number of Meals Consumed by Household Members at Home and Away From Home (Free or on Payment) During a Period of 30 Days by Monthly per Capita Expenditure Class (All-India: Rural) . . . . . . . . . . 5 3. Average Number of Meals Consumed per Household for Entertaining Guests, for Distribution to Employees (Non-Member) and the Average Number of Meals Consumed by Household Members at Home and Away From Home (Free or on Payment) During a Period of 30 Days by Monthly per Capita Expenditure Class (All-India: Urban) . . . . . . . . . . 6 4. Percentage of Households (i) Remaining in the Same Range (ii) Moving Upwards and (iii) Moving Downward in the Calorie Intake Level Scale After Adjustment, For Each per Capita per Diem Calorie Intake Level Group (All India: Rural and Urban) . . 8 5. Unadjusted and Meal-Adjusted Percentage Distribution of Households Over Caloric Intake Level (All India: Rural and Urban) . . . . . . . . 9 6. Percentage Distribution of Households Over Five Broad Groups of per Consumer Unit per Diem Caloric Intake After Adjusting Reported Consumption for Meals (States and All-India: Rural and Urban) 11 7. Per Diem per Consumer Unit Calorie Intake Level at Which Net Displacement of Households Along the Calorie Intake Level Scale After Adjustment is Zero (States and All-India: Rural and Urban) 12 8. Percentage Distribution of Households by Their Own Perception of the Adequacy of Food During 1983 Over Different per Capita Monthly Expenditure (MPCE) Classes (All-India: Rural) . . . . . 20 -ii - 9. Percentage Distribution of Households by Their Own Perception of Adequacy of Food During 1983 Over Different per Capita Monthly Expenditure (MPCE) Classes (All-India: Urban) . . . . . 21 10. Percentage of Households Reporting Self-Perceived Inadequacy of Food (States and All-India: Rural) 23 11. Relationship Between Hunger, Male Unemployment Rates and Female Illiteracy . . . . . . . . . . . . 26 -iii - List of Appendix Tables and Figures Table No. Page No. A.1 Consumer Unit by Age and Sex of Person . . . . . . 29 A.2 Average Number of Meals Distributed by a Household to Employees (Non-Members) During a Period of 30 Days by States: Rural and Urban . . . . . . . . . 30 A.3 Different Estimates of the Proportion of Hungry (Undernourished) Households, Food Shares in Consumer Expenditure, Daily Status Unemployment Rates and Incidence of Illiteracy Among Persons of Age 15 Years and Above Over Different States: Rural 31 A.4 Percentage of Households (I) Remaining in the Same Range (II) Moving Upwards and (III) Moving Downwards in the Caloric Intake Level Scale After Adjustment. for Free Meals: Rural Karnataka . . . . . . 32 Figure No. A.1 Adjusting Intakes for Free Meals . . . . . . . . . 33 -iv - I. Introduction 1.1 Scientifically designed, nation-wide Diet Surveys (DS) for estimating the habitual dietary intakes of communities and nations are far more rare to come across than the large scale household Consumer Expenditure Surveys (CES). The latter are available for very many countries and are repeated at different intervals of time. Although the objectives, concepts, sampling schemes and methods of data collection in CESs are rarely designed specifically to produce readily acceptable estimates of the nutritional intakes of the population, the CES estimates of quantities of food items, reported to have been purchased/consumed by the households, are nevertheless widely used to derive the implied nutritional status of population groups represented in CES samples. These indirectly derived estimates, which are taken to represent the "true" nutritional intakes of the households (or the incidence of hunger among them), often suffer from serious drawbacks which are not always fully appreciated by the users of these data. The main purpose of this note is to explore some of these difficulties with reference to India's Third Quinquennial Survey on Consumer Expenditure, which was canvassed by the National Sample Survey Organization (NSS) from January to December 1983. 1.2 In this note we shall outlines some practical methods for adjusting the "reported" (reported to the CES investigators) consumption of food items by the households and the "implied" intakes of nutrients to correspond more closely to what might have been the "actual" intakes of the surveyed respondents. The differences across regions in the behavior of households in terms of their self-expressed threshold levels of food abundance vis a vis the recommended mean level of caloric intake shall be empirically investigated. We shall also present some preliminary results relating to people's own (subjective) perception of adequacy/inadequacy of their food intakes. II. Consumer Expenditure Surveys in India 2.1 The NSS has been conducting inquiries on household consumer expenditure in India since 1950 - once every year till 1970-71 and once every five years, but with a vastly expanded sample size, since 1972-73. Lately (since 1986-87), in addition to the large quinquennial surveys, a smaller (about 25,000 households) annual inquiry on consumer expenditure has been reintroduced. The basic sampling scheme and methods of data collection in the CESs of the NSSI are designed essentially to produce unbiased estimates of aggregate consumption as well as average per capita consumption expenditure with respect to a specific reference period. These survey results were intended for use in filling the gaps in data needed in the Indian national accounts, analysis and forecasting of demands and, more importantly, for estimating frequency distributions of households and persons by levels of 1/ A brief description of the sampling design and procedures of data collection of the NSS is available, among many other, in Minhas (1988). -1 - living across regions and over different social groups. The questions relating to the nutritional status of the Indian population were never specifically considered by the designers of the CESs in India. Nevertheless the data on food consumption available in some of these surveys have at times been exploited to derive some crude and proximate estimates of the average intake of nutrients per capita and per standard consumer unit per diem in different states and all-India. This kind of crude estimates of nutritional intakes have also been derived from the 1983 CES of India. Before presenting these results, we introduce some definitions and briefly describe the NSS procedure for collection of data on food expenditure. III. Procedure for Collection of Data on Food Consumption 3.1 The NSS collects detailed consumption data in value and quantity terms (on about 200 food items) for the last 30 days preceding the date of inquiry from the sample households by interviewing the heads of households (and other members). A household is defined as a group of persons normally living toget.er and taking food from a common kitchen. A boarding house, a hotel or a hostel is treated as a cluster of households, where each individual boarder (with his dependents or guests) forms a separate household. Household members are grouped into three categories, viz, adult male, adult female and child. All persons of age 15 and above are considered adults. All institutional households (maintained and fed by government), such as those in prisons, police quarters, cantonments, hospitals, asylums and relief camps are excluded from the scope of the consumer survey. The houseless population, which comprised 0.35 percent of the total population in 1981, is also excluded. 3.2 Household Expenditure on Food: The consumer expenditure of a household on food items relates to the actual consumption made by the normal resident members and also the guests. Actual consumption includes all purchased food items, consumption out of homegrown stocks as well as transfer receipts like gifts, loans, etc. Transfer payment of all kinds, however, are excluded. It is pertinent to note that the NSS records all "cooked meals" in the budgets of the serving households. Meals provided to employees, guests visiting the household or invited to ceremonies, cleaning ladies, religious functionaries, etc., whether prepared in the household kitchen, in the field or at some ceremonial location, get included in the domestic consumption of the employer/host household. Also the food served to household pets is booked in the domestic consumption account of the household. Unlike the transfer payments and receipts in other forms, there is a practical difficulty of estimating the quantities and values of individual items used in preparing the meals served to employees and others. To avoid double counting, cooked meals received as perquisites from employer household, or as a gift or charity are not recorded in the account of the recipient household. All cooked meals purchased from the market for the consumption of the normal resident members of for the guests are, however, included in the food expenditure of the purchaser household. 3.3 In order to avoid misunderstanding it may seem necessary to define a meal: A "meal" is comprised of one or more readily eatable (generally cooked) items of food with some cereal as its usual major constituent. The -2 - "meals taken away from home-free" are those received from an employer, or taken as guests in other households on ceremonial and other occasions. The "meals taken away from home - on payment" are clearly counted in the accounts of the purchaser household. 3.4 Some implications of the NSS procedure for recording of cooked meals: The current practice of data collection, in which the meals served to guests (on any occasion) and served "free" to the household employees are accounted for by the NSS in the budgets of host/employer households, (and when such meals are obviously not taken into account in the recipient households), is not only convenient in field work but also avoids any duplication: For large samples, this procedure does not compromise the unbiased nature of the estimates of aggregate consumption and average per capita expenditure on food obtained for the population as a whole2. (Remember these were the two primary objectives in CESs of the NSS). Although this procedure Zor recording of cooked meals may tend to understate the de-facto food intake of the net recipients, who may be more numerous among the poor than among the better-off sections, nevertheless the empirically estimated distributions of food-intakes (from data thus collected) over size classes of consumer expenditure would represent true pictures of the rankings of households both for the rich as well as poor classes as long as the reported consumption of the households is an increasing function of their de-facto (true) consumption. We must nevertheless note that as the households do not eat (themselves) all the cooked meals prepared by them, and whose different ingredients are reported to the NSS investigators as a part of the household's expenditure on food items, the intake of nutrients by the households derived from reported consumption does not conform to the "actual" intakes. 3.5 Distribution of household by caloric intake levels derived from consumption of food items reported in CES 1983: The 1983 survey covered about 123,000 households, divided into two independent, interpenetrating sub- samples, spread over nearly 8,000 villages and 4,400 urban blocks of all states of the Indian Union. The quantity estimates of each of the food items consumed by a household were converted into their equivalent number of calories by using the appropriate conversion factors. The average intake of calories, protein and fat per consumer unit per diem in India as a whole was found to be 2781 calories, 78 grams and 34 grams in the rural sector and 2574 calories, 70 grams and 46 grams in the urban sector respectively. Notice that the average caloric intake in urban areas was lower than national standard requirement of 2700 calories, whereas intake of protein and fat, both in rural and urban India, was considerably higher than the recommended minimum. / This matter is discussed in detail in Minhas (1988), particularly pp. 48-62. -3 - TABLE It PERCENTAGE DISTRIBUTION OF HOUSEHOLDS OVER PER CONSUMER UNIT PER DIEM CALORIE INTAKE LEVEL BY SUB SAMPLES (SS) ALL-INDIA per diem per consumer unit rural urban calorie intake level (percentage) SS-1 SS-II comb. SS-1 SS-II comb. (1) (2) (3) (4) (5) (6) (7) less than 25 1.41 1.20 1.31 1.74 1.82 1.74 25 - 50 2.94 3.14 3.04 2.85 2.76 2.79 50 - 70 11.15 11.06 11.13 13.27 12.86 13.00 70 - 80 10.01 9.80 9.90 12.05 12.08 12.06 80 - 90 12.00 12.36 12.18 14.11 14.65 14.39 90 - 100 12.33 12.48 12.40 14.24 14.40 14.37 100 - 110 11.03 11.10 11.07 11.26 11.44 11.35 110 - 120 9.56 9.13 9.34 7.95 8.53 8.26 120 - 150 16.94 17.03 16.98 13.71 13.23 13.48 150 - 200 9.08 8.98 9.02 6.50 6.01 6.29 200 and above 3.55 3.72 3.63 2.32 2.22 2.27 all levels 100.00 100.00 100.00 100.00 100.00 100.00 Note: For the definition of standard consumer unit, see Appendix table A.1. 3.5.1 All households were arranged in the ascending order of caloric intake per consumer unit per diem and the percentage distribution of households over various ranges of caloric intake, sub-sample wise, is shown in Table 1. The ranges in Table 1 are expressed in terms of percentages of the recommended mean intake of 2700 calories per consumer unit per diem. It is evident from Table 1 that about 25 percent of the rural and 30 percent of the urban households had caloric intakes in 1983 which were lower than 80 percent of the recommended mean level, whereas 30 percent of the rural and 22 percent of the urban households showed (implied) levels of intakes which were higher than 120 percent of the norm. It stands to reason that a certain proportion of food reported to have been consumed by the households, with intakes of, say, 3250 calories and above, may not have been consumed by their members themselves. A part of it may represent the food intakes of their employees and guests in the form of free meals served to them. This problem is investigated in the next section. However, it may be noted here that the sub- sample wise distributions of the caloric intakes of households (derived from CES 1983) presented in Table 1 are quite close to each other, implying that the standard errors of the estimates are indeed small. Table 1 shows that between 1 and 2 percent of the population live on less than 25 percent of the caloric norm. This appears implausible and probably results from an under-reporting of the food consumption of this group as is suggested by their relatively low proportions of expenditures devoted to food (see Table 8 and 9 below). Caloric consumption is greater in the rural area than in the urban area. This is not surprising; rural residents expend a greater number of calories in physical work than do urban residents. The -4 - figures in Table 1, therefore, do not show that rural people are better nourished than their urban counterparts, only that they work harder. IV. Adiustment of Reported Food Consumption of Households by Netting of Free Meals 4.1 In the 38th round of the NSS (CES 1983), two auxiliary (and new for the NSS) sets of information were collected. One of these sets provides data for detailed accounting of all meals taken by the households. This data can be used for adjusting the reported expenditure on food items, particularly for adjusting the estimates of the implied caloric intakes such as those presented in Table 1. The second new set of data in the 1983 survey relates to the self-perception (subjective) of households on the -dequacy/inadequacy of their habitual food consumption. We shall use the results obtained from these two sets of information later to compare them [the "self-expressed" food needs of the households] with the recommended standard requirement of energy intake. 4.2 Accounting of meals: In addition to the usual data required for estimating the total food expenditure of household for the specific reference period, detailed information was collected on the number of meals taken by the household at home or away from home, whether free or on payment; and also on the number of meals served free by the household to its employees and guests, whether during ceremonies or otherwise. This information on average number of meals consumed per household for entertaining guests and feeding (non-member) employees, meals consumed by household member at home or away from home, free or on payment, during a period of 30 days by monthly per capita expenditure class, is summarized for rural and urban India respectively in Tables 2 and 3. TABLE 2: AVERAGE NUMBER OF MEALS CONSUMED PER HOUSEHOLD FOR ENTERTAINING GUESTS, FOR DISTRIAUTION TO EMPLOYEES (NON-MEMBER) AND THE AVERAGE NUMBER OF MEALS CONSUMED BY HOUSEHOLD MEMBERS AT HOME AND AWAY FROM HOME (FREE OR ON PAYMENT) DURING A PERIOD OF 30 DAYS BY MONTHLY PER CAPITA EXPENDITURE CLASS ALL-INDIA RURAL monthly per capita no. of meals per no. of meals per expenditure household household taken class served to away from home (Rs.) guests other employees household free on during guests members payment ceremonies (at home) (1) (2) (3) (4) (5) (6) (7) 0 - 30 6.97 1.41 0.21 295.85 20.49 1.73 30 - 40 - 2.14 0.02 350.96 13.19 0.72 40 - 50 0.08 1.95 0.05 352.33 9.98 0.76 50 - 60 0.06 2.12 0.46 367.86 8.34 0.82 60 - 70 0.13 3.59 0.60 361.16 7.66 0.94 70 - 85 2.87 4.08 0.42 357.17 7.00 0.92 85 - 100 1.79 5.28 0.78 355.91 6.62 1.04 100 - 125 5.55 6.77 1.05 345.87 6.62 1.27 125 - 150 4.23 8.15 1.64 327.24 6.71 1.56 150 - 200 10.69 11.49 2.94 311.38 7.11 1.91 200 - 250 7.67 13.86 5.33 287.06 7.91 3.01 250 - 300 12.62 15.98 6.08 270.64 9.61 3.64 300 & above 60.50 24.46 9.23 242.75 9.21 5.19 all classes 5.97 6.99 1.60 337.87 7.72 1.46 Source: NSS Draft Report No. 356, New Delhi, 1989. -5- TABLE 3: AVERAGE NUMBER OF MEALS CONSUMED PER HOUSEHOLD FOR ENTERTAINING GUESTS, FOR DISTRIBUTION TO EMPLOYEES (NON-MEMBER) AND THE AVERAGE NUMBER OF MEALS CONSUMED BY HOUSEHOLD MEMBERS AT HOME AND AWAY FROM HOME (FREE OR ON PAYMENT) DURING A PERIOD OF 30 DAYS BY MONTHLY PER CAPITA EXPENDITURE CLASS ALL-INDIA URBAN monthly per capita no. of meals per no. of meals per expenditure household household taken class served to away from home (Rs.) guests other employees household free on during guests members payment ceremonies (at home) (1) (2) (3) (4) (5) (6) (7) 0 - 30 - 2.12 2.24 248.77 23.72 2.02 30 - 40 - 2.32 0.44 324.89 37.66 2.53 40 - 50 - 1.64 - 376.63 16.18 0.81 50 - 60 0.05 2.41 0.07 368.74 16.31 1.56 60 - 70 0.08 3.32 0.05 384.78 11.01 1.45 70 - 85 0.19 3.46 0.09 376.06 9.60 1.59 85 - 100 0.28 4.13 0.13 367.17 8.85 1.51 100 - 125 3.74 4.57 0.28 348.30 7.31 2.16 125 - 150 0.85 5.61 0.33 326.97 6.70 2.87 150 - 200 3.10 7.38 1.15 387.39 8.12 3.15 200 - 250 1.41 7.50 1.01 242.88 6.17 5.24 250 - 300 2.54 8.56 1.22 277.48 6.83 6.17 300 & above 12.33 10.74 2.22 175.41 6.39 10.93 all classes 3.33 6.34 0.83 296.41 8.16 4.12 Source: NSS Draft Report No. 356, New Delhi, 1989. 4.2.1 Although the results presented in Tables 2 & 3 are self explanatory, nevertheless some comments would seem in order. One would normally expect that the average number of meals served by households to guests and employees should equal the average number of meals consumed free of cost by them at the state and all-India level. However this expectation is belied by the data: A tendency to under-report the number of free meals consumed by household members away from home is very much in evidence, more so in rural than in urban areas. The average number of free meals received is much higher than such meals given out in the lower monthly per capita expenditure classes (MPCE). The reverse situation obtains in households belonging to higher MPCE classes: The net outgo of free meals among them is far more numerous than receipts. 4.2.2 The procedure for netting of free meals and the consequent adjustment in the derived food intakes over different caloric ranges shall be impacted by this observed tendency of under-reporting of free meals consumed away from home. One does not know whether this tendency is more pronounced among the poor or the rich. However, as the proportion of free meals received (net) by the poor (to total meals consumed by them) is much higher, our adjustment procedure will not err towards overestimating their "actual" food intakes: On the contrary, these derived intakes of the poor households, even after adjustment, might still be lower than what might have actually been consumed by them. This underestimation of food consumption by the poorer households results not only through the above-noted tendency of under- reporting of free meals taken away from home but also because of two additional factors: -6- One: For want of relevant data, we are forced into assuming that "a meal is a meal". Nonetheless informed observation would suggest that while for a rich person every meal might be a feast, the free meals taken by the poor away from home as guests or employees are generally bigger than the usual meals taken by them at their homes. Adequate and meaningful adjustment of the reported fnod intakes cannot be made unless the relative dietary content of free meals is investigated through appropriate measurements. Two: Remember that when quantities of different food items reported by the households to CESs are converted into their caloric content, some uniform allowance is made for food losses which occur both in preparing the eatables for cooking as well as at the dining table stage. This uniform procedure for the treatment of wastages of raw (pre-kitchen dressing, cleaning and preparation) and cooked food introduces a downward bias in the estimates of derived intakes of the poorer households, who on an average would recover more eatable material out of a given amount of raw food (grains, vegetables, meat, fruits, etc.) and also waste less amounts of cooked food in comparison with the rich households, whose intakes, in contrast, get overestimated. In order to estimate the extent of underestimation of energy intakes caused by these two factors, one would need very careful measurements done by nutrition experts. Nonetheless, at this stage, it would seem important to take note of these difficulties rather than completely ignore them. 4.2.3 We might also note that the average number of free meals served to employees per household per month works out to be quite small, both in rural (1.60) and urban (0.83) India, This number is relatively higher in rural areas of Gujarat, Haryana, Jammu & Kashmir, Karnataka, Kerala, Punjab and West Bangall. In general the proportion of free meals (taken away from home) to total meals consumed is substantially higher among the agricultural labor households in comparison with any other social group in the rural sector. In rural Punjab, a household, on an average, distributes about 8 meals to its employees in a 30 day period. For every agricultural labor household, there are approximately 2.5 other households (employers) in the rural Punjab. In other words, every agricultural labor household in Punjab might be receiving approximately 20 free meals in every thirty day period. This should require substantial upward correction in the reported food intakes of agricultural labor households in Punjab. We intend to look into this question of meal adjustment by social groups at a subsequent occasion. 4.3 Adjustment of reported food consumption: The total food consumption reported by a household can be broken up into three components: (i) meals consumed by the household members at home as well as those meals consumed away from home on payment (Mh), (ii) meals consumed by guests (M.), and (iii) meals consumed by employees (M.). I/ The relevant data are given in Appendix Table A.2. -7 - Notice that the three categories enumerated above do not include the food consumption by members of the household away from home (free) as guests and employees of other households which we can call (Mf). Given the information on Mh, Me, Me and Mf, we can derive the "actual" food consumption of a household (C,) from the levels of caloric intake (such as those in Table 1) from reported household consumption of food (C): C,.x Mh+Mt Mh Mg, 4.3.1 Applying this formula to the ordered household data on per consumer unit per diem intakes derived from the reported consumption of food items in the CES, we can work out the percentages of households in each relative caloric range which remain within the same range, or move up or down in the intake level scale after meal adjustment. The results of this adjustment exercise for India as a whole, separately for the rural and urban sector by sub-sample (I & II) as well as the combined (I & II) sample, are presented in Table 4. The corresponding adjustments for free meals in TABLE 4: PERCENTAGE OF HOUSEHOLDS (I) REMAINING IN THE SAME RANGE (ii) MOVING UPWARDS AND (iLL) MOVING DOWNWARD IN THE CALORIE INTAKE LEVEL SCALE AFTER ADJUSTMENT, FOR EACH PER CAPITA PER DIEM CALORIE INTAKE LEVEL GROUP ALL-INDIA perdiemper rural urban consumer unit sub- percentage of households percentage of households calorie intake sample level* moving remaining moving moving remaining moving downwards in the same upwards downwards in the same upwards level level (1) (2) (3) (4) (5) (6) (7) (8) less than 25 1. - 93.81 6.19 - 85.54 14.46 II. - 92.74 7.26 - 93.23 6.77 comb. - 93.39 6.61 - 89.12 10.88 25 - 50 I. 0.62 82.27 17.11 2.47 70.18 27.35 II. 0.82 80.73 18.45 2.22 71.57 26.21 comb. 0.73 81.39 17.88 2.38 71.06 26.56 50 - 70 1. 0.92 87.75 11.33 0.96 87.32 11.72 II. 1.29 87.96 10.75 0.83 86.67 12.50 comb. 1.10 87.85 11.05 0.91 87.09 12.00 70 - 80 I. 5.37 84.37 10.26 6.85 82.52 10.63 II. 4.62 84.34 11.04 5.91 83.07 11.02 comb. 4.98 84.36 10.66 6.40 82.80 10.80 80 - 90 1. 7.88 82.08 10.04 9.33 80.28 10.39 II. 7.61 82.48 9.91 10.47 80.49 9.04 comb. 7.77 82.25 9.98 9.76 80.46 9.78 90 - 100 I. 10.43 80.79 8.78 12.85 79.03 8.12 II. 10.37 81.09 8.54 12.63 78.98 8.39 comb. 10.40 80.95 8.65 12.66 79.12 8.22 100 - 110 I. 13.44 77.77 8.79 15.27 73.48 11.25 II. 13.28 77.83 8.89 15.88 73.47 10.65 comb. 13.38 77.75 8.87 15.52 73.59 10.89 110 - 120 I. 16.87 74.18 8.95 18.97 72.18 8.85 II. 17.55 74.01 8.64 19.22 70.71 10.07 comb. 17.24 74.03 8.73 19.14 71.24 9.62 120 - 150 1. 14.16 81.28 4.56 15.28 78.25 6.47 II. 12.93 82.19 4.88 15.18 78.42 6.40 comb. 13.50 81.80 4.70 15.24 78.28 6.48 150 - 200 I. 19.51 77.43 3.06 19.52 76.24 4.24 II. 19.16 77.69 3.15 21.32 74.17 4.51 comb. 19.40 77.51 3.09 20.33 75.41 4.26 200 and above 1. 32.60 67.40 - 33.31 66.69 - II. 28.76 71.24 - 27.68 72.32 - comb. 30.53 69.7 3 131.15 68.85 Notes: *Percentage to normative requirement of 2700 calories per consumer unit per diem. Data are taken from NSS draft report, No. 356, New Delhi, 1989. -8 - individual states follow the same general picture. However, for illustrative purposes, the numerical results for rural Karnataka for the combined sample are exhibited in the appendix Table A.4. A better appreciation of the nature of adjustment in reported consumption due to netting of free meals can be gained from the diagram (informally drawn without paying any attention to scale), which immediately follows Table A.4. 4.3.2 The adjustment procedure for netting of free meals has produced the expected results. The net movement upwards dominates the downward movement in the lower ranges of unadjusted relative caloric intakes. This dominance gradually diminishes and finally vanishes at 90.58 (2446 calories) in the rural and 89.48 percent (2416 calories) in the urban sector of India as a whole. Beyond these per consumer unit per diem levels of caloric intake (both of which are lower than the recommended mean level of 2700 calories), the downward movement dominates in all the high ranges of caloric intakes. V. Reported Consumption and Simple Meal Adjusted Consumption Distributions 5.1 In this section, we compare the results given in columns (4) and (7) of Table 1 with the meal adjustment distributions of households over per consumer unit per diem caloric intake levels, which are set out in Table 5. TABLE 5: UNADJUSTED AND MEAL-ADJUSTED PERCENTAGE DISTRIBUTION OF HOUSEHOLDS OVER CALORIC INTAKE LEVEL ALL INDIA per diem per rural urban consumer unit unadjusted adiusted unadiusted adiusted caloric intake % z 2 level (%) < 25 1.31 1.54 1.74 2.25 25 - 50 3.04 , 2.71 1 2.79 1 2.19 1 I 14.17 F 13.44 F 15.79 F 14.78 50 * 70 11.13 J 10.73 J 13.00 J 12.59 J 70 - 80 9.90 1 10.11 1 12.06 1 12.26 1 II II so - 90 12.18 1 12.31 I 14.39 1 14.42 1 90 - 100 12.40 F 54.89 12.77 I 56.27 14.37 60.43 14.58 F 61.24 100 - 110 11.07 1 11.45 I 11.35 11.40 1 110 - 120 9.34 J 9.63 J 8.26 J 8.58 J 120 - 150 16.98 1 16.95 1 13.48 1 13.43 , 150 - 200 9.02 29.63 28.75 6.29 F 22.04 6.09 F 21.73 > 200 3.63 J 3.05 J 2.27 J 2.21 J all levels 100.00 100.00 100,00 100.00 5.2 Leaving aside the small fringe group of destitute households (with caloric consumption of less than 25 percent of the recommended norm), where adjustment goes in the unexpected direction, the two low caloric consumption groups (50-70 percent of the norm) and the three very high caloric intake -9 - level groups (over 120 percent of the norm) shrink in size both in the rural and the urban sector. In consequence of this simple meal adjustment, the proportion of households in the five middle groups (between 70 and 120 percent of the recommended standard) goes up in comparison with the distribution of households worked out on the basis of reported (unadjusted) consumption. After adjustment for free meals, 37.40 percent of the rural and 43.71 percent of the urban households remain below the 2430 caloric (90% of the norm) level. The corresponding percentage below the 80% level (2160 calories) are 25.09 and 29.29, respectively, in rural and urban India. 5.3 The distributions of households over five broad groups of per consumer unit per diem caloric intake after adjusting reported food consumption for free meals for 17 major states are presented in Table 6. The proportion of households having meal-adjusted caloric consumption of less than 2160 (<80% of norm), for instance, varies in the rural sector from as low as 11-15 percent in four states to 38, 39 and 43 percent, respectively, in West Bengal, Tamil Nadu and Kerala; the average for rural India as a whole being 25 percent. Approximately 15 percent of the households in rural India had energy intakes of less than 1900 K cals in 1983. The corresponding percentage in 11 states (out of 17) was lower than the national average; in six out of this group of 11 states, only about 9 percent (or less) of the households turned up energy intakes of less than 1900 K cals. The corresponding percentage in Karnataka and Orissa was nearly twice this level, whereas West Bengal (24.5%), Tamil Nadu (26.8%) and Kerala (29.0%) reported very high proportions of rural households whose energy intakes were even less than 70% (<1900 K cal) of the recommended nutritional norm of 2700 K cals. Similar results for the urban sector and other levels of caloric intake can be read from Table 6. 5.4: Points of Net Zero Displacement in Different States (abundance thresholds for energy intakes) 5.4.1 The levels of caloric intakes, by sector and region, at which free meals received and given out by households on an average just balance out (zero net displacement in consequence of meal adjustment), are of special interest to us in behavioral terms: The groups of households with caloric intake levels above the zero net displacement points on the relevant distributions are net givers of free meals; whereas groups below these points are net receivers of free meals. Leaving aside some compressible social and cultural obligations connected with the vital events of birth, marriage and death, it is difficult to imagine the persistence of food altruism among the chronically hungry, who would, in general, be the net receivers of free meals. Nonetheless, not all households, belonging to groups which are net receivers of free meals on the margin, could be regarded to be suffering from inadequacy in their habitual energy intakes. In other words, although the energy requirements of all household groups found to be above the zero net displacement points (which we equate with food abundance thresholds in different states) are clearly more than fully met, nevertheless the inferences that can be drawn on the downside of these abundance thresholds are not symmetrical. In between the food-abundant groups and the food-hungry (those whose caloric intakes fell below a "critical" level - the cut off point for hunger which we shall define in paras 8.6 and 8.7 below), we can identify a category of the food-sufficient. This category - food sufficient/adequate - is bounded by the behaviorally revealed abundance thresholds from above and by the critical cut off point of hunger from below. -10 - TABLE 6: PERCENTAGE DISTRIBUTION OF HOUSEHOLDS OVER FIVE BROAD GROUPS OF PER CONSUMER UNIT PER DIEM CALORIC INTAKE AFTER ADJUSTING REPORTED CONSUMPTION FOR MEALSs STATES AND ALL INDIA Rural Sector State Calories (0) (1) (2) (3) (4) (5) < 1890 < 2160 < 2430 2430-2700 > 2700 1 AP 11.80 21.18 34.66 13.92 51.42 2 Assam 12.62 31.28 44.91 16.87 38.22 3 Bihar 14.41 24.93 37.36 13.27 49.37 4 Gujarat 16.62 29.11 43.20 15.34 41.46 5 Haryana 9.20 14.62 23.02 12.67 64.31 6 HP 6.33 11.31 17.75 9.81 72.44 7 J & K 5.73 11.09 18.13 11.21 70.66 8 Karnataka 18.48 28.42 40.07 11.13 48.80 9 Kerala 29.00 43.05 55.48 11.07 33.45 10 MP 9.18 17.59 29.89 13.35 56.76 11 Maharashtra 13.35 25.84 40.22 14.37 45.41 12 Orissa 18.24 29.99 43.31 12.58 44.11 13 Punjab 8.81 15,08 22.70 3J.79 66.51 14 Rajasthan 10.61 16.28 25.41 10.74 63.85 15 TN 26.83 39.00 51.77 11.02 37.21 16 UP 9.43 17.35 28.15 11.95 59.90 17 W.B 24.47 37,69 51.15 13.53 35.32 18 All India 14.58 25.09 37.40 12.77 49.83 Urban Sector State Calories (0) (1) (2) (3) (4) (5) < 1890 < 2160 < 2430 2430-2700 > 2700 1 AP 17.22 30.36 46.38 14.40 38.77 2 Assam 15.27 27.70 41.48 19.74 38.78 3 Bihar 11.87 21.79 35.48 15.48 49.04 4 Gujarat 18.66 32.11 45.76 17.18 37.06 5 Haryana 11.60 21.17 34.65 11.73 53.62 6 HP 7.97 16.00 19.38 19.12 61.50 7 J & K 6.88 14.66 32.33 17.36 50.81 8 Karnataka 21.62 32.53 42.94 11.89 45.17 9 Kerala 26.48 37.83 50.27 9.87 39.86 10 MP 10.47 20.86 36.48 16.83 46.69 11 Maharashtra 16.88 31.88 48.37 15.15 36.48 12 Orissa 9.03 18.16 30.37 20.09 49.54 13 Punjab 18.06 28.88 40.20 12.35 47.54 14 Rajasthan 12.04 21.00 34.98 12.96 52.06 15 TN 28.14 42.51 55.53 12.00 32.39 16 UP 14.51 27.23 42.91 14.11 42.98 17 1.8 14.50 27.11 42.23 17.19 40.58 18 All India 14.58 25.09 37.40 12.77 49.83 Source: NSS (38th Round, Jan.-Dec. 1983) Draft Report, No. 356, New Delhi, 1989. 5.4.2 It is worth stressing that our results, derived from the demonstrated (self-expressed) food behavior of populations of different Indian states, are based on a very large sample of households, surveyed over a whole calendar year, 1983, in four sub-rounds of equal size and each of three months duration - a procedure which takes seasonal factors fully into account. The behavioral approach to the estimation of food abundance and inadequacy (which our data make possible) in the habitual energy intakes of large populations can, as we shall demonstrate below, throw some new light on the recent -11 - theoretical debates about the question of what the recommended norm for energy intakes might be. 5.4.3 In Table 7, we present the per diem per consumer unit caloric intakes (in parentheses) and the corresponding percentage (derived caloric intake + by the recommended norm of 2700 calories) at which the net displacement of households in consequence of adjustment for free meals is zero in different states and all-India. TABLE 7: PER DIEM PER CONSUMER UNIT CALORIE INTAKE LEVEL AT WHICH NET DISPLACEMENT OF HOUSEHOLDS ALONG THE CALORIE INTAKE LEVEL SCALE AFTER ADJUSTMENT IS ZERO FOR STATES & ALL-INDIA. Per Diem Per Consumer Unit Calorie Intake Level Rural Urban Andhra Pradesh 100.98 (2727) 79.61 (2149) Assam 72.17 (1940) 81.96 (2213) Bihar 73.58 (1987) 90.58 (2444) Gujarat 69.17 (1868) 69.73 (1775) Haryana 94.64 (2555) 80.86 (2183) Himachal Pradesh 65.64 (1772) 95.68 (2583) Jammu & Kashmir 66.40 (1793) 92.05 (2485) Karnataka 70.22 (1896) 78.05 (2107) Kerala 82.24 (2220) 86.51 (2336) Madhya Pradesh 106.93 (2887) 92.62 (2501) Maharashtra 90.51 (2444) 80.62 (2323) Orissa 79.85 (2156) 104.73 (2828) Punjab 105.53 (2849) 85.93 (2320) Rajasthan 95.57 (2580) 106.62 (2879) Tamil Nadu 114.29 (3086) 89.18 (2408) Uttar Pradesh 84.40 (2279) 90.22 (2436) West Bengal 84.83 (2290) 89.49 (2416) All India 90.58 (2446) 89.48 (2416) Note: Net displacement - percentage of households moving upwards minus The percentage moving downwards. All data are taken from various tables in NSS Draft Report No. 356, New Delhi, 1989. These state-specific points of zero net displacement (in consequence of meal adjustment) have been picked from the state-wise distributions of households over caloric intake levels - levels relative to the recommended mean energy intake requirement for the population. As against the all-India level of about 2450 calories in the rural sector, the points of zero net displacement vary from the lowest level of about 1800 calories in Himachal Pradesh and Jammu and Kashmir' [around 1900 calories in Gujarat and Karnataka, 1950 in Assam and 2000 calories in Bihar], to high levels of about 2850, 2900 and 3100 calories, respectively, in the rural sector of Punjab, Madhya Pradesh and Tamil Nadu. Let us remind ourselves that these levels correspond to points beyond which the households, on an average, are not givers of free meals. Behaviorally speaking, these levels represent the &. Because of the small sample size (in the relevant class interval) for these two Himalayan states, these estimates of the abundance threshold for energy intake (on the food adequacy scale) may not reliable. -12 - threshold of energy intake [say, 1900 in rural Gujarat, 2000 in rural Bihar and 2850 in rural Punjab], which could be considered as abundant habitual energy intakes of the rural households in these states. In the urban sector, except for Gujarat where the zero net displacement point occurs at 1775 calories (only 66% of the norm), the behaviorally revealed abundance thresholds do not fall below the 2100 K cals (78% of the norm) level. Most other states have these abundance thresholds ranged between 80 and 90% of the recommended norm. 5.4.4 The abundance threshold point for urban Tamil Nadu almost coincides with the average for urban India, nevertheless the corresponding figure of 3100 calories (114% of norm) for rural Tamil Nadu does look exceptionally high. The computations for rural Tamil Nadu might require some further checking, nevertheless, we must mention one special circumstance of Tamil Nadu in the early 1980s: This state introduced a comprehensive school meals program. All children, whether rich or poor, started receiving free lunch at school and these free meals were included in meal accounts of the recipient households in CES 1983. The point of zero net outgo of free meals in Tamil Nadu may have (temporarily, because of lack of behavioral adjustment to the brand new school lunch program) been pushed up to a level of consumption, which earlier might have been achieved in very rich households only. Gujarat, both rural and urban, is another case that deserves special notice. This is a rich state, with moderate to low incidence of poverty. It also faced a fairly comfortable regime of food availability in 1983. Its behaviorally revealed low threshold level of fully adequate habitual intake of energy (around 1870 calories in rural and 1775 in the urban sectors) can be explained only in cultural terms. Gujaratis are also known for their traditional abstemiousness in matters of food intake, although the distribution of Gujarati population by body size should seem to be no different from, say, Maharashtra or Tamil Nadu. In addition, Gujaratis are also known as a pretty dynamic lot, not given to sloth and low activity levels. 5.4.5 Kerala, it must be said, is still another case of particular importance for our story. The rural-urban divide is difficult to discern in this state: it is a rural-urban continuum. It has the highest levels of literacy and life expectancy and the lowest rate of infant mortality among the Indian states. Yet, 29 percent of the households (highest percentage % age in the country) have energy intakes of less than 1900 K cals (less than 70% of the recommended norm). The abundance threshold for habitual energy intake in rural Kerala works out to be a little over 2200 in contrast with the normative standard of 2700 K cals per consumer unit per day. The food behavior of the population and many other aspects of the socio-economic reality, particularly in states like Gujarat, Himachal Pradesh and Kerala, would seem to confront the norm-driven approach of the nutrition theorists with some really awkward questions. VI. Cross-Validation of Caloric Intakes Derived From Household Expenditure Surveys 6.1 From the consumption of food items reported by households to the NSS during 1983, we have derived the implied distributions over different levels of caloric intake per consumer unit per diem in different states as well as the Indian Union. We have also estimated the threshold levels of habitual energy intakes, which, in behavioral terms, might be regarded as -13 - adequate for achieving energy balance in relation to the normal work and activity patterns prevailing in different states. However, as these empirical results do not come from rigorous measurements of dietary intakes, their validity might be called into question by the nutrition experts. Fortunately our results can be validated against another, totally independent, data source - the product flow estimates of food availability in India based on food production and imports data adjusted for exports and intermediate uses. The results of this cross-check showed that the CES estimates of household consumption of food grains in 1983 were higher (by about 4 percent) than the corresponding estimates of food available for consumption based on the product flow method5. Similarly the CES estimates of aggregate vegetable consumption were also higher, particularly of potato and onion consumption - the only two vegetables whose production is regularly reported by the Ministry of Agriculture every year. Similar validation checks for milk and milk products, edible oils, meat, fish and eggs, sugar, etc., are more difficult to carry out as substantial portions of these items are also used in commercial establishments: The allocation of production of these items between the households and commercial establishments is not easily available. Nevertheless the CES estimates of household consumption of these food items are lower than the corresponding item-wise product flow estimates along the expected lines; however, the difference between them narrows down when the expenditure incurred by households on sweets, tea, coffee and other refreshments consumed in shops and stalls (which is recorded separately in the CES but included under items like sugar, edible oils, milk, etc., in the commodity flow method) is taken into account. 6.2 A more comprehensive cross-validation exercise of this kind was also carried out on the CES data for 1972-73 and 1977-78 (Minhas, 1988), which conclusively showed that the CES estimates of household consumption of food were in reasonable agreement (in fact somewhat higher) with the completely independent, production-based estimate of availability of food for human consumption. In view of all this, one may conclude that the estimates of habitual energy intakes of the Indian population derived from CES data are quite robust and reliable. However the difficulty that one faces is that the NSS estimates for different states cannot be cross-validated in the manner we can do for India as a whole. The state-wise figures of food production cannot be easily adjusted for net inter-state movements, neither can food imports from abroad be allocated among different states. The data problems for this state-level exercise are indeed intractable. VII. Energy Balance and Nutrition Theory 7.1. Combining some notions from the Darwinian theory of evolution with the laws of physics, the nutrition experts often estimate the mean level of energy intake required by a population. According to the mainline nutrition theory, the inter-individual variations (around the population mean) in energy intakes among individuals of the same age, sex, engaged in similar activities and maintaining their body weights are fully explained by the genetic differences among them. In the process of long run adaptation and natural selection, a population becomes increasingly homozygous in its genetic make-up and is able to reproduce itself with its prevailing genetic pool. Under this .1/ See Minhas and Kansal (1989). -14 - theory, the inter-individual variations in energy intakes (and balance) average out without disturbing the theoretically computed mean level of energy requirements for the population. 7.1.1 This view of energy requirements has recently been contested by Sukhatme and Margen (1982) and Sukhatme (1989), who claim that intra- individual variation in daily energy balance is stochastically stationary in nature but much larger than can be explained by chance. In other words, their claim is that there are non-random components in the variance of the population mean which are of significant nature, caused by persistent intra- individual variations in daily energy balance which do not vanish even when within-subject data are averaged over several days. This non-random component is interpreted by Sukhatme and his co-workers to mean that man is endowed with an homeostatic mechanism, which enables him/her to regulate the pattern of variation in energy balance over a range determined by its stationary variance. One implication of Sukhatme's theory (and some experimental work) is that persons with low habitual intakes of energy cannot be considered undernourished (hungry) unless their intakes are below the lower limit of the homeostatic range: This lower limit has been computed to be around 75% of the mean level of energy intake per consumer unit per diem recommended by the FAO/WHO/UNU report of 1985. In underscoring the consequences of interaction between genetic and environmental factors, Sukhatme's work has introduced a new and appealing dimension to nutrition theory. Nevertheless, his computations of the lower limit of the range for energy balance (which are based on too few and somewhat crude experiments), can bear further checking with the aid of better equipment. 7.2 The FAQ/WHO computations of energy requirements proceed as follows: The energy required for basal metabolic functions of the body can be obtained through individual BMR measurements in a sample of the population. These individual BMR measurements can be averaged to get the population mean, designated as BMR. The FAQ/WHO report (1985) proposes that the maintenance (minimal) energy requirement might be set at 1.4 ERf for a given age and sex group6. Beyond this minimal maintenance level, the additional requirements of energy intakes can be figured out in terms of energy costs of activities. For an active life of gainful work and other social activities the mean level of energy requirements would be a higher multiple of MR. VIII. Adequacy of Energy Intakes: Food Consumption Behavior Versus Norm-Driven Requirements 8.1 It is beyond the scope (and competence of the author) of this paper to delve deep into the contentious issues which currently surround the debates on nutrition theory. A crude and brief sketch of some elements of nutrition theory, which was given in the preceding section, should be taken to mean nothing more than a simple-minded attempt to put our empirical results in / Sukhatme's argument has been interpreted to suggest that ER minus 2 standard deviations should represent the lower limit to which all indiiiduals could adjust. The cut-off point for maintenance level of energy requirements, therefore, should be set at 1.2. MBE rather than 1.4. MR. However, this mode of reasoning may not find acceptance with Sukhatme as it implies constancy of mean BMR for the population. -15 - some wider perspective. Our purpose in this section also continues to be modest: We are not in search of a particular theory which might embrace our empirical findings in one of its elegant, scientific folds. Nevertheless, it is hoped that the wide (observed) spread in the behaviorally revealed threshold levels of food-richness across different states might attract some attention from the nutrition experts and thereby assist us towards a correct interpretation of the empirical findings reported in this paper. 8.2 Before we begin to summarize and discuss our findings, some notes of caution are in order: One: Our definition of the abundance threshold - level of food intake beyond which all households are considered (by us) to be food-rich - is based on the notion of freedom of choice in the matter of free meals given away as well as received by any household. This notion needs some elaboration. Let us consider a household which received zero (gross) free meals during the reference period, while this household did give out such meals to members of other households. In our treatment, this particular household would automatically get classified as a net giver (food- rich) of free meals, although the choice of receiving a free meal was never offered to it during the reference period. This is a one-sided limitation on the choice process of this household which might not have been foreseen by it before it began dispensing free meals. The occurrence of such events is undoubtedly possible. However, the worrisome implication of such events for our selection of the threshold (zero net balance of free meals) points on the distribution of households over different levels of habitual caloric intake would arise only if receivers of zero (gross) free meals are concentrated in the neighborhood of those caloric ranges in which the threshold points fall in different states of India. This is a far-fetched possibility: The likelihood of the realization of such worrisome possibilities in our data set may indeed by very small'. Two: In section 4.2.2 we indicated some reasons as to why the procedure for netting of free meals might not capture the full extent of the warranted adjustment in caloric intakes, especially of the poorer (low energy intake) households. However, we wish to point out yet another reason to worry about some degree of underestimation that creeps into the estimates of caloric intakes derived from CES data. It relates to the caloric content of sweets, tea, coffee, soft drinks and other refreshments consumed away from home, particularly in shops and street side stalls, As indicated in section VI, the expenditure on these items is / A special study of the distribution of zero-free-meal-receiving households by levels of caloric intake is already in progress. The data are also being examined to look into the consequences (for the threshold criterion) of differences in the distribution of infants among households. Recalling the definition of a meal in the 1983 CES, a major proportion of the infants become non-consumers of meals by definition. -16 - recorded by the NSS under tt.e miscellaneous foods group and is fully included in the estimates of household expenditure. However, the caloric content of these items is difficult to figure out and, as such, cannot be fully derived from their expenditure estimates in the CES data set. The distributional implications (as among the rich and poor households) of free refreshments may be allowed with the assistance of our results on adjustments for (net) free meals, nevertheless, in the absence of data on the nutritional content of these refreshments, we have to live with some degree of unavoidable underestimation" in the caloric intakes derived from the CES. 8.3 If the food abundance criterion proposed in this paper finds wide acceptance from the scientific community, it would help us separate out unambiguously the proportion of those households (all above the zero net displacement point) whose behavior implies that they are endowed with food abundance in the context of their environment of culture, education and general conditioning to certain regimes of food availability. Nevertheless, it is worth repeating that all groups of households below the abundance threshold cannot be categorized as food-poor or undernourished: groups below the abundance threshold include the food-sufficient as well the food-poor9. This further sorting out can be accomplished by examining the entire cross- section of these abundance threshold points (across different states and regions of India) in conjunction with other facts and arguments, which follow in 8.6 and 8.7 below. 8.4 The economists are often overly concerned with (and are particularly good at deeply exploring) the relationship between income levels and food consumption. Mercifully this paper is not focussed on the economics of tastes. We might nevertheless note that the nexus between income levels and caloric intakes (beyond very low levels of energy intakes) would seem to be somewhat complex and indirect. Rich people do not necessarily consume (only ?) more calories than the not-so-rich; generally they switch to more expensive calories and a greater variety of foods. In certain cultural surroundings, where gluttonous behavior is known to be equated to moral depravity, the adequacy level for average caloric intake may get established at a comparatively lower level of habitual food intake than in other culturally diverse regions, where high social premium is attached to copious consumption of food as a necessary part of "good" living. The population groups in the rich states of Haryana and Punjab belong in the latter category, whereas Gujaratis are both rich and more food-economic. Education also seems in general to work as a moderating influence on the levels of habitual food intakes. In other words the spread of education in population, aided by certain cultural values, might lower the adequacy thresholds in the habitual A/ One must avoid the temptation to overstate the consequences of this source of underestimation as the caloric content of certain standard refreshments is already accounted for in the NSS data on energy intakes. On an average, the intakes of low-energy-intake households are not likely to improve by more than 1-2 percent even if the caloric content of the refreshments consumed by them away from home, but not fully included in estimates of energy consumption, were added on. 2/ See Section 5.4.1. above. -17 - food intakes without necessarily affecting the patterns of gainful work and social activityo. 8.5 As against the normative nutritional requirement of 2700 K cal per standard consumer unit per diem, the behaviorally revealed food-abundance threshold for the rural population of India as a whole appears at the lower habitual energy intake level of 2445 K cal, i.e. at 90.6% of the nutritional norm. [The corresponding point for urban India falls at 2415 K. cal - 89.5% of the norm]. Around this mean value of 2445 K. cal, the rural populations of different states begin to cross food-adequacy thresholds at habitual energy intake levels ranging from lows of about 1800 to 1900 K. cal in Himachal Pradesh, Jammu and Kashmir, Gujarat and Karnataka (between 66 to 69% of the nutritional norm) to highs of about 2850 (and more) K. cal in Punjab and Madhya Pradesh". This wide disparity in the levels of habitual energy intake at which people of different states [say, Gujarat (1870), Kerala (2250) and Punjab (2850 K. cal)] begin to enter the zone of food-abundance, cannot be explained in terms of differences in body size and activity levels of their populations. 8.6 Analogous to this wide disparity in the levels of energy intake at which populations of different states begin to feel food-rich, can one postulate that the cut off points for food-hunger (under-nourishment) would also be different in different states? The answer to this question would seem to be in the negative, ignoring, of course, some minor variations which could be attributed to some shrinkage in discretionary activities forced by scarcity of food. After adjusting for age and sex composition, the critical cut off point for hunger cannot vary much. Nevertheless, it is of some interest to ask what level of caloric intake might represent a rough approximation to this cut off point for hunger for the Indian population as a whole. 8.7 The ho-iseholds residing in the rest of India cannot adjust themselves to the environments of culture and food history obtaining in, say, Gujarat, Himachal Pradesh or Karnataka, whose populations begin to cross the threshold of nutritional adequacy at habitual energy intake levels between 1800 to 1900 K. cal (the lowest among the Indian states). Could one, nevertheless, take this as the lower limit to which people from other states could (hypothetically) adjust before beginning to experience onset of hunger? Thusly contrived, the critical cut off point - the one which divides the food sufficient from the food poor - for the onset of hunger would lie somewhere between 66 to 69 percent of the recommended standard nutritional requirement. Allowing for some underestimation of caloric intakes derived from CES data 10/ The extent of this impact, which can be defined in terms of an (additional) income equivalence, is an empirically testable proposition. We are interested here only in taking note of this proposition rather than seeking its empirical confirmation. 11/ We might note in passing that Himachal Pradesh and Gujarat, although food- economic, are quite prosperous and low-poverty states, Although Punjab and Madhya Pradesh reach their food-abundance thresholds at high levels of energy intake, the former happens to be the richest, outstanding food- surplus and low-poverty state; whereas the latter is slightly below the national average in terms of per capita domestic product, almost self- sufficient in food grains but with about half of its rural population below the poverty line. -18 - (the reasons for this underestimation are enumerated in this paper), our empirical results for India would seem to suggest the hunger threshold being reached at around 70% of the standard nutritional norm. Taking 70 percent (1890 K. cal) as the cut off point for hunger, 29 percent of the rural population of Kerala in 1983 (24.5% in W. Bengal and 26.8% in Tamil Nadu) would still come out to be below the hunger threshold. Kerala, the hungriest state in India, nevertheless, has the highest expectation of life at birth and the lowest rate of infant mortality. However, using the standard norm of 2700 K. cal, two-thirds of all rural households in Kerala come out to be hungry in 1983, in comparison with about 50 percent in rural and 58 percent urban India as a whole. Assuming a hunger threshold of 1890 K. cal, the incidence of hunger in 1983 is estimated at about 15 percent in rural and 17 percent in urban India, without making any allowance (in energy requirements) for the place (rural-urban) of residence of the population. IX. Can We Directly Ask PeoRle Whether They are Hun&ry? (results from an inquiry) 9.1. Besides collecting detailed food consumption data in value and quantity terms for the reference period and constructing meal accounts of the sample households, in its 38th round the NSS also (for the first time in its forty year history) conducted an inquiry, based on self-assessed reporting, on the question whether or not the sampled households considered their food intakes adequate. A single probing question: whether all members of the household got two square meals a day throughout the year? was asked from about 80% of the sample households. This was the parting question put to the head of the household after the detailed expenditure schedule had already been filled in. The percentage distribution of households by type of response to the above-said question over different MPCE classes is presented, separately for the rural and urban sector, in Tables 8 & 9. One must, nevertheless, emphasize that the classification of households, reporting adequacy or inadequacy of food consumption, on the basis of information collected through a single probing question may not always be free from subjectivity and imprecision. Also the probing question was not asked (to avoid embarrassment) from those households who had already been found to be clearly food-rich on the basis of detailed information collected in the household expenditure schedule. In consequence, some amount of subjective selection by the investigators may have crept in. -19 - TABLE 8: PERCENTAGE OF DISTRIBUTION OF HOUSEHOLDS BY THEIR OWN PERCEPTION OF THE ADEQUACY OF FOOD DURING 1983 OVER DIFFERENT PER CAPITA MDNTELY EXPENDITURE (MPCE) CLASSES ALL-INDIA: RURAL NSS: 38TH ROUND, JAN.-DEC.1983 Whether all members of the household got two square meals a day throushout the year? Expenditure _ YES NO NOT Proportion Reporting SorLal mPCE on food Households Population throughout only some REPORTED In adequacy of Food No. (Rs) (%) the year months of the year (Col. 6 + 7) (0) (1) (2) (3) (4) (5) (6) (7) (8) (9) 1 0-30 73.11 0.9 0.9 29.51 38.68 31.19 0.62 69.87 2 30-40 77.48 2.2 2.5 44.23 44.87 10.05 0.85 54.92 3 40-50 77.23 4.6 5.1 56.95 36.26 6.55 0.24 42.81 4 50-60 76.30 7.2 8.0 64.31 20.26 5.11 0.32 25.37 5 60-70 75.45 9.0 9.7 71.01 24.45 4.13 0.41 28.58 6 70-85 74.17 14.5 15.3 76.58 20.93 2.10 0.39 23.03 7 85-100 72.76 13.1 13.6 83.17 15.12 1.37 0.34 16.49 8 100-125 69.79 16.9 17.1 86.91 11.71 1.07 0.31 12.78 9 123-150 66.13 10.4 9.9 91.15 7.95 0.56 0.34 8.52 10 150-200 62.27 10.8 9.7 93.21 6.01 0.42 0.36 6.43 11 200-250 56.84 4.7 3.9 94.96 4.23 0.32 0.45 4.55 12 250-300 52.10 2.3 1.8 96.19 3.14 0.43 0.24 3.57 13 300-above 44.15 3.4 2.5 97.76 1.58 0.27 0.39 1.85 14 all classes 65.46 100.0 100.0 81.09 16.19 2.35 0.37 18.44 Source: Sarvekshana, Vol. XIII, No.2, Oct.-Dec. (1989), Pg. S-179 and S-205 TABLE 9: PERCENTAGE OF DISTRIBUTION OF HOUSEHOLDS BY THEIR OWN PERCEPTION OF THE ADEQUACY OF FOOD DURING 1983 OVER DIFFERENT PER CAPITA MONTHLY EXPENDITURE (MPCE) CLASSES ALL-INDIA: URBAN NSS: 38TH ROUND, JAN.-DEC.1983 Whether all members of the household got two square meals a day throutzhout the year? Expenditure YES NO NOT Proportion Reporting Serial MPCE on food Households Population throughout only some REPORTED In Adequacy of Food No. (Rs) (2) the year months of the year (Col. 6 + 7) (0) (1) (2) (3) (4) (5) (6) (7) (8) (9) 1 0-30 61.95 0.3 0.2 68.50 18.53 12.22 0.75 30.75 2 30-40 65.63 0.4 0.5 57.50 29.22 13.28 - 42.50 3 40-50 74.72 1.0 1.4 67.54 25.48 6.78 0.20 32.26 4 50-60 74.02 2.3 2.9 73.46 21.45 4.76 0.30 26.21 5 60-70 73.40 3.7 4.9 80.76 16.24 2.22 0.78 18.46 6 70-85 75.18 7.6 9.6 85.90 12.03 1.56 0.51 13.59 7 85-100 70.73 8.8 10.6 89.89 8.81 0.92 0.38 9.73 8 100-125 68.75 14.8 17.1 92.03 6.99 0.57 0.41 7.56 9 125-150 66.04 12.0 13.1 94.42 4.47 0.68 0.43 4.15 10 150-200 62.30 17.0 16.3 96.76 2.71 0.19 0.34 2.90 11 200-250 58.27 10.4 8.8 97.82 1.73 0.09 0.36 1.82 12 250-300 55.02 6.8 5.2 98.85 0.77 - 0.38 0.77 13 300-above 44.45 14.9 9.4 98.87 0.51 0.09 0.53 0.60 14 all classes 59.01 100.0 100.0 93.25 5.56 0.77 0.42 6.33 Source: Sarvekshana, Vol. XIII, No.2, Oct.-Dec. (1989). Pg. S-218 and S-245 9.2 Both Tables 8 and 9 show that the proportion of households, reporting inadequacy of food (col.9), falls12 as MPCE rises; so does the share of food in total expenditure (col.2). About 18.4 percent of the rural 6.3 percent of the urban households reported themselves to be hungry. Most of this hunger was seasonal in nature; the chronically hungry households were only 0.77 percent in urban and 2.33 percent in the rural sector. It is indeed true that "two square meals" might mean "different sized squares" to different people within a state or across different states. Why should one grudge this? This is the way things appear to be: Except in the imaginary world of precisely fixed nutritional norms, different human beings (of free will) do perceive their food needs differently. Their own subjective views of the adequacy/inadequacy of their habitual food intakes should deserve at least as much attention13 as the precise (but of doubtful relevance) calculations based on simple formulae. 9.3 Using the WHO/FAQ nutritional norms and procedure (which are endorsed by the Indian Council of Medical Research), about 50% of the rural and 58% of the urban households show up to be hungry and undernourished. However, upon analyzing the 1983 CES data in terms of the behaviorally revealed abundance thresholds (points at which adequacy begins to turn into food-richness) of different states, our estimates suggest that only about 15 percent of the rural and 17 percent of the urban (without making any adjustment for differences between rural-urban activity levels) households might be classified as hungry. In contrast, and as per their own self- assessed adequacy of food, around 18 percent rural and 6 percent urban heads of households reported that they (and other members of their households) experienced chronic or seasonal hunger in the course of the year. 9.4 State-wise data, similar to Tables 8 and 9, is also available. However in Table 10 we present summary estimates of the incidence of hunger in the rural sector14, based on householders' own perceptions of adequacy/inadequacy of food available to them throughout the year. In 12/ Estimates of food expenditure by the lowest MPCE class, which are based on very few sample observations, are better ignored. Village after village in many states and in most of the urban blocks, no household is reported to fall in the lowest MPCE class. In fact, even at the stratum level (The country is divided into hundreds of strata and the NSS estimates are built up first at the village and then at the stratum level only; and the state level estimates are simple additions of stratum level estimates), quite often the estimates for the lowest MPCE class may be based on one or two sample observations only. However, we have resisted the temptation to merge this lowest MPCE class with the second lovest, as some people have expressed interest in this extreme part of the iower tail of the distribution. L3/ Peoples' perceptions on such sensitive questions must be canvassed in the most quiet and least-intrusive manner Prior conditioning of the population through political propaganda can bias the results. Politicians and Indian planners became aware of the importance of the matter only after the results of this inquiry (conducted in a staggered manner over the whole of 1983) were released in the summer of 1989. JA/ To save space, we shall hereafter confine our discussion only to the results for the rural sector of different states. -22 - contrast with the overall estimate of 18.4 percent households in rural India reporting hungry, the corresponding proportion is approximately twice as much or more in West Bengal (39.6%), Bihar (37.2%) and Orissa (36.8%). This proportion in Assam, the only other major state in eastern India, is just 16.1%. In the four southern states, Kerala (19.0%), Karnataka (18.7%), Tamil Nadu (17.4%) and Andhra Pradesh (15.5%), the proportion of household reporting hungry is either a little more or slightly less than the national average. Aside from Maharashtra, Madhya Pradesh and Uttar Pradesh, with incidence of self-perceived hunger between 15-11 percent, the proportion of households reporting hungry is indeed very small in Haryana (0.9), Punjab (1.6), Jammu and Kashmir (1.6), Gujarat (2.9), Himachal Pradesh (3.3) and Rajasthan (3.8) in 1983. TABLE 10: PERCENTAGE OF HOUSEHOLDS REPORTING SELF-PERCEIVED INADEQUACY OF FOODt STATES AND ALL-INDIA RURAL SECTOR NSS: 38TH ROUND, JAN.-DEC. 1983 Proportion of Whether all members of the households household got tvo square meals reporting Serial State a day throughout the year? inadequacy of No. of food YES NO Col. 4 + Col. S throughout only some the ycar months of the year (1) (2) (3) (4) (5) (6) 1 Haryana 98.61 0.73 0.12 0.85 2 Punjab 98.26 1.33 0.24 1.57 3 Jammu & Kashmir 97.81 1.51 0.10 1.61 4 Gujarat 96.92 2.78 0.07 2.85 5 Himachal Pradesh 96.45 3.08 0.24 3.32 6 Rajasthan 95.95 3.11 0.69 3.80 7 Uttar Pradesh 88.24 10.39 0.62 11.01 8 Maharashtra 85.80 13.42 0.68 14.10 9 Madhya Pradesh 84.34 13.35 1.72 15.07 10 Andhra Pradesh 84.32 14.69 0.82 15.51 11 Assam 83.87 12.49 3.58 16.07 12 Tamil Nadu 82.10 16.06 1.36 17.42 13 Karnataka 80.91 17.81 0.93 18.74 14 Kerala 80.80 15.29 3.68 18.97 15 Orissa 62.79 31.80 5.02 36.82 16 Bihar 62.51 31.81 5.42 37.32 17 West Bengal 60.31 31.01 8.60 39.61 18 All India 81.09 16.19 2.35 18.54 Notes: Data in cols. 2, 3 & 4 are taken from various tables in Sarvekshna, Vol. XIII, No. 2, Oct.-Dec. (1989). 9.5 Although estimates of the incidence of hunger in India (based on the standard norm of nutritional requirements) work out to be incredibly large in comparison with the low figure of 18.4 percent of the rural households reporting chronic plus seasonal hunger, nevertheless knowing that, on an average, one out of every 5.4 households reported hungry during 1983 is disturbing enough. However even more disturbing is the situation reported by the rural households living in West Bengal, Bihar and Orissa, where one out of every 2.5 households is afflicted with chronic or seasonal inadequacy of food. -23 - X. Different Estimates of Hun&er and its Correlates: (An Exercise Towards Partial Validation of the Estimates) 10.1 Poverty is a multi-dimensional phenomenon. Food hunger is only one of its components - undoubtedly its ugliest dimension. Nevertheless, among other things, lack of adequate clothing and shelter is also included in the estimation of the incidence of absolute poverty. In other words, the proportion of those afflicted with food hunger alone in a population is expected to be less than the proportion of those who, in addition, are also afflicted with the lack of adequate clothing and shelter. However the Food Policy group of the World Bank (1986) estimated that 50 percent of the total population of South Asia was consuming "not enough calories for an active life", whereas the poverty estimators of the World Bank reported that only 29 percent of the South Asian population was estimated to be "extremely poor" in 198515. Inter-se both of these estimates cannot be correct. The real source of this contradiction lies in the uncritical use of conventional international wisdom on energy intake norms for South Asia. The largest bulk of the hungry in South Asia reside in India and our estimates of the proportion of the hungry in Indian population lie in the range of 15-18 percent only - just about a third of what the World Bank's (following in the foot steps of WHO/FAO/UNU) Food Policy group would have us believe. One might also note that our estimate of incidence of hunger in India lies on the expected (lower) side of the World Bank's estimate of the proportion of "extremely poor" in 1985. 10.2 We have presented three different sets of estimates of the incidence of hunger in India in Appendix Table A.3. The causes of hunger are not only diverse but also interact among themselves in a very complex manner. Although an in-depth analysis of the causes of hunger is beyond the scope of this paper, nevertheless, a preliminary exploration of the relationship between hunger and some of its correlates (selective) is presented below. It is evident from Table 8 that as the average MPCE of rural households rises, the proportion reporting (subjective self-reporting in the form of yes or no) inadequacy (chronic & seasonal) of food falls. A systematic fall in the proportion of food expenditure of households is also noticed. However the rank correlation coefficients between average food shares and self-reported hunger at the state level are rather low (Spearman - .47 and Kendall - .34). The corresponding values of these coefficients are even lower (Spearman - .25 and Kendall - .16), when average food shares are correlated with the estimates of caloric inadequacy worked on the basis of adequacy thresholds. However, the rank correlation between these two measures of incidence of hunger themselves is somewhat higher (Spearman - .76 and Kendall - .56). In other words, ranking different states by the proportion of total expenditure spent on food does a poor job as a proxy for the incidence of hunger whether estimated on the basis of subjective responses of households or through caloric inadequacy derived from the actual food consumption data of U2/ World Bank (1986), Poverty and Hunger: Issues and Options for Food Security in Developing Countries, P.17, Table 2-3; and World Bank (1990), World Development Report 1990, P.29, Table 2-1. -24 - the 1983 CES on the basis of behavioral thresholds16. For instance Haryana and Gujarat have very low incidence of hunger but their food shares are quite high. On the other hand, Andhra Pradesh and Kerala have low food shares but the incidence of hunger is high. 10.3 In Table A.3., data on daily status unemployment rates and the incidence of illiteracy (among persons of age 15 years and above) in 1983 are also presented. Among many other factors, which are not easily tractable, the daily status unemployment rates (those unemployed on different days of the previous week divided by all those who were in the labor force - the most comprehensive measure of the rate of unemployment) have a direct bearing on the incidence of hunger. The relation between literacy and hunger is far more complex. It is of interest to examine how the three respective measures of hunger (dependant variable) stack up against the independently estimated rates of unemployment and incidence of illiteracy (two regressors). As we do not have a clear view of the relationship between hunger and illiteracy, the latter is being used here essentially to control extraneous noise. 10.4 Since the labor force participation rates of rural females in India are known to be quite volatile from year to year and over the states, we shall use daily status unemployment rates among rural males in 1983 as the appropriate independent variable impacting hunger. In our view the incidence of illiteracy among females is a more meaningful indicator of the general socio-economic health of different states rather than male or male plus female illiteracy. We therefore propose to use female (15+) illiteracy as the second regressor variable. Regressing the three different estimates of rural hunger, one at a time, on unemployment rates among rural males (XI) and percentage of illiterate rural females (X2) over the 17 major states of India, we get the following results (Table 11). 10.5 The collage of regression results in Table 11 form an interesting picture. The estimates of hunger obtained through the subjective response (yes or no) of households do not show any relationship with the independent estimates of rural male unemployment rates taken alone or together with the incidence of female illiteracy in different states: Only about 5.7% (3.1%) of the overall variance in hunger over the states is explained by Xz (XI and X2)* Although somewhat larger proportion of variance (29.4 and 27.3%) in state-wise estimates of hunger computed on the basis of WHO nutritional norm is explained by X1, and X, & X2, the regression coefficients are insignificant, except when X, (male unemployment rate) is used alone. Nevertheless, only 29 percent of the overall variance in norm-driven hunger over the states is explained by their corresponding unemployment rates. The situation is much better in case of the estimates of hunger computed by using the caloric adequacy threshold method. The estimate of the elasticity of hunger (using the behavioral threshold method) with respect to rural male unemployment remains highly significant (and stable between 0.948 and 0.999) whether this regressor variable X1 is used alone or in conjunction with the incidence of illiteracy. It is worth noting the 67.4 percent (65.3%) of the total variance in the 16/ It has been heard that the Planning Commission (Government of India) is considering to use food shares as a proxy for relative poverty of different states in preference to head count measures based on levels of consumer expenditure below some norms. It would indeed be bizarre to do so: Not to speak of poverty across states, the food shares do not reflect even the relative incidence of hunger (under-nourishment). -25 - state-wise estimates of the undernourished (threshold method) is accounted for by the independent variations in rural male unemployment rates (and female illiteracy). The fit between the threshold-method estimates of hunger and male unemployment rates is more than twice as good as between the norm-driven estimates and unemployment. Incidence of undernutrition and female illiteracy move in opposite directions, nevertheless the coefficients in both cases are statistically insignificant. For every one percent increase in rural male unemployment, the incidence of undernutrition (threshold method) increase by about one percent - the two estimates of this elasticity being .948 and .999. TABLE 11: RELATIONSHIP BETWEEN HUNGER, MALE UNEMPLOYMENT RATES AND FEMALE ILLITERACY Variable Estimate Standard T-Value Prob. Cor. with R2 Dep. Var. A. Self-Reported Hunser Aaant Xl 2 constant -9.899 25.139 -0.394 0.700 - X1 1,121 0.725 1.546 0.144 0.341 (0.0311 X2 0.209 0.270 0.772 0.453 0.256 S. Self-Reported Hunmer Against X1 constant 9.087 5.153 1.763 0.098 X, 0.746 0.531 1.405 0.180 0.241 [0.0571 C. Hunger Based on WHO Norms Aiainst X &2 constant 54.494 21.853 2.494 0.026 - X1 0.954 0.630 1.514 0.152 0.581 [0.273] X2 -0.179 0.235 -0.761 0.459 -0.510 D. Huner Based on WHO Norms Arainst-Xl constant 38.217 4.477 8.536 0.000 - [0.2941 X1 1.275 0.461 2.765 0.014 0.581 E. Hunger Computed Throuih Threshold Method Against X1h& X2 constant 9.121 8.250 1.106 0.288 - X1 0.948 0.238 3.985 0.001 0.833 (0.653) X2 -0.028 0.089 -0.321 0.753 -0.593 F. Hunger Computed Throuih Threshold Method Against X1 constant 6.534 1.662 3.931 0.001 X1 0.999 0.171 5.834 0.000 0.833 10.6741 10.6 To conclude, the estimates of the proportion of under-nourished households in different states, derived from the 1983 CES data on meal accounts and behavioral thresholds of caloric abundance adequacy in habitual food intakes of households, do pass a partial validation test (partial, because the causes of hunger are too diverse to be covered by the unemployment rates alone) far more successfully in comparison with the norm-driven estimates based on WHO/FAO method. -26 - XI. Some Concluding Remarks 11.1 The science of nutrition is in turmoil tod&y. Physiologists, nutritionists, biometricians and economists are having great difficulties in communicating among themselves. Exchanges among them seem to resemble a dialogue among the deafl?. It is not surprising, therefore, that both conceptual and empirical difficulties are being encountered in assessing the incidence of undernutrition with reference to the traditional approach embodied in the WHO/FAQ nutritional norms. This paper has voiced our own misgivings with the WHO/FAQ approach. We have suggested a new empirical approach, using a behavior-based, caloric-adequacy-threshold cziterion in conjunction with householders' personal opinions of adequacy/inadequacy of their habitual food intakes. We have shown that this threshold approach seems to provide not only more information relevant to the assessment of nutritional status of a population, but the information thus gained is also (in comparison with the norm-driven estimates of hunger) far more consistent with other independently available information relating to an important cause of hunger, i.e., incidence of daily status rural male unemployment across states. 11.2 We did not expect a close correspondence between the estimates of the extent (spread not depth) of hunger (generated from yes or no replies to the question: Whether all members of the household got two square meals a day throughout the year?) and exact measure of the depth of undernourishment worked out (by using the adequacy thresholds) from distribution of households over different ranges of caloric intakes. Nevertheless, one is happy to note that the ball-park estimates of the extent (18.4%) of self-reported hunger, largely seasonal and marginally chronic, in rural India as a whole (and different states) do come out mrch closer to the estimates of caloric inadequacy (14.6%) derived through the behavioral approach. In sharp contrast, the norm-driven estimate of the undernourished turned up an incredibly large figure of 50.2 percent for rural India as a whole; 33.5 percent for Punjab and 66.6 percent for Kerala. We are of the view that properly conducted large scale inquiries on self-perceived adequacy/inadequacy of food intakes are a good supplement to the usual CESs for assessing the incidence of undernutrition and hunger, provided the two inquiries are dovetailed together and canvassed from the same sample population. 11.3 We are not sure whether the current theoretical debate on nutritional adaptation and intra-individual variability will be resolved soon. Nevertheless our empirical results, indicating that the caloric adequacy/abundance thresholds across states range from around 70 percent (or less) of the WHO/FAQ norm in Karnataka, Gujarat, Himachal Pradesh and Jammu and Kashmir to around 106-107 percent in Punjab and Madhya Pradesh, may assist some further speculation on theoretical issues. The prevailing genetic pool of the Indian population seems to be interacting differently with different environments of culture and food histories in different regions. For 17/ For recent evidence of this, see, for example, European Journal of Clinical Nutrition (1989), Vol. 43, pp. 75-87 and 203-210. This volume carries an article by P.V. Sukhatme and three commentaries on this article by J.C. Waterlow, W.P.T. James and M.R.J. Healy. Also see, Durnin, J.V.G.A. (1990), "Is Satisfactory Energy Balance Possible on Low Energy Intakes," Bulletin of Nutrition Foundation of India, Vol. 11, No. 2, pp. 1-4 and comment on Durnin by Gopalan, C., same issue, pp. 5-6. -27 - instance, the abstemious people of Gujarat seem to maintain their energy balance with ease at a much lower threshold level of energy intakes than the people of Punjab where copious food consumption is considered as the primary component of "good" living. However, one would be hard put to find a significant difference between the two populations in terms of patterns of gainful work and social activities in the two regions. -28 - APPENDIX A: TABLES AND FIGURES TABLE A.1: CONSUMER UNIT BY AGE AND SEX OF PERSON age group Sex (in completed years) ma 1 e fema l e (1) (2) (3) under 1 0.43 0.43 1 - 3 0.54 0.54 4 - 6 0.72 0.72 7 - 9 0.87 0.87 10 - 12 1.03 0.93 13 - 15 0.97 0.80 16 - 19 1.02 0.75 20 - 39 1.00 0.71 40 - 49 0.95 0.68 50 - 59 0.90 0.64 60 - 69 0.80 0.51 70 + 0.70 0.50 -29- TABLE A.2: AVERAGE NUMBER OF MEALS DISTRIBUTED BY A HOUSEHOLD TO EMPLOYEES (NON-MEMBERS) DURING A PERIOD OF 30 DAYS BY STATES Serial Rural Urban No. Name of State Sector Secto (0) (1) (2) (3) 1 Andhra Pradesh 1.54 0.41 2 Assam 0.35 0.29 3 Bihar 0.70 0.44 4 Gujarat 2.26 1.14 5 Haryana 2.30 3.06 6 Himachal Pradesh 0.74 3.51 7 Jammu & Kashmir 3.57 1.35 8 Karnataka 3.36 1.10 9 Kerala 3.15 2.46 1u Madhya Pradesh 0.61 0.06 11 Maharashtra 1.04 0.32 12 Orissa 1.28 0.35 13 Punjab 7.95 1.11 14 Rajasthan 0.91 0.12 15 Tamil Nadu 1.76 1.50 16 Uhar Pradesh 0.64 0.43 17 West Bengal 3.31 1.15 18 All India 1.6 0.83 -30- TABLE A.S.t DIFFERENT ESTIMATES OF THE PROPORTION OF HUNGRY (UNDERNOURISUIED) BOUSENOS, FOOD SHARES IN CONSUMER EXPENDITURE, DAILY STATUS UNEMPLOYHENT RATES AND INCIDENCE OF ILLITERACY AMONG PERSONS OF AGE 15 YEARS AND ABOVE OVER DIFFERENT STATES RURAL SECTOR CSS: 38TH ROUND, JAN.-DEC. 1983 Illiteracy souseholds tu Dgnmer Food Share Unemloyment Rates (Personst 154) Serial State Self-Reported Caloric Caloric No. (Chronic & (Behavioral (Nutritional Males Females Males Females Seasonal) Thresholds) Norm) (0) (1) (2) (3) (4) (5) (6) (7) (8) (9) 1 Haryana 0.85 9.20 35.69 63.58 6.69 2.95 47.56 88.15 2 Punjab 1.57 8.81 33.49 58.72 6.97 9.25 51.32 71.09 3 Jama I Kashmir 1.61 5.73 29.34 69.64 8.55 2.85 61.37 87.27 4 Gujarat 2.85 16.62 58.54 66.13 5.15 4.77 45.29 74.16 SHimachal Pradesh 3.32 6.33 27.56 63.01 2.24 0.81 43.06 68.40 6 Rajasthan 3.80 10.61 36.15 60.74 3.50 1.55 63.87 93.05 7 Uttar Pradesh 11.01 9.43 40.10 63.32 3.65 2.46 55.83 88.12 8 Maharashtra 14.10 13.35 54.59 61.49 6.25 7.23 41.43 76.78 9 Madhya Pradesh 15.07 9.18 43.24 66.52 2.07 1.81 57.41 89.27 W 10 Andhra Pradesh 15.51 11.80 48.58 60.30 7.87 10.54 62.44 86.08 11 Assam 16.07 12.62 61.78 73.36 3.47 5.98 33.44 62.61 12 Tamil Nadu 17.42 26.83 62.79 65.05 17.59 20.53 40.39 73.21 13 Karnataka 18.74 18.48 51.20 63.47 6.61 8.32 51.73 79.38 14 Xerala 18.97 29.00 66.55 61.67 24.31 31.01 13.08 26.81 15 Orissa 36.82 18.24 55.89 73.63 7.82 11.79 47.69 79.92 16 Bihar 37.32 14.41 50.63 73.64 7.06 10.66 58.37 90.21 17 West Bfental 39.61 24.47 64.48 74.01 14.36 24.01 38.65 72.04 18 All India 18.54 14.58 50.17 65.50 7.52 8.98 49.76 79.75 Notes: FLgures in cols. 2. 3 and 4, respectively, recalled (derived) from Tables 10 and 6 of this papers Col. 5 has been constructed from various state tables in Sarvekshana, Vol. IX, No. 4, April (1986). Cols. 6, 7, 8 & 9 have taken from Sayrkshanas Vol XI, No. 4, April (1988) pp. 61, 62 and 18. TABLE A.4: PERCENTAGE OF HOUSEHOLDS (I) REMAINING IN THE SAME RANGE (II) MOVING UPWARDS AND (III) MOVING DOWNWARDS IN THE CALORIC INTAKE LEVEL SCALE AFTER ADJUSTMENT FOR FREE MEALS RURAL KARNATAKA (Combined Sample) NSS: 38th Round, Jan.-Dec.1983 Per diem per PERCENTAGE OF HOUSEHOLDS consumer unit caloric intake moving remaining in moving level* downward the same range upwards (1) (2) (3) (4) lers than 25% - 83.33 16.67 25 - 50 1.28 82.55 16.17 50 - 70 1.68 89.72 8.60 **70 - 80 10.03 83.18 6.79 80 - 90 10.26 81.12 8.62 90 - 100 17.46 77.06 5.48 100 - 110 16.40 75.63 7.97 110 - 120 24.31 63.59 12.10 120 - 150 16.12 79.47 4.41 150 - 200 20.76 75.67 3.57 200 - above 38.82 61.18 - Notes: *Percentage to normative requirement of 2700 calories per consumer unit per diem. **Indicates location of the point of zero net displacement (70.22% or 1896 calories) after adjustment for free meals. -32- Figure A.1: Adjusting Intakes for Free Meals (An Illustrative Diagram) Calories Reported Intakes Rural Karnataka 1896 Intakes Adjusted for Free Meals Poor Households Rich Households -33- References 1. Durnin, J.V.G.A. (1990), "Is Satisfactory Energy Balance Possible on Low Energy Intakes," Bulletin of Nutrition Foundation of India, Vol. 11, No.2, pp.1-4; and comments on Durnin's article by Gopalan, C. (1990), "Low Energy Intakes," same issue, pp. 5-6. 2. Minhas, B.S. (1988), "Validation of Large Scale Sample Survey Data: Case of NSS Household Consumption Expenditure," Sankhya, Series B, Vol.50, Part 3, Supplement, pp. 1-63. 3. Minhas, B.S. and Kansal, S.M. (1989), "Comparison of NSS and CSO Estimates of Private Consumption: Some Observations Based on 1983 Data," The Journal of Income and Wealth, Vol. 11, No. 1, pp. 7-24. 4. National Sample Survey Organization (1989), Tables With Notes on the Effect of Adjustment of Energy Intake For Meals Consumed Free and Meals Served to Others, Department of Statistics, Government of India, New Delhi (Mimeographed). 5. Sukhatme, P.V. and Margen, S. (1982), "Auto-regulatory Homeostasis Nature of Energy Balance," American Journal of Clinical Nutrition, Vol. 35, pp. 355-365. 6. Sukhatme, P.V. (1989), "Nutritional Adaptation and Variability," European Journal of Clinical Nutrition, Vol. 43, pp. 75-87; and Three Commentaries on the Paper by Professor P.V. Sukhatme by Waterlow, J.C., James, W.P.T., and Healy, M.R.J. in the same issue, pp. 203-210. 7. World Health Organization (1985), Energy & Protein Reauirements: Report of a Joint FAO/WHO/UNU Expert Consultation, Geneva. 8. World Bank (1986), Poverty_and Hunger:_ Issues and-0ptions for Food Security in Developing Countries, p. 17, Table 2-3. 9. World Bank (1990), World Development Report 1990, p. 29, Table 2-1. -34 - ASIA REGION DISCUSSION PAPER SERIES Title Author Date Originator IDP74 A Case Study of a Gradual Approach to Economic Reform: The Vict Nam Experience of 1985-88 Z. Drabek September 1990 Z. Drabek (80504) IDP85 On Estimating Inadequacy of Energy Intakes: Rcvealed Food Consumption Behavior versus Nutritional Norms B.S. Minhas September 1990 S. Jayanthi (81419) IDP88 Asia Region Seminar on Policy Challenges in India October 1990 C. Chamberlin (81409) Note: Extra copies may be obtained from the Asia Information Service Center.