Choice of Reference Period for Consumption Data

Transcription

1 Report No. 447 Choice of Reference Period for Consumption Data based on NSS 51st Round (July 1994 June 1995) NSS 52nd Round (July 1995 June 1996) NSS 53rd Round (January December 1997) NSS 54th Round (January June 1998) National Sample Survey Organisation Department Of Statistics Government Of India March 2000

2 PREFACE In 1972 the National Sample Survey Organisation (NSSO) began a quinquennial series of all-india surveys on consumer expenditure, employment & unemployment. In addition to this, an annual series of consumer expenditure surveys was started in 1986, which had other subjects of enquiry as their focus. As a result, an annual series of data on consumer expenditure has since become available. For the study on reference period, two separate household consumer expenditure schedules were designed for data collection in the 51 st, 52 nd, 53 rd and 54 th rounds of NSS. The main difference between the two types of schedule was in respect of reference period. Schedule type 1 had a uniform reference period of last 30 days for all items of consumption. Schedule type 2 had as reference period of the last 7 days for food items, pan, tobacco and intoxicants, the last 365 days for clothing, footwear, durable goods, education and institutional medical expenses, and the last 30 days for all other items. The present report contains five chapters and three appendices. Chapter 1 states the problems studied in this report while Chapter 2 gives its background. Chapter 3 explains the concepts and definitions involved in a consumer expenditure survey of NSSO. Chapter 4 summarises the findings of the report. Chapter 5 discusses averages and distributions of overall monthly per capita consumption expenditure level MPCE obtained by the two schedule types. Appendix A gives State-level tables and corresponding all-india tables. Appendix B gives detailed tables at national level. Details of the sampling designs and estimation procedures adopted in the four rounds are explained in Appendix C. The fieldwork for the survey was handled by the Field Operations Division, the data was processed by the Data Processing Division, and the report was prepared by the Survey Design and Research Division of NSSO. I am grateful to the members of the Governing Council of NSSO, the Heads of NSSO s Divisions and their colleagues for their contributions at various stages. The report, I hope, will be useful in generating discussion on sampling and non-sampling errors in the collection of data on different items/groups of consumption expenditure from households. Comments/suggestions from readers will be most welcome. New Delhi March 2000 N. S. SASTRY Director-General & Chief Executive Officer National Sample Survey Organisation

3 Contents Chapter 1 The Problem Stated 1 Chapter 2 The Report and its Background 3 Chapter 3 Consumer Expenditure: Concepts & Definitions 7 Chapter 4 Summary of Findings 12 Chapter 5 MPCE by Schedule Type 1 and Schedule Type 2 31 Appendix A Tables Giving State-Level Estimates A-1 Appendix B Detailed Tables at National Level B-1 Appendix C Sample Design & Estimation Procedure C-1 List of tables Appendix B contains detailed tables at the national level. Appendix A gives summary tables at State-level. Unless otherwise mentioned, all tables give separate estimates for rural and urban areas. State-level tables give estimates for the 15 major States - Andhra Pradesh, Assam, Bihar, Gujarat, Haryana, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Orissa, Punjab, Rajasthan, Tamil Nadu, Uttar Pradesh and West Bengal - and three groups of States and UTs formed from the remaining ( minor ) States and UTs of India. Appendix A: Tables Giving State-Level Estimates Statement 1 Per thousand distribution of persons by MPCE class for 15 major States, three groups of small States and UTs and all-india Statement 1RSE RSE of estimated per thousand distribution of persons by MPCE class for 15 major States, three groups of small States and UTs and all-india Statement 2 Quantity and value of monthly average consumption of different cereals per person for 15 major States, three groups of small States and UTs and all-india Statement 2RSE RSE of estimated quantity and value of monthly average consumption of different cereals per person for 15 major States, three groups of small States and UTs and all-india Statement 3 Average monthly expenditure (Rs) per person on 28 groups of items of consumption for 15 major States, three groups of small States and UTs and all-india Statement 3RSE RSE of estimated average monthly expenditure (Rs) per person on 28 groups of items of consumption for 15 major States, three groups of small States and UTs and all-india A-1 A-17 A-25 A-41 A-49 A-65

4 Appendix B: Detailed Tables at National Level Table 1 Table 1RSE Table 2 Table 2RSE Table 3 Table 3RSE Per 1000 distribution of households and persons by monthly per capita expenditure (MPCE) class and number of persons and economically active persons per 1000 households in different MPCE classes RSE of estimated per 1000 distribution of households and persons by monthly per capita expenditure class and number of persons and economically active persons per 1000 households in different MPCE classes Quantity (kg) and value (Rs) of consumption of cereals, gram and cereal substitutes per person for a period of 30 days for each monthly per capita expenditure class RSE of estimated quantity and value of consumption of cereals, gram and cereal substitutes per person for a period of 30 days for each monthly per capita expenditure class Value (Rs) of consumption of broad groups of food & non-food items per person for a period of 30 days for each monthly per capita expenditure class RSE of estimated value of consumption of broad groups of food & nonfood items per person for a period of 30 days for each monthly per capita expenditure class B-1 B-9 B-13 B-29 B-37 B-53

5 Highlights In the 51 st, 52 nd, 53 rd,and 54 th rounds of NSS, two consumer expenditure schedules were designed to judge the suitability of different reference periods for collecting data on consumption expenditure on 22 groups of items of consumption, including 14 food groups and 8 non-food groups. For nearly all the food groups considered, as well as for pan, tobacco and intoxicants, all-india estimates of per capita consumption with reference period 30 days showed lower RSEs than those based on a reference period of 7 days. However, RSEs arising from schedule type 1 and schedule type 2 do not differ very substantially for these item groups. The RSE criterion does not really clinch the issue in favour of schedule type 1, and the question of bias becomes all-important. For clothing, footwear and durable goods, all-india estimates of per capita consumption with reference period 365 days showed lower RSEs than those based on a reference period of 30 days. The above differences in magnitude of RSEs are seen at State level as well. For food items, pan, tobacco and intoxicants, the all-india estimates of per capita consumption were higher with the 7-day reference period than with the 30-day reference period. The extent of difference was, on the average, about 32% for the food groups and higher for pan, tobacco and intoxicants. The extent of difference was usually under 15% for cereals and also relatively low for milk and milk products, edible oil and sugar, but more than 40%, on the average, for all other food groups (over 50% for meat-fish-egg and vegetables, nearly 60% for beverages, etc. and over 70% for spices and fresh fruits). For 10 item groups, nearly all major States (14 or 15 out of 15) showed higher consumption estimates with the 7-day reference period than with 30 days. For the remaining 7 item groups for which 7 days and 30 days are under comparison, 11 or more states show higher consumption estimates with 7 days. For clothing and footwear, neither of the two reference periods 30 days and 365 days gave consistently higher estimates than the other. For education, medical (institutional) and durable goods, the 365 days period gave consistently lower estimates of all-india per capita consumption, difference being, on the average, 24% lower for education, 36% lower for medical (institutional) and 47% lower for durable goods.

6 For education, medical (institutional) and durable goods, a reference period of 30 days gave higher estimates than a reference period of 365 days in at least 9 of the 15 major States for both rural and urban India in all three rounds. For clothing and footwear, estimates by the longer and the shorter reference period do not differ a great deal, while RSEs are much lower for the 365-days-based estimates. This suggests that it is the longer reference period, which is more appropriate for these two item groups. While the 365-days-based estimator for durable goods undeniably has lower RSE than the 30-days-based estimator, the suspicion of a serious downward bias prevents the acceptance of the former without any modification. ****

7 1 The Problem Stated The objective of the NSS consumer expenditure survey is to ascertain the per capita expenditure on consumption, organized under different consumption heads, of various geographical and socio-economic categories of population during a specified period, usually one year. Apart from average per capita consumption of the population of, say, a State, the survey also investigates the distribution of households by per capita consumption level and the distribution of persons by per capita consumption level of the households to which they belong Instead of asking sample households to report a full year s consumption, the NSS survey technique is to interview different households in different months of the year, and ask each household about its consumption during the last month (sometimes, also during last year for some items). Thus the reference period is usually a month. Often, different reference periods are used for different items, there being no contradiction in asking the informant to report a month s consumption of onions and a year s consumption of refrigerators as long as no mistakes are made in calculation and recording of numbers. Obviously, with shorter reference periods, one encounters fewer recall errors (i.e. reporting errors due to inability to remember) Because the interviews are spread evenly over the year, the estimates do not suffer from seasonality, and can be said to relate to the whole year. For easy comprehensibility, the data based on different reference periods are converted to a common denomenator of duration of a month. The estimates are then expressed as monthly per capita consumption expenditure The purpose of this report is to investigate two specific questions: [1] Which of the reference periods, last week and last month, is more suitable for collection of consumption data on the item groups (a) cereals, (b) milk and milk products, (c) vegetables, and so on for 14 groups of food items and 3 groups of non-food items? [2] Which of the reference periods, last month and last year, is more suitable for collection of consumption data on the item groups clothing, footwear, medical expenses (institutional), education and durable goods? To permit an investigation of this question, two separate household consumer expenditure schedules were designed for use in the 51st, 52nd, 53rd and 54th rounds of NSS. The main difference between the two schedule types was in respect of reference period. Schedule type 1 had a uniform reference period of last 30 days for all items of consumption. Schedule type 2 had as reference period the last 7 days for food items, pan, tobacco and intoxicants, the last 365 days for clothing, footwear, durable goods, education and institutional medical expenses, and the last 30 days for all other items. 1

8 The Problem Stated Before we proceed any further, it is necessary to clarify what we mean by a more suitable reference period A priori, we may expect a very short reference period to yield estimates of very high variability. This is because the true consumption figures of households for a very short period will themselves show a very high variation. Thus if the reference period, to take an extreme example, is last one hour, then it is possible that household H s consumption of, say, milk, will be reported as zero, 200 ml, or 500 ml (usually zero) depending on what time of day is chosen by the investigator for the interview. But, possibly, the same household H s consumption of milk will be reported as 1 litre if the reference period is last 24 hours (and as 30 litres if the reference period is last 30 days ) irrespective of the time of interview. This explains why very short reference periods are not chosen even though reporting is likely to be 100% accurate Why is the longer reference period, then, not always chosen? Because although true consumption becomes less variable as the reference period is lengthened, reported consumption gets affected by recall errors. This is less likely to occur in case of milk or salt, but quite likely for fruits fresh and dry, processed food, and electrical repairs, to name just three item groups Recall errors are of two kinds: [A] the failure to recall the fact that an article was consumed; and [B] the failure to recall the date when an article was consumed. A-type errors occur for items not specifically listed by name in the schedule such as chocolates, plastic containers and water filters. B-type errors occur for all items except those which are consumed daily (or very regularly such as every Sunday), e.g. salt, major cereals and spices. Thus an apple consumed 9 days ago may be reported by mistake as consumed last week. Fortunately, errors of wrong inclusion and wrong exclusion should cancel out for B-type errors, unless the situation is complicated by other psychological factors such as social prestige, which prevents the informant from reporting consumption of fruits as zero and makes her/him give the benefit of the doubt to the orange consumed 7 or maybe 8 or 9 days ago. In other words, very little bias would ordinarily be expected to arise from errors of type B. On the other hand, A-type errors always cause a downward bias in the data and the estimates of consumption. This bias may be quite substantial with a long reference period such as a year, except in case of very expensive, never-forgotten purchases such as TV sets and refrigerators, which are in any case listed in the schedule and ought not to be missed The task, then, is to strike a balance between the short reference period, which causes more accurate reporting but greater sampling fluctuations, and the long reference period, which gives stable consumption figures but a hidden downward bias because of recall lapses. 2

9 2 The Report and its Background In the formative years of NSS surveys, considerable attention was paid to the length of reference period suitable for ascertaining the correct volume of consumption of different items of goods and services In the 2nd round of NSS, conducted during April to June 1951, the reference period for items of consumption, mainly food, which are very frequently purchased, was changed to a week from a month preceding the date of interview. For a group of other items, less frequently purchased, the reference period was taken as a month and for clothing and articles of comparatively durable nature, a reference periods of one year was considered During the 4th round of NSS (April-September 1952), there were, for the first time, two different reference periods - a week and a month - for a number of items A on pattern of consumer expenditure was brought out covering the period April 1951 to March 1954 based on the second to seventh rounds of the survey giving some findings on suitability of reference period. It revealed that the degree of difference between any two estimates with week and month reference period did not manifest any consistency A special investigation on reference period * was carried out in March-April 1952 in 76 villages of West Bengal covering 1254 households. The households were divided into two groups. From the first group particulars of certain food items were obtained by interrogation while from the second group the particulars of consumption were obtained by actual weighing of these items by the field staff. Items under study were clean rice, pulses, sugar and salt. The measurement was done in the following way: the field staff requested households to place the supplies for about 10 days in some suitable receptacles and asked them to draw from these stocks for consumption during the next 7 days. The field staff weighed the opening and came back a week latter and weighed the balance left. The difference gave the quantity of consumption. In one group of households the weighing was done with an interval of 7 days while for another group the weighing was done twice in 7 days, once after 3 days and again after 4 days. These operations were repeated for 3 weeks. The starting days of these 7-day periods for the different households were scattered over the different days of the week. For the sample households where the data were obtained by interrogation, from roughly half of the Report No. 20: Pattern of consumer expenditure * see article by P.C.Mahalanobis and S.B.Sen in Bulletin of the International Statistical Institute, Vol.34, Part II,

10 The Report and its Background particulars were collected with reference to a period of week and from the other half particulars for a month collected. Whatever be the reason of the differences in estimates, chance or otherwise, the figures obtained by interrogation with month as reference period agreed closely with the figures obtained by weighing. However, the figures with reference period week were higher than the figures with reference period month. The results indicate that data collected with month as reference period are more satisfactory The first quinquennial survey on household consumer expenditure was carried out during the 27th round of NSS (July 1972 to June 1973). Since then five such quinquennial surveys have been conducted, the last being in the 50th round ( ). In these surveys, for items of clothing, footwear and durable goods, information was collected from each sample household for two reference periods - last 30 days and last 365 days. In the 50th round, educational and institutional medical expenses were also added to the list of item groups for which data were collected for two different reference periods In between, a pilot enquiry (February 1986 January 1987) was conducted to ascertain the most suitable reference period for collecting data on consumer expenditure and also to examine the effect of a short-cut schedule vis-à-vis the usual NSS consumer expenditure schedule. The enquiry was carried out in 7680 sample households spread over 204 sample villages and 180 urban blocks of five States, viz., Maharashtra, Tamil Nadu, Uttar Pradesh, Haryana, Orissa and two metropolitan cities, Delhi and Calcutta. However, processing problems prevented the study from being completed The foundations of this report were laid with an experiment with two consumer expenditure schedules during the 51st round of NSS. In this round, one-half of the sample of households was subjected to schedule type 1, which had a reference period of 30 days throughout. In the other half of the sample, a schedule with different reference periods for different groups of items was tried out. The reference periods for schedule type 2 were: food, pan, tobacco and intoxicants fuel & light, miscellaneous goods & services and medical (noninstitution-al) education, medical (institu-tional), clothing, footwear and durable goods. : last 7 days : last 30 days : last 365 days This experiment continued in three subsequent rounds: the 52nd, 53rd and 54th As usual, the first-stage units (sample villages and blocks), in these rounds, were selected in the form of two independent samples called sub-sample 1 and sub-sample 2. In the 51st and 54th rounds of survey, schedule type 1 was canvassed in sub-sample 1 and schedule type 2 in sub-sample 2. In the 52nd and 53rd rounds, each schedule type was spread over both sub-samples with a view to getting an estimate of the sampling variance of the estimators of per capita consumption. 4

11 The Report and its Background To release the estimates of consumer-expenditure and employment of the 51st, 52nd, 53rd and 54th rounds as part of the annual series of consumer expenditure surveys (which by convention used, prior to the 51st round, a uniform reference period of 30 days), a separate report * on consumer expenditure (combined with employmentunemployment) was prepared for each of these rounds, the presentation of estimates being largely confined to those based on the 30-day reference period Sample size: At all-india level, the number of sample households with persons surveyed in each round were given in Table T 0. Table T 0 : Number of sample households and persons surveyed all-india number of surveyed households persons round sector stype 1 stype 2 combined stype 1 stype 2 combined 51 rural urban rural urban rural urban Geographical coverage: The geographical coverage of the survey was the same for all four rounds - 51st, 52nd, 53rd and 54th. The entire area of the country - rural and urban - was covered, with the exception of some interior areas of Nagaland and Andaman & Nicobar Islands, and Doda, Anantnag, Pulwama, Srinagar, Badgam, Baramula, Kupwara, Ladakh and Kargil districts of Jammu & Kashmir which were left out owing to unfavourable field conditions Contents of the report: This report contains three more chapters and three appendices. The next chapter, Chapter 3, explains the concepts and definitions involved in a consumer expenditure survey of NSSO. Chapter 4 summarises the findings on which reference period is more suitable, for each group of consumption items on which data were collected for two reference periods. Chapter 5 discusses averages and distributions of overall monthly per capita consumption expenditure level MPCE (widely used as a living-standards indicator based on the NSS consumer expenditure survey) obtained by the two schedule types. Appendix A gives State-level tables and corresponding all-india * Report No. 436(51/1.0/1): Household Consumer Expenditure and Employment Situation in India, Report No. 440(52/1.0/1): Household Consumer Expenditure and Employment Situation in India, Report No. 442(53/1.0/1): Household Consumer Expenditure and Employment Situation in India, 1997 Report No. 448(54/1.0/1): Household Consumer Expenditure and Employment Situation in India 5

12 The Report and its Background tables. Appendix B gives detailed tables at national level. Details of the sampling designs and estimation procedures adopted in the three rounds are explained in Appendix C. Notes 1. The National Sample Survey: The NSS was set up in 1950, with the idea of having a permanent survey organisation to collect data on various facets of the economy through nation-wide sample surveys in order to assist in socio-economic planning and policymaking. The NSS is a continuing survey in the sense that it is carried out in the form of successive "rounds", each round usually of a year's duration covering several topics of current interest in a specific survey period. At present, each NSS round reaches, at the all-india level, about 12,000 to 14,000 villages and blocks in the Central sample (covered by the Central agency NSSO) and an independent sample of about 14,000 to 16,000 villages and blocks in the State sample (covered by the Governments of various States and Union Territories). 2. Household consumer expenditure survey: It was originally the practice to collect data on household consumer expenditure in every round of NSS, starting from the very first round ( ). From , according to a decision of the Governing Council of NSSO, the consumer expenditure survey became a quinquennial feature. It was also integrated with the employment and unemployment survey in the sense that a common sample of households was subjected to both the enquiries. An annual series of relatively small scale consumer expenditure surveys was again launched from the 42nd round ( ) to fill the data gaps which planners and researchers had begun to perceive. The 51st, 52nd and 53rd round surveys of consumer expenditure all belong to this annual series. 6

13 3 Consumer Expenditure: Concepts and Definitions Household: A group of persons normally living together and taking food from a common kitchen constitutes a household. The word normally means that temporary visitors are excluded but temporary stay-aways are included. Thus a son or daughter residing in a hostel for studies is excluded from the household of his/her parents, but a resident employee or resident domestic servant or paying guest (but not just a tenant in the house) is included in the employer/host's household. Living together is usually given more importance than sharing food from a common kitchen in drawing the boundaries of a household in case the two criteria are in conflict; however, in the special case of a person taking food with his family but sleeping elsewhere (say, in a shop or a different house) because of space shortage, the household formed by such a person's family members is taken to include the person also. Each inmate of a mess, hotel, boarding and lodging house, hostel, etc., is considered a single-member household except that a family living in a hotel (say) is considered to be one household only; the same applies to residential staff of such establishments Adult: Any person who has completed 15 years of age is considered an adult Household size: The size of a household is the total number of persons in the household Household consumer expenditure: The expenditure incurred by a household on domestic consumption during the reference period is the household's consumer expenditure. The household consumer expenditure is the total of the monetary values of consumption of various groups of items, namely, (i) food, pan (betel leaves), tobacco, intoxicants and fuel & light, (ii) clothing and footwear, and (iii) miscellaneous goods and services and durable articles For groups (i) and (ii), the total value of consumption is derived by aggregating the monetary value of goods actually consumed during the reference period. An item of clothing and footwear would be considered to have been consumed if it is brought into maiden or first use during the reference period. The consumption may be out of (a) purchases made during the reference period or earlier; (b) home grown stock; (c) receipt in exchange of goods and services; (d) any other receipt like gift, charity, borrowing; and (e) free collection. Home produce is evaluated at the ex farm or ex factory rate. For evaluating the consumption of the items of group (iii), i.e. items categorised as miscellaneous goods and services and durable goods, a different approach is followed. In this case, the expenditure made during the reference period for the purchase of goods and services is considered to be the value of consumption. 7

14 Consumer Expenditure: Concepts and Definitions It is pertinent to mention here that the consumer expenditure of a household on food items relates to the actual consumption by the normal resident members of the household and also by the guests, whether during ceremonies or otherwise. To avoid double counting, transfer payments like charity, loan advance, etc., made by the household are not considered consumption for items of groups (i) and (ii), since transfer receipts of these items have been taken into account. However, the item cooked meals is an exception to the rule. Meals prepared in the household kitchen and provided to the employees and/or others would automatically get included in the domestic consumption of employer (payer) household. There is a practical difficulty in estimating the quantities and values of individual items used for preparing the meals served to employees or to others. Thus, to avoid double counting, cooked meals received as perquisites from employer household or as gift or charity are not recorded in the recipient household. As a general principle, cooked meals purchased from the market for consumption of the normal resident members and of guests and employees will also be recorded in the purchaser household This procedure of recording cooked meals served to others in the expenditure of the serving households only leads to bias-free estimates of average per capita consumption as well as total consumer expenditure. However, as the proportions of donors and recipients of free cooked meals are likely to vary in opposite directions over the expenditure classes, the nutritional intake derived from the consumer expenditure survey data may present a somewhat distorted picture. These derived nutrition intakes may get inflated for the rich (net donors) and somewhat understated for the poor (net recipients). This point has to be kept in mind while using the NSS consumer expenditure data for any nutritional studies relating to the nutritional status of households Value of consumption: Consumption out of purchase is evaluated at the purchase price. Consumption out of home produce is evaluated at ex farm or ex factory rate. Value of consumption out of gifts, loans, free collection and goods received in exchange of goods and services is imputed at the rate of average local retail prices prevailing during the reference period Monthly per capita consumer expenditure (MPCE): For a household, this is its 30 days total consumer expenditure divided by its size. A person s MPCE is understood as that of the household to which he or she belongs MPCE class: For classifying households or persons by monthly per capita expenditure level, 12 classes were formed in each sector - rural and urban. The MPCE classes were formed as follows: MPCE level Rural (Rs) Urban (Rs) MPCE level Rural (Rs) Urban (Rs) 1 below 120 below & above 1055 & above 8

15 Consumer Expenditure: Concepts and Definitions The class limits for rural/urban India were chosen in such a way that the first and last two classes contained 5% of the rural/urban population and the remaining eight classes contained 10% of the population, according to the 50th round survey results Major States: The term will be used to denote the following States of India: Andhra Pradesh, Assam, Bihar, Gujarat, Haryana, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Orissa, Punjab, Rajasthan, Tamil Nadu, Uttar Pradesh and West Bengal. The remaining States and Union Territories of India are divided into three groups, viz. North- Eastern, North-Western and Southern, for the purpose of presentation of estimates. These groups have been formed as follows: North-Eastern: Arunachal Pradesh, Manipur, Meghalaya, Mizoram, Nagaland, Sikkim and Tripura North-Western: Jammu & Kashmir, Himachal Pradesh, Chandigarh and Delhi Southern: Andaman & Nicobar Islands, Dadra & Nagar Haveli, Goa, Daman & Diu, Lakshadweep and Pondicherry Reference period: In collecting data on consumption, the adoption of a reference period always involves a difficult choice (see Chapter 1 for a fuller discussion). Too short a reference period (such as one day) gives rise to irregular data as a household s consumption expenditure does fluctuate a great deal from day to day, or from week to week. Too long a reference period, on the other hand, can introduce a good amount of recall error into the data. The annual series of consumer expenditure surveys, up to the 49th round, used a uniform reference period of last 30 days for all items of consumption. In the bigger surveys of the quinquennial series, an additional reference period of last 365 days was used for some items of consumption - particularly, clothing, footwear and durable goods - but most results were tabulated using the last 30 days data. During the 51st and 52nd rounds, one-half of the sample of households were subjected to schedule type 1, which had a reference period of 30 days throughout. In the other half of the sample, a schedule (schedule type 2) with different reference periods for different items was tried out on an experimental basis. The reference periods for schedule type 2 were: food, pan, tobacco and intoxicants fuel & light, miscellaneous goods & services, and medical (non-institutional) educational, medical (institutional), clothing, footwear and durable goods : last 7 days : last 30 days : last 365 days Milk and milk products: This includes ghee, butter, curd, ice-cream, etc. Milk-based sweetmeats (pera, rosogolla, etc.) come under beverages, refreshments and 9

16 Consumer Expenditure: Concepts and Definitions processed food unless they are prepared from milk, sugar, etc., within the household. In the latter case, consumption of milk, sugar, etc., are accounted separately instead of the entire consumption being accounted under milk products Beverages, refreshments and processed food: This includes all beverages, including tea, coffee, cocoa, cold drinks and various commercially produced beverages, biscuits and confectionery, salted refreshments, sweets, pickles, sauce, jams and jellies, and also cooked meals obtained on payment (see paragraphs and 3.0.6) Clothing: This category includes bedsheets, bedcovers, pillows, curtains, mattresses, blankets, rugs, mats and mattings, cotton yarn, wool and knitting wool. It, however, excludes tailoring charges, which come under miscellaneous goods and services Miscellaneous goods and services: Items of consumption other than food and drink (and substances chewed and smoked), fuel, clothing and footwear, educational and medical goods and services, rent, taxes and durable goods Taxes and cesses: This includes only taxes and cesses which are levied on the household as a domestic consumer. It includes consumer licence fees such as are paid for possession of firearms, etc. It excludes income tax Education: Apart from educational expenses proper, such as school and college fees, private tuition charges, textbooks, etc., this includes expenses on newspapers and magazines, fiction and stationery. Donations to school fund, however, are excluded Institutional and non-institutional medical expenses:.the distinction between institutional and non-institutional medical expenses lies in whether the expenses were incurred for medical treatment received as an in-patient of a medical institution (such as a hospital or nursing home) or otherwise. Medical expenses include expenditure on medicines and medical goods including family planning appliances, payments made for medical treatment, and expenses incurred for clinical tests Durable goods: These include jewellery, vehicles, goods for recreation such as TV sets, cassettes, cameras and musical instruments, furniture and fixtures, household appliances of all kinds including kitchen and bathroom equipment, utensils, clocks, watches and thermometers, and personal medical equipment such as artificial limbs and hearing aids. Normally, durable goods are distinguished from miscellaneous consumer goods by having a longer lifetime (a year or more), but some petty durable articles, such as locks, torches, spectacles and walking sticks, are by convention accounted under miscellaneous goods. Expenditure incurred on repairs and construction of durables used for domestic purpose is included in expenditure on durable goods. 10

17 Consumer Expenditure: Concepts and Definitions Some abbreviations commonly used in the text and appendix tables are: cereal subst. pulse and prod. milk and prod. beverages etc. misc. goods misc. services medical (inst.) medical (non-inst.) total exp. cereal substitutes pulses and pulse products milk and milk products beverages, refreshments and processed food miscellaneous consumer goods miscellaneous consumer services medical (institutional) medical (non-institutional) total consumer expenditure stype1 schedule type 1 stype2 schedule type 2 11

18 4 Summary of Findings RSEs of all-india estimates: 52nd and 53rd rounds We shall first take a look at the variability of estimates based on different reference periods as measured by their Relative Standard Errors (RSEs) A word of explanation is perhaps required to clarify why the estimates themselves are not presented first. Admittedly, we do not have here the textbook situation of a couple of Table T1: RSEs of estimates of per capita consumption for rural and urban India based on 52nd and 53rd rounds of NSS, schedule types 1 and 2 item group LRP* stype1 stype2 for R U R U stype cereals gram cereal subst pulse & prod milk & prod edible oil meat,egg,fish vegetables fruits(fresh) fruits(dry) sugar salt spices beverages etc food total pan tobacco intoxicants clothing footwear education medical(inst.) durable goods non-food total total exp *Length of reference period in days for stype2 is shown in this column. LRP for stype1 is 30 days for all items. 12

19 Summary of Findings unbiased estimators where obviously the less variable one is more acceptable. Still, even when estimators are biased and the question of bias must be investigated to the extent Figure 1: RSE for monthly per capita consumption (quantity) of cereals by schedule type all-india RSE (%) round-sector 52R 53R 52U 53U stype stype possible, it is better to get a preliminary idea of variability to be kept at the back of one s mind while examining the estimates themselves. Further, while small differences between variability are certainly not important in a situation of unknown and possibly large bias, high variability can be a disqualifying factor. For instance, if an estimator of per capita Figure 2: RSE for monthly per capita consumption of clothing by schedule type all-india 6 5 RSE (%) round-sector 52R 53R 52U 53U stype stype consumption is estimated to have a relative standard error of 30%, then estimating this parameter on a year-to-year basis will yield a fluctuating series of very low credibility. If the true changes in the parameter over the years are small by comparison, then the series will not be at all usable for the measurement of these changes Estimates of relative standard error have been computed on the basis of the interpenetrating sub-sampling method for the estimates of the 52nd and 53rd rounds, where the sample of households allotted to each schedule type was drawn in the form of two 13

20 Summary of Findings independent sub-samples. A similar computation was not possible in the 51st round, where the total sample was drawn in the form of two independent sub-samples but schedule type 1 was canvassed in all sub-sample 1 households and schedule type2 in all sub-sample 2 households. RSE (%) Figure 3: RSE for monthly per capita consumption of footwear by schedule type all-india round-sector 52R 53R 52U 53U stype stype Obviously, all-india estimates will have a higher degree of variability than state-level estimates Table T1 gives estimated relative standard errors (RSEs) in percentage form for the 52nd and 53rd round estimates of per capita consumer expenditure on 14 food groups and 8 non-food groups based on stypes 1 and 2, for rural and urban India. Thus a figure of 0.73 Figure 4: RSE for monthly per capita expenditure on education by schedule type all-india RSE (%) round-sector 52R 53R 52U 53U stype stype

21 Summary of Findings against cereals means that the standard error of the estimate E of per capita expenditure on cereals in rural India was estimated at 0.73% of E. Table T2: Percentage increase in RSE of all-india estimate of per capita consumption obtained by adopting schedule type 2 instead of schedule type 1, rural and urban India, 52nd and 53rd rounds of NSS 52R 53R 52U 53U all item group RSE of est. by % incr. RSE of est. by % incr. RSE of est. by % incr. RSE of est. by % incr. av. % incr. stype1 stype2 stype1 stype2 stype1 stype2 stype1 stype2 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) cereals % % % % 43% gram % % % % 22% cereal subst % % % % 3% pulse & prod % % % % 38% milk & prod % % % % 57% edible oil % % % % 75% meat,egg,fish % % % % 25% vegetables % % % % 32% fruits(fresh) % % % % 14% fruits(dry) % % % % 11% sugar % % % % 37% salt % % % % 47% spices % % % % 28% beverages etc % % % % 10% food total % % % % 43% pan % % % % 3% tobacco % % % % 42% intoxicants % % % % 35% clothing % % % % -59% footwear % % % % -50% education % % % % 48% medical(inst.) % % % % -21% durable goods % % % % -44% non-food total % % % % -24% total exp % % % % -6% Round effect. The presence of a round effect is evident: the 53rd round estimates nearly always have higher RSEs than the corresponding 52nd round estimates. (The urban schedule type 2 estimates, however, seem to be an exception to this rule.) A possible reason is the sampling design of the 53rd round, which was worked out with the aim of capturing the salient features of unorganised trading enterprises in different regions of the country. Thus it was attempted, for instance, to give higher probabilities of selection to localities with more such enterprises. In the process, there was no doubt some loss of efficiency in the estimation of parameters relating to household consumption. Had such a round effect not been present, each pair of RSEs (52nd round, 53rd round) could have been treated as a pair of observations on the same phenomenon Rural-urban differences in RSE. For the 53rd round, the rural sector RSEs are almost uniformly higher than the urban sector RSEs for the food item groups. Apart from this, nothing can be said about the direction of the rural-urban effect, except at the level of individual item groups; e.g. rural RSEs are higher than urban RSEs for cereals and sugar, and urban RSEs are higher in the case of tobacco Magnitude of RSEs of all-india estimates. Among the item groups considered, the only one for which RSEs of per capita consumption are uniformly below 2% is cereals (apart from all food ). RSEs of the 52nd round cereals estimates are all below 1%. For pulses, 15

22 Summary of Findings edible oil, vegetables, salt and spices, RSEs are between 1% and 3% in both rounds (though higher in the 53rd round). RSEs are also below 5% for milk-and-milk-products, meat-eggand-fish, fresh fruits, sugar (below 4%), beverages, etc., pan (below 5.1%) and tobacco (below 5.2%). RSEs are above 4% for gram (4-12%), 6-9% for cereal substitutes, and 5-12% for intoxicants. Figure 5: RSE for monthly per capita consumption of durable goods by schedule type all-india RSE (%) round-sector 52R 53R 52U 53U stype stype For clothing and footwear, RSEs are within 1-7% for both schedule types (usually higher for footwear than for clothing). For education, they are within 3-8% except for an over-25% RSE obtained for schedule type 2. For medical (institutional), RSEs are in the range 9-24%. For durables, they are usually between 5% and 15%, but one of the schedule type 1 figures is 24% Schedule type effect. Table T2 recasts the figures in Table T1 so that corresponding stype1 and stype2 figures appear side by side. Percentage increase in stype2 RSE over Stype1 RSE is shown in columns 4,7,10 and 13, and the simple average of the four percentages is shown in col All the food groups, pan, tobacco and intoxicants show, on the average, lower RSEs with schedule type 1 (longer reference period) Clothing, footwear and durable goods show lower RSEs with schedule type 2 (longer reference period) For medical (institutional), the RSE is usually substantially lower with stype2 (longer reference period), though the 53rd round rural data are an exception For education, the evidence on the RSEs is not conclusive On the whole, the data confirm the a priori expectation that stability of the estimates of per capita consumption increases with the length of the reference period. However, for the food groups, the gain in the stability achieved by month in place of week is not very 16

23 Summary of Findings large. State-level RSEs Magnitude. Table T3 shows the lowest and highest RSEs obtained at State level, considering the 15 major States only, for each round-sector-schedule-type combination, along with the all-india RSEs which we have already encountered. Table T3: Lowest and highest RSEs (among major states) obtained for estimates of per capita consumption of specific item groups by sector (R and U), round, and schedule type RSE of estimate of per capita consumption based on item group* schedule type 1 schedule type 2 52R 53R 52U 53U 52R 53R 52U 53U cereals 7 lowest highest India pulse & prod. 7 lowest highest India milk & prod. 7 lowest highest India edible oil 7 lowest highest India meat,egg,fish 7 lowest highest India vegetables 7 lowest highest India fruits(fresh) 7 lowest highest India sugar 7 lowest highest India salt 7 lowest highest India spices 7 lowest highest India beverages etc. 7 lowest highest India food total 7 lowest

24 Summary of Findings highest India pan 7 lowest highest India clothing 365 lowest highest India footwear 365 lowest highest India education 365 lowest highest India medical(inst.) 365 lowest highest India durable goods 365 lowest highest India non-food total - lowest highest India Total - lowest exp. highest India * Length in days of reference period used in stype2 is shown alongside each item It may be observed that the lowest state-level RSE is usually somewhat higher than the all-india RSE (e.g. 53U, Stype1, cereals: lowest state-level RSE = 2.05%; all-india RSE = 1.00%). Figure 6: RSE for monthly per capita consumption of food by schedule type all-india 2 RSE (%) 1 round-sector0 52R 53R 52U 53U stype stype