Assessing Changing Food Consumption Patterns (1981)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

APPENDIX B ackground Papers for A Workshop on Methods for tile Collection of Aggregate Data on Food Consumption

lithe Food System: An Overview HAROLD F. BREIMYER The extensive data that have been amassed and replenished regarding U.S. agriculture and the food system originated in large measure from the needs of operating programs. Their instigation with respect to, among others, food consumption data was more practical or programmatic than intellectual. The emerging agricultural programs of the 1930's revealed a need for national statistics of many kinds that gave impetus to statistical collection programs that have long since been divorced from anything farmlike or rural. I joined the Agricultural Adjustment Administration in May 1936 and found economists there worrying about national income statistics and elas- ticity of demand for oranges and the effect of overseas trade, all with a sobriety never known to classroom academicians. I doubt George Jasny in the Department of Commerce has any idea how much the old operating agency of the Triple-A contributed to the national income data series over which he has exercised fatherly care for so long. Although the National Bureau of Economic Statistics properly gets the first credit for pioneering, I remember vividly how national income estimates Qf the department were spliced with the figures of the young Simon Kuznets in trying to arrive at some background statistical underpinning of demand estimates for farm products. And so it was with data on food production and consumption. To be sure, from its beginning in 1862 the U.S. Department of Agriculture had com- piled certain data on production of farm products. These had gradually been augmented with statistics arising from a variety of sources. Although certain marketing service activities had been earned on for a long time and market news reporting became an early fixture, many of the data can only be de- scribed as of happenstance origin. For example, early in our century, Upton 45

46 HAROLD F. BREIMYER Sinclair set in motion popular support for meat inspection, as the public and especially export buyers did not like to know about workers falling into open-manhole lard-rendering vats. Once federal meat inspection activities were well under way, the managers of the agency had a natural desire to collect and report the figures on how much work they were doing. Therein commenced a series of data on production of federally inspected meat, which could readily be converted into carcass weight data on how much federally inspected meat was being consumed. I remember well a meeting in the USDA when the meat inspection people decided they were being "had" as they performed a major statistical func- tion without receiving recognition or compensation. Kenneth Miller, chief economist for Armour and Company, joined in the profuse tribute to the agency for the signal contribution it was making to the vital industry of meat packing. The inspectors continued to report the data, though grudgingly. I cite this instance not only as an illustration of the nonmethodical origin of many data but to indicate how "consumption" was viewed. The data were the carcass weight of beef and lamb and product weight of pork going through processing channels. No correction was made for losses at any later stage. Then and to this day, to a very considerable extent, wholesale-level bias entered into data on disappearance of food products. To complete my philosophical commentary on how statistical series came into being, the sequence very often has been that some program adminis- trator needed new numbers. His statisticians scrambled to find whatever were available and engaged in some resourceful interpolating for those that were not. If the data proved truly useful their compilation was eventually assigned to a specialized agency, where corrections and refinements became attached. A fully clothed respectable statistical series then became a part of our data fund. DATA FOR PRICE ANALYSIS During the 1930's the commodity data for which program managers begged were those lending to statistical price analysis. Henry Schultz and Henry Moore were savants to whom Mordecai Ezekiel and Louis Bean gave al- legiance. Price-determining forces were examined for a large collection of agricultural products. The quantity variable often was production or even total supply, rather than disappearance. The inadequacy of good distribution data was not regretted too much. THE 1938 AGRICULTURAL LAW I turn now to the writing of the Agricultural Adjustment Act of 1938. My chief, O. V. Wells, had a major hand in drafting the language of the law. In

The Food System: An Overview 47 the previous 5 years a few persons had got the ear and conscience of Secre- tary of Agriculture Henry Wallace, telling him that the interests of consum- ers ought to be taken Alto account in designing and administering farm programs. Secretary Wallace was basically sympathetic but nearly impotent; however, in the 1938 act, language was inserted declaring a concern for protecting the supply of food for consumers. The act contained in its decla- ration of policy an injunction to assist consumers to obtain an adequate and steady supply of agricultural commodities at fair prices. Language in section 304 declared that nothing should be done to discourage producing supplies of food sufficient to maintain normal domestic human consumption as de- termined by the Secretary from records in the years 1920 to 1929, taking into account current trends in consumption and, significantly, the availabil- ity of substitutes. The years of the 1920's were a wise choice because the poorer nutrition of the 1930's, plagued by depression and drought, was thereby excluded. Thereupon began a scramble for data on which to build credible food disappearance and consumption estimates. The first step was clerical. It was to print tabular cards on which supply and distribution data for the various farm commodities could be compiled. The cards had to have many columns. The physical ability to write and read numbers written in miniature longhand was virtually a requirement of employment in Mr. Wells' unit. It would be a gold star on our national history if we could say that thereafter nutrition became an active ingredient in national farm policy. It did not. One reason, however, is that production outran markets so consis- tently that adequacy of food as such was not a serious issue. Another consideration, however, came on the scene at about the same time. It was the proposal to distribute food to lower-income families. Once again the needs of a program, or potential program, were parent to the statistical progeny. In this case, however, a person of outstanding intellectu- ality was the sponsor. He was Frederick V. Waugh. His instincts were both humanitarian and economic. He turned the emphasis from national aggre- gates on food distribution to stratified data. He sought to apply Engel's law about food buying by income class. Demands grew to compile data on food consumption by regional and ethnic groups but especially by family income. The program involved was the new Food Stamp Plan, which was seen as a way both to improve the diets of low-income families and increase the total demand for farm products. A few crude estimates began to appear as to the nutritional adequacy of householders' diets. An interesting offshoot were several studies, one by George Stigler, showing how cheaply a family could obtain adequate nutri- tion. The menu might be plain and dreary but it would have the minimum necessary nutrients. Even the crudest nutritional analyses implicitly required estimates of

48 HAROLD F. BREIMYER waste in distribution of food. As I remember, the waste factor data were extremely poor. Again, getting hold of any kind of estimates required in- genuity more than technical ability. All economists working on those studies were aware of another pitfall, namely, the meaning of "measures of central tendency." We might de- monstrate that families of four with an income of $1,500 (in the valuable dollars of those days) would "consume" a quantity of protein or riboflavin equal to the minimum requirement. Obviously, approximately half would exceed the figure and half would fall short. The skewness of the distribution remained, as I recall, totally unknown. Even if the average exceeded the minimum by 20 percent, many of us suspected underconsumption by a considerable fraction of the families. We had almost no reliable estimates of skewness. Of course, then as now, there were debates on just what figures best represented minimum nutritional requirements. But that is another subject. HOUSEHOLD FOOD CONSUMPTION SURVEYS It would be an injustice to fail to note the variety of statistical studies that had some incidental bearing on food consumption. Almost invariably the compilations had a purpose other than to indicate food consumption. The Census of Business gets expenditure data in its retail trade censuses. The Bureau of Labor Statistics has long assembled data on expenditure patterns for urban worker families as a source of weights for their consumer price indexes. For that matter, the U.S. Department of Agriculture has surveyed farm families for a similar purpose, namely, to update weight formulae for the "prices paid" index. More noteworthy are the decennial nationwide Household Food Consumption Surveys that began, as I remember, in 1955. FOOD DIARIES Probably the most complete data obtainable on families' spending for food are those derived from continuous diaries. I am most familiar with a Michi- gan State University diary study that continued for a number of years and with a Georgia Experiment Station diary enterprise that I believe to have been reinstigated. I regret to admit that I do not know to what extent the data so obtained have been exploited for appraisals of nutritional adequacy. PREOCCUPATION WITH WHOEESAEE DISTRIBUTION DATA In this brief introduction I do not come close to doing justice to the virtual profusion of what I call bits and pieces studies relating to food distribution.

The Food System: An Overview 49 As a selected example, while writing these notes I came across Marketing Research Report No. 1017 of the Agricultural Research Service of USDA. (The agency has since been renamed.) It is titled Marketing Losses of Selected Fruits and Vegetables. I do not know whether Marketing Research Reports truly had reached the number of 1,017 by 1975 (the date of the study), but in any case a scanning of those reports would reveal lots of bits and pieces bearing in some way on quantification of food distribution. The USDA deserves credit, in my opinion, for the great amount of work it has done. But I add quickly, and negatively, that the focus has almost always been on wholesale distribution; and this in turn is explained by preoccupation with assessing price-making influences. Often, the needs of operating programs have been paramount. Marketing orders, for example, are in force for fluid milk and well over 50 fruits, vegetables, and tree nuts. All these are statistically voracious. Manifestly, the wholesale-level orientation implies a similar bias toward the original farm product identification. In distribution data it is easiest to trace through a product such as eggs, particularly if we do not worry too much about stratification of consumption. That is to say, an egg is an egg irrespective of whether it is eaten fried, scrambled, or part of a cake. But if aggregate consumption data are not sufficient and consumption data by classes are sought, we encounter the problem that the various groups of the population do not eat their eggs in the same ratios of fresh versus ingredients of processed foods. The greater the amount of processing, the less meaningful are data ex- pressed in the terms of the farm-identified commodity. This point is so clear, and its implications so complicating, that I need not develop it further. Dr. Van Meir and others on the 2-day program will clarify this problem, I am sure. INSTITUTIONAL FOOD CONSUMPTION I have left until last the topic of institutional food consumption. It has always been left until last. Only in fairly recent times have major projects been undertaken to assemble data on mass or institutional food programs in all their mazelike complexity. In a sense the wholesale bias has continued prominent. I have not had occasion to dig into the institutional food issue in any depth. Persons working on this NRC project on food consumption patterns will doubtless do so. My hunch is that the significance of institutional food services is not minor. Questions can be so simple such as whether noningestion is greater in

so HAROLD F. BREIMYER institutional or household food consumption. My guess is that the ingestion ratio is highest in fast food services. My attention has been directed recently to energy utilization in food distribution and preparation; and here the hypothesis is that institutional services can be energy conserving except that the fast food business is energy-wasteful because of its profligate packaging. REID, BURK, STIEBLING, AND OMAR KHAYAM I caption my final remarks with the names Reid, Burk, Stiebling, and Omar Khayam. I am not sure what individuals most deserve gold medals for developing both data and awareness on food consumption. I know that in the U.S. Department of Agriculture all have had to fight for attention with the political figures who dole out dollars for ccc loans, P.L. 480 foreign food aid, and such. My hazy memory tells me that Jean Mayer was preceded long ago by a man named Henry Sherman. With regard to statistical data my mental association brings forth the names of Margaret Reid and Marguerite Burk; and the underrecognized advocate of giving nutrition more promi- nence in the design of farm programs was that grand lady, Hazel Stiebling. She was head of home economics in USDA. But my favorite citation in this connection is from Omar Khayam. He said he was no great guy; he was only trying to reduce the year to better reckoning. Perhaps we ought to award a few plaudits to those who now and henceforth try to reduce food distribution data to better reckoning. I personally applaud all those who will join the National Research Council in doing so. All I can add is that they will find a mixture of a wealth of bits and pieces, and gaping holes; and fitting the pieces together will prove to be an enormous jigsaw puzzle but an in- teresting and worthwhile one.

Measurement and Forecasting of Food Consumption by USDA ALDEN C. MANCHESTER and KENNETH R. FARRELL INTRODUCTION In the federal government, responsibility for measuring and forecasting food consumption is in the Economics, Statistics, and Cooperatives Service (ESCS) of the Department of Agriculture. Consumption is an integral part of measurement and forecasting for food and agriculture, including produc- tion, stocks, international trade, and prices. This paper discusses the system in which measures of per capita food consumption are developed, the sources and quality of the data used, and some of the uses of the data. It also considers data on food expenditures and some of its uses. It concludes with a discussion of the known deficiencies of the data base and possible steps to deal with the gaps. THE SYSTEM FOR MEASURING CONSUMPTION The basic concept is of a commodity flow data system (Figure 1). Agricul- tural products are produced on U.S. farms, caught by U.S. fishermen, or imported from abroad. Most move to manufacturing plants for processing and/or preservation. Then they move through the distribution system to retail stores or eating places and to consumers. The USDA system measures all food in the commercial system. Farm home production of foods except vegetables also is measured. Rough esti- mates are made of farm garden vegetables and of nonfarm home production. Food consumption is measured at the national aggregate level for 260 foods (see Appendix to this paper). There is no breakdown between food used at home and that used in restaurants and other away-from-home outlets. The use of supplements (vitamins, stabilizers, etc.) is not measured. 51

52 _ a, g 0 C' ~ ._ ' cn V, ~ — ~ ~ s ~ 4 - C C ~ o ~ o Z ~ I c ._ o E 11 j C ~n 4 - a o c ~ Q E ~ —C.D ~ ~ ~ E / . ~ o U. _ E ~ _ - C' / Q o C ~ ~ C s° ~C ~ Q ~C . _ a C) y —m O ~ C I o ~ ~' a, o - ._ . T ~ . ., . ~o . s~ o O ~ ·. C~ c a, ._ c C~ U ·— y Q `,, cn "c a) O ·. o o ._ Cs~ 4_ cn ._ .. C _ ro .- 4 ~ · ·- C53 c a' ._ mm ~ aC, O ~ X ~ C U' ; - C~ Ct - Ct ;^ - ._ o o LU

Measurement and Forecasting of Food Consumption by USDA 53 The basic tool is a supply and utilization balance sheet for each commod- ity (Table 11. Supply of each food consists of beginning stocks, production, imports, and in-shipments from territories (mostly Puerto Rico). Utilization consists of exports, shipments to territories, government purchases for military use and export, nonfood use, food use, and ending stocks. The availability of data is such that civilian food use is calculated as a residual after other measured uses are deducted from the total supply. Thus, it is often called disappearance. Estimates of consumption (disappearance) are prepared at two levels for each commodity. The basic measurement is at the primary distribution level, which is dictated for each commodity by the structure of the market- ing system and the availability of data. For some, it is at the farm gate. For most commodities which are processed, it is at the processing or manufac- turing plant. Once the primary level of distribution has been selected, quan- tities of all other components in the balance sheet for that commodity are converted to the primary weight basis using appropriate conversion factors. For example, the primary distribution level for red meat is the slaughter plant, so all quantities are converted to carcass weight. Most users of USDA consumption data are accustomed to the retail weight figures, which translate from primary distribution weight to retail weight by means of conversion factors that allow for subsequent processing and losses in the distribution system. Fresh beef, for instance, loses 26 percent of its weight from carcass to retail cuts. For some uses, a more desirable basis of computation is edible weight. We have calculated per capita consumption on that basis for special articles (Manchester, 19771. That calculation avoids the problems that arise par- ticularly because of the shift from fresh to processed products such as fruits and vegetables. We are developing such additional consumption series as an adjunct to those using primary distribution weight and retail weight. The ideal system for nutritional analysis and many other uses would be one that measured actual ingestion of foods. No data are available to make such estimates for the U.S. population. While it seems highly unlikely that such data will ever become available except from carefully controlled laboratory tests, it may be possible to move closer to actual consumption. For example, the Market Research Corporation of America has conducted menu surveys that record foods actually served and those present. Such data give no assurance that the food was consumed by all present. And, of course, no attempt is made to measure quantity for each individual. THE DATA SOURCES AND QUALITY The supply and utilization data system for food products is entirely depen- dent upon data that are collected for other purposes. No funds have ever

54 ^ 1 IU) 1 Ct V) ._ o ._ 4 - Ct - ._ Ct o ._ ._ U) o C~ ·fi V) o . U. o U. .~ . . m o o ct 1 LL 1 1 1 ._ 1 ~ I _ 1 — ~ 1 u, ~ 1 0 ~ ~ $ ce c~ 1 cq 0 1 C _ _ `~ _ 1 ~ ._ O ~ 1 1 1 ~ 1 I1 ~1 1 m ~ E~ ~ l C) Ct C) ~ O .= ~ 1 I ~ 00 00 ~ ') 00 ~ ~ O O O 00 ~ - ) _ ~ OX 0\ _ t— ~t _ ~ - } _ t_ V~ ~ ~ ~ U~ __ _____ _____ _ _ _ _ 1 ~ 1 — — — ~ ~ ~ ~ o o 1 ~ — 00 C-l d ~ — ~ O ~D oo \0 ~ ~ ~ 1 ~ CM ~ ~ C~ ~ \0 — t— ~ ~ [_ 1^ ~ . X ~D oo ~ ~ ~ ~ ~ O O O O — ~ ~ ~ _ _ _ _ _ _ I ~ ~ ~ ~ ~ _ ~ ~ ~ 00 00 00 _ 1 ~ O ~ — — \0 ~ ~ ~ ~ — — 00 ~ ~ ~ ~ ~ — ~ ~ \0 0 ~ ~ _ t_ _ O ~ ~ ~ ~ v~ ~ ~D oo ~ ~ oo O - C~ =: ~ ON _ ~ ~ 00 '] ~ ~ - ] ~ ~ ~ — C) ~ ~ ~ ~ ~ ~ O CC ~ ~ ~ O ~ 00 0 ~ ~ t— ~o ~ ~ ~ ~ ) ~ ~ O C~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ _ _ ~ I O e~ O ~ C~ ~ () ~ .= o, ~ ~ ~ _ ~ ~ ~ 00 — ~ ~t aN t— O, C') c =, ~ ~ ~ ~ O ~ ~ ~ — ~ — ~ O ~ ~ ~ O O O ~ ~ — — — ,[— ~ , ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ P~ ~ ~ _ _ _ _ _ _ _ _ _ _ _ _ ~ ~ 1 o ~, ,,, 1 ._ ._ _ _ c~: c5 oo ~ ~ ~ ~ ~ — ~ ~ ~ ~D ~ ~ ~ C) \- 1 ~ ~— ~ — O ~ ~ — C`1 ~ 00 ~1 0 ') ~o ~ ~ ~ ~ ~ ~ Om O~ ~ ~ O O O te4 ~ O O — — — _ ~ ~ ~ ~ ~ ~ ~ ~ ~ C~ ! 1 1 ~ ~ ~ ~D ~ oo ~ O — ~ ~ ~ u~ 1 ~o ~o ~ ~ ~D ~ \0 1 ~ 0N a~ ~ ,> ~ ~ ~ CJ~ C~ 1 — — — — — — — — — — — — _ _

ss 'e En to c) . - ~ ; - - To ct o A' ~ . - Ct 1 I e 04 "c ·= By ~ oo ~ to ~ - a, . ~ o v, c: o v) HI g to | z s~ 1 ~ 1 ~: - o o ~: o . - e ct 'e E~ · _3 i A — ~t \¢ <) ~ 00 ~ O — ~— ~ — ~1 00 O — ~ X 00 0 00 ~ Ct ~ _ _ C~ O O _ O 0\ 00 r- [- \0 Z ~ ~ _ _ _ _ _ _ _ O ~ '_ Ct ~ ~ ~ ~ ~ 00 0—t- t- 00 0 0 t- C-l ~ C~ ~ . _ ~ ~ ~ ~ 00 ~ ~ — 00 ~ O 0N ~ ~ — ~— — \0 mIm=o oO o~ooc'c~ oO=oooo~ 00- c'} ~ ~ ~ ~ X 0` O — ~ ~ ~ ~ \0 Ce ~} ~ ~ ~.0 ~ ~ \0 t~ 1 a~ ~ ~ ~ ~ ~ ~ ~ ~ 0- _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ~D ~ t— C~ _ o ~ oo _ ') _ ON C~ _ _ ~ O O ~ ~ — ', ~ d. ~D — — c~ — ~ O - ~ ~ ~ ~ In ~ u~ ~ ~ ~ ~ U~ ~ 0\ oo ~ ~ ~ ~ oo O U~ o ~ ~ U~ ~ \0 ~ ~ ~ ~ O oo oo oo _ t_ ~ O~ t_ _ oo oo ~ — O ~ O O ~ U) ~ ~ O ~ ~ oo ~D oo ~ ~ ~ oo oo ~— ~ =) t— ~ ~D _ oo c~ ~ =\ u~ _ oo t~ oo O ~ ~ — O O U~ ~ _ ~ oo O ~ ~ ~ C~ — O ~ oo ~ oo ~ — ~ oo O O — O O ~C\ 00 0 ~ X ~ ~ C-l oo ~ U) oo ~ ~ ~ oo ~ ~ ~ C~ oo ~ ~ ~ ~ O O O O ~ — — — — __ _~___ __ O O ~ U~ C~ ~ O ~ O ~ ~ _ X oo t_ ~ ~t — — ~t ~ ~ ~1 0 ~ t— ~ X ~ oo ~ 0N 0 — ~ ~ ~ — C~ ~ C~ C~ ~ ~ C~ O O O ~ V) _ _ _ _ - , oo O _ o~ ~ 0N — C~ ~ O ~ ~ — ~ d — ~— <~\ O ~ ~ oo — ~ oo ~ O ~ ~ ~ ~ ~ oo oo oo X ~ O _ _ _ _ _ _ _ _ _ ~ ~ ~ O oo ~ ~ ~ _ O ~ Ct~ =~ — ~ \~:~ \O ~ u~ d v0 ~ ~ ~C) oo Ct .c ca D b ~ ~ .~: =O ~ ~o o o eS . ~ ~ =: C,) ~ 't =4 ~ X .O ;^ .= ~ . _ U Ct ·— C~ U] ;^ o - , U, ;^ ^ ~ ~ o o ~q C) ", _ o ~ o ~ ~ o ~ ce' c) ~ c) o c) =~= — — c-) 3 ~ rL1 G ~ ~ ;^ .= - 'G

56 ALDEN C. MANCHESTER and KENNETH R. FARRELL been available to obtain data specifically on food consumption over time. Periodic surveys of food consumption or expenditures do provide useful checks. (For a detailed comparison of the disappearance data with House- hold Food Consumption Survey data, see LeBovit, 19681. Data on farm production and stocks come primarily from the statistics program of ESCS. The ESCS statistics program also provides information on the uses of some products and the production of manufactured dairy prod- ucts. Data on production of other processed products are obtained from other government and private sources, including Current Industrial Reports of the Bureau of the Census (for flour and fats and oils) and sugar utilization from the Agricultural Marketing Service. Where comprehensive data are available from trade associations, such as the National Food Processors Association, they are used. Foreign trade data are compiled by the Bureau of the Census from Cus- toms Service reports. Military use is reported by the Department of Defense for products procured through the central procurement system. Local pro- curement is estimated by USDA, primarily for milk and bread. Other data are used where available. A much more detailed account of the methods used and data sources is provided in Major Statistical Series of the U.S. Depart- ment of Agriculture, How They Are Constructed and Used, Vol. 5: Con- sumption and Utilization of Agricultural Products, U.S. Department of Agriculture, Agriculture Handbook No. 365, April 1972. Major gaps exist because of the dependence on data collected for other purposes than measuring food consumption. A principal problem area is the measurement of food produced and consumed outside the commercial system farm and nonfarm gardens, sport fish, and game. Survey data- such as the Household Food Consumption Survey (now the Nationwide Food Consumption Survey), which provides data once every 10-12 years is now the principal source of such data. The difficulty is in de- veloping satisfactory methods for making annual estimates with such data. In the last few years, ESCS has obtained better data on the number and size of home gardens through annual consumer surveys. When the results of the 1977-78 Nationwide Food Consumption Survey become available, we should be able to make reasonably good estimates for the years since the last survey in 1965-66. In terms of aggregate food consumption, gardens are fairly minor, accounting for only a few percent of total food. But, for some individual commodities of nutritional importance, they are far from trivial. In 1965-66, for example, home garden tomatoes accounted for 23 percent of all fresh tomatoes consumed at home. Sport fish are similarly important, accounting for over half of all fish consumed in 1965-66. We have developed annual estimates of consumption of sport fish using the number of fishing licenses as a mover from the base

Measurement and Forecasting of Food Consumption by USDA 57 provided by the 1965-66 survey. Again, the 1977-78 survey will provide a check on the reasonableness of these estimates. Game animals are not a major source of meat, compared to that from farm production, although they contribute about as much to the diet as game fish. They are estimated similarly. The other major problem area is also a result of the dearth of data. Most of the available current data are concentrated near the farm and primary pro- cessing levels. There are little or no data available for many further- processed products, especially comminuted or fabricated products such as bread, other bakery products, and soup. In short, relatively good data exist for the ingredients but not for many final products. Some of the firms represented at this workshop have data in more commodity detail for por- tions of the food market. No one has a data set with both the commodity detail we would all like to have and the detail by type of outlet. NUTRITIVE VALUE OF THE FOOD SUPPLY The Consumer and Food Economics Institute of the Science and Education Administration, USDA, utilizes the per capita consumption figures derived by ESCS to compute the nutritive value of the U.S. food supply for energy and 11 nutrients. The series shows trends in supplies of major nutrients that are related to changing food use patterns- the net effect in terms of nutrients of decreases in consumption of some foods and increases in others. To obtain estimates of the nutritive value of the food supply, quantities of food consumed per capita per year are multiplied by the appropriate food composition values. Most of the composition values are from Agriculture Handbook No. 8 (Watt et al., 1963) and its successors. No deduction is made for loss or waste of food in households or for destruction or loss of nutrients in the preparation of food. Like all time series, regardless of what is being measured, the data are more useful as indicators of change over time than of absolute levels at any one time. In other words, this series provides an indication of whether or not Americans on the average are improving their diets over time. It is not a measure of nutritional adequacy. EXPENDITURES In 1978, ESCS introduced a new data series on total expenditures for food in the United States (Manchester, 1978; Manchester and King, 1979~. Such a series has never been available before. The Department of Commerce, in the National Income and Product Accounts, has for many years estimated per- sonal consumption expenditures for food, but this series intentionally

58 ALDEN C. MANCHESTER and KENNETH R. FARRELL excludes foods not paid for directly by individuals i.e., meals on business expense accounts and in hospitals and institutions. The new series differen- tiates between food for use at home and that purchased in restaurants and furnished in hospitals and institutions. With appropriate adjustments for differences in price levels which have been made one can quantify the proportion of all food eaten at home, in public eating places, and elsewhere. The new series is based on Bureau of the Census data on retail sales by type of establishment grocery stores, drug stores, restaurants, etc. where available. These account for the biggest part of the basic data in the system. Thus, it is possible to estimate total food expenditures for regions and states, an advantage for many forms of economic analysis and probably also for analyses of food programs and other nutrition-related problems. The new series on total food expenditures is consistent with the quantity figures supplied by the consumption data discussed in the preceding sec- tions. The coverage of both series is all food and the basic equation price x quantity = value- holds. In other words, consumption (quantity) times price equals expenditures, when retail store prices are used both in the price variable and in expenditures. FORECASTING Basically, forecasting is an extension into the future of the system previ- ously described. The obvious difference is that we are no longer attempting to measure what has happened but are attempting to indicate what is most likely to happen next month or next year, given what we know about the current and prospective situation and the economic relationships involved. The basis is forecasts of production, exports, imports, and changes in stocks. These are analyzed in terms of their effects on prices and the effects of those prices on consumer willingness to purchase. DETERMINANTS OF CONSUMPTION AND DEMAND In the short run, it is a truism that what is produced is consumed; in economists' terms, that supply equals demand. The great equilibrator is price. When supplies go up, price goes down and consumers buy and con- sume more. Conversely, smaller supplies bring higher prices and smaller purchases. Demand for food in the aggregate is inelastic i.e., not very responsive to price changes. Demand for individual foods is more elastic. Over time, rising incomes increase expenditures on more expensive foods. Thus, short-period changes in consumption reflect mostly changes in supply rather than changes in consumer tastes or wants. The demand schedule, in economists' terms, may not have changed at all, although because of

Measurement and Forecasting of Food Consumption by USDA 59 larger supplies and lower prices consumption may be much larger than it . . was in a previous year. Sources of Change, 1954-76 Analysis of the sources of changes in food expenditures between 1954—the first year of the ESCS Total Food Expenditure series and 1976 provides a background for consideration of the factors affecting food expenditures and consumption. Food expenditures increased from $366 per person in 1954 to $443 in 1965 and $939 in 1976 (Table 2~. Food-at-home expenditures more than doubled. Food away-from-home more than tripled. The biggest increase was in expenditures for meals and snacks in public eating places up four times. Prices doubled in grocery stores (Table 3) and rose substantially more in restaurants. Quantities consumed (in pounds) went down a bit, although the price-weighted per capita food consumption index rose 10 percent, reflect- ing the shift to higher-valued foods, especially beef. Over the same 22-year period, disposable income per capita rose 250 percent—much more than the increases in food expenditures. A combination of consumption and expenditures data allows us to sort out the components of change. Out of a $573 increase in per capita food expen- ditures from 1954 to 1976, the biggest share was due to price change, $416 (73 percent) (Table 4~. The shift in outlets primarily from home to away- from-home—increased per capita food expenditures by $62 (11 percent), while changes in quantities accounted for $34 (6 percent), and shifts be- tween foods for $61 ~ 10 percent) per person. Demographic factors change in household size and in the age-sex distribution would have brought about an increase of about $28 per person over this period, even if nothing else had changed. TABLE 2 Food Expenditures Per Capita, 1954, 1965, and 1976 (in dollars per person) 1954 1965 1976 _ _ Food at home: Purchased 252 293 584 Home-produced and donations 22 18 32 TOTAL 274 311 616 Food away-from-home Public eating places 59 93 246 Limited clientele 15 20 32 Meals furnished 18 19 45 TOTAL 92 132 323 All food 366 443 939

60 Change due to: Shift in outlets Changes in prices Changes in quantities Shifts among food groups Shifts among foods within food groups TOTAL CHANGES Highly Processed Foods ALDEN C. MANCHESTER and KENNETH R. FARRELL TABLE 3 Factors Affecting Food Expenditures, 1954, 1965, and 1976 1954 1965 1976 Consumer price indexes for food: Grocery store prices 100 111 209 Restaurant prices 100 130 265 Farm value 100 103 185 Wholesale prices 100 107 199 Per capita food consumption: Index 100 101 110 Pounds 100 97 99 Disposable income per capita: Current dollars 1,574 2,430 5,511 Index 100 154 300 TABLE 4 Sources of Change in Food Expenditures Per Capita from 1954 to 1965 and 1976 (in dollars) Change from 1954 to: 1965 1976 +19 +52 + 2 + 62 +416 + 34 - 1 + 55 + 6 +573 + 5 +77 The available data can be used to shed additional light on the changes in food consumption patterns. As an example, let's consider the consumption of highly processed foods. For this analysis, data provided annually by Supermarketing Magazine, which give detail for products in final form as they appear in the retail store or enter the restaurant kitchen, are deflated by appropriate price indexes so that each year is in effect calculated at 1971 prices (Manchester, 1977a,b) (Table 51. Thus, approximations of change in quantities not measures of actual quantity are obtained. The strongest impression is stability. Products that are unchanged from the raw state except for having been cleaned, slaughtered, dressed, or preserved fresh meat, poultry, milk, eggs, fresh and processed fruits, and vegetables- accounted for two-thirds of the value of all food in each of the

Measurement and Forecasting of Food Consumption by USDA 61 TABLE 5 Food Consumption by Degree of Processing, 1952, 1971, and 1975 (percept a) Item 1952 1971 1975 Products unchanged from raw state except that they are: Cleaned, slaughtered, ordressed 36.6 38.7 39.8 Preserved 30.4 28.8 27.3 SUBTOTAL 67 .0 67. 5 67.1 Manufactured products: Derived mainly from a single raw product 10.5 10.1 11.0 Combine ingredients 22.5 22.4 21.9 SUBTOTAL 33.0 32.5 32.9 TOTAL 100.0 100.0 100.0 a Percent of value of total domestic food consumption at 1971 retail store prices. SOURCE: Basic data from Supermarketing Magazine. Price level adjust- ments and classification by ESCS. years; and manufactured products those that combine a number of ingredients represented virtually the same share of the value in 1952 and 1971 and then dropped 0.5 percentage points between 1971 and 1975. Thus, it would appear that the much-touted turnaway from highly processed pro- ducts by consumers when prices rose rapidly in 1973 and 1974 did occur, although on a fairly modest scale. Bigger shifts occurred between products that are preserved by canning, freezing, heating, or drying than those that are merely cleaned, slaughtered, or dressed. Here the changes between 1952 and 1971 toward less preserva- tion continued in the same direction to 1975. DATA GAPS Because the ESCS food consumption data system is dependent upon the availability of data collected for other purposes, there are major data gaps. The most serious occur as we get nearer to the consumer in the food chain. Farm production is well covered by an increasingly sophisticated statisti- cal system. Reduction in the coverage of production caused by budget con- straints has created problems in a few areas notably fresh vegetables. Production of some minor vegetables is no longer reported annually, nor is the production of major vegetables in minor states. This makes very little difference in the total farm picture or in total food consumption, but may be significant for monitoring the consumption of nutrients for which these vegetables are important sources.

62 ALDEN C. MANCHESTER and KENNETH R. FARRELL Imports and exports are covered almost completely as a by-product of customs service activity. Manufacture of primary products is now covered relatively well by a combination of agencies. Meat slaughter and dairy product production is reported by USDA. Fats and oils, confectionery, and flour are reported by the Bureau of the Census in their Current Industrial Reports. The pack of canned and frozen fruits and vegetables are covered by the National Food Proces- sors Association and the Frozen Food Institute. Manufacture of secondary products is covered much less completely. The production of products that combine a number of food ingredients is most commonly reported only by the Census of Manufactures every 5 years. Products in this category include canned and frozen prepared foods; jams, jellies, and preserves; baked goods, including bread; breakfast cereals; baby foods; condiments and sauces; and ready-to-mix desserts. Soups are not reported at all, even in the Census, because of the domi- nance of one company. The problem of disclosing the operations of that company led to a decision by the Bureau of the Census not to report all information for soup. The last complete statistics for soup are for 1946. Manufacture at retail is not reported at all. The largest gap here is for retail bakeries. We know how much flour is produced monthly. We know how it is used in manufacturing bakeries and other plants (e.g., breakfast cereals) 1 year in 5. We can only infer the amount of flour going into retail bakeries from the unaccounted-for use of bulk flour, assuming that smaller packages are all used in households. What the retail baker makes from that flour we have no way of knowing at any time. With retail bakeries making annual sales of $2 billion, it is not insignificant. And these sales do not include in-store bakeries in supermarkets. Away-from-home food consumption has been studied only once, in 1969 (Van Dress and Freund, 1968, 1971, 1972a, 1972b), although the Institu- tional Food Manufacturers Association is now updating the 1969 survey. We are now purchasing data from National Diary Panel, Inc., based on a quarterly diary from a sample of households which provides expenditure data on away-from-home purchases of food and some limited data on the foods purchased, primarily for the main course. Household food purchases are reported regularly by a number of com- mercial survey organizations, but no one covers all foods. A number of firms operate panels that report purchases of selected foods on a regular basis. The long-time operator in this field is the Market Research Corpora- tion of America. The foods covered are primarily packaged items, because those are the ones their clients are interested in. For a number of years, the Economic Research Service and now the Economics, Statistics, and Cooperatives Service have been interested in

Measurement and Forecasting of Food Consumption by USDA 63 obtaining data from a national consumer panel that reports purchases of all foods on a regular basis. Such a panel, using a national probability sample of adequate size, would give us information on the use of foods in the household by various income and family size groups, food stamp recipients, the poor, families with and without children, and many other groups. The main obstacle is money. Such a panel would be expensive. Such panels for all foods have been operated by universities in several cities in the past. The Georgia Experiment Station is currently operating a small panel in Griffin, Ga. The Bureau of Labor Statistics (BES) has plans to start a Continuing Consumer Expenditures Survey. It is planned to utilize a 2-week diary each quarter for food purchases, with the sample of households remaining in the survey for probably five quarters and a portion rotating each quarter. While the diary will provide space for recording quantities of food as well as expenditures, BES experience in the past has been that only a portion of panel members actually record quantity data. Present plans provide for a delay of approximately a year between data collection and publication. For several years, ESCS attempted to establish a probability panel of re- tailers and wholesalers who would report current movement of all foods into retail stores. This is similar to Selling Areas Marketing, Inc., but would cover all foods, including perishable items and those delivered directly to the retail store and not moving through the warehouse. This effort was initiated because of an interest in retail prices and marketing margins, but it would yield substantial information on food sales by retail stores. Unfortu- nately, we have been unable to obtain cooperation from the retailers, be- cause of their concern over possible disclosure of financial data. Stocks Per capita consumption data are now available annually for all foods and quarterly for animal products. The major gap preventing compilation of quarterly figures for crop products is the lack of stocks data on a quarterly basis, so we can tell how much of a given product has moved into consump- tion from the current season's pack. We are working on methods of over- coming this problem short of expensive additional data collection. Expenditures A number of items in the expenditures series are estimated by less-than- satisfactory methods. These include food use in colleges and universities, hospitals, and institutions. Home food production is also measured poorly, both for quantities and value.

64 SUMMARY ALDEN C. MANCHESTER and KENNETH R. FARRELL ESCS compiles national food consumption statistics for 260 different foods and food expenditure statistics covering all food. The basic data that enter these calculations are largely a by-product of other interests. No funds are or have been available to USDA specifically to obtain food consumption or expenditure statistics on a continuing basis. There are major gaps in these statistics once we leave the farm or first manufacture level. Thus, consumption of each food is measured at the last level for which data obtained for other purposes are available. Flour con- sumption is measured, not bakery and cereal products, for example. Filling these gaps will be an expensive process, but it needs to be done. We badly need comprehensive data on consumption or, at the minimum, purchases by households of food both for use in the home and away from home. With such data, the system can be closed measuring production or manufacture at each level where it occurs and sales to the final consumer at the other end. A consumer panel would provide data for different types of households on food purchases for home use and some information on away-from-home consumption, although not in the same detail as for that used at home. This would improve the national consumption statistics and have the additional benefit of permitting the analysis of consumption by individual households of different size and composition, income levels, and other characteristics. With a large enough sample, such data would also permit separation of food stamp recipients from other households of similar characteristics who were not participating. Detailed data on food store sales (or receipts) of individual goods would be almost as useful as data from a consumer panel in many ways. It would provide better detail on individual products than it is possible to obtain from households. It would not, of course, be helpful in analysis of consumption or purchases by different types of households. The BLS Continuing Consumer Expenditures Survey will be helpful, al- though it is probably not realistic to expect high-quality data, if any, on the quantities of individual foods purchased. Its other chief drawback, from our point of view, is the 1-year time lag between collection and publication. REFERENCES LeBovit, Corinne. 1968. U.S. food consumption: Annual disappearance and household survey data. Natl. Food Sit., November 1968, pp. 32-41. Manchester, Alden C. 1977a. Are highly processed foods taking over the market? Natl. Food Sit., March 1977, p. 34.

Measurement and Forecasting of Food Consumption by USDA 65 Manchester, Alden C. 1977b. Eating more vegetables. Natl. Food Sit., March 1977, pp. 34-35. Manchester, Alden C. 1978. Total food expenditures A new series. Natl. Food Rev., April 1978, pp. 16- 17. Manchester, Alden C., a~RichardA. King. 1979. U.S. food expenditures, 1954-78: New measures at point of sales and by type of purchaser. Econ. Res. Rep. Agric. Econ. Rep. No. 431, August 1979. 26 pp. Van Dress, Michael G., and William H. Freund. 1968. The food service industry: Its structure and characteristics, 1966. U.S. Econ. Res. Serv. Stat. Bull. No. 416, February 1968. 370 PP. Van Dress, Michael G., and William H. Freund. 1971. The food service industry: Type, quantity, and value of foods used. U.S. Econ. Res. Serv. Stat. Bull. No. 476, November 1971.451 pp. Van Dress, Michael G., and William H. Freund. 1972a. Separate eating places: Type, quantity and value of foods used. U.S. Econ. Res. Serv. Stat. Bull. No. 487, June 1972. 223 pp. Van Dress, Michael G., and William H. Freund. 1972b. The market for food consumed away from home: Dollar value statistics. U.S. Econ. Res. Serv. Stat. Bull. No. 491, September 1972. 105 pp. Watt, Bernice K., Annabel L. Merrill, et al. 1963. Composition of foods Raw, processed, prepared. U.S. Agric. Res. Serv. Agric. Handb. No. 8, Rev., December 1963. 190 pp. APPENDIX: FOOD COMMODITIES COVERED IN CONSUMPTION STATISTICS Primary Distribution Level MEAT Beef Veal Pork Lamb Edible offal (variety meats) POULTRY Young chickens (broilers) Other chickens Turkey EGGS Shell Processed DAIRY PRODUCTS Fluid milk products: Whole milk Lowfat and skim milk Cream and half and half Sour cream products Yogurt Butter Slaughter Slaughter Slaughter Slaughter Slaughter Slaughter Slaughter Slaughter Farm Manufacture Manufacture

66 ALDEN C. MANCHESTER and KENNETH R. FARRELL Cheese (16 varieties) Cottage cheese Nonfat dry milk Evaporated milk Condensed milk Dried whole milk Dried buttermilk Ice cream Sherbet Ice milk Mellorine Water ices Dry whey FISH AND SEAFOOD (National Marine Fisheries Service) Fresh and frozen: Edible Weight Salmon Other fish Shrimp Northern lobster Spiny lobster Oysters Clams Crabs Scallops Canned: Salmon Sardines (pilchards and herring) _ Tuna Shellfish Other Cured FRUITS Fresh Canned Frozen Dried Chilled Citrus: Oranges x Tangelos x Tangerines x Lemons x Limes x Grapefruit x Citrus sections x x Citrus juice: Orange juice x x x Grapefruit juice x x x Citrus concentrate x Blended orange and grapefruit juice x x Lemon and lime juice x ~ . . . 1 angerlne Julce x x

Lemon juice Lemonade Limeade Noncitrus: Apples Apricots Avocados Bananas Bushberries Cherries Cranberries Figs Grapes Nectarines Peaches Pears Pineapple Papayas Plums and prunes Strawberries Miscellaneous fruit Salad and cocktail Olives Apple juice Grape juice Pineapple juice Prune juice Dates Raisins VEGETABLES Artichokes Asparagus Beans, lima Beans, snap (green) Broccoli Brussels sprouts Cabbage Carrots Kale Lettuce and escarole Peas, green Peppers Spinach Other vegetables Beets Cauliflower Celery Corn x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x Fresh x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x Measurement and Forecasting of Food Consumption by USDA 67 Fresh Canned Frozen Dried Chilled x x x Canned Frozen Dehydrated

68 Cucumbers Eggplant Garlic Onions and shallots Tomatoes Pumpkin and squash Catsup and chili sauce, tomato paste and sauce Tomato pulp and puree Tomato and other vegetable . . Julces Pickles Sauerkraut Peas and carrots Southern greens Potatoes Potato chips and shoestrings Sweet potatoes MELONS Watermelons Cantaloupes, honeydew PEANUTS Peanut butter Salted peanuts Peanut candy Peanut butter sandwiches Unshelled peanuts FATS AND OILS Margarine Lard Shortening Cooking and salad oils Other edible (salad dressings, mayonnaise, etc.) BEANS, PEAS AND SOY PRODUCTS Dry edible beans Dry field peas Soya grits and flour CEREAL PRODUCTS Wheat flour: White and whole wheat Semolina and durum flour Wheat cereal Rye flour Rice ALDEN C. MANCHESTER and KENNETH R. FARRELL Fresh Canned Frozen Dehydrated x x x x x x x x x x x x x x x x x x x x x x x Manufacture Manufacture Farm Manufacture

Measurement and Forecasting of Food Consumption by USDA 69 Corn products: Flour and meal Cereal Starch Hominy and grits Oat food products Barley products SWEETENERS Sugar, cane Sugar, beet Corn, dextrose (refined corn sugar) Honey Syrups: Corn, high fructose Corn, glucose Maple Sugar cane Sorgo Refiners Edible cane molasses Saccharin (sugar sweetener equivalent) CONFECTIONERY COFFEE Roasted Instant TEA COCOA SOFT DRINKS TREE NUTS Almonds Filberts Pecans Walnuts Macadamia Other tree nuts Coconut SPICES AND HERBS Pepper Mustard seed Chili peppers Anise seed Manufacture Manufacture Manufacture Imports Imports Manufacture Imports or production Imports or production

70 Caraway seed Celery seed Cinnamon Cloves Coriander seed Cumin seed Fennel seed Ginger root Mace Nutmeg Paprika Pimento (Allspice) Poppy seed Sage Sesame Seed Turmeric Vanilla beans Other spices ALDEN C. MANCHESTER and KENNETH R. FARRELL Aggregate Measures of Food Consumption Retail Weight of Food Consumed Per Capita The main purpose for de- veloping a uniform measure for consumption at the retail level is to facilitate the development of the price-weighted index. However, these data are use- ful by themselves for measuring trends and shifts, especially at an unaggre- gated level of food groups. For example, the substitution within the fresh fruit or fresh vegetable categories can be reasonably identified with this measure. Shifts within the canned product categories can be accurately measured. The substitution of beef for pork on a poundage basis can be traced over time. This measure of civilian per capita disappearance is computed from pri- mary weight consumption data. For example, beef is measured on a carcass weight basis, a form in which it leaves the packing plant. A factor that reflects cutting loss, trim, and bones is applied to the carcass weight. The primary weight of fresh fruits and vegetables is a farm weight. Different factors are used to equate the farm weight to a retail weight basis. A total food consumption measure in terms of total (retail-weight) pounds is published. This series has been quite stable at a little over 1,400 pounds in recent years. However, the total pounds consumed peaked out in 1945. The consumption of crop products by this measure has shown the most decline as consumers have shifted to more processed products. Also, there may be some double counting in this series. Adjustments are made for sugar and butter. Sugar in canned fruits for example is included in the canned fruit total but subtracted from the sugar and other sweetener total.

Measurement and Forecasting of Food Consumption by USDA 7 Index of Per Capita Consumption The index of per capita food consump- tion is probably the best economic measure of the food consumed at the retail level. Pounds of food consumed on a retail weight basis are combined with retail store prices in a base period to measure annual or quarterly changes in food consumption. While the index primarily measures quantity changes, it also reflects shifts such as those from a low priced to a higher- priced food. This could indicate the shift from cereal products to animal products and the shift from fresh to processed forms. The consumption data include the significant portion of food consumed away from home as well as food consumed on farms where produced. The shift to away from home is not reflected in the index. This index can rise though the quantities consumed may be unchanged from year to year. For example, as family incomes rise, shifts to more expensive items such as chicken to beef would result in a higher index of consumption. This index has generally increased during the past decade as consumption has switched to higher-priced foods. It reached a record high in 1976 and has declined slightly during the past 2 years as red meat supplies have tapered off. This index is published for 27 different groups, which include animal product and crop product categories. A similar food use index is also available. It measures changes in domes- tic consumption at the farm level. Farm prices are used to weight the index and even the value of processed products is factored back to the farm level. This index has risen less than the per capita consumption index, since the shifts to more processed forms are washed out by the indexing procedure. However, shifts in the mix of food used would have an influence on the index. Nutritive Value of Food Consumed The retail weight equivalent of food consumed is the basis for developing the nutrient values for food energy, protein, fat, carbohydrate, calcium, phosphorus, iron, magnesium, vitamin A, thiamin, riboflavin, niacin, vitamin Be, vitamin Bit, and ascorbic acid. Quantities of food consumed are multiplied by appropriate food composition values. The information is published as nutrients available for per capita consumption per day and percentages of total nutrients contributed by 24 major food categories.

A. C. Nielsen Company Services OLI VER S . CAS TLE The Nielsen Food Index (NFI) service is the comprehensive national service covering products sold through grocery stores in all 48 contiguous states. All data are projected to national and regional totals and shown separately by the 10 Nielsen geographic areas (Figure 1), by store type (chain grocery versus independents), and the latter by volume classifications (Table 1~. Other divisions, such as client sales areas, are available as ordered. Consumer sales and shares (Table 2) are reported bimonthly by brand and product group in dollars and by a physical measurement either pounds, units, gallons, etc. Retail inventories and retailer purchases are reported in the same fashion. Retail distribution, for each brand and size, provides a measure of the stores handling each brand and size and are computed in two ways. Store count distribution, as the name implies, reports on the actual number of stores handling a brand. All-commodity distribution is weighted by store dollar volume. As such this measure reflects the extent to which each major brand is exposed to all grocery activity based upon the total dollar sales of the stores handling the brand. Measures of out-of-stock are also reported on a store count and all- commodity basis and are projected to national and regional areas. Out-of- stock when computed on an all-commodity basis reveals the relative impor- tance of stores without the item or items in question on the date of audit, but handling the item during the 2-month period between audits. Average retail prices are available for each brand, size, and type by area and nationally. These prices, keep in mind, are prices charged to consumers by retailers. 72

A . C. Niel.sen Company Servic es )~_ MEN LOS ANGE LES 73 ~ WEST CENTRAL PACIFIC ~ SOUTHWEST METRO X-7 ~ ' - NEW YORK CHICAGO | EAST CENTRAL ~ SOUTH EAST ~ ~ \ r #'\; - ~ FIGURE I Nielsen territories. We also produce sales influencing data based on observations in each sample store relative to Crucial nrices and in-store disolaYs. All advertising that originates with the sample stores, including that of corporate chain stores, is also checked every audit period. These data are reported for each brand on an all-commodity basis nationally and by Nielsen territory. The means are essentially some 650 full-time field representatives, 62 district managers and 8 regional managers. All field representatives are TABLE 1 NF] Services Data Breakdowns Available Brand, Type, and Client County Size Territory Areas Sizes Store Types Private New England 1 10 Metro New York Grocery Label Metro New York 2 11 Metro Chicago Chains (4 or more stores) Brands Mid. Atlantic 3 12 Metro Los Angeles Large Over 500 M East Central 4 13 A Counties annual sales National Metro Chicago 5 14 B Counties Medium—under 500 M and West Central 6 15 C Counties annual sales regional Southeast 7 16 D Counties Independents brands Southwest 8 17 Large Over 500 M by name, Metro Los Angeles 9 18 Medium 100-500 M type, and Pacific Small—Under 100 M size

74 O LI VE R S . C A STLE TABLE 2 NF! Services Consumer sales and shares reported bimonthly in- Dollars By a Physical Measurement: Units, Pounds, Gallons, etc. college educated, then trained from 6 to 9 months in Nielsen's field training school, and finally given several more months of training in the field with experienced field representatives selected for this purpose. The end result is a field staff that is unique in terms of caliber, training and size and one not duplicated by any other market research firm. This staff, supervised by district managers, enables Nielsen to handle a variety of complex projects in addition to the painstaking audits of the sample stores. What then is the primary function of each member of the field staff relative to the distribution and sales of grocery items? The field representatives' goals are to determine the movement of each and every grocery item by brand, size, and type in the product classes under audit and in the stores included in the sample (Table 34. The table is based on a bimonthly period, but monthly or weekly audits are performed in the same fashion. TABLE 3 Principles of Nielsen Retail Index Auditing ("Alpha" Brand of Dry Soup Mix in Super X Market) For June-July: Pkgs. Value Inventory May 30 July 30 Change Purchases From manufacturer ( 1 order) From wholesalers (4 orders) Total Consumer sales Packages Price, per pkg. Dollars, total Adv. 1 2 3 4 5 6 7 8 9 Display X 114 93 21 12 48 60 81 Selling price 39¢ Special price 35¢ $ 3.72 15.00 $18.72 $ 0.39 $3 1.59

A. C. Nielsen Company Services 75 In the example, the 114 units were the inventory the retailer had on hand at the end of the previous audit the 93 units represent the current inven- tory. Total sales of the item during the period amounted to 81 units a 21- unit change in inventory plus the purchases of 60 units during the period. In addition, the brand was on display, as noted by the X, and had been advertised six separate times by the retailer during the past 60 days. The item was selling for 39¢ on the date of audit but had been advertised at the special price of 35¢ during the period. Since the auditor separates the in- ventory of each item on a reserve versus the selling area of the store, it is possible to show the percentage of the inventory which is visible to custom- ers (i.e., selling area). As you can see, the audit is simple in itself but to obtain the correct numbers inventories, purchases, etc. for every brand, size, and type in a giant supermarket, it becomes a complex and meticulous task hence the careful screening of field personnel and extensive training and supervision. Nielsen is currently auditing 152 food and beverage product classes, including frozen and refrigerated products. This includes a wide variety of packaged, nonperishable foods in cans, cartons, bottles, jars, etc. primarily warehouse items. But Nielsen, by concentrating its efforts at the store or retail level, is able also to cover store-door delivered items. These include the great majority of soft drinks, crackers, cheese, and significant amounts of snack items and refrigerated and frozen foods that do not pass through either chain or wholesale warehouses. It further means that we can provide complete data on drop-shipments of any other packaged food pro- ducts that are delivered directly from the manufacturer to the store and thus bypassing all warehouses. There are many other standard grocery product groups—such as deter- gents, toilet soap, floor wax, paper towels, and so forth that are not edible but have been handled in grocery stores for as long as these outlets have been in existence. This would add another 61 groups for a total of 213. In addition to the national service, we also cover 38 major markets by using supplemental samples specifically designed for and projectable to the designated market. Each market includes the central city, the suburbs, and remaining television area. The areas are large enough to ensure that products in question are exposed to a broad range of people, demographically, and the close conformity to the television areas allows for measurement of television promotional activity. The identical types of data produced for the national service are available, that is consumer sales and shares, retail inventories, distribution, sales rates, etc., measured on both a dollar and physical basis. Figures 2 and 3 show two typical areas: Rochester, N.Y., and Miami-Ft. Lauderdale, Fla. Table 4 lists the entire 38 markets. Combined, they account for 61.4 percent of the population and 62.3 percent of total grocery sales.

76 OLIVER S. CASTLE LA KE ONTA R10 Niagara Orleans Genesee 1 Wyoming , Allegany Taraugus ~ ROCHESTER ; ;' I'm' '''''''1 :§ Monroe Livingsto NEW YORK ~ Ontario 1 1 1 1 1 .~ ~~J \ ~ S;t~llhPn ~ ~ ~- I Chemung ~ PENNSYLVANIA . FIGURE 2 Rochester, N.Y. De Soto ~ | St. Lucie) Lake Martin Glades (Okeechobee . FLORIDA Collier . I. :. ~ ..... . Brov~ Ird i ~—~ ~ Palm Beach .. _ ........................ FT. LAUDERDALE _ -O FIGURE 3 Miami-Ft. Lauderdale, Fla.

A. C. Nielsen Company Services TABLE 4 Nielsen Major Markets % of U.S. % of U.S. % of U.S. Metro Population Grocery Sales Drug Sales Areas (Jan. 1, 1976) (1975) (1975) New York 7.5 6.9 5.9 Los Angeles 4.7 4.9 5.8 Chicago 3.6 3.4 4.8 . Philadelphia 3.6 3.7 3.2 ''. ~ Baltimore-Wash., D.C. 3.2 3.2 4.3 Boston 3 .2 3.2 2. 8 Detroit 2.9 3. 1 3. 7 cat San Francisco-Oakland 2.3 2.6 3.2 4,, Q ~ Cleveland 2.1 2.3 1.9 & ~ Pittsburgh 2.0 1.8 1.7 Dallas-Ft. Worth 1 .5 1 .6 1 .6 Miami-Ft. Lauderdale 1.4 1.8 1.9 Minneapolis-St. Paul 1.4 1.3 1.2 St. Louis 1.4 1.4 1.3 Atlanta 1.3 1.4 1.3 c ~ Cincinnati 1.3 1.2 1.1 ~ A Houston 1 .3 1 .5 1.2 38 ~ Indianapolis 1.3 1.1 1.4 in. Seattle-Tacoma 1.1 1.3 1.3 O '' Kansas City 1.0 0.9 0.9 ~ ~ ' Memphis 1 .0 0.9 0. 7 it, oQ ~ Milwaukee 1.0 0.9 0.8 _ Portland 1.0 1.2 1.0 Buffalo 0.9 0.9 0.9 Denver 0.9 0.9 1.0 Nashville 0.8 0.9 0.8 Phoenix 0.8 0.9 0.9 Sacramento-Stockton 0.8 0.9 1 .2 Birmingham-Anniston 0.7 0.7 0.6 Charlotte 0.7 0.7 0.7 Grand Rapids-Kalamazoo 0.7 0.8 0.6 Louisville 0.7 0.6 0.6 Albany-Schenectady -Troy 0.6 0.7 0.5 . Oklahoma City 0.6 0.6 0.5 In. Omaha 0.6 0.5 0.6 ~ ~ O San Antonio 0.6 0.6 0.5 it, Q ~ Jacksonville 0.5 0.5 0.7 _ Rochester 0.4 0.5 0.5 TOTAL MAJOR MARKET AREAS 61.4 62.3 63.6 77 NOTE: All population and household data are from Sales and Marketing Management magazine estimates and, unless otherwise stated, refer to the condition as of January 1, 1976. Store count and volume data are Nielsen estimates and refer to 1975.

78 OLIVER S. CASTLE Back in 1959, when the research and development of new products were becoming a way of life for large grocery manufacturers, the need for a data source that was both broad in scope, i.e., encompassing virtually all products sold through supermarkets and yet reliable enough to identify and evaluate growth opportunities in the grocery field became apparent. To fill this need, we developed a new system of services called Nielsen Early Intelligence System (NElS). The basic service in this system Data Services is based on a summary of warehouse-to-store shipments of all grocery items from a nationwide sample of 150 supermarkets and reported every 60 days. First, all packaged grocery items are merged into 600 logical product groups but each individual brand within the group is reported separately by brand, size, type, flavor, etc. Some 430 product groups are edible foods, and the remaining 170 are assorted product categories such as detergents, soaps, health and beauty aids, and other general merchandise. We originally thought the primary use, and perhaps sole use of Data Services, would be to identify and quickly appraise product areas of oppor- tunity for manufacturers interested in developing new products and markets outside of their areas of interest. Even though the service was eminently successful in this regard, we soon discovered, as did our clients, that the data had additional uses or applications in a variety of other situations. Table 5 shows a sample report page from Data Services. The product class used in the table is Canned Beef Stew. Note that the unit movement for the total market for October-November 1977 compared to the same period of 1976 reveals an increase of some 2,900 units or over 7 percent versus a dollar gain of about 8 percent. Also provided are data for each brand of beef stew by size, type, flavor, etc. Note that Dinty Moore Beef Stew had a 1977 unit share of 48.7 percent of the total market, up four points from the previous year, and a 53.5 percent share of the dollars spent for beef stews. During this 1977 period, 99 percent of the sample stores purchased at least one size of the brand indicating wide distribution, and average unit movement for the brand was 12.7 cans per week. These same data are provided in identical form for each and every brand in a product group—however many there are. Thus, it is readily apparent that the entire range of food products can be monitored in order to detect any major changes or shifts in consumer preferences or product innovations. Table 6 provides a "shopping list" of the types of information our clients have obtained from Nielsen Early Intelligence Systems. As you can see, the real strength of this service is its ability to answer straightforward questions with direct answers relative to marketing trends in terms of prices, pack- ages, types, etc. Point three may need a moment's explanation. By "controlled brands"

79 3 m an l Be: . o Cal 5 - ._ l ad Cal ._ V) Cot ~ .= ~ cat V) as .c~ A: ~4 ._ - - Ct Cal is ;^ ~ A: Cal > ~ O 3 ~ o ~ CQ Pa Cat o - o ~ ~ Al o U' o ._ ;: ~ em z o 1 EN O . ~ m _ Sit ~ ~ 00 ~ ~ . . . . . . . \0 ~ ~ ~ ~ V) ~ ~ _ ~ ~ ~ ~ O . . . . . . . 00 US ~ _ 00 ~ _ _ O _ 00 ~ O ~ ~ ~ O ~ V) ~ ~ Cat \0 O · O O O 00 ~ ~ 00 ~ _ _ 00 _ _ · C~ O 0\ ~ ~ ~ \D ~ ~ _ ~ C~ · x · ~ ~ ~ ~ - · c~ oo ^ ~ ^ ^ ^ ^ (~) ~ o ~ ~ - oo ~ - ~ - - oo ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ o ~ O U~ ~ ~ u~ ~ ~ ~ d. u~ ~ r~ oo ~ ~ O · ^o ^ ~ o x ~ ~ oo ~ - ~ - - ~ ~ ~ ~ o ~ ~ ~ ~ ~ ~ o o o ~ ~ ch · ~ · ~ oo - oo ~ . ^ ~ ^ ^ ^ ^ - o ~ ~ ~ ~ ~ - ~ - - t - vo ~ ~ oo ~ ~ c~ ~ ~ ~ . . . . . o o o - o o u~ c-) 'e cl . - v) o . - c) c~ m _ 0 ~ ~ ~ 0 0 0 u~ ~ o — ~ ~ — 0 0 ~o _ 3 3 3 3 v~ v~ c~ v~ 3 3 mmmm 0 ~ 0 0 0 0 0 ~ 0 ~ ~ m" ~ ~ m" ~mm — = — s ~ ~ ;^ ~ — ;~. ;~ O g ·~ — O ~ ~ ~ E~ ~ ~ ~ ~ E~ ~ ~

80 O LI VE R S . C AS T LE TABLE 6 NElS Data Services for Identifying Significant Consumer Buying Trends 1. Price (Lower/Higher) 2 . P. a c k a g e ( C o n v e n i e n c e / D u r a b i 1 i t y / N o v e 1 t y ) 3. Brand (National, Regional vs. Controlled) 4. Type (Moist/Dry Chunky/Regular) 5. Size (Larger/Smaller) 6. Formula (With Borax/Ammonia With Raisins/Dates) 7. Nutrition (Natural Ingredients/All Beef) 8. Convenience (Instant Breakfasts/Spray Cleaners) 9. Flavor/Color/Design (BBQ Flavor/Floral Print Design) 10. Diet (Low-Calone/Cholesterol) 11. Ecology (Regular vs. Recyclable) we mean private label, chain's own, or brands such as TOPCO, which, for example, may be marketed by two different chains one in Cleveland and one in Kansas City but are exclusive to these chains in their respective areas. These controlled brands are reported in the same detail as other brands size, type, etc. but not identified by chain—only as "controlled brands. " Other trends that are also identifiable through our Data Services report may relate to nutrition, convenience, diet, or be ecologically based on container type. Although clients often purchase Data Services alone, many utilize it in conjunction with Nielsen Product Pickup Service, since by means of the pickup service it is possible to obtain samples of new products just intro- duced into the market. As the name indicates, Product Pickup Service is just that; it's a facility for obtaining samples of any grocery product introduced into the market—at any place or at any time. By means of 350 field agents, covering all principal cities and the areas between, we can supply one sample unit, or many hundreds of cases, of a specific brand to our clients. We are prepared to handle all perishable products frozen, fresh, or other fragile items and ship them anywhere, in special containers whenever necessary. We can pick them up on a regular schedule weekly, monthly, quarter- ly or on a one-time-only basis. We can also secure samples in 23 addi- tional countries through our overseas companies. Over the years, we have obtained retail samples for an ever-widening variety of reasons or purposes. Table 7 lists a few of the more common reasons. They vary from wanting to check a competitive product for any number of reasons to qualitative assurance reasons relative to taste, flavor, color, freshness, nutrition maintenance, and so forth.

A. C. Nielsen Company Services TABLE 7 Some Reasons for Using Nielsen Product Pickup Services 1. Competitive Products- New end established 2. Average Age of Shelf Product 3. Taste and Flavor 4. Color 5. General Appearance Label, Contents, etc. 6. Nutrition Maintenance 81 One of the more interesting programs we are now conducting is for the Food and Drug Administration, which utilizes the resources of not only Data Services and Product Pickup, but also our Food Index Service. The key here was Nielsen's unique ability to provide accurate market data coupled with a highly trained, nationwide field force capable of handling in-store assign- ments. Here's how these services are being used to solve a problem facing the FDA. They wanted to assess the extent to which nutritional labeling had been accomplished, on a category-by-category basis, and then to measure the growth of the nutritionally labeled items versus competing items not so labeled. The FDA contracted for the Data Services Directory, and matching com- puter tape, covering the progress of 430 edible grocery product classes in which they were concerned with nutritional labeling. These in turn were regrouped into very broad categories such as products containing flour, canned vegetables, dairy products, and so forth. Using the flour category as an example, they grouped all items in which flour was the primary ingredient thus including pancake mixes, cake mixes, all-purpose flour, cookie mixes, etc. These broad categories were classified relative to con- tribution to total food sales, then each product class was ranked relative to its importance within the broad category. And finally, each individual brand was ranked relative to its importance within its product class. Once these listings were completed, the FDA, by means of a sampling procedure, developed a brand listing of retail samples from each category to be secured from supermarkets. This listing was passed on to Nielsen's Product Pickup Service's central headquarters in Northbrook for fulfillment. A number of leading brands on the list with virtually 100 percent dis- tribution could be found in almost any supermarket and presented no prob- lem. On the other hand, a good many others were regional or were fairly obscure brands with limited distribution. To secure retail samples of these brands, we returned to the data bank and found the cities and areas where these brands were in distribution. From then on it was a simple matter to transmit pickup orders to the field agents

82 OLIVER S. CASTLE in the respective areas and cities to secure the brand samples needed. Here is another example of the advantages of store-by-store data. In this case it was simply to retrieve shelf samples quickly and at low cost. But most often it is to determine what is happening, in what marketplace, and often why. From these samples, the FDA iS currently sorting out: ~ . The brands requiring nutritional labels and, in fact, so labeled, from 2. The brands requiring nutritional labels but not bearing them. 3. The brands not requiring nutritional labels but bearing them anyway. 4. Those not requiring nutritional labels and not having them. Once all of the brands are placed in the proper groups, the FDA will go back to the Data Services Directory to establish the current dollar volume of each classification. To determine future growth, they will continue to monitor each group for the next year to assess gains and losses for each of the four. In addition to the above, the FDA iS also using the product samples to check whether the contents match the contents' label in terms of water, solids, etc. And, if the label carries a picture of the contents, does the picture truly represent the contents? Another use of the product samples by the FDA iS to run a nutrient assay of each of the products bearing nutritional labels to see if the nutritional values claimed by the label are in fact true. A fourth use of the samples, and related to the one above, is an in-depth analysis of all of the products labeled as "dietetic. " If the label claimed the product contained only four calories and no sugar, the contents were analyzed to verify the label claim. As we understand it, the USDA iS going to perform similar analyses on any products containing meat sandwich spreads, chili, canned meat, etc. since the FDA has no jurisdiction over meat products. I'd like to make one final comment on this overall study. As you probably know, many food products are delivered directly to the store and never move through a wholesaler's or chain warehouse. For example, close to 90 percent of all soft drinks are delivered directly to the store; many dairy items milk, butter, cheese, ice creams—are handled in the same way. Most crackers, cookies, some frozen foods, and other specialty items are also store-door delivered. In order to provide data on all of these items, we utilized Nielsen Food Index Services since they cover all items, regardless of source or type of distribution. Since what is happening in the marketplace is such an important factor in the marketing of any product, we added another service, Store Observation Service, to assess the elements that so often influence the customer's deci- sion to buy or not buy.

A. C. Nielsen Company Services 83 Availability or distribution is perhaps the most critical of all from everyone's point of view manufacturer, retailer, and customer. Is the pro- duct there, on the shelf, where it can be purchased? But also important are the selling prices, your product and your com- petitors'; the number of shelf facings (i.e., Is it easily visible?; shelf location What shelf, what section, and where is the section in the store; and, Is the product being promoted or featured, and if so, how? These are the questions most often asked, but many others of equal importance can be handled by this service. To answer questions such as these, Nielsen employs a nationwide field staff specifically trained for this work. The staff, 80 in total, are located in 50 standard metropolitan statistical areas, in all areas of the country. The administration of the service is centralized in Northbrook, as are all Nielsen services. In your invitation offering us the opportunity to present "our wares" before you this afternoon, you also requested that we address ourselves to certain questions, the first being, What gaps exist in the data base? Nielsen Food Index, Nielsen Major Markets, and Nielsen Early Intelli- gence Service samples are all drawn from a universe with the following definition: all grocery stores, as defined by the U.S. Census Bureau, plus supermarket departments of general merchandise stores, i.e., the full-line grocery sections in mass merchandisers. This definition leaves out fringe- type stores handling small amounts of packaged groceries. This measurement gap, expressed in terms of consumer grocery products, could be estimated for a typical product category to be 5 percent. Nielsen Early Intelligence System's Data Service's samples do not include small grocery stores. Since these small stores account for approximately 15 per- cent of total dollar volume, this could raise the gap for Nets to close to 20 percent. How complete are the data? I believe this can be answered in the follow- ing summary of Nielsen services: · Nielsen Food Index Services including Major Market Service- provide consumer sales plus retail inventories, purchases, distribution, out-of-stock, the prices paid by consumers, in-store promotional data, and retailer advertising. · Data Services provide product movement to a sample of supermarkets. From this source, data are produced covering brand movement in both dollars and units per week per store, shares, distribution, and retail prices, and all brand details are broken down by size, flavor, type, and so forth. · Product Pickup Service is a complete service and limited only by the client's desires. Store Observation Service, as the name implies, is based on observable data in the selling area of the store.

84 OLIVER S. CASTLE How reliable are the data? Samples for these services are drawn from a sampling frame defined in Table 1. Practically maximum geographical dis- persion is achieved since, as an example, the 1,300-store national sample is spread through 606 counties. Similar dispersion is achieved in smaller areas sampled. Quinquennial Census material, updated with material from the Census Bureau Current Retail Trade Programs, is used both to determine the sam- pling frame and to estimate universe data from sample results. Sampling list material on a current basis is obtained from: · progressive grocer lists, · chain lists, and · gathered by survey. Sample units are selected with disproportionate sampling rates with larger stores having the greater chance of selection over small stores. This differ- ence in selection rates is taken into consideration when sample data are projected to universe estimates. At the same time that stores are audited to obtain data on individual grocery products, total retail dollar movement is obtained for each store for the audit period. This dollar movement is combined with the individual grocery product data to form a ratio. This ratio, having less variance from store to store than the absolute product data to be measured, gives the estimating process higher efficiency. In summary, the sample design and estimation process has the following characteristics: a stratified sample design using practically maximum geographic dis- pers~on; 2. disproportionate sampling with greater chance of selection for large stores; and 3. ratio estimation to achieve greater precision. Finally, I would like to comment on what I believe are the advantages that distinguish Nielsen marketing research services from any other. First, in most instances, the data produced flow from the marketplace and result from consumer actions. As such, when we report consumer sales, it means the products in question have been purchased by consumers, not merely moved out of a warehouse. It's the same with prices these are the prices paid by consumers. The distribution measurements are made in the store, as are the retail inventories and the data relating to sales influencing factors.

A. C. Nielsen Company Services 85 Second, nearly all the services are based on scientifically drawn samples. We believe that samples provide the greatest value in terms of dollars ex- pended, accuracy achieved, and breadth of data. And last, because of our access to the stores making up the samples, we can achieve a flexibility that's impossible to provide otherwise. I believe the study we are currently doing for the FDA iS a good example of that.