TECHNICAL PAPER NO. 46 Reevaluation a Further Of the 1 972-73 Based on Revised U.S. Consumer Estimates of _. ■- - Personal Consume Expenditure Expenditures Survey U.S. DEPARTMENT OF COMMERCE Bureau of the Census Digitized by the Internet Archive in 2012 with funding from LYRASIS Members and Sloan Foundation http://archive.org/details/reevaluationof1900pear TECHNICAL PAPER NO. 46 Issued July 1979 Reevaluation of the 1972-73 U.S. Consumer Expenditure Survey A Further Examination Based on Revised Estimates of Personal Consumer Expenditures Robert B. Pearl Survey Research Laboratory University of Illinois I] f fit spates & U.S. Department of Commerce Juanita M. Kreps. Secretary Courtenay M. Slater, Chief Economist BUREAU OF THE CENSUS Daniel B. Levine, Acting Director BUREAU OF THE CENSUS Daniel B. Levine, Acting Director Daniel B. Levine, Deputy Director DEMOGRAPHIC SURVEYS DIVISION ACKNOWLEDGEMENTS This evaluation of the results of the 1972-73 Consumer Expenditure Survey has been carried out under a joint statistical agreement between the Survey Research Laboratory of the University of Illinois and the Research Center for Measurement Methods of the Bureau of the Census. The author, Robert B. Pearl, of the Survey Research Laboratory, wishes to acknowledge the major contributions to this undertaking of such present and former members of the staff of the Census Bureau's Demographic Surveys Division as Thomas Walsh, Gail Hoff, Jeanne Griffith, Charles Cowan, and Judy Wiener, and an intrepid group of computer programmers headed by Jack George, Hal Stone, and Stephen Phillips. Appreciation is also extended to Robert Ferber and Seymour Sudman of the Survey Research Laboratory and Barbara Bailar of the Research Center for Measurement Methods for their advice, suggestions, and support. Daniel B. Levine, Associate Director for Demographic Fields, Earle J. Gerson, then Chief, Demographic Surveys Division, Bureau of the Census, and Harold Nisselson, then Chief, Research Center for Measurement Methods, played an important role in facilitating implementation of the project. The survey data used in the evaluation were derived principally from reweighted original survey data tapes and do not incorporate editing changes which may have been made at later stages of processing by the Bureau of Labor Statistics (BLS), which is responsible for the official survey results. The data in this report may, therefore, differ somewhat from those already published or to be published by BLS or which may be derived from public-use data tapes issued by that Agency. The author takes full responsibility for any errors or misinterpretations in the use of the data as well as for the conclusions and recommendations in the report. Library of Congress Cataloging in Publication Data Pearl, Robert B. Reevaluation of the 1972-73 U.S. consumer expenditure survey. (Technical paper - U.S. Bureau of the Census ; no. 46) 1. Cost and standard of living— United States. 2. Consumption (Economics)— United States. 3. Prices- United States. I. Title. II. Series: United States. Bureau of the Census. Technical paper ; no. 46. HD6983.P423 339.4'1'0973 79-15548 For sale by Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402, or any U.S. Department of Commerce district office. Postage stamps not acceptable; currency submitted at sender's risk. Remittances from foreign countries must be by international money order or by a draft on a U.S. bank. Stock Number 003-024-01938-8 Contents Page Introduction 1 Chapter 1 . Summary of findings 3 Chapter 2. Food and beverage expenditures 9 Chapter 3. Small nonfood expenditures 13 Chapter 4. Clothing expenditures 15 Chapter 5. Household appliances 18 Chapter 6. Household furnishings 21 Chapter 7. Automobile and vehicle expenses 23 Chapter 8. Housing costs 26 Chapter 9. Health expenditures 28 Chapter 10. Education, travel, and miscellaneous expenses 31 Chapter 1 1. Conclusions and recommendations 33 Appendix A. Personal consumption expenditures 37 Appendix B. Variations in completion of diaries by different population groups 39 Appendix C. Use of special check questions in the diary procedure for sensitive items 42 Appendix D. Timing biases in diary reporting 46 Ml Introduction In 1972-73, a large-scale national Consumer Expenditure Survey was conducted by the Bureau of the Census, on behalf of the Bureau of Labor Statistics (BLS), primarily for the purpose of updating the commodity weights and the selection of items for the Consumer Price Index (CPI). A completely different methodology was employed in this survey from that utilized by BLS in prior periodic undertakings in this field. In previous surveys, the most recent in 1960-61, the basic procedure entailed extremely lengthy interviews in which an effort was made to determine the expenditures of each family for an entire calendar year. The procedure introduced in 1972-73, in con- trast, consisted of two principal elements: 1. A national panel of about 10,000 families each year visited at quarterly intervals to obtain information by retrospective interview for the larger items of expenditure and certain repetitive items (rents, utilities, etc.). Specific expenditure categories were covered either each quarter, or at semiannual or annual intervals, depending primarily on expenditure size. 2. A rotating sample of about 200-250 families each week were asked to keep a diary, or record, of all expenditures for the subsequent 2-week period. Although the diary operation was intended primarily to cover smaller items of expenditure, the fact that all categories were included provides a number of options in compiling estimates as well as many research opportunities. Because of the importance of the survey results in con- junction with the CPI, and the far-reaching mothodological changes instituted this time, an evaluation of the data has been underway for some time. Such an assessment may be especially pertinent in view of the Government's plans to inaugurate a continuing expenditure survey in the near future using similar techniques. A detailed report, presenting the preliminary findings of the evaluation, has recently been published in the Census Technical Paper Series. 1 The basic procedure used in the evaluation was to compare the results from the quarterly panel with those from the diary operation, where the same categories were covered in both, and to relate either or both sets to various independent sources of expenditure data. The findings were regarded as preliminary partly because the data were not always in sufficient detail for precise comparisons, but primarily because of uncertainty concerning the validity of some of the independent data used as a standard. Although a variety of independent sources were consulted for this purpose, principal dependence was placed on the Personal Consumption Expenditure (PCE) estimates prepared by the Department of Commerce in conjunction with the Gross National Product Accounts. These essentially represent the market value of goods and services purchased by private individuals and nonprofit institutions in the United States. The preparation of these estimates is an extremely detailed and complex matter which cannot be appropriately described here. 2 In brief, the principal starting point consists of data on the production value of goods from the quinquennial censuses of manufactures and other sources, and the value of services from a variety of sources. Through input-output analysis, the flow of commodities and services is traced through the various channels in the economy and cost and profit margins are added to the production value at each stage to arrive at the final market value. Although the initial production data probably constitute one of the more reliable bodies of statistics, the numerous transformations required to convert to a consumer basis make it difficult, if not impossible, to assess the accuracy of the final estimates. However, these estimates represent the only compre- hensive set of consumer expenditure data available on a more or less continuing basis. On a current basis, PCE estimates are prepared only for relatively broad categories of expenditures. These essentially represent projections from the most recently available bench- mark source, in particular the quinquennial censuses of manu- factures and service industries. Estimates for detailed categories are prepared only at benchmark dates. At the time this evaluation was undertaken, PCE estimates for broad categories were available for the years 1972 and 1973, based mainly on projections from the quinquennial benchmark year of 1967. The detailed estimates, however, were available only for 1967. In order to examine the survey data in sufficient detail, it was 'See U.S. Bureau of the Census, Technical Paper No. 45, "United States Consumer Expenditure Survey: 1972-73, A Preliminary Evalua- tion," issued October 1978. 2 For a detailed description, see U.S. Department of Commerce, Office of Business Economics (now Bureau of Economic Analysis), "Development of National Income Measures, Supplement to Survey of Current Business, Washington, 1954. • necessary for the author to project the 1967 estimates to 1972, using available census production and foreign trade data. Since the methods used for this purpose were, of necessity, rather crude, additional uncertainty attended the detailed comparisons between the survey and independent data. Recently, the Department of Commerce has completed a revision of the detailed PCE estimates on the basis of the benchmark data from the 1972 Economic Censuses and other sources. The primary purpose of this present report is to reexamine the preliminary findings of the evaluation and the various recommendations made for methodological improve- ments, in the light of the revised PCE data. Although the new estimates should reduce the degree of uncertainty in the comparisons, the PCE estimates continue to be subject to various errors and biases and considerable caution must be exercised in interpreting the results. As discussed further in the appendix to this report, a number of incomparabilities remain between the survey and the PCE data for which only rough allowances can be made. Moreover, in some important sectors, particularly health, education, and housing expenditures, the conceptual disparities are such that reliance had to be placed primarily on independent sources other than PCE. A summary of the revised findings is presented in chapter 1. Chapters 2 through 1 examine the implications of the new PCE estimates for detailed categories. Final conclusions and recom- mendations appear in chapter 11. A good deal of material (including various statistical tables) presented in the preliminary report (Technical Paper No. 45), but not affected by the PCE revision, is not repeated here. However, references to the earlier material are made wherever appropriate. For example, a rather detailed description of the survey design and methodology of the 1972-73 Consumer Expenditure Survey was included in the preliminary report. Illustrations of the survey questionnaires and explanations of the various independent sources used in the evaluation were provided in the appendixes to that report. On the other hand, the present report includes the results of various special analyses of the diary procedure— differential completion rates (appendix B), results of special check questions (appendix C), and timing biases (appendix D)— which had not heretofore been published. Chapter 1. Summary of Findings Table 1-1 presents a summary of the comparisons between the survey data and the independent sources for the various categories of expenditures, which are covered in greater detail in subsequent chapters. For purposes of summarization, some of the detailed categories have been combined and averaged. The table designates the "best" survey source, that is, the one- quarterly panel or diary operation— which was generally closest to the independent data, in cases where the two survey estimates were significantly different. The ratios of the "best" survey estimates to the independent levels are indicated in terms of broad class intervals, although the actual ratios, both on the basis of the revised and preliminary PCE figures, are also shown. 1 Although the magnitude of these ratios may provide some indication of the adequacy of the survey estimates, the margins can be regarded as only approximations, for the reasons already stated, and have been used primarily for purposes of detecting any consistent and meaningful patterns which might have a bearing on the methodology. It should be noted that the survey results used in this evaluation are derived from special tabulations of reweighted original data tapes. They do not reflect editing changes which may have been made at later stages of processing by BLS. Asa result, the figures may differ somewhat from those already published or to be published by that Agency or which may be compiled from the public-use data tapes recently issued. Certain differences in conceptual approach and time reference would also contribute to the disparities. For example, the BLS estimates for specific product categories generally exclude gifts purchased for persons outside the household which are, instead, grouped together as a separate aggregate. For the evaluation, however, the gifts are included in their respective categories in order to achieve closer correspondence with the independent estimates. For the same purposes, the estimates for ap, liances,' furniture, and certain other items in the evaluation are restricted to those acquired in new form, whereas the BLS estimates often combine new and used products. In a number of instances, the revised PCE estimates were substantially higher than the preliminary ones, which had the effect of widening the disparities in relation to the survey data. In perhaps an equal number of cases, the revised PCE levels were lower, resulting in the opposite effect on the comparisons. Most commonly, the changes in the PCE estimates were relatively insignificant. There was also little or no effect on the compari- sons, of course, in those categories in which sources other than PCE provided the main basis for evaluation. On balance, it was only in rare instances that the preliminary assessments affecting a broad expenditure category were significantly altered as a consequence of the PCE revision. Even within major categories, the principal patterns were, most often, relatively unaffected. A summary of the revised findings follows for the various expenditure categories. Food and Beverage Expenditures (Chapter 2) The revision made virtually no difference in the PCE levels for total food expenditures for home use, and there were only moderate changes for the majority of specific food categories. The data, therefore, continued to exhibit the reasonably close agreement previously observed for this category, after allowance insofar as possible for conceptual incomparabilities. The fact that the homemaker— the usual diary keeper for the family— ordinarily does the food buying for the family undoubtedly contributed to this outcome. In addition, about half the space on the diary record was devoted to food purchases, so that this category received much more attention than any other. 2 Considerable differences were observed, however, in the precision with which the various food categories were reported. The reporting on both the revised and preliminary basis was apparently most complete for relatively costly items, such as meat and poultry, and for perishable products such as milk, eggs, and bread, which are commonly used on a very frequent basis. On the other hand, pronounced disparities in relation to the independent estimates continued for food staples such as flour, sugar, and shortening, which are bought less frequently, with each purchase used over a considerable period of time. A likely explanation for these differences given in the preliminary report, which still seems to hold, is that many respondents may not start keeping their diaries promptly— or do not make entries, as requested, on a daily basis— but later attempt to reconstruct the omitted periods by memory. In doing so, items which constitute the main course in a meal, or which are purchased and used frequently, might be more readily remembered. Other comparisons lend some support to this thesis. For example, the generally smaller disparities for fruit products as compared to vegetables may reflect the fact that the former are more likely 'The class intervals, which are in terms of letter codes A to E, may not always be entirely consistent with the actual ratios shown. The former sometimes allow for conceptual and other differences between the survey and independent data for which numerical adjustments cannot be made. 2 The diary record contained two facing pages for each day of the reporting period. The left-hand page was devoted entirely to food and beverage expenditures for home use and was subdivided into several sections covering broad categories (bakery and dairy products; meat, poultry, and fish; fruits and vegetables, etc.) to be eaten as snacks or as separate courses and could, thus, stand out more distinctly than items usually blended into a meal. The PCE estimates for meals in restaurants and other eating places were raised somewhat as a consequence of the revision. Nevertheless, the unexpectedly close correspondence between the survey and independent sources reported in the preliminary findings continues to be observed on the revised basis. Some deficiencies had been anticipated in this sector because so large a proportion of these expenditures could be made by individual family members other than the housewife. The prominent positioning of the section for recording purchased meals on the diary form and some special attention to this item in the diary checking procedure apparently served to overcome this diffi- culty in large measure. Certain conceptual and reporting differences between the sources, discussed in the detailed chapter on food expenditures, could have narrowed any differences between the survey and independent data. It is possible, therefore, that the precision of the survey results, while still relatively high, could be a notch lower than that shown. The revised estimates did not achieve any improvement in the drastic understatement in the survey estimates for alcoholic beverages. This almost inevitable survey finding is mainly attributed to sensitivity in reporting expenditures of this kind. Small Nonfood Expenditures (Chapter 3) The pattern for small nonfood expenditures, for which the diary was the principal if not the only source, was generally similar to that reported in the preliminary findings. The revised PCE figures were somewhat higher in some instances and lower in others, but the implications for the survey estimates remained about the same as before. The predominant factor in the survey reporting continued to be the role of the various family members in making purchases. Where the responsibility was principally that of the home- maker— such as for cleaning supplies and other household products or for household services— the reporting was con- siderably more complete than in cases where other household members were substantially involved, as for toiletries or hair care. Even for those expenditures where the homemaker predominated, however, the reporting appeared to be generally less adequate than for food purchases, probably partly a reflection of the much smaller amount of space and attention given nonfood items on the diary record. Clothing Expenditures (Chapter 4) The picture also did not change materially for clothing expenditures as a result of the revised PCE estimates. As previously reported, the quarterly panel and diary estimates were about equivalent on an overall basis, in spite of the fact that this category was explored in exhaustive detail in the interview survey but accorded scant attention in the diary procedure. Both sets of estimates, however, fell appreciably below the PCE levels and those from other independent sources. The total PCE figure for clothing expenditures was not markedly affected by the revision. More complete and accurate estimates are available, however, for the various components. As a result, there are now somewhat larger differences between the survey and independent estimates for most specific categories. Only the larger items— coats and suits— continue to exhibit a close correspondence between the sources. Major disparities are apparent for most of the others. The more detailed comparisons with the independent data are limited primarily to the quarterly panel data, since the diary results contained a substantial proportion of incomplete or inadequately described entries which could not be assigned to specific categories. In spite of this fact, the diary apparently provided the more complete data for a diversified category such as clothing accessories. More surprisingly, the diary estimates also exceeded those from the quarterly panel, and were thereby closer to the independent levels, for a more clear-cut category such as footwear. For the broad range of middle and lower priced articles, the disparities in comparison with the independ- ent data varied a good deal among categories, without any apparent relationship to the cost or importance of the items. Household Appliances (Chapter 5) The primary findings issued earlier revealed a rather close correspondence between the survey estimates and the PCE data for both major and minor household appliances. This result was attributed, in part, to the use of the so-called inventory approach in the quarterly panel. Instead of asking directly about purchases for a given period, the procedure used was to inquire about possession of the articles in question and about the date of acquisition. The initial inventory was then updated in the course of subsequent interviews. The procedure provides various options in preparing estimates, which are described in chapter 5. Comparisons with the revised PCE figures still indicate a generally favorable outcome for most major appliances. Defi- ciencies of some magnitude now appear, however, for several of the minor appliance categories, as a consequence of substan- tially increased PCE levels. This result is now more consistent with reporting differentials for smaller as compared to larger products in other expenditure categories. Household Furnishings (Chapter 6) This is one category where the revised PCE estimates generally resulted in a somewhat closer correspondence with the survey data. As before, the comparisons present a rather classic example of the relationship between the size of an item and the likelihood that it will be reported. The closest correspondence was for the largest items— furniture and floor, window, and furniture coverings. The gap widened considerably for house- hold linens and especially for smaller items such as dinnerware and cookware, decorative items, and luggage. The quarterly panel appeared to be the superior source for the larger expenditures, partly because of lower sampling variances. The diary results seemed to be at least as complete for the smaller items, although both sets were deficient in these respects. Automobile and Vehicle Expenses (Chapter 7) In most categories, the survey data corresponded closely with the independent sources, on the revised as well as the preliminary basis. The main disparities were for tires and other accessories. Although the homemaker would normally have less responsibility for vehicle expenses than most other items, the diary estimates held up surprisingly well for some of the smaller categories. The quarterly panel, as expected, was clearly the superior source for vehicle purchases and also appeared to provide the more reliable data for vehicle registration, insurance, and similar items. Housing Costs (Chapter 8) Once again, with one possible exception, the survey and independent estimates corresponded closely on a revised as well as unrevised basis. The exception was fuel expenditures where the gap between the survey and the PCE estimates was, nevertheless, narrowed considerably as a result of the revision. In several sectors, moreover, the diary estimates were in substantial agreement with the quarterly panel data, although the latter were initially considered to be the primary source for this information. Health Expenditures (Chapter 9) The unexpectedly close correspondence between the survey results and the independent estimates in the complex health sector was highlighted in the preliminary report. Since the comparisons for the major elements of health care— physician and dental services, hospital care, etc.— were based on independ- ent sources other than PCE, the findings are not much affected by the revision. As before, the only significant disparity for these basic services occurred in the case of hospital care and was attributed to the complications arising from insurance reim- bursements and other so-called third-party payors. The main sector involving PCE data comprised drugs and medicines and various medical supplies and appliances. The previous agreement was not altered appreciably for drugs and medicines, in spite of an increase in the PCE levels as a result of the revision, but a gap of some magnitude between the survey and independent data was observed for other medical products. As noted in the preliminary report, the general correspondence between the quarterly and diary results for some of the more important health expenditures was one of the more surprising survey findings. Education, Travel, and Miscellaneous Expenses (Chapter 10) As might be expected, the findings were quite variable for this diverse group of expenditures. In some instances, the results were not affected by the PCE revision because other sources were the main basis for comparison. Only very limited comparisons were possible for education expenses, a sector where the PCE data are conceptually different from the survey estimates. On the basis of other independent sources, there appeared to be some deficiency in the survey results, possibly because of the difficulty of covering students attending college away from home in a household survey. The quarterly panel was evidently the better survey source for the larger payments, such as tuition, but it seemed desirable to alter the survey procedures, whereby students away at college would be interviewed directly at their school quarters instead of obtaining proxy information, in some instances, from their parents at home. Although there was not direct evidence on this point, there is no reason to believe that the smaller items such as books, supplies, and meals eaten out— like other small expenditures— would not be better reported in the diary operation. Comparisons for trips and vacations— which were covered in detail in the quarterly panel— could be made only in terms of the number of trips taken rather than for expenditures. On this basis, the survey results appeared to be consistent with other sources. Since the validity of expenditure reporting may depend more on whether or not an expenditure (in this case, a trip) is reported than on the precise amount of the expenditure, even this limited finding with respect to trips taken could be significant. In a related sector, the survey data, in this case from the diary operation, for expenditures for public transportation continued to exhibit serious disparities in comparison with the revised PCE estimates. In fact, the differences were widened as a result of increased PCE levels for taxis and airfare. The apparent understatements in this sector could possibly be attributed, in part, to the fact that most such expenditures are made on an individual basis. For various miscellaneous products and services, a few generalizations may suffice since the patterns were reasonably consistent with those already cited for items of a similar nature. For larger expenditures (pianos, organs, funeral expenses, etc.), the quarterly panel was the superior survey source and the results corresponded closely with the independent estimates. A major deficiency for watches and jewelry could have resulted, in part, from some sensitivity in reporting this item. A similar disparity for moving expenses— a category for which there was a sharp increase in the PCE estimate as a result of the revision- might be attributed to some undercoverage of recent movers in the sample. Other General Findings Although unrelated to the PCE revision, mention should be made of some other general findings which have or could have a bearing on the methodology. Timing biases in diary keeping. Perhaps the most persistent survey finding was confirmation of a characteristic phenomenon in diary operations, that is, for a higher level of expenditure reporting in the earlier as opposed to the later stages of the recordkeeping period. For 2-week diaries, the outcome is almost invariably a higher level in the first compared with the second week. This result was observed for virtually every expenditure category in the 1972-73 survey, and detailed comparisons for the 2 weeks are provided for food purchases and small nonfood expenditures in the preliminary report (Technical Paper No. 45). The magnitude of first-week/second-week differences among categories appeared to be almost random in nature and unrelated to other characteristics, such as completeness of reporting in relation to the independent data. A more detailed analysis based on day-by-day reporting is presented in appendix D. The data reveal that the higher levels for the first week are primarily a consequence of the exceed- ingly large amount of expenditures reported for the first day of the recordkeeping period (some 40 to 50 percent above the daily averages). Reporting was also above average on the first day of the second reporting week, although not nearly to the extent found in the first week. 3 There was a general tendency for a gradual decline in reporting in the course of both weeks. Although estimates on a daily basis were not computed for detailed products and services, information based on various combinations of days suggests that this pattern was persistent among categories. Several theories have been offered concerning the reasons for this reporting phenomenon. Perhaps the most reasonable, on the basis of the evidence, is that reporting in the early days may be exaggerated as a result of "telescoping." According to this version, many respondents may not start their diaries immedi- ately but omit the first day or two and then attempt to reconstruct that period by memory. In doing so, they may inadvertently include some expenditures made prior to the reporting period. 4 The gradual decline in reporting in the course of each week also suggests that some fatigue or boredom may occur as the period lenthens. Still another hypothesis— that temporary alterations in buying habits may occur at the start because a diary is being kept— does not appear to be supportable as a major factor because the differences are so widespread among categories. 5 Telescoping in the interview panel. The phenomenon of "telescoping," that is, the tendency to report an event (in this 3 Under the procedure followed, interviewers made three visits to the sample households. At the first visit, certain background information was obtained and the diaries were placed for the first week. The interviewer returned a week later to pick up and check the first diaries and place the record books for the second week. The final visit was made the following week to pick up the second diaries. 4 The likelihood of exaggeration for the first day is supported by the fact that a projection of the levels for that day would produce aggregates far in excess of the independent data. s Certain mechanical factors could have contributed to the "first day" phenomenon. For example, some persons may have disregarded the fact that separate pages were provided for each day and, instead, continued making entries on the first and subsequent pages until the space was exhausted. case, an expenditure) as having occurred more recently than was actually the case, has previously been observed mainly in interview surveys. 6 This tendency was controlled in the 1972-73 survey by means of a "bounding" procedure in the quarterly panel. Interviewers always had possession of information of expenditures reported at previous visits, so that they could detect and exclude any items reported a second time because of telescoping. The bounding process, however, controlled only for the standard survey periods, that is, on a quarterly basis for items covered each quarter, on a semiannual basis for those asked every 6 months, etc. In recording expenditures, the actual month of purchase was specified, so that alternative estimates could be computed, if desired, for varying periods. The results indicated, however, that considerable telescoping occurred within the standard periods. For example, data for the most recent month of each quarter, or for the second quarter of each half-year, considerably exceeded the levels for the earlier parts of those standard periods. The differences appeared to be wider than could be attributed to improved reporting for shorter recall periods due to memory factors. Some of these differences are indicated in the detailed discussion in both the present and preliminary reports. Response rates. One of the principal concerns in initiating a complex new system of this kind was the matter of achieving adequate levels of cooperation in the survey. As previously reported, this was one concern that was almost entirely allayed. The response rate in the quarterly panel started at close to 95 percent at the first interview and dipped to only slightly under 90 percent by the final quarterly visit, even though the interviews ranged from VA to 3 hours in length depending on the phase of the cycle. In the diary operation, because of initial resource and administrative problems, the response rate started at only around 75 percent, but gradually increased to and held at almost 90 percent for most of the survey period once the difficulties were resolved. 7 6 Actually, telescoping can run in either direction but the main tendency is to move events forward in time. 7 For further details, see Walsh, Thomas C, "Selected Results from the 1972-73 Diary Survey," Journal of Marketing Research, Special Edition, August 1977. ,oj Table 1-1. Summary of Findings for Expenditure Categories: to Independent Sources 1972-73 Expenditure Survey Estimates Compared Category Best survey source 1 QP = Quarterly panel D = Diary opera- tion N = No significant difference Ratio of Best survey estimates to independent sources Best judgment as to range (allowing for conceptual and other differences between sources) 2 A = . 9 or higher B = .8 to .89 C = .7 to .79 D = .6 to .69 E = Less than .6 Survey results Actual ratio of survey to independent estimates 3 Prelim- inary Estimated standard error of ratios Independent sources used* Food purchases for home use Meat or poultry Eggs Fresh milk Bread and other fresh baked items Food staples (flour, sugar, shortening, canned milk, etc.) Fruits — fresh or processed Vegetables--f resh or processed Purchased meals or snacks Alcoholic beverages Small nonfood expenditures Products Items purchased mainly by homemaker (laundry and cleaning products, paper goods, etc.) Items likely to be purchased by various members (toiletries, film, reading material, etc.) Services Mainly responsibility of homemaker (laundry services, household help, etc.) Dispersed responsibility (hair care, shoe and watch repairs, sporting events, etc.) Clothing expenditures Larger items (coats, suits, etc.) Medium and smaller articles (dresses, shirts, underwear, hosiery, etc) Accessories (ties, handbags, gloves, etc.). Footwear Household appliances Major appliances (refrigerators, washers, television, etc.) Minor appliances (toasters, hair dryers, radios , etc . ) Household furnishings Furniture Other larger items (floor coverings, drapes, slipcovers, etc.) Household linens (sheets, tablecloths, towels, etc.) Smaller items (dinnerware, cookware, luggage, decorative items, hand tools tools, etc.) Automobile and other vehicle expenses Vehicle purchase Gasoline and oil Tires and accessories Vehicle repairs and maintenance Vehicle insurance Housing expenses Rent Mortgage payments and taxes* Home repairs and alterations* Utility costs (electricity, gas, water, te lephone , etc.) Fuel costs (fuel oil, bottled gas, coal, etc.) Footnotes at end of table. N QP N D D **QP **QP N QP N QP b A.l D,E B C A,B E D,E A A C A,B A A A A A »B,C .85 1.02 .95 .92 .59 .86 .74 1.07 .36 .82 .55 1.04 .54 .73 1.01 .66 .72 .77 .96 .75 .91 .81 .77 .59" 1.01 .98 .75 .91 .93 1.06 1.02 1.04 1.01 .76 .85 1.02 .87 .93 .98 .63 .85 .74 1.11 .39 .72 .57 1.04 .68 .72 1.11 .76 .72 .75 1.00 1.01 .50" 1.01 .98 93 1.03 1.02 1.04 1.01 .61 .02 .015 .03 .015 .03 .015 .02 .03 .04 .04 .02 .02 Cen-Bus , PCE Cen-Bus,PCE PCE Cen-Serv,PCE Cen-Bus, PCE Cen-Bus, PCE Cen-Bus, MRCA, PCE Cen-Bus, MRCA, PCE PCE .04 AHS , PCE 015 PCE .03 PCE .03 PCE .02 A.M. Best .03 .04 .06 .03 .03 PCE AHS SORAR PCE pa: Table 1-1. Summary of Findings for Expenditure Categories: to Independent Sources— Continued 1972-73 Expenditure Survey Estimates Compared Category Best survey source 1 QP = Quart pane D = Diary tion N = No sig diffe erly 1 opera- nif leant re nee Ratio of Best survey estimates to independent sources Best judgment as to range (allowing for conceptual and other differences between sources) 2 A = . 9 or higher B = .8 to .89 C = .7 to .79 D = .6 to .69 E = Less than . 6 Survey results Actual ratio of survey to independent estimates 3 Prelim- inary Estimated standard error of ratios Independent sources used 6 Health expenditures Hospital services* Physician, dental, and other professional services* Drugs and medicines Medical supplies and appliances Health insurance premiums* Other expenditures Education tuition and fees. Trips and vacations Public transportation Miscellaneous Large items (pianos, organs, funeral costs) Moderate items (musical instruments, sporting equipment, appliance repairs). Watches and jewelry Moving and storage costs WP B,C E E .76 .98 .96 .72 1.10 .82 1.02 .53 .96 .81 .48 .51 .76 .98 1.02 NA 1.10 .82 1.02 .71 .93 .86 .52 CHAS SSA CHAS SSA PCE SSA PCE HIS NCES NTS PCE PCE ^Signifies that comparisons are based on 2 years of data. For other categories, comparisons were based on only 1 year, usually 1972. *The "best" survey estimate is defined as the one—quarterly panel or diary estimate — which was closest to the independent figures. Where a double asterisk (**) is appended to the code, this indicates the specified source was the only one for which a comparison with the independent estimates could be made in the required detail. In most of these latter cases, the specified source was also the only realistic source for the data in question. 2 The ranges in this column are not always entirely consistent with the computed ratios in the next column, but attempt to make allowances for conceptual differences between the survey and independent sources, disparities among the independent sources themselves, and other factors for which numerical adjustments could not be made. 3 The ratios are based on weighted averages of the survey data (numerator of ratio) and of the independent data (denominator of ratio) for the various individual categories combined on a given line of the table. Where the two survey estimates were not significantly different (Code "N" in the "best" survey source column), the quarterly panel and the diary data were averaged. Where more than one independent source is specified, those data were averaged for this purpose. & The independent sources, in alphabetical order, are as follows: AHS--Annual Housing Survey, Bureau of the Census A. M. Best — Insurance company premimum data compiled by A. M, Best and Co. Cen-Bus- 1972 Census of Business Merchandise Line Data Cen-Serv — 1972 Census of Selected Service Industries CHAS--Center for Health Administration Studies, University of Chicago ERS--Economic Research Service, U.S. Dept. of Agriculture HIS — Health Interview Survey, National Center for Health Statistics, U.S. Dept. of HEW MRCA--Panel data from the Market Research Corp. of America NCES — National Center for Education Statistics, U.S. Dept. of HEW NTS — National Travel Survey, Bureau of the Census PCE--Personal Consumption Expenditures, GNP Accounts, U.S. Dept. of Commerce SSA — Social Sercurity Administration, U.S. Dept. of HEW SORAR--Survey of Residential Alterations and Repairs, Bureau of the Census See appendix A of the preliminary report, Technical Paper No. 45 , for a description of these sources. Uncertainty about the magnitude of a conceptual difference precludes a more precise judgment for this borderline value. The ratio of the survey to the independent estimate is understated because the PCE figure used as a base relates to "off premise" consumption and includes (in addition to food purchased for home use) snacks purchased in stores (and other establishments which are not eating places) but consumed outside the home. The survey estimate, however, relates only to food purchased for home use. The reverse situation holds for the PCE and survey estimates for "purchased meals and snacks" where the ratio is overstated . 6 This is an average value for the eight major categories included in this bracket. Of these, three were significantly more completely reported than indicated by the coded mean value, but there was no clear-cut distinction among these by expenditure size. 7 Using optimum estimate (including value of gifts received and averaging the two alternative estimates for the year), see text. Under the stand- ard estimation, the rating for major appliances would be "A, B" but that for minor appliances would be "d". e Survey estimates used in the computation were diary estimates for dinnerware-cookware and hand tools and average of the two estimates for decorative items and luggage. 'Computed ratio is understated (magnitude uncertain) because the PCE base figure (but not the survey estimate) includes fuel used in rental quarters which is supplied and directly paid for by the landlord. 10 Comparison based on college tuition only. Comparison based on number of trips taken rather than expenditure. Chapter 2. Food and Beverage Expenditures Although some rough approximations of overall food and beverage expenditures were obtained in the quarterly panel, the diary operation, as might be expected, was the main source of information in this sector. As indicated elsewhere, the left-hand page of the two facing pages provided for each reporting day was devoted to food expenditures for home use. A separate section at the top of the right-hand page was provided for recording expenditures for meals and snacks purchased and eaten out. The main purpose of the inquiry in the interview panel— which was asked each quarter— was to produce an annual overall figure for each individual family, which was not obtainable from the rotating sample used in the diary operation. Overall Comparisons Between the Survey and Independent Estimates Table 2-1 presents the overall comparisons between the survey estimates and the independent sources for food and beverage expenditures. In addition to the PCE estimates, comparisons are made with aggregates developed by the Economic Research Service (ERS) of the U.S. Department of Agriculture and with data from the Census of Business. As may be seen, the PCE estimate for food purchases for home use was unchanged by the revision, but the levels were raised for purchased meals and snacks and for alcoholic beverage expenditures. The Census of Business estimate excludes sales by enterprises without paid employees, which could be substantial for certain food items as a result of seasonal produce stands, small bakeries and groceries, and the like. For purposes of these comparisons, the diary estimates were adjusted upward to include expenditures made while families were away on trips and vacations. Under the survey procedure followed, diaries were not to be maintained at those times. However, an estimate was available for such expenditures from an inquiry on trips and vacations made in the interview panel. The adjustment for food expenditures for home use (in this case, in vacation homes) was about $1 billion. It was much larger ($3 billion) for purchased meals and snacks. Also included in the survey estimates, for greater comparability with the independent sources, were expenditures for pet food (close to $2 billion). The higher levels shown for the quarterly panel as compared with the diary estimates for food expenditures for home use are fairly characteristic of differences between global and detailed data of this kind. Although an effort was made to avoid this, the summary quarterly panel estimates probably included various nonfood products, such as cleaning and paper supplies, which are customarily purchased together with food. Although the survey estimates for food expenditures for home use are appreciably below those from the two principal independent sources, even on the revised basis, part of the disparity can be attributed to conceptual incomparabilities. The survey estimates relate more or less literally to food purchased for preparation and/or consumption at home. This would include prepared foods purchased in carryout places which were also eating places, but intended for home consumption. The independent estimates shown for comparison relate to off- premises consumption. They would include food bought in any establishment, not an eating place, which was to be consumed elsewhere, whether at home or some other place. 1 Food bought in carryout places which were also eating places, however, would not be considered off-premises purchases in the inde- pendent estimates, no matter where consumed. The main difference between the concepts relates to such items as candy, crackers, soft drinks, etc., bought in establish- ments which were not eating places but consumed outside the home. These would tend to raise the independent off-premises estimates in relation to the survey home-use estimates, probably by as much as several billion dollars. There would be some offset as a result of the different treatment of food bought for home consumption in carryout places which were also eating places. No precise estimates can be made of the net effect of these differences, but it is likely that the survey and inde- pendent levels would vary by no more than 10 percent or so on a strictly comparable basis. The reasonably close correspondence for food purchases for home use is consistent with the experience in other survey undertakings using the diary method. 2 The fact that the housewife, who would normally keep the diary for the family, buys most of the food would undoubtedly contribute to this end. Also, the greater amount of space and attention devoted to food purchases on the diary record would tend to improve the level of reporting. 1 Under the classification system used in PCE, vending machines are considered to be eating places, so that such purchases would not be off-premises consumption. 2 Such as the expenditure surveys in the United Kingdom, Canada, and elsewhere. 10 • A less expected finding was the close agreement between the diary and the independent estimates for purchased meals and snacks. Previous studies have indicated the likelihood of an underage in this sector, where individual family members other than the homemaker may account for a large proportion of the outlays. One factor contributing to the present outcome may have been the prominent positioning of the section for recording such expenditures on the diary record. Some special inquiries were also specified in the diary checking procedure for these expenditures. The conceptual incomparabilities cited above, which widened the disparities for food expenditures for home use, would, in contrast, have narrowed the differences for purchased meals and snacks. Also, the survey figure for meals could have been exaggerated somewhat by inclusion of some expenditures for alcoholic drinks which were not separately identified. It is possible, therefore, that the survey estimate could actually be as much as several percentage points under the revised PCE level, which would, however, still be a good showing in relation to previous experience. Not too much can be added about the drastic and charac- teristic underreporting of alcoholic beverages in the survey, which is usually attributed principally to sensitivity in reporting these expenditures. The figure for purchased alcoholic drinks could be understated somewhat, because of the difference cited above, but would still be very much below the independent estimate whatever allowance is made for that factor. Some special checks made at the time of diary pickup in an effort to improve the reporting would have added only slightly (about 15 percent) to the survey levels. 3 Detailed Comparisons Table 2-2 provides comparisons between the survey and independent levels for detailed food expenditures for home use. Although some of the specific margins were altered as a result of the PCE revision, the reporting differentials observed in the preliminary comparisons remained essentially unchanged on the revised basis. As before, agreement with the independent data was close for costly products such as meat and poultry which constitute the main element in a meal. There was also considerable correspondence for other perishable products such as eggs, milk, ice cream, bread, and fruit juice, which are purchased frequently and used more or less on a daily basis. At the same time, some of the widest disparities were found for staple products such as flour, sugar, mayonnaise, and shortening which are bought infrequently, with each purchase used over a considerable period of time. The result was not much better for certain "shelf" items, which people may buy just to have on hand, such as canned milk, canned soup, baked beans, or gelatine desserts. A likely explanation for these differences, which was given in the preliminary report and still seems to hold on the revised basis, is that they may be largely attributable to the manner in which many respondents may keep their diaries. The theory holds, and there is some evidence on this score, 4 that many do not start their diaries promptly and may not keep them up on a daily basis. Instead, they may skip days and later attempt to reconstruct the omitted periods by memory. In addition, the expenditures had to be obtained entirely by retrospective interview for about 15 percent of cases, where respondents did not keep the diaries at all, after having agreed to do so initially. 5 Where detailed information of this kind is obtained or recon- structed by memory, it is likely that respondents would better recall those items which represent the main course in a meal or which are purchased and used on a frequent basis. At the same time, they could easily forget items purchased infrequently which go into the household food inventory. Among other possible reasons for the differences is that staple items may be more heavily consumed by low-income families who might hav£ greater difficulty in maintaining the diaries. Certain other comparisons lend credence to the notion that many of the reporting differences can be attributed to memory biases. For example, the generally closer agreement for fruit products than for vegetables could reflect the fact that the former are more likely to be consumed as snacks or as separate courses and, thus, be more distinguishable than items usually blended into a meal. Relatively high ratios for breakfast cereaU, peanut butter, and jam and jelly could reflect the frequency of consumption of these items and their importance in the diets of children. An interesting finding was the overage in relation to the independent estimate for butter and the corresponding under- reporting for margarine. If these two were combined, the ratio would be closer to unity. Thus, it appears that the tendency still existed a few years ago for respondents to upgrade margarine purchases, consciously or otherwise, to the more expensive spread. The apparent deficiency in the survey estimates for both frozen fruits and vegetables, in relation to canned and dried products, is probably spurious and attributable to inadequate descriptions in the diary. A major problem in the diary, in general, was that a considerable proportion of entries (consti- tuting around 4 percent of all food expenditures) were not sufficiently descriptive for classification purposes and had to be allocated among categories in a more or less uniform basis. The conceptual factor described earlier— that is, for certain purchases to be included in the PCE off-premises estimates but not in the survey at-home purchases— could be seen in a number of the comparisons. In particular, the wide disparities between the survey and independent data for cookies and crackers, potato chips, candy and chewing gum, and soft drinks could be ascribed to this conceptual incomparability. Although the precise magnitudes cannot be determined, there would un- doubtedly be sizeable differences for these categories aside from the conceptual factor. 3 No actual adjustment was made on the basis of the special checks. The relatively unproductive nature of the checking procedure may have been due to certain inadequacies in the method used. See appendix C for a de'^ilpd presentation of these results. 4 See, for example, Quackenbush, G. G. and J. D. Shaffer, "Collecting Food Purchase Data by Consumer Panel," Technical Bulletin 279, Michigan State University, Department of Agricultural Economics (undated). s See appendix B for detailed data on diary completion rates. 11 Table 2-1. Aggregate Expenditures for Food and Beverages: Compared to independent Sources, 1972 Quarterly Panel and Diary Operation (Billions of dollars) Source Food and nonalcoholic beverages for home use Purchased meals and snacks Alcoholic beverages for home use Purchased alcoholic drinks Survey estimates 1 Diary operation Quarterly panel Personal consumption expenditures (PCE)-- (GNP Accounts) 2 Revised Preliminary Economic Research Service (ERS) — Dept . of Agriculture 3 Census of business—merchandise line data 3 ... 91 100, 100. 98.6 86.4 $31.7 23.3 30.0 28.1 (NA) 29.3 $5.8 6.2 14.8 12.4 (NA) 11.3 $1.6 2.9 7.5 7.4 (NA) 7.3 NA Not available. 1 These estimates are based on special reweighted tabulations from the original survey data tapes. They do not reflect editing changes which may have been made by the Bureau of Labor Statistics at later stages of processing and may, therefore, differ somewhat from estimates which have already or will subsequently be published by that Agency. Quarterly panel estimates relate to calendar year 1972, diary estimates to period, July 1972 - June 1973. Diary-based estimates were increased to include food and beverage expenditures while on trips or vacations (diaries were not kept at those times), as reported in a special inquiry made in the quarterly panel. Also included for comparability with the independent estimates are expenditures for pet food. 2 See appendix A for description. 3 See appendix A of Census Technical Paper No. 45 for description. Census of Business estimates exclude sales by enterprises without paid employees, which would be substantial in the case of seasonal produce stands, small bakeries and groceries, etc. Note: Coefficients of variation for the diary estimates are about 1 percent for food expenditures for home use, 2 percent for purchased meals and snacks, and 3 percent for alcoholic beverages for home use. Quarterly panel estimates for food for home use have less than a 1 percent coefficient of variation. 12 Table 2-2. Aggregate Expenditures for Food and Nonalcoholic Beverages for Home Use, by Major Category: Diary-based Estimates Compared to Independent Sources, 1972 Product Survey estimates (diary operation) l Annual amount (millions of dollars) 2 Estimated coefficient of variation (percent) Ratio of survey estimates to: Personal Consumption Expenditures (PCE) 3 Revised Preliminary Economic Research Service (ERS) estimates 4 Meat (all types). Poultry Canned seafood.., Eggs Dairy products Fresh milk Cream Canned or dry milk. Butter Cheese Ice cream Other Fruit Fresh Processed Frozen Canned or dried. Fruit juice Fresh or frozen. . . Canned or bottled. Vegetables Fresh Processed Frozen Canned or dried. Cereal products Breakfast cereals Flour and mixes Macaroni and spaghetti. Bakery products Bread and other fresh baked. Crackers and cookies Fats and oils Margarine Mayonnaise and salad dressing. Shortening and cooking oil.... Peanut butter Jams and jellies. Sugar Candy and chewing gum. Beverages Coffee Tea Soft drinks. Frozen prepared foods Baby foods, except cereals. Canned or dehydrated soups ., Baked beans Catsup and sauces Pickles and relishes Syrup, molasses, and honey. Spices and salt Gelatine desserts Potato chips and snacks.... $23,744.0 3,560.5 760.4 1,889.2 11,120.0 6,310.0 263.5 414.4 522.4 2,187.7 1,210.2 220.6 3,020.4 1,244.8 58.0 1,186.8 1,458.8 849.5 609.3 $3,715.0 2,455.0 684.1 1,770.9 1,030.5 707.3 299.5 5,884.8 1,332.3 628.3 451.7 497.9 263.6 253.5 602.2 1,201.3 1.718.6 357.4 3,458.5 096.4 410.2 654.9 197.5 613.7 250.6 271.9 322.6 183.3 791.6 3.0 2.7 4.4 2.3 2.0 2.3 6.4 4.2 4.1 2.5 2.9 5.7 3.1 2.9 10.0 2.9 2.5 3.0 2.8 2.3 2.6 5.0 3.1 3.1 3.0 5.0 2.6 2.8 3.6 4.0 7.5 5.0 3.2 .974 .998 1.066 .890 .955 .554 .552 1.177 .800 1.041 .693 .918 .330 1.005 .854 .820 .903 .759 .476 .984 .811 .524 .662 .875 .544 .707 .679 .671 .884 .874 .616 .344 .799 .752 .584 .589 .759 .553 .566 .893 .655 .063 .874 .518 .576 .970 1.051 1.043 .892 .979 .634 .639 .932 .809 .856 .763 .888 (NA) (NA) .896 .871 .931 .775 (NA) (NA) .838 .535 .670 .977 .541 1.088 .671 .681 .878 .850 .703 .318 .761 .775 .598 .697 .754 (NA) .516 .809 .630 (NA) (NA) .535 .595 1.037 1.183 .948 .873 .986 .842 .724 *The revised PCE estimate for sugar was considerably higher than the one used in this computation (which would have lowered the ratio a good deal). That estimate, however, contained a somewhat arbitrary assumption concerning allocation of a large undis- tributed bulk sugar residue. The figure used here was based on 1967-72 changes in production of consumer size packages of sugar. (NA) Not available. ^ased on diaries for period June 1972 - July 1973. Some categories are excluded because comparative data were not available. Unlike table 2-1, no adjustment is made for food purchases while on trips or vacations. includes a uniform allocation for reported expenditures where the product category was unknown or where the categories were combined. Those amounted to about 4 percent of all food expenditures. 3 See appendix A for description. *See Appendix A of Census Technical Paper No. 45 for description. Chapter 3. Small Nonfood Expenditures Although some small nonfood expenditures were covered in summary form in the quarterly panel, the diary constituted the main source of information for these items. The categories covered under this heading include laundry and cleaning supplies, toiletries and other personal care products, drugs and medicines, reading materials, tobacco products, small recrea- tional items, and a variety of personal, household, and recrea- tional services. Although expenditures are small in unit terms, these outlays aggregated to some $60 billion a year or more at the time of the survey. Several small sections were provided for recording these items on the diary record. Comparisons with Independent Data Table 3-1 presents the comparisons between the diary estimates and independent sources for small nonfood expenditures. The PCE estimates represent the independent source for product outlays and comparisons are made with both the revised and preliminary PCE figures. For services, data from the 1972 Census of Selected Service Industries as well as PCE are used for comparative purposes. It should be emphasized again that considerable uncertainty exists with regard to the accuracy and precise comparability of the independent data, especially at the level of detail shown. Although the revised PCE estimates are higher than the preliminary ones in some instances and lower in others, the patterns have not been altered appreciably by the changes. The survey estimates, in the majority of cases, still appear to be substantially understated in relation to the independent data. As before, the correspondence was closer for items likely to be purchased by the homemaker, who would usually be the recordkeeper for the family. Examples are laundry and cleaning supplies and paper products, where the ratios of the survey estimates to the PCE levels were around 80 percent or more. An even closer agreement (close to unity) was found for pet food, where regularity of purchase could be an additional factor in achieving more complete reporting. In sharp contrast is the exceedingly low ratio for toiletries, which might be purchased by almost any individual family member. Even within a category such as toiletries, certain clear-cut distinctions were apparent. About the lowest ratio was found for shaving preparations, presumably purchased mainly by the male members. (A large deficiency was also found for shaving articles, such as razors and blades.) On the other hand, higher ratios were observed for toilet soap, dental preparations, and other such products more likely the prerogative of the home- maker. The large disparities for categories such as "other hair preparations" (which include hair coloring) and "other prepa- rations" (which include deodorants and depilatories) could reflect some sensitivity in reporting products of this kind. The pattern continued for other categories of small products. The ratios were comparatively low for reading materials, stationery and greeting cards, toys and games, and particularly for film, where purchases are not only individual matters but likely to be intermittent. In the case of film, the fact that diaries were not maintained when families were away on vacation- when this item might be in especially great demand- undoubtedly contributed to the deficiency. The characteristic disparity for tobacco products can be attributed, in addition, to possible reluctance in reporting such expenditures, especially these days. Drugs and medicines represented one case where the revised PCE estimates were higher than the preliminary ones. The correspondence with the survey estimates was close on the basis of the preliminary figures and remained so in spite of the revision. Possibly, regularity of purchase and the prominent role of the homemaker in family health matters served to overcome the difficulties usually associated with determining such expenditures. This is one sector, however, where the PCE figures, revised or otherwise, are subject to considerable uncertainty. The comparisons for services generally substantiated the findings for product expenditures. Correspondence with both sets of independent estimates remained close for laundry and dry cleaning services and household help, normally the respon- sibility of the homemaker. It is interesting that the agreement was better for laundry services than for dry cleaning, where some members might take care of their own requirements. Differences were much greater for individual-type services, such as hair care and small repairs. In the entertainment sector, ratios were higher for activities, such as movies and bowling, which may be attended on a fairly regular basis, than for more intermittent and individually oriented ones such as sporting events. _The_cje_neraL deficiencies observed for small nonfood items can probably also be attributed, in part, to lack of space and less 13 14 prominence on the diary record, at least in comparison to that accorded food products. Although the picture was mixed, it appears that those specific nonfood items given as illustrations on the diary form tended to be more completely reported. This type of prompting might have occurred especially in the 15 percent of cases where the family failed to keep the record, after initially agreeing to do so, and the interviewer had to reconstruct the expenditures through retrospective questioning, using the diary form as a guide. 1 1 See appendix B. Table 3-1. Aggregate Expenditures for Small Nonfood Products and Services: Diary-based Estimates Compared to Independent Sources, 1972 Category Survey estimates (diary operation) Annual amount (millions of dollars) Estimated coefficient of variation (percent ) Ratio of survey estimates to: Personal consumption expenditures (PCE) 2 Revised Prel iminary Census of selected service industries estimates 3 PRODUCTS Laundry and cleaning supplies. Soaps and detergents........ Laundry supplies Cleaning supplies.. Paper products Light bulbs Personal care preparations.... Toilet soap Cosmetics Dental preparations ......... Toothpaste ................ Other.. ................... Hair preparations ........... Shampoos .................. Sprays Other Shaving preparations Other preparations.......... Shaving articles. ............. Drugs and medicines........... Prescription Nonprescription ............. Pet food Tobacco products. ............. Cigarettes.................. Other products .............. Reading materials ............. Stationery and greeting cards. Records and tapes . ............ Film (excluding development).. Toys and games SERVICES Laundry and dry cleaning Laundry and diaper service... Dry cleaning Rugs and carpet cleaning Paid household help....... Barber and beauty shop care.... Shoe repair Watch and jewelry repair Film development Admission charges Motion picture theatres Legitimate theatres-concerts, Sporting events Bowling alleys $3, 1, 1, 1, 449.8 832.4 563.6 053.8 358.9 326.6 522.9 315.5 621.2 537.4 301.6 235.8 909.1 353.6 223.2 332.3 134.1 005.6 166.0 238.0 696.2 541.8 810.4 651.3 759.2 892.1 161.7 766.8 774.7 348.9 153.2 ,328.2 ,539.8 ,683.2 105.1 ,018.4 ,633.5 160.7 125.2 356.0 ,463.4 407.3 592.3 870.6 2.5 3.0 5.0 4.5 4.5 6.0 5.0 9.0 3.0 (*) 5.0 5.5 5.0 3.0 2.5 2.5 4.5 2.5 6.0 6.0 (*) 5.0 .811 .760 .840 .899 .836 .840 .570 .720 .602 .806 .740 .910 .620 .786 .700 .477 .389 .449 .550 .955 .953 .958 1.001 .628 .616 .732 .555 .672 .735 .375 .672 .780 .775 .756 .811 .719 .959 .533 .764 .594 .760 .701 .851 .587 .748 .672 .447 .366 .438 NA 1.020 1.032 .998 .967 .627 .628 .618 .577 NA .963 .333 .583 3.0 .916 .935 4.5 1.074 NA 4.5 .787 NA (*) .879 NA 3.0 1.125 1.125 3.5 .611 .679 (*) .418 .670 (*) .243 .222 (*) .354 NA 4.0 .883 .890 (*) .770 .644 (*) .496 .489 5.0 1.145 1.070 .917 1.158 .790 .613 .674 .798 .728 (*■) Not available but probably large - 10 to 15 percent or more. 'These estimates are based on special reweighted tabulations from the original survey data tapes. They do not reflect editing changing which may have been made by the Bureau of Labor Statistics at later stages of processing and may, therefore, differ somewhat from estimates which have already or will subsequently be published by that agency. Survey estimates are for the period, July 1972-June 1973. See Appendix A for description. 3 See appendix A of Census Technical Paper No. 45 for description. Chapter 4. Clothing Expenditures There was a good deal of uncertainty, at the outset, concerning the optimal approach for measuring clothing expenditures. The decision for the 1972-73 survey was to cover this subject in considerable detail in the interview panel and, moreover, to inquire about such outlays on a quarterly basis. A small section was also provided for recording clothing expenditures in the diary. However, this section was located in an especially inconspicuous place on the form. In spite of the difference in treatment, the overall estimates of clothing expenditures from the two sources were not significantly different. A major problem in the diary procedure was the existence of a substantial proportion of entries with incomplete or inadequate descriptions, which could not be assigned to specific categories (about 15 percent of the total). Lack of sufficient recording space contributed greatly to this deficiency. As a result, it is difficult to make comparisons between the two sources in any detail. 1 In spite of this problem, it appears that the diary provided the higher, and presumably more complete, estimates for the diversified category of clothing accessories (hats, ties, belts, handbags, etc.). It was apparently difficult to communicate the full range of such items in an interview situation. More surprisingly, the diary estimates were appreciably higher for a more clear-cut category such as footwear, despite the considerable probing for these expendi- tures in the quarterly panel. Comparisons with Independent Sources Table 4-1 presents a comparison between the two survey sources and PCE and Census of Business data for overall clothing and footwear expenditures. As is apparent, the survey estimates are markedly lower in both cases. Although a reduction had been anticipated, in line with previous benchmark adjustments in this sector, the summary PCE estimates were only slightly changed as a result of the revision. One possible reason the Census of Business estimates are below those for PCE is that sales to consumers by wholesale establishments or transitory venders may be omitted. Some limited information on sex differences provide rather unexpected results, although uncertainties exist in matching some of the categories. The correspondence with the PCE estimates appeared to be closer for clothing for men and boys combined than for clothing for women, girls, and infants. On the one hand, since women are the usual survey respondents— and would normally do the buying not only for themselves but for the children as well— it might be expected that their reporting would be more complete for these cate- gories of clothing. On the other hand, the range of items and frequency of purchase are probably much greater for women's and children's clothing, so that recollection could be more difficult in these instances. Also, teenage girls would probably be more likely than boys of comparable age to purchase their own clothing. The deficiencies in the survey data would be less troublesome if they were distributed uniformly among the various clothing categories. Unfortunately, the more detailed comparisons between the quarterly panel and PCE estimates in table 4-2 indicate that this is probably not the case. It may be noted that the ratios between the survey and the independent estimates are generally lower with the revised PCE figures than with the preliminary ones used in the earlier evaluation. 2 The main reason is that more complete and reliable estimates were possible with the revised PCE data. The method used to develop the preliminary PCE estimates for clothing apparently tended to understate the levels in most cases. 3 The revised data continue to exhibit a close correspondence for the large clothing items, such as coats and suits, which is probably a reflection of the usually more complete reporting for more important expenditures. At the other end of the scale is the exceedingly large disparity for the diverse clothing acces- sories category. Although a check list of several such items was used for these expenditures, it was apparently insufficient to remind respondents of the numerous different products involved. 4 The differences among the ratios for the various other categories are difficult to assess. Some disparities could have resulted from difficulties in achieving a precise match between the sources in this detail. Although smaller items are usually less 'A comparison, assuming a uniform distribution of the undifferen- tiated diary entries, is included in chapter 5 (table 5-1) of Technical Paper No. 45. 2 In the preliminary report, comparisons were also made with certain data from the National Consumer Panel operated by the Market Research Corporation of America (MRCA) because of uncertainty about the validity of the preliminary PCE estimates for specific categories. The MRCA estimates corresponded somewhat more closely with the survey data but are likely to be understated. 3 A summary of the detailed preliminary estimates fell about 10 percent short of the overall PCE clothing figure prepared by the Commerce Department at that time. The detailed figures were not adjusted upward, however, because of the belief that the overall PCE estimate would be lowered at the time of revision. This did not turn out to be the case. 4 The correspondence with PCE appeared to be much closer for accessory items specifically included on the checklist. In fact, the addition of one important item (neckties) to the checklist in the 1973 survey appreciably boosted the overall accessory total. 15 16 well reported, the hosiery category was apparently an excep- tion. The greater regularity of purchase of an item of that kind could have helped to stimulate the memory of respondents. The relatively close correspondence for infant clothi.ng could be partly spurious, since the PCE estimate may be understated because of the problem of identifying all of the components which belong in this category. The rather disappointing results for clothing expenditures can probably be traced to the scope and complexity of this group. In the course of the interviews, some 13 broad clothing categories were probed in detail. A considerable element of fatigue probably resulted, especially since questions about clothing represented only part of a lengthy interview. In addition to their greater item cost, the fact that coats and suits appeared at or near the beginning of the clothing list could have contributed to the more complete reporting. Footwear expendi- tures, on the other hand, were located near the end of the list. Timing Differences Because month of purchase was recorded in the interview survey, it is possible to prepare estimates for shorter periods than a quarter. One alternate set relates only to purchases in the month prior to the interview, which would essentially represent a 1 -month recall period. As a result of the manner in which the survey sample was operated, all months of the year would be equally represented in these alternate estimates, so that there would be no seasonal biases. 5 One problem with such alternate estimates is that they are likely to be affected by "telescoping," since (as indicated in chapter 1) the "bounding" procedure used to control for this tendency applied only to the standard (in this case, quarterly) 5 A systematic one-third of the sample was enumerated in January, April, July, and October; the second in February, May, August, and November; and the third in March, June, September, and December. The questioning (for quarterly items such as clothing) covered the three calendar months preceding the interview. survey periods. Thus, it is likely that the estimates for the final month of each quarter included some expenditures which had actually been made earlier. At the same time, however, there was undoubtedly some omission of purchases even for the most recent month because of memory lapses or lack of knowledge on the part of the survey respondents. It is possible, therefore, that the estimates for the most recent month may be the more accurate ones, if only because of offsetting errors. In ajiy event, alternate estimates for the most recent month and their relationship to the PCE levels are presented in the final columns of table 3-2. It is apparent that this approach would result in some exaggeration for a relatively well reported category such as coats and suits, unless a bounding procedure could be applied on a monthly basis (which would be very costly). For other categories, some appreciable, even dramatic, improvements are apparent. In general, however, the changes are not sufficient to overcome the previous deficiencies. Also, the procedure does not appear to achieve greater uniformity of reporting among categories. Some alternatives are also possible for the diary procedure as a result of the first- vs. second-week differences discussed in chapter 1. If only the first week's diaries were used, the overall clothing estimates would be increased about 4 percent. The increase would be especially large for accessories, whereby the survey figure would be boosted to about 80 percent of the revised PCE level. Although more far-reaching improvements are needed, some advantage could be gained by a composite use of the available data from the 1972-73 survey, rather than from any single estimation procedure. The best composite, in terms of corre- spondence with PCE, would appear to be to use the standard quarterly panel estimates for coats and suits and perhaps a few other larger items, the estimates for the "most recent" month from the quarterly panel for most other categories, and the "first-week" diary estimates for clothing accessories. 17 Table 4-1. Aggregate Expenditures for Clothing and Footwear: Quarterly Panel and Diary Operation compared to Independent Data Sources, 1972 (Billions of dollars) Clothing Source Total Mens and Boys Women, girls, and infants Footwear Survey Estimates 1 $30.9 31,6 44.8 45,2 40.9 $12.5 (NA) 16.4 15.9 15,0 $18,3 (NA) 28.4 29.3 25.9 $5,7 6.7 8,7 Personal consumption expenditures (PCE) 2 8.9 Census of Business — merchandise 7,7 NA Not available. 1 These data are based on special reweighted tabulations from the original survey data tapes. They do not reflect editing changes which may have been made by the Bureau of Labor Statistics at later stages of processing and may, therefore, differ somewhat from estimates which have already or will subsequently be published by that agency. Quarterly panel data are for calendar year 1972; diary estimates for fiscal year, July 1972-June 1973. 2 See Appendix A for description 3 See appendix A of Census Technical Paper No, 45 for description. Table 4-2. Aggregate Expenditures for Clothing and Footwear, by Major Categories: Quarterly Panel Estimates Based on Data for All Months and on Reported Purchases in Final Month of Each 3-Month Period, Compared to Personal Consumption Expenditure (PCE) Estimates, 1972 Quarterly panel estimates based on data for all months Quarterly panel estimates based on purchases in final month of each 3-month period Clothing category Annual amount (millions of dollars) Estimated coefficient of variation (percent ) Ratio of survey estimate to: Annual amount (millions of dollars) Estimated coefficient of variation (percent ) Ratio of survey Revised PCE estimate Prel im- inary PCE estimate estimate to revised PCE estimate Total clothing 1 Untailored jackets and sweaters Dresses and skirts ..................... Trousers and slacks.................... Shirts and blouses ..................... Nightwear, Infant clothing Hats, gloves, and accessories..... Total footwear. .................. $30,850.0 7,427.1 1,824.4 4,203.5 5,319.8 3,787.1 1,937.4 1,151.2 2,020.0 1,124.1 304.0 1,276.4 5,735.1 1.0 2.5 2.8 1.8 1.1 1.3 1.1 1.7 1.3 5.9 3.4 1.7 1.0 .69 1.01 .61 .61 .78 .56 .59 .69 .81 .85 .66 .44 .66 1 68 11 63 71 84 68 59 71 88 NA NA 52 64 *36,864.8 8,499.1 2,063.3 5,091.9 6,5 94.5 4,779.1 2,327.2 1,399.1 2,408.2 1,553.8 370.4 1,215.9 7,084.4 1.7 4.3 4.8 3.1 1.9 2.3 1.9 2.9 2.3 10.2 5.9 2.9 1.7 .82 1.15 .69 .74 .97 .70 .71 .84 .97 1.17 .81 .42 .82 NA=Not available. Detail will not add to total because certain categories for which comparisons could not be made are omitted, Chapter 5. Household Appliances Although expenditues for household appliances represent only a small percentage of consumer outlays (some $15 to 20 billion at the time of the survey), there is considerable interest in this category because it seems to be particularly sensitive to cyclical fluctuations in economic activity. Because of the infrequency of purchase of individual appliances, sampling variances are especially large. The quarterly panel, where a full year's experience was obtained for each family, was regarded as a more realistic source than the diary operation, where only 2 weeks' coverage was provided for each sample case. Moreover, previous experience had indicated that underreporting was likely in a short-term operation such as the diary procedure, because of uncertainty as to when to record an acquisition— when a deposit was made, a contract signed, delivery received, etc. 1 A special procedure known as the "inventory" approach was used for this category in the quarterly panel. A detailed discussion of this procedure was included in the preliminary evaluation report, but the essential elements will be repeated here for convenience. Respondents were asked at their first interview about possession of a lengthy list of appliances and for the date of acquisition and other details (including cost) for those which were present. Items in the household inventory were differentiated into two classes— those the family had purchased for its own use (or which came with the residence), and those it had received as gifts from persons outside the household. The inventory was brought up to date at various intervals, with any new acquisitions divided into the same two classes. The updating was done twice, at semiannual intervals, for small or minor appliances but only once, at the end of the year, for major appliances. Questions were also asked at the updatings about appliances which the family had purchased since the previous inquiry for gifts to be given to persons outside the household. This was done because it was believed the more reliable expenditure estimates could be derived for items the family had actually purchased, whether for itself or as gifts for others, since they would know the actual prices. In any event, the procedure provided two estimates for gifts— the value of those received and the cost of those given— either of which could be included as part of the overall expenditure total for appliances. Comparison with PCE Estimate Preparation of the PCE estimates for appliances used in the preliminary report was a particularly difficult and hazardous undertaking because of the large proportion of imports involved, the proliferation of small products, the adjustments needed for those installed in rental units, etc. It is not surprising, therefore, that the revised estimates are appreciably different in a number of instances. The effect has been greatest for small appliances whereby substantially increased PCE levels have resulted in much greater disparities in relation to the survey data. Table 5-1 compares the survey results from the quarterly panel with the revised and preliminary PCE estimates. The two survey estimates reflect the different treatment of gifts described above. Method 1 is the more conventional approach, combining appliances the family had purchased for its own use (including those installed in purchased homes) and those it had bought as gifts to be given to persons outside the household. Method 2 substitutes the value of gifts received from outside the household for those given to others. Although the two estimates for gifts are theoretically measures of the same thing, it may be seen that the second estimate almost invariably provides the higher figures. Some of the difference could possibly derive from problems of estimating the value of gifts received. In a substantial proportion of cases (about 15 percent), respondents could not supply an estimate, and a value was imputed from other information. The more important cause, however, was the fact that many more appliances were reported as gifts received than as gifts given. Respondents were apparently better able to report appliances in their possession, including those received as gifts, than to recall items they had purchased and given to others. Except for television sets, the estimates from the two methods did not differ very much for major appliances, because only a small percentage of these represent gifts. Also, the correspondence with the PCE estimates remains close for most groups. 2 A sharp increase in the revised PCE estimate for clothes washers and dryers created a considerable disparity between the two sources. Since there is no logical reason this 1 The diary estimates actually fell about 20 to 30 percent below those obtained from the quarterly panel (see the preliminary evaluation report for details). This was a surprisingly good showing since no special mention of these items appeared on the diary form and the entries were, thereby, relegated to the catchall section. 2 Comparisons for two major appliance groups are omitted for various reasons. For sewing machines, where imports dominate the field, the PCE estimate was exceptionally low (only about half of the survey estimate) and believed to be understated. The reverse was true for typewriters, but the PCE estimate includes purchases by nonprofit institutions (schools, hospitals, etc.), which probably buy a large proportion of these (no adjustment possible). 18 19 category should be less completely reported in the survey than any other, the difference can probably be ascribed to some kind of statistical aberration. Somewhat wider differences for tele- vision sets than for some of the others can probably be attributed to the existence of multiple sets in many households, some of which might be overlooked by respondents. More attention to this possibility in compiling the inventory is indicated. As for most other product categories, the picture is less favorable for small than for large appliances. Since gifts are a more important element here, especially for small kitchen appliances, the second survey method provides substantially higher estimates. Even these higher levels, however, fall sub- stantially below the revised PCE totals in most instances. The inventory method is apparently less successful where a multi- plicity of small products is involved (as in the case of kitchen appliances) or where the family may possess several items of the same kind (radios, cameras, etc.). The fact that expenditures for small appliances were collected at 6-month rather than 3-month intervals could also have contributed to the shortfall. The timing of interviews for items covered semiannually in the survey does not provide a basis for obtaining seasonally unbiased estimates for the first compared to the second 3-month period. There would also be a tendency for "telescoping" within the 6-month semiannual period, whereby the estimates for the final 3 months would tend to be exaggerated. Some rough estimates for the first half of the year (thus excluding the extreme Christmas buying period) indicate that reported expenditures for the final 3 months were perhaps twice as large as those for the first 3 months for virtually every small appliance group, a difference which probably exceeds anything that could be attributed to seasonal or telescoping influences. The inference is that quar- terly updating for small appliances might overcome much of the deficiency apparent in the 1972-73 survey results. A consid- erable element of respondent fatigue could be engendered by such a step, however, unless commensurate reductions were made in other parts of the interviews. Use of Prior Year's Data In addition to the different ways of estimating gifts, another feature of the inventory approach is that it is possible to derive two separate estimates for the same year. The first represents the direct approach, whereby the data are obtained by updating the inventory to identify purchases and acquisitions in the course of the year. The second is based on the initial inventory for the following year, where items added in the previous year can be identified from the date of acquisition. Since there is probably very little or no correlation in the extent to which individual families would buy the same appliances in successive years, the two aggregates would essentially constitute inde- pendent estimates, even if the same sample were used in both years. 3 Thus, it would be possible to combine the two estimates and achieve a reduction in sampling variance about equivalent to doubling the sample size. Some question might be raised, however, as to the adequacy of the previous year's estimates obtained in this manner. Those data do not contain the "bounding" safeguards which apply to the current year's inquiry, and other slippages are possible, such as loss, theft, or destruction of items acquired during the previous year but no longer present when the inventory is taken. It should be noted that since the prior year's estimates are based entirely on the current inventory of items on hand, they represent the second kind of combination discussed earlier, that is, purchases for the family's own use plus the value of gifts received from others. The two sets of estimates for 1972 are presented in table 5-2, in relation to the revised PCE figures. 4 The two survey estimates are statistically different for only about half of the categories. In all these cases, however, the prior-year estimates are higher. Also, for both major and minor appliances as a group, the prior-year estimates are significantly greater. Evidently, there is some element of telescoping in the prior-year estimates which partially offsets the general tendency for understatement in the current-year figures. For major appliances, there would appear to be some advantage to combining the two sets of estimates, from the standpoint of reducing both sampling variability and reporting bias. Such a process would not be sufficient, however, to resolve the disparities for small appliances. 3 The two samples were actually independent in the 1972-73 survey. In a continuing survey program, however, at least part of the sample would probably extend from one year into another. 4 For greater comparability, the estimates for the current year are also based on the second kind of combination— purchases for own use plus value of gifts received. 20 Table 5-1. Aggregate Expenditures for New Appliances: Quarterly Panel Estimates (Two Methods) Compared to Personal Consumption Expenditures, 1972 Category Quarterly panel estimates 1 Annual aggregates (millions of dollars) Method 1 Purchases for own use and for gifts given to persons outside household Method 2 Purchases for own use plus value of gifts received from persons outside household Estimated coefficient of variations (either method) (percent) »tio of survey estimates to Personal Consumption Expenditures (PCE) 2 Preliminary Method 1 Method 2 Major appliances Cooking stoves and ranges Refrigerators and freezers Clothes washers and dryers Dishwashers and disposalls Vacuum cleaners and other floor cleaning equipment Room airconditioners Television sets Black and white Color Yard machinery Minor appliances 3 Small kitchen appliances Portable heating and cooling equipment. Electrical personal care equipment Sound equipment Photographic equipment Power tools $9,834.6 978.0 1,636.1 1,453.4 546.9 587.4 587.4 535.6 2,866.7 643.1 4,305.1 587.9 237.9 322.5 2,260.9 541.5 354.4 $10,254.4 1,005.8 1,707.8 1,513.1 563.0 616.0 605.1 588.5 3,012.4 642.7 4,767.4 779.9 243.9 368.3 2,417.3 584.4 373.6 4.8 3.6 7.3 6.1 .868 .925 .981 .769 1.048 .878 .891 .737 .834 .916 .648 .476 .842 .619 .678 .589 .994 .905 .951 1.024 .801 1.079 .921 .918 .810 .876 .916 .717 .631 .863 .707 .725 .636 1.048 .884 .874 .879 .944 1.098 1.013 1.105 .722 .810 .865 .868 .661 1.135 .955 .974 .636 1.008 .922 .898 .917 .983 1.130 1.063 1.138 .793 .858 .865 .961 .877. 1.164' 1 1.091 1.042 .687 1.06?,, J These estimates are based on special re-weighted tabulations from the original survey tapes. They do not reflect editing changes which may have been made by the Bureau of Labor Statistics at later stages of processing and may, therefore, differ somewhat from estimates which have already or will subsequently be published by that agency. In addition to direct purchases and gifts, the estimates under both methods include the value of new appliances installed in and included in the price of a home. 2 See appendix A for description. 3 Minor appliance categories include the following: Small kitchen appliances — toasters, mixers and blenders, electric can openers, cof feemakers , broilers, electric pans, electric knives, electric food warmers, electric irons, etc. Portable cooling and heating equipment — portable heaters or stoves, dehumidif iers, humidifiers or vaporizers, electric window or portable fans, etc. Electric personal care equipment — hair dryers, electric shavers, electric hair setters, electric toothbrushes, etc. Sound equipment— radios , phonographs, tape recorders, record players, stereo components, etc. Photographic equipment — still cameras , slide projectors , movie cameras , etc. Power tools — electric drills, electric saws, electric sanders, etc. Table 5-2. Aggregate Expenditures for New Appliances from the Quarterly Panel: Current-Year vs. Prior- Year Based Estimates, Compared to Personal Consumption Expenditures, 1972 Category Current-year based estimates (1972 estimates based on 1972 survey) 1 Annual amount (millions of dollars) Ratio to revised PCE estimates Prior-year based estimates (1972 estimates derived from 1973 survey) 1 Annual amount (millions of dollars) Ratio to revised PCE estimates Estimated standard error of the differences between the two estimates (millions of dol lars) Major appliances Cooking stoves and ranges Refrigerators and freezers Clothes washers and dryers Dishwashers and disposalls Vaccuum cleaners and other floor cleaning equipment Room air conditioners Television sets Black and white Color Yard machinery Minor appliances 2 Small kitchen appliances Portable heating and cooling equipment Electrical personal care equipment.... Sound equipment Photographic equipment Power tools $10,254.4 1,005.8 1,707.8 1,513.1 563.0 616.0 605.1 588.5 3,012.4 642.7 4,767.4 779.9 243.9 368.3 2,417.3 584.4 373.6 .905 .951 1.024 .801 1.079 .921 .918 .810 .876 .916 .717 .631 .863 .707 .725 .636 1.048 $11 ,480.5 1 ,067.6 2,057. 1 1,777.7 652.1 771.8 614.5 656.4 2,888.5 994.8 5,228.2 882.1 271.9 423.7 2,669.5 638.9 342.1 1.013 1.009 1.234 .941 1.250 1.154 .932 .903 .840 1.418 .786 .714 .962 .813 .801 .695 .959 323 85 131 107 60 i 50 44 42 150 85 170 31 18 16 119 55 31 'Method 2 estimates, purchases for own use plus value of gifts received from others. 2 For inclusions, see footnote 3 of table 5-1. Chapter 6. Household Furnishings This chapter covers a wide range of household furnishings and related items: furniture, floor coverings, household linens, dinnerware and cookware, decorative items, and various others. Expenditures for these items totaled some $25 to $30 billion annually at the time of the survey. Because of the diversity of products, the survey procedures differed somewhat for various categories. Most were covered in the quarterly panel, with the information collected earh quarter for certain ones (e.g., household linens), and semiannually for the remainder. Unlike the procedure for household appliances, direct questions were asked about expenditures in the previous 3-month or 6-month periods, rather than starting with an inventory of items on hand. The large variety of items possessed by most households made it impracticable to use an "inven- tory" approach throughout. All of the categories were implicitly covered in the diary procedure, where there was a small section set aside for housewares, furnishings, hardware, and garden supplies. Comparisons with Independent Estimates The PCE estimates were not the only independent sources available for comparative purposes for this category of expendi- tures. Also used to some extent were estimates from the 1972 Census of Business and from the National Consumer Panel of the Market Research Corporation of America (MRCA), adjusted to client production and sales data. The revision generally lowered the PCE levels and brought them more closely into agreement with the other independent sources in most instances. However, there are still some disparities among the sources, and the general complexity of this sector warrants special caution in interpreting the findings. Table 6-1 presents the comparisons between the survey data from the quarterly panel and diary operation and the various independent sources. The most surprising finding, initially noted in the preliminary report, is the great similarity between the quarterly panel and diary results for most categories. Although sampling variances are much lower for the quarterly panel data— and they should probably be used for this reason for the larger items— the biases are sometimes smaller for the diary results. This was especially the case for the highly diversified category of dinnerware, glassware, and related items, where the quarterly panel data were notably deficient. As would be expected, the correspondence with the inde- pendent sources was closest for the larger and more costly products. The survey estimates of furniture expenditures, in particular, closely matched the independent data, whether the revised or p reli mi n ary^C E estim ate^wefe^ised for comparison. With the revised PCE estimates there was considerably better agreement than before for floor coverings and for window and furniture coverings. For furniture and window coverings, how- ever, there is a considerable gap between the survey estimates and a second independent source, the 1972 Census of Business data. It is possible that the PCE estimates, which largely derive from production sources, would be understated in a sector where much of the output is custom made. Whichever sources are used, the ratios of the survey to the independent data recede rapidly as the comparisons proceed to the smaller and more diversified categories. The relationship still holds up reasonably well for household linens but the disparities become increasingly greater for the remainder. The differences, large as they are, could even be understated for a category such as decorative items, where it is difficult to fully reflect such items as original art works in the PCE estimates. Timing Differences In considering ways of improving the survey reporting for some of the more deficient categories, one possibility would be a reduction in the time reference. As noted earlier, the inquiry was made quarterly for household linens but only semiannually for the other categories. In the preliminary report, some alternate estimates were presented for various categories, using shorter time references. For household linens, the alternate estimates related to expendi- tures in the most recent month of each quarter. As indicated in the chapter on clothing expenditures, such estimates would be unbiased seasonally, since all months of the year would be equally represented. However, there would likely be some telescoping in the estimates for the most recent month which could exaggerate the levels. In any case, the alternate estimates increase to about 87 percent of the independent figures for household linens, which suggests that some improvement might be achieved by shortening the recall period. A similar exercise was carried out for other categories of household furnishings, in this case developing alternate esti- mates based on expenditures in the most recent 3 months of each semiannual recall period. For semiannual items, it was not possible to develop seasonally unbiased estimates. In addition, telescoping would also be involved, leading to some exaggera- tion in the alternate estimates. Although the alternate estimates were some 30 percent or more higher than the original ones, it did not appear— after allowing for the various elements of exaggeration— that very much of the deficiencies for the smaller products could be overcome solely by reducing the recall period. 21 22 Table 6-1. Aggregate Expenditures for Household Furnishings and Equipment, Quarterly Panel and Diary Operation Compared to Independent Sources, 1972-73 Survey estimates 1 Ratio of survey estimates to: Quarterly panel Diary operation Personal Consumption Expenditures (PCE) 3 Census of Bus iness merchandise line data 4 Adjusted National Consumer Panel data (MRCA) 4 Expense category Annual amount (millions of dollars) Estimated coefficient of variation (percent) 2 Annual amount (millions of dollars) Estimated coefficient of variation (percent) Revised Preliminary Quar- terly panel Diary opera- tion Quar- terly panel Diary opera- tion Quar- terly panel Diary opera- tion Quar- terly panel Diary opera- tion $9,494 3,136 2,860 276 1,907 1,878 1,554 879 1,200 263 2.7 3.2 3.2 3.0 3.0 3.5 4.5 4.1 7.2 $9,390 2,350 2,080 270 1,934 2,292 1,444 1,976 1,034 340 339 10 7 7 7 5 10 7 10 14 10 .897 .857 .756 .915 .258 .668 .447 .887 .642 .767 .850 .581 .576 .578 .640 .899 .645 .576 .800 .231 .567 .287 .889 .484 .584 .744 .520 .488 .371 .579 .926 .795 .862 .790 .579 .632 .916 .595 .629 .801 .707 .588 .860 .736 .625 Household linens 5 Sewing materials 6 Window and furniture coverings 7 .. Dinnerware , glassware, and .747 Lamps, mirrors, clocks, and Luggage and closet storage items. Hand tools (except garder tools). - *Survey and PCE data for these categories are 1972-73 averages. All other estimates in table are for 1972 except for National Consumer Panel (MRCA) data which are for 1973. 1 These estimates are derived from special tabulations from the original survey data tapes. They do not incorporate editing changes which may have been made by the Bureau of Labor Statistics at later stages of processing and may, therefore, differ somewhat from estimates which have already or may be subsequently published by that Agency. 2 Sample sizes used in these computations are roughly twice as large for these categories marked with an asterisk (*) in the first column as for ^ other categories. 3 See appendix A for description. ''See appendix A of Technical Paper No. 45 for a description of these sources. The National Consumer Panel is operated by the Market Research Corporation of America (MRCA) . 5 Includes bedsheets, pillowcases, towels, washcloths, tablecloths, etc. 6 Includes needles, pins, thread, and other notions as well as broadcloth. Includes window shades, Venetian blinds, drapes, curtains, slipcovers, decorative pillows, etc. 8 PCE estimates used to compute survey-to-PCE ratios in table were adjusted downward by about $700 million to exclude plastic products and a few other items not included in this category in the classification system used in the expenditure survey. 9 Decorative items include ceramics, original paintings, art reproductions, decorative glassware, etc. Chapter 7. Automobile and Vehicle Expenses Expenditures for the purchase, maintenance, and operation of automobiles and other vehicles represent one of the largest items in the family budget, aggregating to over $100 billion annually at the time of the survey. Because of the diversity of products and services, dependence was placed on both the quarterly panel and the diary operation for purposes of measuring these expenditures. The interview panel was almost automatically assumed to be the appropriate source for vehicle purchases, in view of the large sampling variances associated with this item. There was less certainty, however, with respect to the optimal source for the remaining categories. An inquiry was made each quarter in the interview panel regarding expenditures for the maintenance and operation of vehicles during the preceding 3 months. At the same time, the diary record contained a small section for gas, oil, tolls, parking fees, and other vehicle expenses. Vehicle Purchases The inventory approach, described earlier, was used to obtain the information on purchases of automobiles and other vehicles. Respondents were asked at the first interview about possession of various kinds of vehicles and for details about those on hand. The inventory was updated at the final interview, a year later, to identify any new acquisitions during the year. Table 7-1 compares the estimates from the quarterly panel on expenditures for the purchase of vehicles with similar data from independent sources. In addition to the PCE estimates, data from the Census Bureau's Annual Housing Survey for 1973 are used for comparative purposes. Since the PCE figures were revised only slightly, comparisons are shown only with the new estimates. Three different estimates are presented for automobile purchases. The gross value (before allowance for trade-in) and the net value (after trade-in) estimates are reasonably com- parable with the Annual Housing Survey data. Both surveys exclude vehicles used entirely for business purposes but include those used partly for business. The third estimate— gross value after adjustment for business use— is the one most comparable with the concept used in PCE. The adjustment was based on information provided by respondents on the percentage of total mileage devoted to business purposes. PCE data on used cars relate only to the profit margin on such sales and are not, therefore, comparable with the survey data. As is evident, the survey estimates both for new and used automobiles correspond closely to the independent data. The result is not surprising since respondents are not likely to overlook the purchase of so major an item as an automobile. Comparisons are also reasonably close for other vehicles. Classification difficulties could possibly account for part of the difference between the ratios for trucks as opposed to campers and trailers. The considerably lower ratio for boats could be attributable to the variety of small sailboats and other such vessels possessed by the younger family members, some of which could be overlooked by respondents. In addition, some persons purchase boats for use as their main living quarters, and such acquisitions could be omitted in a household survey because of the difficulty of locating and covering that segment of the population. Other Vehicle Expenses Table 7-2 compares the survey data from the quarterly panel and the diary operation with the revised and preliminary PCE estimates and other sources (in the case of insurance). The revised PCE estimates were higher than the original ones for tires but lower for overall repair and maintenance costs. There was not much difference for gasoline and oil. As the table indicates, the correspondence with the PCE estimate is close except for tires and batteries, and a small item such as toll charges. Whereas reasonably accurate reporting might be expected for a regularly purchased item such as gasoline, the close agreement for repairs, maintenance, and other such expenses is rather surprising. Because of the wide diversity of products and services included in the latter category, however, there could be some uncertainty about the completeness of the PCE figures. The general agreement between the quarterly panel and the diary results for many categories was also unexpected, since the homemaker would be less involved with these expenditures than with most other kinds. The existence of a large unspecified group of expenditures in the diary results, however, adds considerable uncertainty to the comparisons. The quarterly panel was evidently the superior source for vehicle insurance expenditures, a pattern which persisted for other kinds of insurance as well. Although there were no available independent data on vehicle registration and licensing fees, the much higher quarterly panel estimates are likely to be the more reliable figures. 23 24 Table 7-1. Aggregate Expenditures for Purchase of Automobiles and Other Vehicles: Quarterly Panel Estimates Compared to Independent Sources, 1972-73 Vehicle type Quarterly panel estimates Annual amounts (billions of dollars) 1972-73 Average 1972 1973 Estimated coefficient of variation for a single year (Percent) Ratio of survey estimates to:' Personal consumption expenditure (PCE) estimates 1973 Annual Housing Survey (Bureau of census ) New automobiles Gross value (before trade-in).... Net value (after trade-in) Gross value after adjustment for business use Used automobiles Gross value (before trade-in).... Net value (after trade-in) Other vehicles (gross value after adjustment for business use) 3 Trucks Campers, trailers and other recreational vehicles Motorcycles Boats $35.7 29.5 33.3 19.0 17.0 3.4 1.5 1.0 0.9 $34.7 28.8 32.2 19.0 16.8 $36.6 30.2 34.4 18.9 17.2 3.5 3.5 3.5 3.5 3.5 (#) (#) (#) (#) 1.011 .988 .985 1.024 1.042 .865 .975 .934 .749 # Not available but likely to be large, 10 to 15 percent or more. 1 These estimates are based on special reweighted tabulations from the original survey data tapes. They do not reflect edit- ing changes which may have been made by the Bureau of Labor Statistics at later stages of processing and may, therefore, differ somewhat from estimates which have already or may subsequently be published by that Agency. Computations are shown for survey value most comparable with independent source (see text). Comparisons for vehicles other than automobiles are based on 1972-73 averages, both in the case of survey and PCE data, for greater statistical reliability. *See appendix A for description. 5 See appendix A of Census Technical Paper No. 45 for description. 25 Table 7-2. Aggregate Expenditures for Maintenance and Operation of Automobiles and Other Vehicles: Quarterly Panel and Diary Operation Compared to Independent Sources, 1972 Expense category Survey estimates Quarterly panel Annual amount (millions of dollars ) Estimated coefficient of variation (percent) Diary operation Annual amount (mi 1 1 ions of dollars) Est imated coef f ic ient of variation (percent) Ratio of survey estimates to personal consumption expenditures (PCE) 1 Revised Quar- terly pane 1 Diary opera- t ion Pre 1 iminary Quar- terly panel Diary opera- t ion Gasoline and oil Tires and batteries: Tires and tubes Batteries Repair, maintenance, rental and parking , total Repair and maintenance , total Vehicle repair Tune -up, lubrication, etc.. Combined expenses 3 Vehicle rental Parking fees Other expenses: Toll charges Registration, license, and inspection fees Vehicle insurance 4 $25,443 4 ,162 574 1 ,635 666 8 260 2 406 304 666 (NA) 3,036 14,292 1.5 3.0 10.0 2.0 2.0 14.0 9.0 4.5 1 .0 2 $24,215 3 246 424 1 116 186 7 366 794 2 026 367 563 444 1,635 9,343 2.5 10.0 20.0 7.0 7.0 25.0 15.0 15.0 7.0 3.0 1.000 .781 .713 .928 .927 (NA) (NA) . 147 .871 (NA) .927 .952 .609 .527 .887 .885 (NA) (NA) (NA) 1.383 .737 .581 (NA) .606 1 .023 .923 (NA) .836 (NA) (NA) (NA) (NA) .752 (NA) .927 .973 .720 (NA) .798 (NA) (NA) (NA) (NA) (NA) .636 .595 (NA) .606 NA Not available. 'See appendix A for description. 2 Diary estimates were adjusted upward by around $2.5 billion to include gasoline expenditures on trips and vacations. Under the procedure used, no entries were made in diaries during periods in which the family was away overnight or longer. Adjustment was based on a special inquiry in the quarterly panel on itemized expenses incurred on trips or during vacations. 3 Consists primarily of cases where diary entries were not sufficiently precise to allocate the expenditures to any of the categories shown. Also includes a small amount ($150 million) for car washing. May also include some purchases of gasoline made in combination with other expenses. 4 Because PCE data on insurance premiums have a different conceptual basis (see Appendix), the comparison is based on insur- ance industry figures instead. The source is aggregate premium data for private vehicles compiled by A. M. Best and Co. The figures and ratios in the tables are for 1973. Chapter 8. Housing Costs Housing costs now represent the largest single element in the American family budget, having replaced food expenditures in this regard. This category includes rent, mortgage payments, real estate taxes, fuel and utility costs, and property maintenance and improvement. Although the quarterly panel was regarded as the principal source of this information, a parallel set of estimates can be constructed from the diary results. The inventory approach was used in the interview panel to identify real property possessed by the family at the outset and acquired during the course of the survey year. At the final quarterly visit, sufficient information was obtained on indebted- ness, size and frequency of periodic payments, etc., to compute annual mortgage outlays. Information was also obtained at that time on real estate taxes, special assessments, and ground rent. For renters, questions were asked at the first visit on the monthly (or other periodic) rent, including a listing of the kinds of facilities and services covered by the rent. The same kind of information was obtained at the final visit to identify any changes that might have occurred. These data, together with information on prior rental quarters for in-movers during the year, provided the basis for calculating annual expenditures for rent. At each quarterly interview, an inquiry was made on expenditures for repairs, alterations, and maintenance of owned property. The procedure involved identifying any projects of this kind undertaken during the previous 3 months and obtaining a detailing of any contract costs or direct expendi- tures for materials and supplies. A special technique, known as the last-payment principle, was used each quarter to determine utility costs. Instead of asking directly about expenditures in the previous period, information was obtained on the amount of the last bill and for the period to which it related (month, 2 months, etc.), thus providing a basis for converting the information to a consistent time frame. For fuel expenditures, however, direct questions were asked about purchases during the preceding 3 months. Only minimal provision was made in the diary operation for housing expenditures. A small section was set aside on the diary record for rent, utilities, fuel, phone, and insurance. Comparisons with Independent Data Table 8-1 presents a comparison of the survey data from the quarterly panel and the diary operation with various independ- ent sources. Although PCE estimates are used in several instances, data from that source are not comparable for some of the largest categories of housing expenditures— home mortgage payments, real estate taxes, and repair and maintenance expenses. For PCE purposes, an artificial measure theoretically representing the rental value of owner-occupied properties is substituted for these kinds of ownership costs. Comparisons are made with data from two Census Bureau sample surveys— the Annual Housing Survey (AHS) and the Survey of Residential Alteration and Repair (SORAR)— where the PCE estimates are not applicable. The only change of any consequence in the revised PCE estimates was for fuel purchases, where the level was reduced by about 20 percent. Comparisons in the table are, thus, limited to the revised estimates. The close correspondence between the survey and independ- ent estimates exhibited in the table is consistent with previous experience on housing expenditures. The regularity of many of these outlays was undoubtedly a contributing factor. Also, special techniques were used for utilities and home repair expenditures which had proved successful in the past. The substantial deficiency in fuel expenditures exhibited in the preliminary report was considerably reduced as a conse- quence of the PCE revision. Even the gap which remains is probably exaggerated as a result of conceptual factors affecting rental quarters. The PCE estimate includes all fuel used for rental quarters regardless of whether the landlord or the tenant was responsible for providing heating and hot water. The survey estimates, on the other hand, would include fuel purchases for rental quarters only where the tenant paid directly for those services (mainly for rented one-family houses and older tene- ments). No precise allowance can be made for this factor, but it could easily account for half of the difference. A less expected development was the close agreement between the quarterly panel and diary estimates in a number of instances, in spite of the cursory attention accorded these expenditures on the diary record. The much lower diary estimate for mortgage-related payments could be attributed, in part, to the fact that such disbursements are sometimes made directly from bank accounts and could be forgotten by respondents. No direct comparison was possible with the data presently available for home repairs and alterations, since the quarterly panel estimates relate to all real property owned by the family, whereas the diary estimates theoretically exclude property used for business purposes or rented to others. Data from SORAR suggest, however, that the difference between the quarterly panel and diary estimates, as shown, is not incon- sistent with that expected from the difference in coverage. Although the overall match with SORAR was close, disparities were observed in various of the components (e.g., alterations as opposed to repairs) which suggest caution in interpreting the results. 26 27 Table 8-1. Aggregate Estimates of Housing Expenditures: Compared to Independent Sources, 1972-73 Quarterly Panel and Diary Operation Expense category Years covered by compari- sons 1 Survey estimates' Quarterly panel Annual amount (bil lions of dollars ) Estimated coefficient of variation (percent ) 3 Diary operation Annual amount (bil lions of dollars ) Est imated coefficient of variation (percent ) 3 Ratio of survey estimates to independent data Quar- terly panel Diary opera- tion Source of indepen- dent data 4 Contract rent 5 Mortgage payments 7 Property taxes 8 Home repairs and alterations All owned property Own home ' Utilities, total Electricity and gas Water, sewerage, and trash collection Telephone and telegraph Utility type not specified.. Fuel purchase (fuel oil, coal, tank gas , etc . ) 1972 1973 1973 1972-73 1972 1972 $37.8 52.1 19.1 18.9 (NA) 34.1 18.3 4.0 11.9 4.0 3.0 3.0 3.5 5.0 2.0 2.0 3.0 1.5 3.0 $36.6 35.1 (NA) (NA) 12.5 37.0 17.1 3.2 13.4 3.3 3.5 4.5 5.0 15.0 3.0 3.0 4.5 3.0 7.0 1.062 1 .044 .9 54 1 .048 .972 .975 .986 .953 1.028 .703 1.037 'l .054 .703 PCE AHS AHS SORAR SORAR PCE PCE PCE PCE PCE NA Not available . 'Survey years selected so as to achieve greatest comparability with available independent data. 2 These estimates are based on special reweighted tabulations from the original survey data tapes. They do not reflect edit- ing changes which may have been made by the Bureau of Labor Statistics at later stages of processing and may, therefore, differ somewhat from estimates which have already or may subsequently be published by that Agency. 3 Coef f icients of variation of the two AHS base figures used in computing ratios of survey to independent estimates were about 2 percent (mortgage payments) and 2.5 percent (property taxes). The coefficients of variation of the SORAR base figures used for home repairs and alterations were about 3 percent in both cases. The PCE estimates are not constructed in a manner which would permit calculation of sampling variances. * Independent sources are as follows: PCE - Revised Personal Consumption Expenditures, GNP Accounts (See appendix A) AHS - Annual Housing Survey, Bureau of the Census SORAR - Survey of Residential Alterations and Repairs, Bureau of the Census See appendix A of Census Technical Paper No. 45 for a description of AHS and SORAR. 5 Includes any utilities or facilities covered as part of periodic rent payment. Excludes utilities paid separately by con- sumer unit . 6 PCE estimates adjusted upward by about $4.1 billion to restore amounts deducted by Commerce Department in converting Census contract rent figure to "space" rent concept used for PCE purposes. 'includes principal, interest, and, if covered by periodic payment, property taxes and insurance. 8 Includes both property taxes paid as part of periodic mortgage payments and those paid separately by home owners. There is some duplication, therefore, between the "mortgage payment" and "property tax" categories in the table. 9 For diary estimates, excludes expenditures for business purposes (presumably on business property or property rented to others). SORAR estimates used to compute ratios refer to 1-to 4-unit owner-occupied property. 10 Ratios for diary estimates shown only for total utilities because of large residual group (utility type not specified) in distribution by utility type . Chapter 9. Health Expenditures Considerable difficulties had been anticipated in the measure- ment of health expenditures on the basis of previous experience in this field. Aside from the general complexity of the field itself, the reporting is complicated by the pervasive influence of health insurance and other so-called third-party payors involved in the process. The objective was to measure out-of-pocket expenditures, exclusive of reimbursements by insurance or any other source. For this purpose, a detailed series of questions was included in the interview panel on a semiannual basis, covering the use of health services in the previous 6 months, the total cost of such services, and the actual or anticipated reimburse- ment, if any, from insurance or other sources. A small section was set aside on the diary form for personal care, drugs, and medical supplies. There was no specific mention of the main body of health expenditures (hospital charges, physician services, etc.) so that these were presumably relegated to the catchall section. Various previous efforts to measure health expenditures were reviewed at some length in the preliminary report, and data from several of those undertakings were used as independent sources in assessing the survey results. Although PCE data are available for some aspects (primarily for drugs and related items), the concept used in that series for the principal kinds of health expenditures relates to total costs, including outlays by insurance companies and other third-party payors. The main independent data used in place of PCE were the annual aggregates of health expenditures, by type, prepared by the Social Security Administration, which provide separate esti- mates for out-of-pocket costs. Also used were data from the continuing Health Interview Survey (HIS) of the U.S. Public Health Service and the periodic detailed surveys on health expenditures by the Center for Health Administration Studies (CHAS) of the University of Chicago. Since various transforma- tions of these data were needed for comparability with the 1972-73 survey results, the cautions cited earlier with regard to PCE estimates apply with equal or greater force in these instances. Overall Comparisons with Independent Sources Table 9-1 provides a comparison between the survey results and various independent sources for total out-of-pocket health expenditures, exclusive of drugs and health insurance premiums. This is the same information that was included in the preliminary report (since PCE data are not used for these composites) but is repeated here for convenience. These particular aggregates are presented because they can be derived for all of the sources shown. As is evident, the survey estimates are the smallest of those cited, although reasonably within the range of the Social Security Administration (SSA) and the Center for Health Administration Studies (CHAS) levels. The difference with SSA almost disappears, however, when an adjustment is made to exclude nursing home care, which represents a conceptual difference between the sources. The 1972-73 survey was confined to the civilian noninstitutional population, and expenditures for nursing home care would mainly represent payments made by the family on behalf of relatives who were resident in nursing homes. The SSA estimates cover the entire population and would include nursing home expenditures whether paid for by private families, by public agencies, or by the residents of the institutions themselves. As a result, the reduction is much greater in the SSA estimates when this category of health care is excluded. Also apparent is the fact that the third source, the Health Interview Survey (HIS) estimates, exceed the 1972-73 survey figures by a wide margin. There is some reason to believe, however, that the procedure followed in HIS— using self- administered questionnaires to obtain annual estimates for very broad categories of services— could lead to an exaggeration of expenditures. In fact, previous analyses of HIS estimates reveal that they consistently exceed the levels in SSA and other independent sources and often by appreciable amounts. Comparisons for Health Categories Comparisons between the survey and independent sources for broad categories of health expenditures are presented in table 9-2. The ratios of the survey to the independent estimates are provided for the "best" survey source, that is, the one (quarterly panel or diary) which was generally closest to the independent sources. Actually, although the space and attention accorded health expenditures on the diary record was minimal, the two survey procedures provided surprisingly similar results for the prime categories of physician and dental services. The quarterly panel results are designated as the "best" source because of smaller sampling variances. The quarterly panel also emerged as the "best" source for hospital services probably because it was difficult to identify from sometimes incomplete diary entries of all the subsidiary services (x-rays, laboratory tests, etc.) provided in hospitals. (Such services were assumed to be rendered outside unless specifically indicated as given in hospitals.) An obviously understated diary estimate for health insurance premiums is understandable, since so large a pro- portion of these are paid through payroll deductions and might be unknown to or easily overlooked by the family recordkeeper. 28 29 On the other hand, the diary was clearly the "best" source for the diversified group of drugs and related products, for which only a partial, and largely unsuccessful, effort was made in the interview panel. Again discounting the HIS levels as probably exaggerated, the survey data appear to hold up reasonably well in relation to the independent sources for most categories. The exception in the case of hospital services could possibly reflect the difficulties of obtaining accurate measurements through the survey mechanism in a sector where insurance reimbursements are particularly prevalent. The relatively low ratios for medical supplies and appliances are probably attributable to the great diversity of products involved and the intermittent and irregular nature of such purchases. Table 9-1. Annual Aggregate Out-of-Pocket Health Expenditures, Excluding Cost of Drugs and Sundries or Health Insurance Premiums: Quarterly Panel and Diary Operation Compared to Independent Data Sources, 1972-73 Total expenditures Estimated expenditures for nursing home care (millions of dollars) Source Amount (millions of dollars) Estimated coefficient of variation ( Percent ) Total excluding nursing home care (millions of dollars) Survey estimates 2 $18,916 18,682 22,566 26,903 20,823 2.0 5.0 3.0 5.0 $ 110 250 3,332 (*) (*) $18,806 $18,432 19,234 Social Security Administration ( SSA) estimates 3 Health Interview Survey (HIS) estimates (National Center for Health Statistics) 4 Center for Health Administration Studies (CHAS) estimates (University of Chicago) 5 (NA) (NA) NA Not available . *Amount of nursing home care included in total expenditure estimates from these sources is probably quite small (of the order of magnitude shown for expenditure survey estimates) but no specific information is available on the amount. 1 Includes hospital and nursing home care, physicians' services in and out of hospital, dental care, other professional ser- vices, and eye care and services. 2 Quarterly panel data based on average for calendar years 1972-73, diary estimates on average for fiscal years, July 1972 - June 1974. 3 Based on aggregates prepared by that agency for fiscal year, July 1972-June 1973. Unlike the survey estimates, which relate to the civilian non-institutional population, the SSA estimates cover the entire population, which accounts for the relatively large amount shown for nursing home care. The estimates are not derived in a manner which permits calculation of variance measures . Based on average of HIS survey data collected for calendar years 1970 and 1974. Based on CHAS survey data for 1970 updated to 1972-73 by two methods, (1) percentage change for each major expense category as indicated in SSA estimates, and (2) percentage change for each major category as indicated in HIS estimates. CHAS estimates in table represent average of data derived from the two methods, which essentially assumes that the relationship between CHAS and the two other independent sources was the same in 1972-73 as it was in 1970. 30 Table 9-2. Annual Aggregate Out-of-Pocket Health Expenditures: Independent Sources: 1972-73 'Best" Survey Estimates Compared to Category "Best" survey estimate 1 Annual amount (mi llions of dollars ) 2 Source (QP=quarterly panel ; D=diary ) Est imated coefficient of variat ion ( Percent ) Ratio of survey estimates to: 4 Social security adminis- tration (SSA) estimates Health interview survey (HIS) estimates Center for Health Adminis- tration studies (CHAS) estimates Revised personal consumpt ion expenditure (PCE) est imates Hospital services 5 Physician, nursing, eye care, and other professional services Dental care. Drugs and sundries 8 Prescription drugs.... Nonprescription drugs. Medical supplies 9 Ophthalmic and orthopedic appliances 8 10 Health insurance, including medicare $ 2,713 10,809 5,284 7,689 4,696 2,632 361 1,273 11 ,811 QP ■QP "QP D D D D QP 5.0 2 2 7 7 9 (NA) (NA) 2.0 .752 1 .049 .993 1.034 (NA) (NA) (NA) (NA) (NA) .541 .734 .738 (NA) .862 (NA) (NA) (NA) 1.100 6 . 777 .963 .866 1 .323 (NA) (NA) (NA) (NA) (NA) (NA) (NA) (NA) .951 .953 .992 .709 .720 (NA) ^Quarterly panel and diary estimates not significantly different; former chosen because of lower sampling variances. 'Where the two survey estimates were significantly different, the one closest to the independent estimates was chosen as the "best" estimate. Where they were not significantly different, the quarterly panel estimate was selected because it had smaller sampling variances; such cases are denoted by asterisk (*). 2 Estimates based on special reweighted tabulations from the original survey data tapes. They do not include any editing changes which may have been made by the Bureau of Labor Statistics at later stages of processing and may, therefore, differ somewhat from estimates which have already or may subsequently be issued by that Agency. 3 Coef f icient s of variation would be slightly higher than those for the 1972-73 survey for comparable estimates from HIS and perhaps two to three times higher for the CHAS estimates used in deriving the ratios shown in the table. SSA and PCE estimates are not derived in a manner which permits computation of variances. 4 See appendix A for a discussion of the PCE estimates. Descriptions of the SSA, HIS, and CHAS estimates can be found in appendix A of Census Technical Paper No. 45. 5 Includes room and board, X-rays and laboratory tests in hospital, professional services included as part of regular hos- pital charges , ect . 6 CHAS figure used in this computation represents the smaller of two estimates derived from that survey. The higher estimate was exaggerated by inclusion of one extreme sample report. The hospital services category in CHAS includes nursing home care, which was relatively small for the population covered. 'includes physician services in hospital where billed separately, physician services outside of hospital (including eye ex- aminations), private nursing services, chiropractor services, other medical services outside of hospital such as X-rays, labo- ratory tests, physical therapy, etc. 8 Survey estimates shown relate only to 1972 for greater comparability with the revised PCE figures (thus accounting for the relatively large coefficients of variation shown). In the computations of ratios to the SSA, HIS, CHAS figures for drugs and sundries, however, survey averages for 1972-73 were used. 'includes bandages, gauze, plaster, heating pads, syringes, rubber goods, and various other products. 10 Includes eyeglasses and contact lenses, hearing aids, artificial limbs, orthopedic appliances, etc. Chapter 10. Education, Travel, and Miscellaneous Expenses This chapter deals with a variety of miscellaneous expenses. Not all of these were affected by the PCE revision, since it was necessary to use other independent sources for some of the comparisons. Education Expenses PCE data on education are conceptually different from the survey results and were not used for comparative purposes. In fact, the comparisons that could be made were largely limited to college tuition and fees for which some independent data were available from the National Center for Education Statistics (NCES). Data on education costs, including room, board, and other expenses, as well as tuition, were obtained on a semiannual basis in the quarterly panel, by type of school. Except for some mentions of school supplies in one of the detailed sections on the diary record, there was no special provision for education expenses in that survey. Probably as a consequence, the diary estimate for school tuition— the only category that could be readily compared— was less than half as large as that from the interview survey. As indicated in the preliminary report, the quarterly panel estimate for college tuition costs amounted to only about 80 percent of the independent NCES level, in spite of the sizeable impact of this item in the family budget. It is possible that the problems of covering college students in household surveys accounted for part of the deficiency. Unmarried students away at college were considered as part of the family unit. Although the estimates of college expenses obtained from the family in these cases might be reasonably accurate, at least for tuition payments, there could be some slippage where the student contributed part or all of the cost. Married students away at college but living off campus were enumerated wherever they were staying. However, the semiannual inquiries on educational expenses happened to be scheduled during the three summer months and the three winter months when many, if not most, might be situated elsewhere. The results suggest the desirability of altering the survey procedure, whereby all students away at college would be interviewed at their school quarters, whether in dormatories or off campus. This approach would be especially important in measuring living and miscellaneous expenses which could be largely unknown to the family. Also, the scheduling of the interviews should obviously take account of the times when students would most likely be found at school. Travel and Transportation The interview panel contained an inquiry each quarter on trips taken in the previous 3 months which lasted overnight or longer. For each such trip, detailed information was obtained on expenditures for transportation, housing, food, and other costs. There is no way of deriving from the PCE data a set of estimates comparable with the survey expenditures for trips. In fact, there were no available independent sources for expendi- tures of this kind. The only comparisons which could be made were in terms of the number of trips taken rather than expenditures. In the preliminary report, it was indicated that the aggregate number of trips taken according to the quarterly panel (some 138 million in 1972) was within a couple of percentage points of a reasonably comparable figure from the National Travel Survey conducted in conjunction with the 1972 Census of Transportation. The correspondence was also close for certain broad categories of trips. Although this finding is not conclusive, the accuracy of expenditure reporting is often more dependent on whether or not a particular expenditure (in this case, a trip) is reported than on the precise amount of the expenditure. In a related sector, table 10-1 compares the survey results from the diary operation for public transportation expenses with the revised and preliminary PCE estimates for this sector. In this case, major deficiencies are apparent in every category, which can possibly be attributed, in part, to the fact that such expenditures are usually made on an individual basis. The disparities are even greater than in the preliminary findings, because of sharp increases in the revised PCE estimates for taxi and airline fares. These last two categories are difficult to estimate in PCE, as well as elsewhere, because business expenses represent so large a proportion of the outlays. Miscellaneous Expenses Table 10-2 compares the survey data with the revised PCE estimates for various miscellaneous products and services. All of these categories were covered, in one way or another, in the quarterly panel. Inquiries were made each quarter for watches and jewelry and for equipment repairs. An annual check, using the inventory approach, was made for pianos and organs. The remainder were covered semiannually. No special mention was made of any of these items on the diary record, so that any entries would be relegated to the catchall section. With a couple of exceptions, the patterns are similar to those observed for other expenditure categories. In general, the 31 32 quarterly panel appeared to be the superior source for the larger items of expenditure and the correspondence with the PCE estimates was reasonably close in these cases. The sizeable deficiency for watches and jewelry could be attributed to a variety of factors, among them the irregularity of such purchases and the fact that a substantial proportion may constitute gifts. Also, there could be some reluctance to report what may appear to be an extravagance as well as some concern about the security of expensive possessions. The similarly low ratio shown for moving expenses represents a change from the preliminary findings and occurred as a result of a large increase in the revised PCE estimate. The new comparison may be more reasonable, however, as many persons who have recently moved, or are in the course of moving, could be located in temporary quarters and might be underrepresented in the survey. The diary estimates were not significantly different from the quarterly panel data in several instances, but the sampling variances are quite large in these cases. For major sport equipment, the semiannual inquiry in the quarterly panel did not produce usable results. The diary estimates, although rather deficient themselves, were clearly the superior source at least for the 1972-73 survey. Table 10-1. Aggregate Expenditures for Public Transportation from the Diary Operation Compared to Personal Consumption Expenditures, 1972 Survey estimates (Diary operation) * Ratio of survey estimates to personal consumption expenditures (PCE) 2 Category Annual amount (millions of dollars) Estimated coefficient of variation (percent) Revised Preliminary Total. $4,232.7 1,015.0 439.3 173.1 1,992.6 612.7 8.5 11.0 14.0 20.0 11.0 .530 .446 .279 .442 .532 (NA) .713 .481 .522 .490 .756 (NA) 1 Survey data are derived from re-weighted tabulations of original survey data tapes and do not incorporate editing changes which may have been made at later stages of processing by the Bureau of Labor Statistics. The estimates may differ somewhat, therefore, from those already published or to be published by that agency. Diary estimates relate to period, July 1972-June 1973. 2 See appendix A for description. 3 Includes intercity as well as local bus. ^Includes commuting as well as intercity travel. Table 10-2. Aggregate Expenditures for Miscellaneous Products and Services: Revised Personal Consumption Expenditures, 1972 Quarterly Panel and Diary Operation Compared to Category Survey estimates 1 Quarterly panel Annual amount (millions of dollars) Estimated coefficient of variation (percent) Diary operation Annual amount (millions of dollars) Estimated coefficient of variation (percent) Ratio of survey estimates to revised personal consumption expenditures (PCE) 2 Quarterly panel Diary operation PRODUCTS Pianos (new) •••••••••••< Organs (new) Other musical instruments (new), Bicycles (new) , Watches and jewelry Major sport equipment 3 , SERVICES Moving and storage. .............. . Funeral expenses 4 . ................ Furniture repair and upholstering. Electrical appliance repair 5 ...... $179.5 252.1 285.2 637.2 2,514.5 (NA) 698.1 2,461.9 546.9 1,940.9 20.0 20.0 9.0 4.0 4.0 10.0 7.0 10.0 7.0 (NA) (NA) (NA) (NA) 2,290.2 1.191.0 429.6 831.4 649.2 1,668.1 6.0 9.0 (*) (*) (*) 13.0 1.095 .946 1.187 .810 .543 .510 .972 .859 .852 .494 .703 .314 .328 .020 .732 (NA) Not available. (-*) Not available but likely to be large, 15 to 20^. J See footnote 1 of table 10-1 for qualifications concerning data. 2 See appendix A for description. includes equipment for hunting and fishing, camping, golf, water sports, and winter sports, which were categories for which comparisons could be made. ^Includes cemetery plots as well as funeral parlor and burial expenses. 5 Includes radio, television, and other sound equipment repairs, as well as those for other electrical appliance; survey figures include cost of service contracts. Chapter 11. Conclusions and Recommendations Although a number of differences arose as a result of the PCE revision— some operating to widen and others to narrow the disparities in relation to the survey data— the general conclusions based on this reevaluation of the 1972-73 data are not materially changed from the preliminary findings. If allowance is made for the almost inevitable problems that will occur in any new and complex survey undertaking, the results can be judged as certainly tolerable and, in some instances, surprisingly successful. Perhaps more important than the immediate results is whether enough has been learned from the process to help shape an improved product for the future. Although many uncertainties remain, there are indications from the evaluation which offer some prospects for advances. Many of the findings were predictable on the basis of previous experience in this field. For example, larger expendi- tures and recurrent outlays were the better reported items in the quarterly panel. Expenditures made primarily by the home- maker or which were incurred on a frequent basis represented the more successful diary results. An unexpected outcome was the relatively effective per- formance of the diary procedure in some sectors, such as health and housing, where the quarterly panel had been assumed to be the only realistic option. The diary results also appeared to be about as effective as those from the quarterly panel in cases where both sets were deficient, including some situations (such as for clothing) where a great deal of attention had been devoted to expenditure reporting in the interview procedure. These findings suggest a general rule of thumb for future efforts in this field, to wit, that unless a clear-cut reason exists for using an interview procedure such as the quarterly panel, dependence might better be placed on the diary approach for a given expenditure category. The diary would be a dubious source for items with exceptionally large variances (vehicles, appliances, furniture, etc.) or where unusual payment arrange- ments might require special questioning (insurance paid through payroll deductions, mortgage payments made automatically through bank accounts, hospital bills paid largely but belatedly by insurance, etc.). The diary might also be inferior to an interview process in cases where a composite set of questions may be necessary in obtaining complete responses, as may be the case, for example, in cataloging trips and alteration and repair work to housing. In most other instances, the diary procedure appears to be at least as good a risk as the interview approach, and probably a less costly one as well. A number of deficiencies were revealed in the diary procedure, however, for which solutions should be attempted before considering wider use of the method. Several recom- mendations were made in the preliminary report with regard to possible improvements in both the diary procedure and the quarterly panel operation. These are repeated here, with minor modifications, for the convenience of readers. Recommendations for Diary Procedure The main possibilities seen for improvement in the diary procedure are as follows: 1. Limit the range of items any one family would be asked to report. The use of a diary covering all items of expenditure, as was done in the 1972-73 survey, may have certain theoretical benefits, but considerably proscribes the ability to improve the overall process sufficiently to satisfy the expanded requirements just cited. Evidently, as stated earlier, one of the reasons for the more successful coverage of food purchases than other small items in the 1972-73 diary was the much greater amount of space and attention accorded the former. Moreover, for all categories including food, there was a considerable undifferentiated residual group, resulting mainly from incomplete or inadequate entries which could not be classified in detail, which detracted materially from the usefulness of the results. The general lack of space and the inability of interviewers to focus on so wide a range of items in reviewing the diaries probably accounted largely for this latter deficiency. Since it would be impracticable to provide adequate space and annotation for all categories on a single form, one rather evident solution would be to limit the range of items which any one family would be asked to report. For example, one subsample might be asked to report only on food and other supermarket products, a second on clothing and household linens, a third on health-related expenditures, etc. There is obviously some practical limit to the number of subsamples that could be simultaneously operated and a good deal of thought and some experimentation would be necessary to devise a workable plan. Probably even more important than space considerations, the use of this kind of specialized approach would make it feasible to provide for a more focused set of check questions and procedures at the time of diary pickup to overcome some of the disparities noted in the present survey (such as underreporting of certain food items relative to others). In fact, a modified procedure would likely entail much more of a combination of interviewing and recordkeeping than is now 33 34 the case. A good deal of developmental work would be required to devise effective checking procedures. 2. Vary length of recordkeeping period. As previously noted, the 1972-73 survey provided for 2 weeks of recordkeeping for each sample family, covering all items of expenditure. If specialized subsamples are developed as proposed above, it is obvious that either a larger overall sample would be needed or much higher sampling variances would have to be accepted. One way out of this dilemma would be to vary the length of the recordkeeping period depending on the variances of the subjects covered for a given subsample. For example, for a low variance category such as food, it might even be possible to reduce the recordkeeping period to 1 week or 10 days. For most categories, however, such as clothing expenditures or health costs, an increase in the period of recordkeeping to up to 3 months or longer might be considered. The fact that only a limited set of items is covered might reduce the reporting burden sufficiently to secure extended cooperation of this kind. In such a system, use of less costly collection methods such as having respondents mail in completed diaries on a periodic basis (monthly, semimonthly, etc.) would be more practicable. Provision would, of course, have to be made for followup (by mail or telephone, where possible) for nonrespondents or to carry out special checking for returns which did not meet prescribed standards. 3. Provide separate diaries, where indicated, for individual members. As noted, only one person, usually the home- maker, probably maintained the diary for the entire family in the 1972-73 survey. Not surprisingly, the results were clearly more favorable for the kinds of expenditures for which the homemaker was mainly responsible than for those likely to be made by other members. One possible way of obtaining more consistent results, where the expenditures to be reported are of a more dispersed nature, would be to provide separate diaries for all family members above a certain age (perhaps 12 and over) on which to record their individual disbursements. For this purpose, the diaries could be briefer and less formal than the main record for the family. Experimentation with various versions would obviously be important in developing a procedure of this kind. 1 4. Reconsider matter of providing monetary or other incentives for cooperation. An experiment was conducted in the early stages of the 1972-73 survey on whether an offer of cash payments would materially improve cooperation in main- taining diaries. The results were inconclusive in this regard and the incentives were dropped from the procedure. However, most previous experience supports the notion that both cooperation and adequacy of reporting are benefited by some inducements of this kind. If greater dependence is to be placed on diaries in a continuing operation, as is being proposed, it would seem Individual diaries have been used successfully for some years in the continuing consumer expenditure survey in the United Kingdom. especially important to reconsider this matter of offering incentives and to experiment with alternative approaches. The rewards should, of course, be attuned as closely as possible to achieving the main objectives. If, for example, it is important for individual family members to keep separate records, the incentives should be offered only if all such members agree to cooperate, which would presumably promote intrafamily pressure for individual compliance. Similarly, if cooperation over extended periods is requested, a useful approach is to offer a small reward for each subperiod with a sizeable bonus for completing the entire cycle. Special consideration for prompt and complete returns by mail (where used) might be still another element. Payment of incentives might also be made contingent on retention by respondents of cash register tapes, bills, and other evidences of expenditures which could be consulted in reviewing the diaries. The incentives would not have to be monetary in nature, but could include gifts of various kinds, publications, chances for prizes, etc. 5. Continue exploration of timing biases. As noted earlier in this report, perhaps the most conclusive survey finding was affirmation of the traditional bias found in diary operations, whereby a higher level of expenditures is reported in the earlier as opposed to the latter stages of the recordkeeping period. Various proposals have been made for purposes of overcoming this problem, on the likely assumption that it is attributable to some element of telescoping at the outset and to some tailoff in reporting (possibly due to fatigue) toward the end. One proposal has been to lengthen the reporting period, for example, to 17 days for 3-week diaries, and then to exclude the entries for the early (first two) and final days from the estimation. One problem is that exclusion of the exceptionally high first day would likely accentuate the tendency for understatement in the overall survey results. Another possibility which has been suggested would be to obtain some information at the outset by interview for the previous day or two, which could be used as a "bounding" device in examining the diary entries for the first few days. The net result, however, would probably be the same as for the truncated procedure described above. The best choice would appear to be to aim at improvements in diary reporting by various of the measures suggested in earlier sections, in order to be able to afford the luxury of some of these techniques for controlling telescoping. Recommendations for Quarterly Panel Although the deficiencies in the quarterly panel results could be attributed to a variety of causes, the fatigue caused by the voluminous questionnaire was undoubtedly a prime factor. Perhaps the single most important step that could be taken to improve the results would be a significant reduction in the scope and content of the inquiry. The extended use of the diary procedure, already discussed, would accomplish a good deal in this direction. Such a change, however, would impair one of the :< : , auxiliary objectives of the survey, that is, the ability to analyze the consumption patterns of individual families across a wide range of products and services. From the standpoint of the operation as a whole, the main recommendations affecting the quarterly panel are as follows: 1. Restrict coverage to essential items. Mention has already been made of the kinds of items that should probably be retained in the interview procedure: high variance items, those requiring special questioning, and perhaps certain topics for which there were efficient composite interviewing packages. On this basis, the quarterly panel would probably cover not only vehicles, appliances, and furniture, but larger clothing items (e.g., coats and suits), other major furnishings (floor coverings, drapes, etc.), hospital services, school and college tuition, mortgage payments, moving and funeral expenses, insurance, subscriptions and membership fees, travel and vacation costs, and the like. Obviously a more detailed review and analysis would be necessary in selecting the optimum set. 2. Extend "inventory" approach. The generally effective use of the so-called inventory approach for major appliances sug- gests that it be extended to other categories as well. Examples would be major furniture items, floor coverings and drapes, and certain kinds of durable equipment pre- viously excluded. The evidence also suggests that considera- tion be given to pooling the two estimates obtainable for a given year (or other time period) under this approach, in the interest of reducing sampling variances. For minor appliances, the inventory approach is probably more effective than using a direct recall procedure, in spite of the deficiencies found in the 1972-73 results. More attention should be given in the interviews, however, to the likelihood of multiple possession of these items within the family. In addition, consideration might be given to reducing the time reference for coverage of these items from a semiannual to a quarterly basis. This would be more feasible if a marked reduction were made, as proposed, in other aspects of the quarterly panel interviews. 3. Retain flexibility in reference periods. The procedure used in the 1972-73 survey drew oh the well-documented principle that the larger the expenditure, the longer it can be reliably recalled. Thus, inquiries were made for specific categories in the interview panel on either an annual, semiannual, or quarterly basis, depending mainly on expenditure size. For the most part, the range of those choices was confirmed by the results. In a few cases, such as just described for small appliances, it appeared that a reduction in the recall period from 6 months to 3 months might be beneficial for some of the smaller items. However, there did not appear to be any need to curtail the time reference for the larger expenditures, where retention of longer recall periods would serve to minimize the number of occasions any given family would be asked about the same items. 2 4. Retain analytical opportunities, where possible, with respect to individual family behavior. The proposed reduction in the content of the quarterly panel interviews and the possible introduction of specialized diaries means that only partial expenditure data would be available for individual families, either for annual or shorter periods. It would be desirable, if this could be done without jeopardizing the basic expendi- ture data, to retain as much of the ability as possible for studying individual consumer behavior. Some of the largest items, which distinguish one family's expenditures from another's, would be retained on the quarterly panel under the proposal, but most categories would not be covered. A first step to close the gap might be to add a few questions, perhaps on an annual basis, to identify exceptional expenditures in other sectors, such as large orthodontic bills, major house additions, purchase of expensive paintings, and the like. Beyond that point, a further although admittedly experi- mental step would be to obtain certain limited information in the quarterly panel on behavioral patterns which would help in imputing data from the diary operation or other sources to the individual family record. For example, in the health sector, questions could be asked on frequency of doctor or dental visits, consistent usage of costly drugs, etc., which, together with demographic and socioeconomic char- acteristics, could provide a basis for imputing a value for health expenditures to the family from another source. The topics covered could vary from one quarter to another, so that these behavioral questions would not add excessively to the quarterly panel interviews. Obviously, considerable research and testing would be needed before embarking on a large-scale endeavor of this kind. Possible Use of Other Data Sources In collecting independent estimates for purposes of comparison with the survey results, it became apparent that there are a variety of such sources which sould be taken into account in developing a continuing data system in this field. For example, the Health Resources Administration of the U.S. Public Health Service is instituting a detailed national household survey on health expenditures. The Census Bureau operates a national Annual Housing Survey containing some expenditure data and another survey on expenditures for home repairs and altera- tions. That agency is also launching a household survey on trips and vacations. There are also many market research and other private operations of this kind, although of rather uneven quality. 2 Since appliances, furnishings, and similar items are purchased very infrequently, repetition of the same questions on a frequent basis (e.g., quarterly), where this is not clearly necessary, is likely to create a considerable annoyance (and waste of survey capacity) without achieving any productive results. 36 Some of these operations are or have been on a continuing basis, but the status of others is uncertain. The main point is that, in a system such as that suggested in this report, there is no reason that data from other sources of this kind could not be substituted for those from a continuing comprehensive expendi- ture survey, provided a set of matching demographic and socioeconomic characteristics is available. The advantage of specialized surveys, assuming they are of adequate size, con- tinuity, and technical quality, is that particular subjects can generally be explored in considerably greater detail, using methodologies especially designed for that purpose. Need for Methodological Research Several instances have already been cited whereby further methodological research would be essential in shaping a more effective data system for consumer expenditures. Although the needs are varied, it appears that the highest priority would be to experiment with modifications in the diary procedure such as testing the use of specialized diaries and diaries for individual family members, development of special checking procedures to overcome reporting deficiencies, exploring the feasibility of varying lengths of diarykeeping periods, consideration of less expensive collection methods, and reexamining the use of monetary or other incentives for cooperation. Some longer range possibilities should also be investigated. For example, respondents might be asked to record the "universal product codes" now appearing on most canned and packaged super- market and drug store items (and likely to extend to many others) in the interest of achieving more precise and consistent product classification. In a different vein, interviewers might seek permission to take brief shelf inventories when collecting diaries in an effort to locate the kinds of products which tend to be overlooked inadvertently by respondents. In more sensitive areas, such as alcoholic beverages, tobacco products, jewelry acquisitions, and the like, experimentation with techniques used in other sensitive fields would be in order. Since the survey findings support retention of an inter- viewing capability as a companion to the diary procedure, some further methodological research relating to the quarterly panel would also be in order. One fruitful area might be testing the extension of the inventory approach used for appliances in the 1972-73 survey to additional sectors. Another might be explora- tion of less costly collection procedures in this instance also, such as use of mail questionnaires or telephone interviews for certain of the quarterly contacts following the first visit. Mention has already been made of the need for research in the event that consideration is given to imputing expenditure data from various sources to individual family records. Appendix A. Personal Consumption Expenditures The principal source of independent estimates used in this report is the Personal Consumption Expenditures (PCE) com- ponent of the Gross National Product Accounts. These estimates, in effect, represent the market value of goods and services purchased by persons and nonprofit institutions in the United States. A description of the methods used in deriving the PCE estimates occupies 100 or more pages of text 1 and only a brief summation will be presented here. The principal starting point is data on the production value of goods from the quinquennial Censuses of Manufactures and other sources and of services from a variety of sources. Through input-output analysis, the flow of commodities and services is traced through the various channels in the economy and cost and profit margins are added to the production value at each stage to arrive at the final market value. Numerous assumptions are needed to achieve these complex transformations and there is no way of assessing the accuracy of the processes. PCE estimates are prepared by the U.S. Department of Commerce on a quarterly basis for a limited number of broad product and service categories. More detailed estimates are prepared only for benchmark years— those years ending with digits "2" and "7"— which coincide with the occurrence of the Census of Manufactures and the other major economic censuses. These benchmark data, however, are usually not available until some years following completion of the censuses. At the time the preliminary evaluation was started, the latest available detailed PCE estimates were for 1967. To permit more detailed comparisons, the author performed an updating of the 1967 PCE estimates, using a rather straightforward approach. The procedure was to project the detailed 1967 PCE figures to 1972 by applying the percentage change in production value for the corresponding items as shown in the 1967 and 1972 Censuses of Manufactures. A similar procedure was followed for exports (shown separately in the detailed PCE estimates) in this case using Census foreign trade statistics for the two dates. For services, the detail provided by the Commerce Department for 1972 was usually sufficient for comparative purposes, although some subdivision was done, in a few instances, using 1972 Census of Service Industries data. The main assumption in this method for products was that the marketing margins (per- centage difference between production and market value), as calculated in PCE, did not change between 1967 and 1972; this is probably a conservative approach since marketing and related costs have reportedly risen especially rapidly. The procedure 'See footnote 2 in the introduction to this report for the source reference. probably had greater validity in those instances where most of the production was eventually purchased by private individuals (e.g., clothing) and where no major portion was exported. The least success would be in cases where much of the production went into business use (e.g., office supplies). The recent completion by the Commerce Department of the benchmark revisions to the 1972 Economic Census data for detailed PCE components permits substitution of these 1972 estimates for those projected by the author. Certain adjustments are still needed, however, for comparative purposes. For example, one of the detailed PCE components comprises "women's outerwear, not elsewhere classified" which includes aprons, swimwear, sweaters and untailored jackets, slacks, shorts and playsuits, etc. Since the analysis required a split of these into the separate components for comparison with the survey data, a subdivision was made on the basis of the Census of Manufactures data for 1972 and the comparable data on imports. This type of subdivision was done in a number of other instances in order to obtain data for analytical purposes in even finer detail than presented in this report. The main effect of the revision actually was not in these detailed categories projected by the author. Rather, it was in several of the broader categories for which the Commerce Department had been producing projections from the 1967 benchmark. In some important categories, the PCE concepts were sufficiently incomparable with those used in the expenditure survey so that other independent sources were used as the main or only basis for evaluation. For example, there are no PCE estimates for some of the major elements of housing costs: mortgage interest, real estate taxes, home repairs or alterations, etc. Instead, PCE provided an artificial measure of these ownership costs, which essentially represents the rental value of owned homes. For comparison with the survey data on housing costs of this kind, information was used from various census surveys cited in the report. In the health sector, the PCE estimates for the major kinds of health services include the contributions from health insurance and other third-party payors, as well as direct payments by private individuals. Since the survey data related to out-of- pocket costs (exclusive of reimbursements), the comparisons were made with other independent data from the Social Security Administration and various health surveys. Education costs in the PCE series relate to -the actual expenditures of the educational institutions and, moreover, only to private schools. The comparisons for the survey data on 37 38 out-of-pocket costs had to be made with other sources, such as the National Center for Education Statistics. Insurance costs represented still another conceptual incon- sistency. The PCE estimates relate to insurance premiums less reimbursements to individuals, whereas the survey data relate to premiums paid. Some insurance industry data and health survey data were used, instead, for comparison with the 1972-73 survey results. The inclusion in PCE of expenditures by nonprofit institu- tions would not appreciably affect the comparisons except in rare instances. Another small element of difference is that the survey estimates, unlike the PCE data, might not fully reflect sales taxes. The combined effect of these two differences would be to widen slightly the margins between the survey and independent data, but not to such a degree as to alter any conclusions. Appendix B. Variation in Completion of Diaries by Different Population Groups The diary operation was conducted in essentially three steps. Interviewers made initial visits to obtain background informa- tion about the families and to place the diaries for the first 7 days. They returned a week later to collect the completed record books and place the diaries for the second 7 days. A third visit was made the following week to pick up the second week's diaries and to ask some concluding questions. At the time the diaries for either period were collected, the inter- viewers were instructed to review the entries for completeness, clarity, and legibility and to ask any additional questions necessary to satisfy the data requirements. In the event the diaries were completely blank, the interviewers were required to obtain and record the necessary expenditure information for the previous week by interview. A check item was provided on each diary form in which the interviewer was to indicate the state of completeness, that is, whether the record book had been filled entirely by the respondent (no substantive entries made by interviewer), the diary had been filled partly by the respondent and partly by the interviewer, or the diary had been filled entirely by the interviewer because the respondent had failed to enter any of the information. This check item was marked for all but 2 percent of the completed diaries. Some ambiguity probably is attached to the second of the possibilities (joint participation of the two parties) since it could theoretically entail anything from minimal to rather significant intervention on the part of the interviewer. In any case, the results of this check item may be useful as an indicator of the relative success of the diary procedure among various demographic and socioeconomic groups in the population, at least in terms of enlisting 'their participation. Summary of Results Table B-1 presents the distribution of the diaries by the various degrees of completeness for different geographic, demographic, and socioeconomic groups. On an overall basis, about 75 percent of the diaries for both survey years combined were reportedly filled completely by the respondent whereas, at the other extreme, close to 15 percent had to be completed entirely by the interviewer. The completeness rate (in terms of the respondent's contribution) was evidently slightly higher for the first year of operation than for the second. Ironically, this difference undoubtedly reflects the much greater success in recruiting proportionally more families to participate to some extent in the diary operation in the second year than in the first. 1 Apparently, the additional recruits included a dispropor- tionate number of those who experienced difficulty or less desire in maintaining the diaries, once they agreed to participate in the survey. The differentials in respondent completeness rates among the various population groups were about as predictable as any set of survey variables in the entire spectrum. With only rare exceptions, the differences could be attributed to such evident factors as educational level, economic status, age or infirmity, and similar traits. Since these same factors impinge on most of the population characteristics included in the analysis, there is a good deal of redundancy in the results. It would, therefore, be useful to refine the analysis through multivariate techniques in order to assess the relative importance of the various elements. Starting with the geographic residence categories, there were no particular advance expectations except in the case of residents of central cities of standard metropolitan statistical areas (SMSA's). The lower respondent completeness rates in that instance clearly mirror the greater concentration of older people and persons of lower socioeconomic status in the inner cities. The comparisons by the type of family also essentially reflect socioeconomic circumstances. The more or less "normal" families, consisting of married couples with or without minor children, would be expected to have the higher completeness rates, since they essentially represent the broad middle class. The far lower rates for families headed by unmarried men can be explained both by their economic status and by the characteristically lower cooperation obtainable in such under- takings from male respondents. At the same time, the group of families headed by women comprises many elderly women living alone (the term "family" being used rather loosely for this purpose) whose difficulty with recordkeeping on the whole would be a natural consequence of age or infirmity, as well as younger broken families where a similar result would ensue because of cultural or economic deprivation. Differentials by family size appear only at the extremes. The "one-person" families represent a composite of disparate groups, elderly people living alone or unattached persons of various younger ages, all of whom might be expected to exhibit lower completeness rates, although for different reasons. The largest families, in turn, are disproportionately those in lower 1 For detailed data on diary recruitment rates, see Walsh, Thomas C. op cit, pg. 6. 39 40 economic circumstances. Among the largest differences of all were those between white and black families, for reasons already adequately stipulated, with families headed by persons of other races falling somewhere in between. Since the latter group includes many persons of Oriental origin, language problems could have created an additional barrier to record- keeping. Age variations in respondent completeness rates are only clearly evident at the upper extreme. The higher rates for male-headed as compared with female-headed families at every age level is just another reflection of socioeconomic differences. Of course, most male-headed families are husband-wife families wherein the diaries would usually be maintained by the wives. Turning next to perhaps the most directly associated variable— educational attainment— there is an almost classic monotonic upswing in respondent completeness rates as we move up the educational ladder. The pattern is similar for income classes although there appears to be some tendency for the rates to level off beyond the median point. 2 The differences between owners and renters fall into the usual socioeconomic mold. The final three distributions— work experience of wives, date of diary placement, and starting calendar day for record- keeping— have a little different focus, that is, to examine the extent to which presumably busy people are less likely to keep their diaries. If there is a sizeable impact of this kind, families where the wife works virtually year round might be expected to exhibit lower respondent completeness rates. As the table indicates, there are no significant differentials of this kind. However, since full-time working women tend to have higher educational attainment, on the average, that factor might have served to offset the effect of busier schedules. In terms of seasonal distinctions, another expectation might be that diaries placed close to Christmas would be less fully completed. December diaries reveal somewhat lower respondent complete- ness rates, but the difference is not statistically significant. Lastly, there might be some likelihood that diaries placed toward the end of a calendar week would be less completely filled, since many of the individuals involved would be persons difficult to reach during the week. In this case, the expectations appear to have been realized to some degree. 2 The much lower completeness rate for the sizeable proportion of families (about one-seventh) who did not report their incomes is probably indicative of a generally less cooperative attitude toward the survey. Conclusions Although the results concerning respondent diary completion rates turned out to be largely predictable, they offer little direct insight into ways of improving the situation. Perhaps the only realistic option is to continue to provide for a standby procedure for the conduct of interviews in the event respond- ents fail to acquit their commitments. Actually, there is as yet no evidence— although some may be forthcoming later— that the information developed by interview is less complete or accurate, especially for the lower socioeconomic groups, than the results of diarykeeping. However, it is noteworthy that the majority of families in even the population groups with the poorest overall performance did manage to complete their own diaries. If some means could be found to increase that proportion, this would result in conserving interviewer resources for more essential tasks, such as more exhaustive diary checking procedures across the board. In order to pursue this matter, the following kinds of research may be helpful: 1. Refine the analysis through such statistical techniques as discriminant functions to identify categories whose combina- tion of characteristics might suggest a relatively small chance of completing a diary. Based on such determinations, critical problem cases might be flagged in advance and assigned to special alternative procedures, such as telephone or personal interviews on a frequent, short-term basis, which might provide more reliable results than obtainable from retro- spective interviews covering the entire week. 2. Conduct followup interviews with persons who fail to complete diaries in subsequent methodological tests or even ongoing operations to obtain more insight into their diffi- culties with recordkeeping and to solicit their views on the kinds of alternatives that might reduce their problems or otherwise induce them to actively participate in the survey. 3. Based on these findings or on the experiences of other researchers, experiment with various other options aimed at minimizing the difficulties of and inconvenience to problem groups, such as providing less formal record books, simple tape recorders in lieu of written diaries, enlisting the aid of relatives, friends, or neighbors where individuals are clearly incapable of self-reporting, etc. 4. Examine the impact of incentive payments on improving compliance. A positive effect might be more likely for these problem cases than for others since they are largely concentrated in lower socioeconomic groups. 41 Table B-l. Diary Completion Rates, by Population Group, 1972-73 Consumer Expenditure Survey (Percentages based on the approximately 98 percent of all diaries for which completion status was indicated by interviewers. Percentages may not add exactly to 100 percent because of rounding) Population group Percent of all diaries Completed entirely by respondent Completed partly by respondent ind partly by interviewer Completed entirely by interviewer Population group Percent of all diaries Completed entirely by 1 .■■.[!■ Hl()l' 111 Completed partly by respondent and partly by interviewer Completed entirely by interviewer Total, all diaries 1972 diaries 1973 diaries Geographic location In standard metropolitan statistical area (SMSA) In central city Outside central city Not in standard metropolitan statistical area Urban places Rural areas Type of family Male head, married spouse present... With children under 18 years No children under 18 years. Other male head Female head With children under 18 years No children under 18 years Size of family 1 person 2 persons 3 persons 4 persons 5 or 6 persons 7 persons or more Race of head White Black Other Sex and age of head Male, total 14 to 24 years 25 to 34 years 35 to 44 years 45 to 54 years 55 to 64 years 65 years or over Female , total 14 to 24 years 25 to 34 years 35 to 44 years 45 to 54 years 55 to 64 years 65 years or over 74.7 76.5 73.1 73.9 69.5 77.9 76.5 75.3 77.3 79.2 80.7 77.4 63.9 67.0 65.5 67.4 65.8 76.3 78.1 80.0 78.1 71.7 77.4 52.3 66.1 77.0 77.1 81.8 78.3 78.1 75.3 69.1 67.0 72.2 70.9 67.7 69.9 70.8 60.9 10.8 11.7 10.1 10.9 12.2 9.8 10.6 10.9 10.4 10.2 9.9 10.6 11.2 12.3 14.4 11.7 11.5 10.4 10.3 10.0 11.5 13.3 9.8 19.5 12.7 10.4 9.3 8.5 10.7 10.3 11.9 11.9 12.3 11.9 10.1 14.4 11.4 12.6 12.9 14.4 11.9 16.9 15.1 18.3 12.3 12.9 13.8 12.3 10.6 9.4 12.0 24.9 20.7 20.1 20.9 22.7 13.3 11.6 10.1 10.5 15.0 12.8 28.3 21.2 12.6 13.6 9.7 11.0 11.6 12.8 19.0 20.7 15.9 19.0 17.9 18.7 16.6 26.1 Education of head 8 years of schooling or less 9 to 11 years 12 years (high school graduate) 13 to 15 years 16 years or more (college or higher) Family income Less than $5,000 $5,000 to $6,999..... $7,000 to $9,999 $10,000 to $14,999 $15,000 to $24,999........ $25,000 or over I ncome not reported Housing tenure Owne r Renter Rent-free occupant Work experience of wife (previous year) All husband-wife families Wife did not work Wife worked 1 to 13 weeks Wife worked 14 to 26 weeks Wife worked 27 to 39 weeks Wife worked 40 to 52 weeks Month of diary placement January to March April to June July to September October to December December Starting day for diary Monday Tuesday Wednesday Thursday Friday Saturday Sunday 62.8 72.5 78.1 80.2 85.2 63.5 74.1 77.8 82.4 83.2 84.6 67.9 77.9 70.1 68.3 79.1 80.0 79.7 80.2 78.8 74.9 74.3 76.4 73.4 71.8 75.9 77.4 75.0 75.8 75.1 71.5 70.3 14.9 12.1 9.9 9.2 6 . ■', 13.5 10.8 11.0 9.4 9.1 7.2 11.4 10.1 12.1 12.5 10.0 9.2 10.1 10.1 11.0 10.5 9.5 11.2 12.0 11.8 10.5 11.3 10.7 10.6 11.4 11.8 22.3 15.3 11.9 10.6 8.5 23.0 15.1 11.2 8.2 7.8 8.2 20.7 12.0 17.8 19.2 10.9 10.9 10.2 9.7 10.2 14.6 16.2 12.4 14.5 16.3 14.2 12.2 13.7 13.4 14.3 17.1 18.0 Appendix C. Use of Special Check Questions in the Diary Procedure for Sensitive Items Experience in previous surveys had indicated the likelihood of a serious understatement of certain expenditures, partly because of sensitivity about reporting the items involved. The most striking examples are alcoholic beverages and tobacco products where only a fraction of the expenditures are ordinarily reported regardless of the method of collection. Some difficul- ties have sometimes been experienced also with outlays for meals in restaurants and other eating places, although this has been regarded as more likely a matter of oversight than of deliberate misreporting. Some special check questions were included in the diary procedure in an effort to overcome these anticipated defici- encies. At the time the diaries for the second week were collected, interviewers addressed an additional series of questions for families whose diaries for that week did not reveal expenditures of at least $1 each for alcoholic beverages, tobacco products, or purchased meals. For each category, an inquiry was made as to whether the family ever made expenditures for that purpose, even occasionally. If so, they were asked if any such expenditures were made during the previous week and for the amount, if any. Overall Findings As reported in chapters 2 and 3, the original survey estimates for alcoholic beverages and tobacco products affirmed the continued failure of undertakings of this kind to measure these particular expenditures. The estimates for alcoholic beverages amounted to only about 35 percent of the levels indicated by the independent sources and those for tobacco products were only moderately better (60 percent). The shortfall exhibited for alcoholic drinks purchased in restaurants, bars, and similar places was probably exaggerated, because some of those expenditures were included in the cost of purchased meals and not separately identified by the respondents. However, the disparity would undoubtedly have been very large regardless of whatever reasonable allowance could be made for this factor. A less anticipated finding was the relatively close corre- spondence between the survey and independent estimates for purchased meals and snacks although the precise amount of difference is obscured by certain conceptual incomparabilities. 1 'The independent (GIMP Personal Consumption Expenditures) esti- mates used in this comparison for purchased meals and snacks related essentially to expenditures in restaurants and other eating places. The survey estimates would also include snacks purchased in stores, refresh- ment stands, etc. for consumption off the premises but outside the home. The size of this latter group cannot be precisely determined, but a "ballpark" estimate would place it at several billions of dollars. Among the possible reasons for this favorable outcome mav have been the prominent positioning of a section devoted to these expenditures on the diary record. The special attention accorded this category in the diary checking procedure could also have played a role. Table C-1 presents the original survey data for these expenditure categories in relation to the independent estimates, together with the results of the special check questions. As is apparent, the gains from the checking procedure are far from impressive and contribute little to overcoming the deficiencies for alcoholic beverages and tobacco. The rather negligible pickup for purchased meals may merely substantiate the relatively good showing for this category in the original diary data. The seemingly more substantial gain for alcoholic drinks outside the home is deceptive, however. Many persons who had included the cost of alcoholic drinks as part of the price of purchased meals when completing the diaries evidently again reported the purchase of such drinks when asked specifically about them in the checking procedure. Although only relatively few additional cases were picked up by means of the check questions, it is interesting that they were not purely marginal cases, that is, persons with small expendi- tures. As shown in table C-2, the average expenditures for the newly identified cases were not markedly different from those for the original group. While this experience with special check questions is not particularly encouraging, there is some new evidence from other sources to indicate that the problem may have been due more to the kind of inquiry made than to the process itself. In a joint study by the Illinois Survey Research Laboratory and the National Opinion Research Center of different approaches to asking about gambling, drinking, and sexual behavior, it was found that the more intensive approach did not significantly increase the proportion reporting the activity but had a striking impact on the indicated volume of activity among reporters. 2 In the case of drinking, the proportions admitting some indulgence during the previous year were almost unaffected by the approach used, but the aggregate amount reportedly consumed was 2 to 3 times larger for the group asked longer and more informal questions than for those subjected to a less detailed inquiry. In fact, in comparisons against independent data on 2 Blair, Ed, Seymour Sudman, Norman Bradburn, and Carole Stock- ing, "How to Ask Questions About Drinking and Sex: Response Effects in Measuring Consumer Behavior," Journal of Marketing Research, Special Issue, August 1977. 42 43 consumption of alcohol, much of the usual gap was apparently closed by use of the more detailed approach. Although the situations are not quite parallel, one implication of these findings is that the checking procedure used in the 1972-73 survey was deficient in confining the questioning to those who reported virtually no expenditures in their diaries, rather than also attempting to obtain more complete informa- tion for those who had recorded some such outlays. In fact, in further experimentation of this kind, the preferable approach might be to omit categories such as alcohol and tobacco entirely in the diaries but to include a set of appropriate probing questions, perhaps patterned after some of the research just described, to be asked at the time the diaries are collected. Differences for Demographic and Socioeconomic Groups Although the pickup in participation was small, it may be interesting to examine the results of the checking procedure for various demographic and socioeconomic groups. Table C-3 presents the percentage increases resulting from the check questions over the proportions originally reporting the expendi- tures in their diaries for various population groups. Most of the differences are small and probably not statistically significant, but a few stand out. For example, the proportionate increases are generally larger for blacks than for whites, for older families compared to younger ones, and for families whose heads had lower educational attainment and income compared to those in higher socioeconomic strata. These differences may reveal some further indications of disparities in the adequacy with which diaries were maintained by the various population groups. It may be reasonable to assume that the larger the relative yield of special check questions, the less adequate were the original diary entries in the first place. It may be significant, in this regard, that the larger relative yields were precisely for those population groups who would be expected to have the greater difficulty in keeping diaries. Of course, another interpretation could be that check questions of this kind are just more effective for certain groups than for others regardless of the validity of the original data. Table C-l. Aggregate Expenditures for Purchased Meals, Alcoholic Beverages, and Tobacco Products: Original Diary Estimates and I Estimates Derived from Special Check Questions Compared to Independent Sources, 1972 (Billions of dollars) Category Original diary estimates 1 Additional expenditures reported from special check questions Independent estimates (Personal consumption expenditures PCE-GNP accounts ) 2 $31.7 7.4 5.8 1.6 7.7 $0.5 1.2 0.5 0.7 0.3 $30.0 22.3 14.8 Alcoholic drinks purchased outside 7.5 12.2 These estimates are based on special reweighted tabulations from the original survey data tapes and do not reflect editing changes which may have been made at later stages of processing by the Bureau of Labor Statis- tics. They may, therefore, differ somewhat from estimates which have already or may subsequently be published by that agency. Since diaries were not maintained during periods the entire family was away overnight or longer, the diary estimates have been adjusted upward (by $3 billion in the case of purchased meals) to reflect expenditures while on trips or' vacations, as indicated by information on such outlays obtained in the quarterly panel . 2 See appendix A for description. 44 Table C-2. Average Weekly Expenditures for Purchased Meals, Alcoholic Beverages, and Tobacco Products: Families Reporting Expenditures in Diaries Compared to Those Indentified Through Special Check Questions, 1972-73 (Dollars per week) Category Families reporting expenditures in diaries Families reported as making expenditures according to special check questions $11.32 6.24 5.37 4.01 $8.50 5.87 Alcoholic beverages Alcoholic drinks purchased outside 6.90 3.24 45 Table C-3. Increase Resulting from Special Check Questions in Proportion Reporting Purchased Meals, Alcoholic Beverages, and Tobacco Products, Expenditures by Population Group: 1972-73 (Increase as a percentage of original diary reporting rates) Population group Purchased meals and snacks Alcoholic beverages for home use Alcoholic drinks purchased outside home Tobacco products All families. White Black. ...... Other races. Male head, married spouse present. With children under 18 years.... No children under 18 years Other male head. Female head With children under 18 years.... No children under 18 years Age of head Under 25 years......... 25 to 34 years......... 35 to 44 years......... 45 to 54 years.. , 55 to 64 years.,......, 65 years and over Education of head 8 years or less 9 to 11 years ...... 12 years 13 to 15 years. 16 years or more... Family income Less than $5,000 $5,000 to $6,999 $7,000 to $9,999.... $10,000 to $14,999.. $15,000 to $24,999.. $25,000 or over Income not reported. 3.9 3.6 5.8 2.5 3.5 2.6 4.6 4.5 5.0 4.4 5.0 4.4 3.0 3.0 <. '• 4.1 7. I 8.1 4.0 2.8 3.1 2.6 6.8 3.2 11.7 10.5 22.8 31.9 11.0 12.0 9.4 14.6 13.7 11.9 13.7 14.3 11.3 10.0 9.3 13.5 15.5 19.1 16.3 9.8 10.2 7.2 23.1 10.3 10.8 10.9 9.0 8.8 13.2 26.0 23.6 53.3 31.3 25.2 26.0 24.7 28.7 23.5 31.0 22.6 19.0 26.9 27.1 22.9 25.8 40.6 54.8 33.3 21.4 20.5 20.8 43.2 29.9 33.7 25.0 22.8 15.1 24.2 7.0 6.7 9.1 28.1 7.3 7.2 7.2 8.6 7.0 5.2 7.2 9.8 6.7 7.8 7.0 6.4 6.5 6.8 9.4 To illustrate this calculation, about 20.9 percent of all white families reported expenditures of $1 or more in their original diaries for alcoholic beverages for home use. In response to the check questions, an additional 2.2 percent of all white families reported such expenditures . Dividing 2.2 percent by the original 20.9 percent, we obtain a proportionate increase of 10.5 percent attributable to the check questions. Appendix D. Timing Biases in Diary Reporting As reported in Technical Paper No. 45, one of the most clearcut survey findings was affirmation of the characteristic timing bias found in diary operations, that is, for a higher level of expenditures to be reported in the earlier as opposed to the latter stages of the recordkeeping period. For 2-week diary operations, the outcome is nearly always a higher level in the first as compared with the second week. Differences of this kind were found for virtually every expenditure category— food as well as nonfood— in the 1972-73 survey. Overall, the first-week expenditures were about 8 percent higher for food products and 10 percent higher for nonfood items. Several hypotheses have been advanced to explain this persistent phenomenon. One theory is that the differences are mainly attributable to "telescoping" of expenditures at the initial stages. According to this theory, many respondents do not begin their diaries immediately upon placement 1 but skip the first few days and later attempt to reconstruct the earlier period from memory. In doing so, they may inadvertently include some expenditures made prior to the reference period. A countervailing theory is that the lower levels at the latter stages primarily reflect the increasing fatigue or disenchantment of respondents. Temporary changes in buying habits because a diary is being kept has also been offered as a possible explanation. These could occur, according to this version, either because respondents initially buy additional or different products in order to impress the interviewer or, conversely, later refrain from making certain purchases to avoid some of the burden of recordkeeping. In order to explore this matter further, special tabulations have been prepared on expenditures reported in the diaries for each separate day of the 14-day recordkeeping period. Various alternate groupings of 7 consecutive days within the 14-day period have also been constructed for purposes of assessing some of the differences. These special computations are possible because separate pages in the diary record were provided for each day of the reporting period so that expenditures can be allocated to specific days. The estimates relate to families who participated in the survey in both weeks, about 95 percent of all cooperators. Methodological Problems Various problems arose in the course of developing the detailed estimates for individual days of the recordkeeping period. First of all, the placement of diaries was not randomized over all calendar days of the week because such a procedure was deemed to be unduly restrictive and costly from an administrative standpoint and because it was not considered feasible to contact families on Sundays. Since the volume of consumer purchases varies a good deal according to the day of the week— gradually reaching a peak by the weekend— some means of standardizing for imbalances in placement was necessary in making the calculations. 2 The procedure followed for this purpose is described below: 1. Average expenditures per family were computed for each reporting day for each calendar day. Thus, an average was calculated for Monday expenditures for families for which Monday was the first reporting day, those for which Monday was the second reporting day, etc., up through those for which Monday was the 14th reporting day. The same procedure was followed for each other calendar day. 2. Next, average daily expenditures were computed for each calendar day for all reporting days combined. Thus, an overall average was computed for Monday expenditures, for Tuesday expenditures, etc. 3. An index number was then calculated for each reporting day for each calendar day of the week. This was done by dividing the reporting day averages in step 1 above (e.g., Monday expenditures where Monday was the first reporting day) by the corresponding calendar day averages computed in step 2 (e.g., all Monday expenditures). The end product was an index number for each of the 14 reporting days for each of the 7 calendar days of the week. 4. Finally, the average index was computed for each reporting day. This was done by averaging the indexes computed in step 3 for each reporting day, e.g., averaging the indexes for reporting day 1 for all of the 7 calendar days. The final 'As mentioned earlier, the diary operation was conducted in three steps. Interviewers visited the households to obtain background informa- tion about the families and to place the diaries for the first 7 days. They returned a week later to pick up and check the first diaries and to place the diaries for the second 7 days. A third visit was made the following week to collect the second diaries. 2 Actually, as it turned out, the placement of diaries was spread reasonably uniformly over the days of the week with a sharp reduction only on Saturday (and no placement on Sunday). The instructions called for asking respondents to start the recordkeeping on the day following placement. However, where diaries were placed early on Monday, it was permissible to start the diaries that same day. 46 47 product was an average index number for each of the 14 reporting days. The entire procedure could, of course, be carried out for any subgroup, such as an expenditure class, a population group, etc., to the extent that a sufficient number of observations were available. Another problem in constructing those measures arose because of the survey procedures specified in cases where the diaries were not entirely completed by respondents. For the 15 percent of families where the respondents made no entries at all in the diaries (see appendix B), interviewers were instructed to obtain the expenditure information for the previous week by interview and enter it in the record books. However, in order to avoid unnecessary complications— such as the need to determine the precise day on which particular expenditures were made— the instructions were to record the information on the pages provided for the first reporting day for that week and to go beyond that point only if additional space was required. The outcome is that estimates derived from the information on the first recording pages would exaggerate the actual expenditures for the first reporting day each week unless some way could be found to identify and correct for the excess interviewer entries. The course followed was to bypass this problem by com- puting the estimates for individual reporting days for three categories: all diaries, those completed entirely or partly by respondents, and those completed entirely by respondents. The final category, diaries completed entirely by respondents, which accounted for about 75 percent of the total, should theore- tically be unaffected by a first-day exaggeration attributable to interviewer entries. 3 The effect of exaggeration due to the interviewer supplemen- tation discussed earlier is clearly evident in the first-day indexes for all diaries combined, 5 but the values are impressively high even for those completed entirely by respondents, which should be unaffected by this distortion. Moreover, the pattern holds for each of the broad classes of expenditures. The higher than average levels for the next two reporting days would also be consistent with various of the theories as is the generally downward movement of the indexes for the remainder of the first week. The revival at the beginning of the second week can be attributed partly to the stimulus provided by the interviewer's visit at that point to pick up the first diaries and place the second ones. The fact that the initial second-week indexes are considerably lower than those at the start of the first week may be indicative of some reduction in the extent of telescoping. This could occur because the first week's record- keeping would provide some element of "bounding," that is, a point of reference for starting the second week diaries. The pattern for the remainder of the second week appears to repre- sent some leveling off at a somewhat lower plateau than in the initial week. These differences among days do not, per se, indicate which levels are the more accurate. If the levels for the first reporting day were sustained, however, the survey estimates would exceed expenditures data from independent sources by a wide margin. On the other hand, the general deficiency in the overall survey results, in relation to the independent sources, would be accentuated if levels comparable to those for the latter days each week prevailed thoughout. Reporting Day Differences A comparison of expenditure levels for each of the 14 reporting days, in index number form, is presented in table D-1. Besides the different computations by degree of respondent completion of the diaries, separate indexes are shown for total food expenditures for home consumption, all other expenditures combined, and a group of selected expenditures other than food for home use. This latter category comprises products and services for which the diary was initially regarded as the primary source, mainly the numerous smaller items of expenditure. Excluded are categories such as durable goods, housing costs, health outlays, clothing expenditures, insurance, etc., for which principal dependence was placed on the quarterly panel. The results in table D-1 confirm the general pattern which has been predicted on the basis of previous experience in diarykeeping but on which little direct evidence has heretofore been available. 4 Overshadowing everything else, of course, are the strikingly higher levels exhibited for the first reporting day. 3 The intermediate category (combining diaries filled entirely and those filled partly by respondents) was included in the computations initially in order to increase the sample of observations. However, as it turned out, the "partial" group added only about 10 percent to the total number of cases (see appendix B). 4 Some limited information of this kind had been developed during the experimental stages of the British Family Expenditure Survey. See Kemsley, W.F.F. and Nicholson, J. L., "Some Experiments In Methods of Conducting Consumer Expenditures Surveys", Journal of the Royal Statistical Society, Series A, Vol. 123, Part 3 (1960), pp. 307-328. Differences for Specific Categories The samples are not sufficiently large to compute reliable day-by-day indexes for specific expenditure categories. How- ever, some insights are obtainable from various 7-day com- posites derived in order to explore possible alternate record- keeping periods. Among those developed for this purpose were composites based on reported expenditures for the 3rd through the 9th days of recordkeeping and those comprising outlays for the 6th through the 12th days. The ratios of expenditures for the first 7 days to those for the 3rd through the 9th days— to the extent that they exceed unity— will largely reflect the excess reporting in the early days of recordkeeping. On the other hand, ratios of the expenditures for the second week (8th through 14th days) to those for the 6th through the 12th days) to the degree that they are below unity— will indicate the extent of dropoff in reporting toward the very end of the cycle. These two sets of ratios are presented in table D-2 for those categories of products and services for which the diary was considered to be the primary source, that is, food products and small nonfood expenditure items. As is evident, the first set of ratios (first week to 3rd-9th day expenditures) are well above unity with only a couple of exceptions. The variations among these ratios appear to be largely random in nature and without any substantive meaning. The main inference is that the levels 5 The same kind of deviation affects the 8th day, or the first day of the second week. 48 for certainly the first day and possibly the next day or two appear to be above average for virtually all expenditure categories. In the case of the ratios of second week to 6th-12th day expenditures, in contrast, the numbers are not significantly different from unity except in a few instances. Thus, although there seems to be a general dropoff in the course of each week, the detailed data essentially confirm the overall results, that is, that the pattern of reporting is much the same at the end of the second as at the close of the first week. Even if the most likely explanations— some element of telescoping in the early days and some dropoff in reporting later on— are accepted for purposes of discussion, no simple solution emerges. One approach which has been suggested from time to time, but rarely used, is to extend the recordkeeping period by a few days but to eliminate the reported expenditures for the earliest and latest days at the tabulation stage. For example, if the design called for 2 weeks of information, a 17-day recordkeeping period might be specified with the intention of later excluding the entries for the first two days and the final day. Conclusions Although the patterns are rather evident, the conclusions to be drawn from these data are far from clear. One of the popular theories surrounding these differences, however, may now be largely discounted, that is, the possibility of a significant albeit temporary alteration in buying habits because a diary was being kept. The only way of reconciling that theory with the empirical evidence would be to suggest that respondents went on a reckless spending spree during the first day or two or recordkeeping. Since virtually all products and services exhibited large initial excesses, it would have been necessary, furthermore, for them to have purchased virtually everything on the market, hardly a physical let alone an economic possibility for our hardpressed homemakers and other consumers. This does not mean that keeping a diary had no effect at all on consumer behavior, only that is was not likely a major influence. The hypothesis concerning telescoping of expenditures which actually occurred prior to the reporting period is definitely supportable by the data. So is the notion of some letdown in recordkeeping in the course of any given week. Other, more mechanical explanations come to mind as well. For example, it is conceivable that some respondents may not have recorded information on the separate pages provided for each day, but rather started at the beginning of the record book and filled only as many pages as needed to accomodate their number of entries. A physical examination of a sample of record books would be needed to determine the extent of possible aberrations of this kind. The problem with this kind of truncated procedure is that it could result in a further understatement in expenditures. As reported earlier, there is a general tendency to understate expenditures in household surveys for many, if not most, classes of products and services. Although overall food expenditures were fairly well reported, there were some evident deficiencies for certain items, mainly food staples (flour, sugar, etc.). Moreover, underages were clearly apparent for various nonfood expenditures, especially those likely to be made by family members other than the homemaker. Exclusion of the high-level early reporting days would obviously serve to heighten these disparities. Another possible technique would be for the interviewer to obtain expenditure information at the time of the first visit by recall for perhaps the previous day or two. This information could then be left with the respondent to serve as a "bounding" device in filling the diaries or, conversely, be used later at the data processing stage to eliminate items which were obviously reported a second time because of telescoping. The outcome, however, would likely be the same as for a truncated procedure, that is, some increase in net underreporting of expenditures. A more sanguine approach would be to retain the high-level days, telescoping or no, on the justification that any exaggera- tions arising from that source merely offset omissions resulting from unavoidable lapses in recordkeeping. Of course, the best solution of all would be to attempt to overcome reporting deficiencies through some of the procedural modifications suggested earlier, or other methodological improvements and, thereby, permit the luxury of controlling telescoping by various of the devices mentioned. 49 Table D-l. Indexes of Relative Expenditures for Each Day of the 14-Day Diary Keeping Period, by Broad Class of Expenditures and Degree of Respondent Completion of Diaries: 1972-73 (Average for all days = 100. Computations based on families who completed diaries for both weeks, roughly 21,000 cases or 95 percent of all cooperating families in the 2 years combined. The data on daily expenditures were standardized to correct for imbalances in diary placement on the various calendar days. The more valid comparisons among days would be for the columns labelled "Diaries completed entirely by respondents". See text). Reporting day Food expenditures for home use All diaries Diaries completed entirely or partly by respondent Diaries completed entirely by respondent All other expenditures All diaries Diaries completed entirely or partly by respondent Diaries completed entirely by respondent Selected other expenditures 1 All diaries Diaries completed entirely or partly by respondent Diaries completed entirely by respondent All days. Week 1 1st day. . 2d day. . . 3d day. . . 4th day. . 5th day.. 6th day. . 7th day. . Week 2 8th day.. 9th day. . 10th day. 11th day. 12th day. 13th day. 14th day. 100 187 100 95 88 92 84 82 151 91 87 90 85 83 86 100 153 106 101 94 100 90 86 119 96 90 96 91 87 91 100 148 106 102 94 101 92 87 116 96 91 98 91 87 90 100 196 105 99 88 81 84 8 b 129 86 85 86 85 c >6 93 100 165 111 105 93 87 91 92 100 8 l > 8 8 91 88 102 98 100 160 110 105 92 88 92 90 98 90 8'J 93 87 106 100 100 177 103 98 91 90 87 85 136 94 91 8 8 85 86 8') 100 146 108 105 97 97 93 90 109 97 94 4 1 89 91 93 100 139 108 106 99 '18 93 92 108 '18 95 91 90 90 94 'Represents expenditure categories (other than food for home use) for which the diary was considered to be the primary source, mainly the smaller items of expenditure (see table D-2 for listing.) Comparisons among days are probably more meaningful for this group than for the broader category "All other expenditures" which includes many high variance items for which the quarterly panel rather than the diary was the intended source. 50 Table D-2. Ratio of Expenditures for the First and Second Weeks of Diary Keeping to Other 7-Day Composites During the 2-Week Record Keeping Period, for Food Products and Selected Nonfood Items: 1972-73 (Based on families who completed their diaries entirely or partly) Category Ratio of expenditures for 1st week (1st 7 days) to those for the 3d through the 9th days combined Ratio of expenditures for 2d week (8th-14th days) to those for the 6th through the 12th days combined Food purchases for home use . Cereal products Bakery products . . . „ . . . Meat, poultry, and fish . . Dairy products . Fruits, fresh and processed . ... Vegetables, fresh and processed Sugar and sweets . . Carbonated beverages Other nonalcoholic beverages All other foods Purchased meals and snacks Small nonfood expenditure items Personal care products (toiletries, etc.) Personal care services (beauty shops, barbers, etc. ) Household supplies (laundry-cleaning supplies, paper products, etc.) Household help Laundry, dry cleaning, etc Gasoline and oil Vehicle repair and maintenance Public transportation Drugs and medicines Reading materials Sporting equipment and toys Admission fees (sporting events, movies, etc.) Tobacco products Alcoholic beverages 1.061 1.074 1.044 1.060 1.060 1.076 1.064 1.060 1.110 1.077 1.048 1.026 1.067 1.094 1.09 2 1.082 .911 1.113 1.077 1.063 .992 1.192 1.077 1.108 1.072 1.065 1.070 1.005 .997 .999 1.005 1.007 .998 .999 1.011 .992 .997 1.015 .985 1.004 1.028 .995 .980 .993 1.056 1.011 1.069 1.003 1.020 1.017 1.025 1.008 1.020 1.008 JJU.S. GOVERNMENT PRINTING OFFICE: 19 7 9-281-049/1078 working papers * The titles listed below are available in printed copies or on microfiche. The date in parentheses following the title indicates year of publication. Copies of recent papers may be reviewed and/or ordered at district off ices of the Department of Commerce located in principal cities throughout the United States. Additional information, including prices, may be obtained by writing to Subscriber Services (ASD), Bureau of the Census, Washington, D.C. 20233. No. 1 No. 2 No. 3 No. 4 No. 5 No. 6 No. 7 Raw Materials in the United States Economy: 1900-52 (Preliminary) (1954) The Role of the 1954 Census of Manufactures in Overcoming Problems of Industry Data (1956) Tract Data Compared for a 25-Percent Sample and a Complete Census (1956) Sampling in the 1950 Census of Population and Housing (1956) Occupational Trends in the United States: 1900 to 1950(1958) Raw Materials in the United States Economy (1964) Papers Presented at the Census Tract Conference, December 29, 1958 (1959) No. 8. Materials on the Preparation and Conduct of the U.S.S.R. All-Union Population Census of 1959 (1959) No. 9. Historical Comparability of Census Manufactures Industries: 1929-1958 (1959) No. 10. Papers Presented at the Census Tract Conference, December 29, 1959 (1960) No. 11. Papers Presented at the Census Tract Conference, August 25, 1960(1960) No. 12. Papers Presented at the Census Tract Conference, August 28, 1961 (1962) No. 13. Papers Presented at the Census Tract Conference, September8, 1962(1962) No. 14. The Spectral Analysis of Economic Time Series (1963) No. 15. Methodology and Scores of Socioeconomic Status (1963) No. 16. Procedural Report on the 1960 Censuses of Population and Housing (1963) No. 17. Papers Presented at the Census Tract Conference, September 5, 1963(1964) No. 18. The Measurement of Performance Potential in Manu- facturing Establishments (1965) No. 19. Tests of Use of Post Office Resources to Improve Coverage of Censuses (1965) No. 20. Industry Classification and Sector Measures of In- dustrial Production (1965) No. 21. A Spectral Study of "Overadjustment" for Sea- sonality (1965) No. 22. Papers Presented at the Census Tract Conference, December 29, 1964 (1965) No. 23. Spectral Analysis and Parametric Methods for Sea- sonal Adjustment of Economic Time Series (1965) No. 24. Self-Enumeration as a Method for the 1970 Census of Housing (1966) No. 25. Measuring the Quality of Housing (1967) No. 26. Changes in the Structure of Manufacturing Employ- ment (1968) No. 27. Methodology of Consumer Expenditures Survey (1968) No. 28. Metropolitan Area Definition (1968) No. 29. Survey Applications of Social Psychological Ques- tions (1969) No. 30. Raw Materials in the United States Economy: 1900-1966(1970) No. 31. Price Variation in New FHA Houses: 1959-1961 (1971) No. 32. Pretests and Dress Rehearsals of the 1970 Census of Population and Housing: A Procedural History (1972) No. 33. An Estimate of A Quasi-Stable Age-Sex Distribution for Ghana in 1960(1972) No. 34. Investigation of Census Bureau Interviewer Char- acteristics, Performance, and Attitudes: A Summary (1972) No. 35. Raw Materials in the United States Economy: 1900-1960(1972) No. 36. Response Variance in the Current Population Survey (1972) No. 37. Who's Home When (1973) No. 38. Economic Censuses of the United States: Historical Development (1973) No. 39. Population and Housing Inquiries in U.S. Decennial Censuses, 1790 -1970 (1973) No. 40. Studres in Occupational and Industrial Classification (1974) No. 41. Information Privacy and Statistics: A Topical Bibli- ography (1978) Superintendent of Documents U.S. Government Printing Office Washington, D.C. 20402 Official Business PENN STATE UNIVERSITY LIBRARIES A000Q7inb2321 ~««.x • US.MAIL COM-202 V^ ^ M — technical papers The titles listed below are available in printed copies or on microfiche. The date in parentheses following the title indicates year of publication. Copies of recent papers may be reviewed and/or ordered at district offices of the Department of Commerce located in principal cities throughout the United States. Additional infor- mation, including prices, may be obtained by writing to Subscriber Services (ASD), Bureau of the Census, Washington, D.C. 20233. No. 1. No. 2. No. 3 No. 4 No. 5 No. 6 No. 7 No. 8. No. 9. No. 10. No. 11. No. 12. No. 13. No. 14. No. 15 No. 16 No. 17 No. 18 No. 19 No. 20 No. 21 No. 22 The Sample Survey of Retail Stores: A Report on Methodology (1953) Description of the Sample for the Monthly Retail Trade Report (unnumbered revision) (1955) Accuracy of Census Statistics With and Without Sampling (1960) Farm Population, 1880-1950 (1960) The Post Enumeration Survey: 1950 (1960) Tests and Revisions of Bureau of the Census Methods of Seasonal Adjustments (1961) The Current Population Survey Reinterview Program, Some Notes and Discussion (1963) The Current Population Survey, A Report on Methodology (1963) Trends in the Income of Families and Persons in the United States: 1947-1960(1963) Reconciliation of the 1958 Census of Retail Trade with the Monthly Retail Trade Report (1963) Population Trends in the United States: 1900 to 1960(1964) Response Errors in Collection of Expenditures Data by Household Interviews: An Experimental Study (1965) Estimating Trading-Day Variation in Monthly Economic Time Series (1965) Sampling Application in Censuses of Population and Housing (1965) The International Standard Industrial Classification and the U.S. Standard Industrial Classification (1965) The X-1 1 Variant of the Census Method II Seasonal Adjustment Program (1967) Present Value of Estimated Lifetime Earnings (1967) Trends in the Income of Families and Persons in the United States: 1947 to 1964 (1967) Changes Between the 1950 and 1960 Occupation and Industry Classification (1968) The Current Population Survey Reinterview Program, January 1961 through December 1966 (1968) Correlation Between United States and International Standard Industrial Classifications (1969) Characteristics of America's Engineers and Scientists: i960 and 1962 (1969) Measures of Overlap of Income Distributions of White and Negro Families in the United States (1970) No. 23. The Position of United States in World Commodity Exports in 1968 (1970) No. 24. The Annual Survey of Manufactures: A Report on Methodology (1971) No. 25. Demographic Computer Library (1971) No. 26. 1970 Occupation and Industry Classification System Showing Sources (1972) No. 27. Sampling Applications of the 1970 Census Publication, Maps, and Public Use Summary Files (1972) No. 28. From the Old to the New 1972 SIC for Establish- ments (1972) No. 29. Scientific and Technological Development Activities of the Census (1973) No. 30. Census County Division, Past and Future (1973) No. 31. Consistency of Reporting Ethnic Origin in the Cur- rent Population Survey (1974) No. 32. Standards for Discussion and Presentation of Errors in Data (1974) No. 33. Characteristics of Persons in Engineering and Scientific Occupations: 1972 (1974) No. 34. Indexes to Survey Methodology Literature (1974) No. 35. Family (Money) Income: 1947 to 1971 : Summarizing Twenty-Five Years of a Summary Statistic (1974) No. 36. Census Bureau Programs to Measure Consumer Pur- chase Expectations: 1959-1973 (1974) No. 37. The Census Bureau: A Numerator and Denominator for Measuring Change (1975) No. 38. Comparison of Persons of Spanish Surname and Per- sons of Spanish Origin in the United States (1975) No. 39. Guic'e for Local Area Population Projections (1977) No. 40. The Current Population Survey: Design and Method- ology (1978) No. 41. An Evaluation of 1970 Census Occupational Classi- fication (1978) No. 42. 1976 Survey of Institutionalized Persons: Methods and Procedures (1978) No. 43. Graphic Presentation of Statistical Information (1978) No. 44. Standard Statistical Establishment Program (1978) No. 45. The 1972-73 U.S. Consumer Expenditure Survey: A Preliminary Evaluation (1978) No. 46. Reevaluation of the 1972-73 U.S. Consumer Expend- iture Survey: A Further Examination Based on Re- vised Estimates of Personal Consumer Expenditures.