1992 CENSUS OF RETAIL TRADE Coverage and Methodology =>STRUCTURE AND METHOD OF ENUMERATION Firms in the 1992 Census of Retail Trade were divided into the nonmail universe and mail universe. The coverage and the method of obtaining census information from each follow: The nonmail universe consists of firms which were not required to file a regular census return and includes: All nonemployers, i.e., all firms with no paid employment during 1992. Sales information for these firms was obtained from administrative records of other Federal agencies. Although consisting of many firms, nonemployers account for less than 5 percent of total retail sales. The census included only those retail nonemployer firms which reported a sales volume of $1,000 or more during 1992. Data on nonemployers are provided in subsequent 1992 Census of Retail Trade reports; Nonemployer Statistics and Selected Statistics. Selected small employers, i.e., single-establishment firms with payroll below a specified cutoff. (The term "employers" refers to all business firms with one or more paid employees at any time during 1992 as shown in the active administrative records of other Federal agencies.) Although the payroll cutoff varied by kind of business, small employers generally included firms with one to four employees and represented about 10 percent of total retail sales of establishments covered in the census. Data on sales, payroll, and employment for employer firms below the payroll cutoff were derived or estimated from administrative records of other Federal agencies except for a sample of small employer firms. This sample was included in the mail universe. The mail universe consists of firms for which information was obtained by means of a mail canvass and includes: Large employers; i.e., all multiestablishment firms and all employer firms above the payroll size cutoff referred to in the nonmail universe section above. Within this category, a report of company organization is conducted periodically to identify firms which operate establishments at more than one location and to obtain information on payroll and mid-March employment at each location. The 1991 Report of Company Organization was used as a coverage check in the census. In the 1992 census, all multiestablishment firms were asked to notify the Census Bureau of any establishments for which a form was not received. Report forms were subsequently provided to the firms for these establishments. A sample of small employer firms referred to in the nonmail universe section above. These firms were sent the census mailing packages containing the appropriate 1992 questionnaire. For the retail trade sector, the overall sample of small employer firms was 20.6 percent which varied by kind of business. =>METHOD OF CLASSIFYING KINDS OF BUSINESS The retail trade classifications for all establishments were based on the Standard Industrial Classification (SIC) Manual (Standard Industrial Classification Manual: 1987. For sale by Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402. Stock No. 041-001-00314-2). However, the method of assigning these classifications, and the level of detail at which establishments were classified, differed between the nonmail and mail universe as follows: Nonemployers were classified on the basis of information obtained from administrative records of other Federal agencies. Selected small employers were classified on the basis of the most current census kind-of-business classification available from one of the Census Bureau's current sample surveys or the 1987 census. Otherwise, the classification was obtained from administrative records of other Federal agencies. If the census or administrative record classifications proved inadequate (none corresponded to a 1992 census classification in the detail required for employers), the firm was sent a brief inquiry requesting information necessary to assign a 1992 census kind-of-business code. Establishments in the mail universe were classified on the basis of their self-designation, answers to questions on sales by merchandise line, and other special inquiries. =>COMPARABILITY OF THE 1987 AND 1992 CENSUSES The 1987 and 1992 censuses were conducted under similar conditions and procedures except for the following: The boundaries of a number of areas for which data are shown in the 1992 census are not the same as in the 1987 census because of annexations; other boundary changes; and redefinitions of metropolitan statistical areas (MSA's), primary metropolitan statistical areas (PMSA's), and consolidated metropolitan statistical areas (CMSA's) (Newly defined metropolitan areas (MA's) were announced by the Office of Management and Budget (OMB) effective June 30, 1993). In addition, data for consolidated cities are included in the 1992 census. Data for special economic urban areas (SEUA's) with 10,000 inhabitants or more in Michigan, Minnesota, New York, and Wisconsin qualified for publication in the 1992 census. This included townships in Michigan and Minnesota and towns in New York and Wisconsin. However, Minnesota did not have any townships that met the publication criteria. =>MERCHANDISE LINE SALES Merchandise line inquiry composition - The merchandise line inquiries on 1992 retail questionnaires were tailored to the kinds of business that would receive them. That is, a broad merchandise line was listed on a particular report form only if it accounted for at least 0.1 percent of sales reported by the kind-of-business categories receiving that form in 1987. Because a complete set of broad merchandise lines was not present on any particular retail questionnaire, respondents sometimes found that part of their sales did not fit any available merchandise line category. When this occurred, they were asked to report these sales on lines for "all other merchandise" and to describe the kind of merchandise represented. Census personnel subsequently attempted to classify this merchandise based on the respondent's description, and to assign the sales to the appropriate merchandise line category. A small percentage of sales could not be classified and is therefore summarized in this report in a category called "unclassified merchandise." The effect of excluding insignificant broad merchandise lines on particular report forms is an understatement of the number of establishments handling each merchandise line and, to a lesser extent, the corresponding sales figure. The magnitude of this understatement for all merchandise lines combined is indicated, at least in part, by the data presented for the "unclassified merchandise" category. Limitations in reporting sales by merchandise lines - Respondents often failed to report sales for detail lines for their particular business and only reported sales for broad lines. This deficiency causes an understatement in the number of outlets for detail merchandise lines and, to a lesser extent, affects the measurement of the sales volume of detail lines. Merchandise line categories by which individual retailers group their sales are not uniform. These categories do not always correspond to the categories established by the census of retail trade for collecting and presenting merchandise line data. In addition, some retailers have little if any recorded information on sales by line of merchandise. A related reporting problem for retail firms is the absence of merchandise line records on an individual establishment basis. Some firms have information available only for the group of stores within a warehouse district or some other grouping used by the firm. In such cases, the firm may estimate sales for individual stores by using the pattern of sales shown by the entire group of stores. The effect of individual reporting differences and the use of approximation is assumed to be negligible in summary tabulation. Differences between 1987 and 1992 merchandise lines - The 1987 census presented data for 41 broad merchandise lines. For 1992, these merchandise lines were restructured into 43 categories. The two new broad lines are: a. Children's wear (ML 240) - Previously reported under "Men's and boys' wear" (ML 200) and "Women's, girls', infants', and toddlers' wear" (ML 220). b. Books (ML 420) - Previously reported under "All other merchandise" (ML 850). Treatment of nonresponse - Reporting was incomplete or inadequate for establishments representing about 16.7 percent of the total dollar volume of establishments with payroll. However, merchandise line data were expanded to estimate the sales of all retail establishments with payroll. The expansion is based on the premise that the merchandise line data for those establishments not reporting this information are similar to merchandise line data for those establishments in the same kind of business that reported this information. Merchandise line data were expanded at the lowest published level of geography and kind-of-business classification based on reported data at that level. If there were no reported merchandise line data for a particular combination of geographic area and kind of business, expansion factors developed at the United States level were used to produce an estimate. These estimates for the most detailed level of geography and kind of business were then summed to produce higher level geography and kind-of-business estimates. Coverage - File 1 of the Merchandise Line Sales Series, "RC92L1.dbf Merchandise Lines by Kind of Business," presents coverage percents for each kind of business shown. Coverage percents indicate the degree to which establishments in each kind of business acceptably reported sales for broad merchandise lines. Coverage was determined by dividing total sales of establishments reporting acceptable data for broad lines (whether or not additional detail lines were reported) by total sales of all establishments classified in that particular kind of business. In addition, coverage percents are presented for selected broad merchandise lines where additional detailed merchandise line information within the broad line was requested. This data indicates the degree to which detail lines within that broad line were acceptably reported. The table presents each broad line under which additional detail line reporting was requested and shows the degree to which businesses that reported the specified broad merchandise line gave the additional detail breakdown. The coverage percent was computed by dividing total sales of establishments reporting detail within the particular broad line by the estimated sales of establishments reporting that broad line within the particular kind of business. Except when precluded by the census disclosure rules (see the "Introduction" menu choice), data are shown for individual kinds of business when the dollar volume of reporting coverage accounts for 60 percent or more of sales after weighting merchandise line sales of the sample of "small employers" described in the Census Coverage and Methodology section above. Measures of sampling variability - Because the merchandise lines estimates are based in part on a sample, exact agreement with the results that would be obtained from a complete census of establishments using the same enumeration procedure should not be expected. However, because each establishment's chance of being selected for the sample was known, it was possible to estimate the sampling variability of the estimates made from the sample. The standard error of the estimate is a measure of the variability among the estimates from all possible samples of the same size and design and, thus, is a measure of the precision with which an estimate from a particular sample approximates the results of a complete enumeration. The coefficient of variation (expressed as a percent) is the standard error of the estimate divided by the value being estimated. Note that measures of sampling variability, such as the standard error or coefficient of variation, are estimated from the sample and are also subject to sampling variability. Tables of coefficients of variation for the broad and detail line estimates are shown in File 3, available upon request. The coefficients of variation presented in File 3 allow certain confidence statements about the sample estimates. The particular sample used in this survey was one of a large number of samples of the same size that could have been selected using the same design. In about 2 out of 3 (67 percent of these samples, the estimates would differ from the results of a complete enumeration by less than the corresponding percentages for that estimate shown in the coefficients of variation file. In about 9 out of 10 (90 percent) of these samples, the estimates would differ from the results of a complete enumeration by less than 1.65 times the percentages shown. In about 19 out of 20 (95 percent) of the samples, the estimates would differ from the results of a complete enumeration by less than two times the percentage shown. To illustrate the computations involved in the above confidence statements as related to the dollar value estimates, assume that an estimate of sales is $10,750 million for a particular broad line, and that the coefficient of variation for this estimate is 1.8 percent or 0.018. First obtain the standard error of the estimate by multiplying the estimate by the coefficient of variation. In this example, $10,750 million times 0.018 yields a standard error of $194 million. The upper bound of the 67-percent confidence interval can then be formed by adding the standard error to the estimate and the lower bound formed by subtracting the standard error from the estimate. Thus, the 67-percent confidence interval for this example is $10,556 million to $10,944 million ($10,750 million plus or minus $194 million). If corresponding confidence intervals were constructed for all possible samples of the same size and design, approximately 2 out of 3 (67 percent) of the intervals would contain the figure obtained from a complete enumeration. Typical practice is to construct a 90- or 95-percent confidence interval. Using the same illustration as above, a 90-percent confidence interval would be $10,430 million to $11,070 million ($10,750 million plus or minus $320 million). Similarly, a 95-percent confidence interval is $10,362 million to $11,138 million ($10,750 million plus or minus $388 million).