CHAPTER XXXI Monthly Retail Sales Abstract.......................................XXXI-3 Introduction...................................XXXI-5 Reliability of Data............................XXXI-9 Adjustment for Seasonal and Other Variations...XXXI-11 Glossary.......................................XXXI-13 File Specifications............................XXXI-15 Database Structures............................XXXI-17 User Notes.....................................XXXI-19 MONTHLY RETAIL SALES Abstract CONTENTS Citation---------------------XXXI 3 Type of File-----------------XXXI 3 Universe Description---------XXXI 3 Subject-Matter Description---XXXI 3 Geographic Coverage----------XXXI 3 Technical Description--------XXXI 3 Reference Materials----------XXXI 4 Related Printed Reports------XXXI 4 Related Machine-Readable Data Files XXXI 4 Availability-----------------XXXI 4 CITATION Monthly Retail Sales and Inventories, 1967 to 1992, on CD-ROM [machine- readable data files] / prepared by the Bureau of the Census. Washington: The Bureau [producer and distributor], 1993. TYPE OF FILE Summary statistics. UNIVERSE DESCRIPTION: The universe of these files is all retail trade establishments with or without payroll. The Monthly Retail Trade estimates are developed from samples representing all sizes of firms and kinds of business in retail trade throughout the nation. The primary component of the sample is the list sample a probability sample selected from the retail employers who made Federal Insurance Contributions Act (FICA) payments. The secondary component of the monthly retail trade sample is the area sample a probability sample of land segments across the United States. The area sample covers all retailers not subjected to sampling in the list sample. SUBJECT-MATTER DESCRIPTION These files present monthly estimates of retail sales and inventories for selected kind-of-business categories and geographic areas. Estimates for the United States as a whole are also shown with adjustments for seasonal variation and trading-day and holiday differences. GEOGRAPHIC COVERAGE The most detailed estimates by kind of business are shown at the national level. Less detail is shown for 4 geographic regions, 9 divisions, 19 selected States (CA, FL, IL, IN, LA, MD, MA, MI, MN, MO, NJ, NY, NC, OH, PA, TN, TX, VA, WI), 25 MSAs and CMSAs, 15 PMSAs and 4 cities. TECHNICAL DESCRIPTION: File Structure: dBase III+ RSALES92 Monthly retail sales estimates, 1967 to 1992 1. File sequence: By table (see below) by geographic area by kind of business by year. (Each record includes 12 months of data and a total.) Table 1 U.S., unadjusted and adjusted, 1967 to 1992 Table 2 U.S., Group II companies (selected with certainty) only, unadjusted and adjusted, 1978 to 1992 Table 3 Regions, unadjusted, 1978 to 1992 Table 4 Divisions, unadjusted, 1978 to 1992 Table 5 Selected States, unadjusted, 1978 to 1992 Table 6 Selected metropolitan areas and cities, unadjusted, 1978 to 1992 (some areas are shown only from 1987 to 1992) 2. Available indexes: a. RSALTA92: First record for each table (type of geography) b. RSALST92: First record for each State, region, division c. RSALMA92: First record for each MSA, PMSA and city d. RSALCI92: First record for each city e. RSALKB92: By kind-of-business by State by MSA/PMSA 3. Linkable label files: a. RMGEO (Area names) b. RMKB (Kind-of-business titles) c. RMMISC (Titles for table, series_id, and year) REFERENCE MATERIALS 1987 Economic Censuses CD-ROM 1E, Technical Documentation. This documentation has general information, glossaries, record layouts for all files, and other reference material. One copy accompanies each CD-ROM order. Additional copies are available for $10 from Customer Services, Bureau of the Census, Washington, DC 20233. Guide to the 1987 Economic Censuses and Related Statistics. Describes the scope, coverage, classification systems, data items, and data products for each of the economic censuses and related surveys. Data comparability and uses are also discussed. Single copies are free from Customer Services, Bureau of the Census, Washington, DC 20233. Standard Industrial Classification Manual, 1987. Prepared by the Executive Office of the President, Office of Management and Budget, and available from the Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402. Price is $24. RELATED PRINTED REPORTS: Revised Monthly Retail Sales and Inventories, January 1983 through December 1992. This report does not include years prior to 1983, as are included on CD-ROM, but does include statistics on inventories and inventory/sales ratios not included on CD. For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402. For information about availability and prices of individual reports, contact Customer Services, Data User Services Division, Bureau of the Census, Washington, DC 20233. RELATED MACHINE-READABLE DATA FILES Monthly Retail Sales and Inventories on Diskette. Includes the same sales data as are included on CD-ROM, plus statistics on inventories and inventory/sales ratios and additional unpublished detail on sales. The diskettes are updated each spring to include the latest full year. Price $110 from Customer Services, Data User Services Division, Bureau of the Census, Washington, DC 20233. Census of Retail Trade on CD-ROM, 1987: Selected Statistics. The monthly retail sales figures are most comparable to the 1987 census statistics in files RC87N1US and RC87SS3, since they include sales by both employers and nonemployers. Most other census files include only employers (establishments with payroll). (See chapter XXVI.) AVAILABILITY These files are included on Economic Census CD-ROM 1E (price $150). MONTHLY RETAIL SALES Introduction CONTENTS Description of Sales and Inventories Samples---------XXXI 5 Benchmarking-----------------XXXI 7 Estimation Procedures--------XXXI 7 Percent Change---------------XXXI 8 Unpublished Data-------------XXXI 8 The Monthly Retail Trade report is produced by the Bureau of the Census to provide (1) estimates of sales of retail stores by kind of business for the United States and geographic areas- regions, divisions, selected States, metropolitan areas, and cities (2) national estimates of end-of-month inventories of retail establishments by kind of business. The retail sales and inventories estimates in this report are developed from samples representing all sizes of firms and kinds of business in retail trade throughout the nation. The samples have been designed so that estimates can be evaluated in terms of their sampling variability. DESCRIPTION OF SALES AND INVENTORIES SAMPLES Components of Sales Sample Primary component. The primary component is the list sample a probability sample selected from the retail employers (Standard Industrial Classification (SIC) 52-59) contained in the Census Bureau's Standard Statistical Establishment List (SSEL) which effectively covers all employers who made social security payments for their employees under the Federal Insurance Contributions Act (FICA) during 1989. The list sample originally was drawn from the SSEL as updated to December 31, 1989. The initial SSEL consisted of two lists. The first list was made up of all employer identification (EI) numbers (assigned in connection with FICA) with reported payroll in at least one quarter of 1989. The second list consisted of all establishments of known multi-establishment companies as of December 31, 1989. These lists contained information on sales/receipts, payroll, employment, name and address, kind-of- business classification, etc. Before the sampling frame was available, a study was made of the universe of retail businesses using files from the 1987 Census of Retail Trade. This study determined the stratification of the sampling units based on 1987 sales and kind of business, and also determined the optimal allocation of the sample necessary to meet specified sampling variability objectives for sales estimates for different kind-of- business groups. The primary stratum boundary determined in the study was the certainty cutoff to be used for each kind of business. The cutoff, which ranged from annual sales of $2.5 million to annual sales of $100 million, was particularly important since it also determined the type of sampling unit. Sampling units for the list sample consisted of both companies and EI's. For an EI to be eligible for the initial list sample, the EI had to be active, i.e., had payroll in 1989 and was on the latest available Internal Revenue Service (IRS) mailing list for FICA taxpayers. If a known company had total retail sales (on a 1987 basis) above the cutoff for its major kind of business, the company was selected for the sample with certainty (i.e., probability = 1.0). The company, which might consist of two or more EI's, was then the sampling unit; therefore, any new retail establishments that the company might acquire, even if under new or different EI's, were in the sample with certainty. Single- establishment companies, whether or not selected with certainty, were considered as EI sampling units. All retail companies not selected with certainty were treated on an EI basis; that is, the EI was the sampling unit. The EI's were stratified according to their major kind of business and their estimated sales (on a 1987 basis). Within each stratum, a simple random sample of EI's was selected. The sampling rates for these EI's varied between 1 in 3 and 1 in 1,000. For all EI 'births' after the initial selection, a two-phase selection procedure was used. EI births are new EI numbers recently assigned by the Internal Revenue Service (IRS), with a kind-of-business classification assigned by the Social Security Administration (SSA), and currently on the IRS mailing list for FICA taxpayers. In the first phase, births were stratified by kind of business and size (expected employment or quarterly payroll). A relatively large sample was drawn and canvassed for a more reliable measure of size (sales/receipts in 2 recent months) and a more detailed kind-of-business code, if needed. Using this more reliable information, the births selected in the first phase were subjected to probability-proportional-to- size sampling with overall probabilities equivalent to those used in drawing the initial sample from the 1989 SSEL. Because of the lag in reporting births to the IRS and the SSA and the time needed to accomplish the two-phase birth-selection procedure carried out quarterly, births were actually added to the list sample approximately 12 to 18 months after they began operation. During this period they were represented by the area sample. To be eligible for list sample canvass and tabulation in a given month, a retail noncertainty business must meet both of the following requirements: 1. It must be on the latest available IRS mailing list for the FICA taxpayers. 2. It must have been selected from either the SSEL or the file of employer births. For this purpose, a universe file of all retail businesses ever given a chance for selection is maintained. This file shows, among other things, which EI numbers actually have been selected. For businesses selected into the sample with certainty, the first requirement is changed. These certainty sampling units are not dropped from canvass and tabulation if they are no longer on the IRS mailing list. Rather, these businesses are contacted, and if there are successor businesses, they are added to the survey. This is a tighter form of control used for these larger businesses. The list sample is updated quarterly to account for business births and deaths and represents approximately 94 percent of retail sales. Canvass of this component is primarily by mail. Secondary component. The secondary component is the area sample a probability sample of land segments. All retailers not subjected to sampling in the list sample are represented by similar retail business establishments found in these land segments. Personal enumeration is used for this component. The area sample is a multistage sample. In the first stage, 12 primary sampling units (single counties or small groups of contiguous counties) were selected with certainty, and 46 were drawn with probability-proportional-to-size (population). In succeeding stages, a sample of small land segments was selected in each primary sampling unit (PSU) and was then divided into 6 or 12 area panels to be used in different months on a rotating basis. These land segments contained an average of four retail establishments at the time of selection. The probability of selection of the segments is usually 1 in 1,000 (1 in 2,000 or 1 in 3,000 in the smallest PSU's). Approximately 420 land segments are enumerated each month. All retail establishments in these selected land segments are canvassed. Sufficient information (mainly concerning the EI number) is obtained to determine whether the business has had a chance of selection in the list sample component. If it has no EI number or if its EI number fails to match either the list sample universe or the current IRS FICA mailing list, it is tabulated in the area sample. The area sample thus includes businesses without employees and employers, mainly recent EI births, not represented in the list sample. Inventory Sample The list sample portion of the inventory sample is a subsample of the EI's and companies included in the monthly retail sales survey. All retail establishments included in the area sample component of the sales sample are also included in the inventory sample. Studies similar to those performed for the sales sample were conducted prior to selecting the inventory sample. Those studies determined the stratification (based on sales from the 1987 Census of Retail Trade), allocation (based on inventory from the 1987 Annual Retail Trade Survey), and certainty cutoffs required to meet desired inventory sampling variability constraints for various kind-of-business groups. The initial inventory list sample was selected from the initial sales list sample by first stratifying EI's and companies by kind-of-business groups and estimated sales (on a 1987 basis). To meet reliability constraints, the required number of cases in each size stratum were then selected on a probability- proportional-to-size basis. Some companies and EI's because of their relatively large impact on the inventory estimates, were selected with certainty in order to reduce the sampling variability associated with the estimates. After the initial inventory sample selection, all births selected into the sales sample were subjected to probability- proportional-to-size sampling with overall probabilities equivalent to those used in drawing the initial inventory sample. BENCHMARKING Retail sales in this report reflect the results of a benchmarking operation which developed revised monthly sales estimates for the period January 1989 through March 1992. The benchmarking process equated the 12-month sum of the estimates for 1990 to the annual sales estimate derived from the Annual Retail Trade Survey while minimizing the differences between the month-to-month trends of the estimates from the original and the revised series. Estimates of the original series were derived by using the composite estimation procedures described in the ``estimation procedures'' section. For those kinds-of-business subjected to a change in kind-of-business definition, the estimates were revised back to the beginning of the series. Estimates prior to January 1989 for those kinds of business with no change in definition are unchanged by this year's process. End-of-month inventory estimates were revised for the period January 1980 through March 1992 with the exception of the furniture group stores and food group stores, which were revised back to January 1989, and building materials group stores and the apparel and accessory stores, which were revised back to January 1990. The benchmarking process equated the December 1990 estimates to the end-of-year inventory estimate derived from the Annual Retail Trade Survey while minimizing changes to the month- to-month trends of the original inventory series. Estimates of the originally tabulated inventory series were also derived by using the composite estimation procedures. For kinds-of-business subjected to changes in definition, inventory estimates were revised back exactly as the sales estimates above. Carry-forward factors for sales and inventories were derived by dividing the revised December 1990 estimates by the original composite estimates for December 1990. Through the use of these carry-forward factors, the effects of the revision process were applied to composite sales and inventory estimates subsequent to December 1990. These factors were derived at the most detailed kind-of-business levels and will be held constant until the next benchmarking process. Aggregate level estimates were derived by summing appropriate detail level estimates, thereby assuring additivity. Ratios of the revised-to-original composite sales estimates for all stores in each kind of business were developed for each month subsequent to January 1989. These ratios were multiplied by the composite Group II and geographic area sales estimates to derive benchmarked levels of sales for those series. Relationship of Group I and II Components of Sample The Group II component consists of companies that had 11 or more retail establishments as of December 31, 1989, and qualified for certainty selection. To qualify for certainty, total annual sales of these companies (on a 1989 basis) had to exceed specified dollar volume cutoffs, that varied by kind of business. The certainty cutoff ranged from annual sales of $2.5 million to annual sales of $100 million. The Group I component consists of all other retail establishments. Although separate tabulations are made for the Group I and Group II components of the list sample, only the data developed for the Group II component are published separately (table 3). Group I estimates are included in the ``retail trade'' figures shown in this report. ESTIMATION PROCEDURES Rotating sample panels are used for both the list and area sample components of the sales and inventory samples (except for very large retail businesses which are selected with certainty and canvassed each month). For sampling units in the rotating panels, 2 months of data are obtained at each enumeration. For example, February list sample rotating panel cases and February land segment cases each report data for February (the current month) and January (the previous month). In the following month, different rotating panels (the March panels) report figures for March and February. Three list sample panels and 6 or 12 area sample panels are used. This permits the use of a composite estimation procedure that provides estimates more reliable than those that would be obtained from a completely fixed sample of about twice the size and also reduces the reporting burden for selected businesses. The first estimate issued each month, based on the full sample, is called the preliminary estimate. The preliminary composite estimate for the current month at each three-digit SIC level for sales and at each major kind-of-business grouping for inventory is a weighted average of two estimates: (1) the current month unbiased estimate (weight 0.25) and (2) a ratio estimate (weight 0.75) obtained by multiplying the current-to-previous month ratios developed from the current month report forms by the preliminary composite estimate for the previous month. A final estimate is developed 1 month later using information available from the following month's enumeration. The final composite estimate for individual kind-of-business levels is also a weighted average of two estimates the preliminary composite estimate for a given month (weight 0.8) and the unbiased estimate (weight 0.2) for the same month as obtained from the next month's reporting panel. The aggregate level estimates are obtained by adding the individual kind-of-business components. The preliminary composite estimates provide a measure of the dollar volume level that is considerably more reliable than the unbiased estimates, and the final composite estimates are generally more reliable than the preliminary composite estimates. The most reliable indication of month-to-month trend is the ratio of the current month's preliminary composite estimate divided by the previous month's final composite estimate. This is true because the numerator and denominator of this ratio are more highly correlated than would be the case if the numerator and denominator were two preliminary or two final estimates. The preliminary-final ratio, therefore, has less sampling variability than one based on two preliminary or two final estimates. Another advantage gained by using the rotating panel system is that the very large non-certainty cases in each enumeration are identified and recanvassed in the following month or months. This makes it possible to reduce their weights without biasing the results and thus lessen their influence on the estimates. This increases the reliability of the estimates, particularly the measure of month-to-month change. PERCENT CHANGE Month-to-month and year-to-year comparisons for sales and inventory shown in this report are based on the total dollar volume of receipts or value of inventory for each of the periods being compared. These trends thus include the effect of stores starting operation or ceasing operation and are not limited to ``identical stores,'' i.e., those in operation in both of the periods being compared. The percent changes shown in table 2 for sales for kinds of business not marked with an asterisk are derived directly from the dollar volume estimates in table 1. The sampling variability of dollar estimates for those kinds of business with an asterisk is relatively large. Therefore, such estimates are not published in table 1 for sales. However, the sampling variability of the percent changes derived from such estimates are relatively small and are, therefore, published in table 2 for sales. UNPUBLISHED DATA Selected additional data, such as dollar volume estimates for some kinds of business not separately shown in this report, are produced as a byproduct of the regularly published statistics. These additional data have not been included in this publication because of high imputation or sampling variability (relative to the changes from month to month or between other periods), so as to make them potentially misleading. For a fee, the Bureau of the Census will release such figures for individual use, though not for publication. Selected additional data providing greater detail by geographic area and by kind of business are also available for a fee. It should be noted that some unpublished figures can be derived directly from this report by subtracting published data from their respective totals. However, the figures obtained by such subtraction would be subject to the high imputation rates or high sampling variability described previously for unpublished kinds of business. MONTHLY RETAIL SALES Reliability of Data CONTENTS Measures of Sampling Variability XXXI 9 Nonsampling Errors -----------XXXI 10 An estimate based on a sample survey will differ from the population value because of sampling variability and nonsampling error. Sampling variability occurs because observations are made only on a sample, not on the entire population. Nonsampling errors can be attributed to many sources in the reporting, collection and processing of data. The accuracy of a survey result is determined by the joint effects of sampling variability and nonsampling errors. MEASURES OF SAMPLING VARIABILITY The particular sample used in these surveys is one of a large number of all possible samples of the same size that could have been selected using the same design. Estimates derived from the different samples would differ from each other. The average of these estimates would be close to the estimate derived from a complete enumeration of the population. This assumes that a complete enumeration has the same nonsampling error as the sample survey. For sales and inventories, the average of the estimates differs from a complete census because of the composite estimation technique. Disregarding this difference, 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 times 100 divided by the value being estimated. Note that the coefficients of variation are estimates derived from the sample and are also subject to sampling variability. Tables B-1 through B-4 give the estimates of coefficients of variation in percent for recent monthly dollar-volume sales and inventory estimates prepared by the Bureau of the Census and shown in this report. The coefficients of variation presented in the tables permit certain confidence statements about the sample estimates. As noted before, the particular sample used in this survey is 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 estimate would differ from a complete enumeration by less than the corresponding percent for that estimate shown in the sampling variability tables. 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 percentage shown. In about 19 out of 20 (95 percent) of these samples, the estimates would differ from the results of a complete enumeration by less than twice the percentage shown. To illustrate the computations involved in the above confidence statements, as related to dollar volume sales estimates, assume that an estimate of sales published in table 1 is $10,750 million for a particular month and that the median coefficient of variation for this estimate, as given in table B-1, is 1.8 percent, or 0.018. Multiplying $10,750 million by 0.018 yields 194 million. Therefore, a 67-percent confidence interval 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 these intervals would contain the figure obtained from a complete enumeration. Typical practice is to construct 90- or 95-percent confidence intervals. Continuing with the illustration, a 90-percent confidence interval using 1.65 x 0.018 x 10,750 million to yield 320 million. A 95-percent confidence interval is $10,362 million to $11,138 million ($10,750 million plus or minus $388 million). Any comparisons made of estimates, for which measures of sampling variability are available are accompanied by a indication of the 90-percent confidence interval. Thus, a statement such as ``+ .8( + or -1.3)'' indicates a 90-percent confidence interval from -0.5 to + 2.1. If the confidence interval contains zero, it is uncertain whether there is an increase or decrease. NONSAMPLING ERRORS As calculated for this report, the coefficient of variation measures certain nonsampling errors but does not measure any systematic biases in the data. Bias is the difference, averaged over all possible samples of the same size and design, between the estimates and the true value being estimated. Nonsampling errors can be attributed to many sources: (1) inability to obtain information about all cases in the sample, (2) response errors, (3) definitional difficulties, (4) differences in the interpretation of questions, (5) mistakes in recording or coding the data obtained, and (6) other errors of collection, response, coverage, and estimation of missing data. These non-sampling errors also occur in complete censuses. Although no direct measures of the biases have been obtained, precautionary steps were taken in all phases of the collection, processing, and tabulation of the data in an effort to minimize their influence. A major source of bias in the published estimates is due to imputing data for nonrespondents, for late reporters, and for data which fail edit. For all kinds of business combined, imputed sales amount to about 20 percent of the national sales estimates. MONTHLY RETAIL SALES Adjustment for Seasonal and Other Variations Seasonal factors for adjusting data in this publication have been derived by the use of the X-11 ARIMA program developed by Statistics Canada. The program produces factors by using the method described in the X-11 Variant of the Census Method II Seasonal Adjustment Program, U.S. Bureau of the Census Technical Paper No. 15, Revised 1967. The forecasting options were not used as input to the X-11 ARIMA program. This adjustment program develops more accurate computations and diagnostics than the previously used program. Trading-day factors for adjusting retail sales estimates were also derived from the X-11 program. Adjustment of estimates is an approximation based on current and past experiences. Therefore, the adjustments could become less precise if current competitive pressures, changes in consumer buying patterns during holiday periods, and other elements introduce significant changes in seasonal, trading-day, and holiday patterns. A description of trading-day adjustment factors may be found in Estimating Trading-Day Variation in Monthly Economic Time Series, Bureau of the Census Technical Paper No. 12, 1965. Holiday adjustment factors were developed by a method similar to that described in Seasonal Adjustment on Electronic Computers, pp. 356-359, Organization for Economic Cooperation and Development, Paris, 1961. Additional details concerning the adjustment factors may be obtained from the Chief, Business Division, Bureau of the Census, Washington, DC 20233. Concurrent seasonal adjustment uses all available unadjusted estimates (including the latest preliminary and advance estimates) as input to the X- 11 program. When unadjusted advance, preliminary, and final estimates become available, all estimates will be used as input to the X-11 program and new factors will be applied to the advance, (one month before the preliminary) preliminary, and final (one month after the preliminary) estimates and to the previous year estimates that correspond to the advance and preliminary months. MONTHLY RETAIL SALES File Specifications CONTENTS General Data Specifications--XXXI 15 GENERAL DATA SPECIFICATIONS Data Fields. All files are recorded in dBase III+ format. Numeric data fields contain no alphabetic information. Data items are scaled in millions, and this is noted only in the file layouts. Record Sequence and Geographic Coding. Records are arranged with like records together (e.g., U.S. records, followed by State records, followed by MSA records). MSA records do not include State codes. Names for geographic variables are carried in files separate from the data files. Each is indexed, so that geographic names can be accessed in conjunction with the corresponding data values. See the Introduction to this CD-ROM for more information on geographic variables. Kind-of-Business Labels. Records in the file are specific to a particular kind of business, as identified by the KB_CODE code. The term ``kind of business'' is used instead of ``standard industrial classification'' (SIC) because some kind-of- business categories represent either combinations or subdivisions of SIC categories. Titles for each kind of business are available in the file RMKB.dbf on the CD-ROM. Kind-of-business titles can be linked to data files using the dBase SET RELATION command. Index Files. Each data file has one or more index files which can be used to quickly jump to desired records (e.g., using the dBase FIND command) or to change the sequence of access. Available index files are listed in the abstract. Where noted in the abstract that an index is to the ``first record for'' a particular type of area, the index can be used during dBase FINDs, but the index must be shut off with SET INDEX TO in order to get access to records other than the first for each area. 1987 Monthly Retail Sales on CD-ROM 1987 Monthly Retail Sales on CD-ROM DATABASE STRUCTURE File RSALES92: Monthly Retail Sales, 1967 to 1992 --------------------------------------------------------------------------- Field Field Name Type Size Decimal Field Description --------------------------------------------------------------------------- SERIES_ID C 2 0 Series: 1st char:L/D=w leased, 2nd:U=unadjusted,A=adjusted TABLE C 1 0 Table: 1=US, 2=GII,3=region,4=division, 5=state,6=MSA/city ST C 2 0 FIPS State Code (plus R1-R4 regions, D1-D9 divisions) MSA C 4 0 Metropolitan Statistical Area/Consolidated MSA code PMSA C 4 0 Primary Metropolitan Statistical Area code CITY C 1 0 City (1=Chicago, 3=Los Angeles, 4=NewYork,5=Philadelphia) GEO_CODE C 4 0 Consolidated geographic code KB_CODE C 6 0 Retail Kind-of-Business code (500005=Retail trade, total) YEAR C 4 0 Year (1967 to 1992) JANUARY N 6 0 January retail sales ($ millions) FEBRUARY N 6 0 February retail sales ($ millions) MARCH N 6 0 March retail sales ($ millions) APRIL N 6 0 April retail sales ($ millions) MAY N 6 0 May retail sales ($ millions) JUNE N 6 0 June retail sales ($ millions) JULY N 6 0 July retail sales ($ millions) AUGUST N 6 0 August retail sales ($ millions) SEPTEMBER N 6 0 September retail sales ($ millions) OCTOBER N 6 0 October retail sales ($ millions) NOVEMBER N 6 0 November retail sales ($ millions) DECEMBER N 6 0 December retail sales ($ millions) TOTAL N 7 0 Total annual retail sales ($ millions) Record size: 108 MONTHLY RETAIL SALES User Notes This section will contain information relevant to the Monthly Retail Sales on CD-ROM that indicates specific problems with the data, or that becomes available after the file is released. The cover letter to the updated information should be filed behind this page. User Notes will be sent to all users who (1) purchased their file (or technical documentation) from the Census Bureau and (2) returned the coupon following the CD-ROM File Information.