Economics is just a fancy word for mathematical masturbation ...
>>REVISIONS TO PREVIOUSLY PUBLISHED ESTIMATES
Annual Estimates
Sales, end-of-year inventories, purchases, and accounts receiveables for 2002 were estimated from the 2002 Annual Retail Trade Survey (ARTS), For 2001, 2000, and 1999, revisions were applied to the previously published sales, end-of-year inventories, and purchases. There were no revisions to the previously published 19986 annual sales estimates. However, there were revisions to the previously published 1998 annual purchases and end-of-year inventories. Because these revisions were not significant, we did not revise purchases or end-of-year inventories7 prior to 1998. Accounts receivable estimates were revised for 2001.
Estimates of Monthly Sales
For December 1997 through January 2004, monthly retail and food services sales estimates were input to the benchmarking program. The December 1997 through February 2001 estimates were developed by restating the SIC-based estimates on a NAICS basis. Estimates from March 2001 through January or February 2004 were derived from the NAICS sample. Also, for selected kinds of business, corrections were applied to the estimates. The estimates were revised in a manner that -
constrained the sum of the 12 monthly sales estimates for a particular year to equal the corresponding annual sales estimates (benchmarks) derived for the years 1998 through 2002.
minimized the differences between the month-to-month changes of the input and revised estimates.
used the previously published December 1997 estimate to link the revised estimates to the previously published estimates.
A mathematical result of the benchmarking methodology is that all revised estimates following the end of the last benchmark year (2002) are derived by multiplying the corresponding input estimates by the ratio of benchmarked-to-input estimate for the last month of the last benchmark year. Therefore, for a specified sales series, a ratio of the December 2002 benchmarked estimate to the December 2002 input estimate was computed. Estimates for periods following December 2002 were multiplied by these constant ratios (called carry-forward factors) to derive published sales estimates. The carry-forward factors remain constant until the next benchmarking operation.
Estimates of End-of-Month Inventories
For December 1997 through January 20048, monthly retail end-of-month inventory estimates were input to the benchmarking program. December 1997 through February 2001 estimates were developed by restating the SIC-based estimates on a NAICS basis. Estimates from March 2001 through January 2004 were derived from the NAICS sample. Also, for selected kinds of business, corrections were applied to the estimates. The estimates were revised in a manner that -
equated the December or end-of-year inventory estimates for 1997 through 2002 to the end-of-year inventory estimates derived from the annual survey.
minimized the differences between the month-to-month changes of the input and revised estimates.
used the previously published December 1997 estimate to link the revised estimates to the previously published estimates.
Estimates subsequent to December 2002 were derived by multiplying the input estimates by the ratio of the December 2002 published estimate to the input estimate for December 2002. This ratio is the carry-forward factor for inventory and remains constant for all series until the next benchmarking operation.
REVISIONS OF SALES AND INVENTORIES
The following table shows a comparison of the revised sales and inventories to the previously published estimates for 2003:
2003 Retail and Food Services Sales and Retail Inventories Comparison of the Revised
Estimates to the Previously Published Estimates
(Millions of dollars)
NAICS code
Kind of Business
Revised
Previously published
Percent difference
SALES
Retail and food services sales, total 3,756,688 3,780,616 -0.6
Retail, total 3,399,544 3,421,870 -0.7
441 Motor vehicle and parts dealers 895,703 904,352 -1.0
442 Furniture and home furnishings stores 97,977 100,220 -2.2
443 Electronics and appliances stores 94,561 95,368 -0.8
444 Building mat., garden equip., & supplies dealers 321,134 323,049 -0.6
445 Food and beverage stores 505,933 508,581 -0.5
446 Health and personal care stores 192,191 191,611 0.3
447 Gasoline stations 268,519 263,258 2.0
448 Clothing and clothing access. stores 178,435 178,674 -0.1
451 Sporting goods, hobby, book, & music stores 79,447 81,520 -2.5
452 General merchandise stores 471,078 476,143 -1.1
453 Miscellaneous store retailers 104,865 104,969 -0.1
454 Nonstore retailers 189,701 194,125 -2.3
722 Food Services and drinking places 357,144 358,746 -0.4
INVENTORIES*
Retail, total 446,637 453,551 -1.5
441 Motor vehicle and parts dealers 155,953 162,516 -4.0
442,443 Furniture, home furnishings, electronics, and appliance stores 27,112 28,395 -4.5
444 Building mat., garden equip., & supplies dealers 46,093 45,917 0.4
445 Food and beverage stores 35,495 35,339 0.4
448 Clothing and clothing access. stores 34,411 33,907 1.5
452 General merchandise stores 63,528 63,577 -0.1
* End of December, not adjusted for seasonal variations.
Reasons for Monthly Revisions
There are several reasons for revisions. The main contributors to the revision from the previously published estimates are:
Timing. The respondents have more time to prepare their annual reports than they do for their monthly reports. The annual responses are requested at a time when many firms have already compiled audited book figures for their own use. The timing of the annual survey is such that we are also able to obtain independent verification of the reported data from such sources as a company's annual report. On the other hand, respondents to the monthly survey have just a few weeks to provide reports of their sales and end-of-month inventories. Sometimes these reports are based on incomplete or unaudited records. These include estimates by the respondents to represent their understanding of their business.
Sampling. As described in more detail in Appendix A of this report, the estimates derived from the annual survey are based on a sample that is much larger than the samples used to produce the monthly sales or inventory estimates. Furthermore, the monthly inventory estimates are based on a sample with fewer units than the sample used to produce the monthly sales estimates. This often results in larger revisions for inventory estimates than for sales estimates.
Response. The annual estimates are based on more reported data than are the monthly estimates. The response to the Annual Retail Trade Survey is required by law. This requirement results in a dollar volume response rate for sales above 90 percent. The response to our monthly survey is not mandatory. The dollar volume response rates for the monthly surveys are usually around 80 percent for sales and 72 percent for inventories. The sales and inventories for the non responding firms have trends similar to the responding firms in their respective kind of business.
<< |
