Did the Census Bureau modify its Advance Monthly Retail Trade and Food Services Survey non-response or estimation procedure?
No. Ordinarily, no imputation is done for individual nonrespondents. The MARTS survey accounts for missing data when estimating industry totals. The estimate makes use of data from MARTS respondents that report sales data for both the current and prior month and applies the resulting month-to-month change to the prior month sales estimates from the Monthly Retail Trade Survey (MRTS). The MRTS imputes missing data for nonresponding firms based on reported data from firms that are in the same industry and the same size category as the nonrespondents.
The estimation procedure assumes that for a given industry the month-to month change for firms that report is similar to the month-to-month change for firms that do not report. This assumption is supported if sufficient reports from the affected area are received or if the reports received are representative of the area.
Retail sales for the most affected area are ordinarily a small percentage of the national total, so any bias that may be incurred because sales for responding firms are substantially different from sales for nonresponding firms will be negligible in the context of the Nation's total sales. As part of our review, we evaluated the effect of nonresponse on the national totals and felt no change to the methodology was necessary.
For more information, visit the Monthly and Annual Retail Trade website.