All Times ET
Program is Subject to Change
Evaluating the Link-Relative Estimator Used by a Monthly Economic Indicator Survey (308017)*Laura Terese Bechtel, US Census Bureau
Nicole Czaplicki, U.S. Census Bureau
Keywords: Link-relative estimator, synthetic estimator, weighting estimator, advance survey
The Economic Directorate within the U.S. Census Bureau is currently in the process of researching potential methodology updates that can be implemented to minimize the revisions between the Advance Monthly Retail Trade and Food Services Survey (MARTS) and the Monthly Retail Trade and Food Services Survey (MRTS) sales estimates. The MARTS estimates serve as a forecast of the MRTS estimates by releasing early total sales estimates approximately two work weeks after the reference month. One month later, MRTS releases its preliminary estimate for the same reference month. Large revisions between the two corresponding estimates are undesirable, especially when a revision reverses the direction of the month-to-month change. In this paper we present an evaluation of the link-relative estimator currently used by MARTS. The link-relative estimator is a synthetic estimator that multiplies a benchmark total from the prior month by an estimate of the current month to prior month change. We examine several weighting estimators in addition to modifications to the currently used link-relative estimator. For each considered estimator, we assess the predictive effectiveness of the advance (MARTS-based) on the preliminary (MRTS-based) estimates in terms of levels, month-to-month changes, and year-to-year changes. Additionally, we present mean absolute revisions and mean squared revisions to examine the magnitude of revisions. We use these statistics to compare each estimator and make a recommendation accordingly.