690 – Topics in Statistical Computing
Statistical Procedures for Reconciling Time Series of Large Systems of Accounts Subject to Low-Frequency Benchmarks
Baoline Chen
Bureau of Economic Analysis
Tommaso Di Fonzo
National Institute for Statistics
Marco Marini
International Monetary Fund
In this paper we study statistical procedures to reconcile large systems of annual time series subject to low-frequency benchmarks (e.g. available every five years). Our aim is to reconcile the preliminary levels of the series such that they (i) are consistent with the low-frequency benchmarks available, (ii) fulfill all the accounting relationships for any given year, and (iii) show movements that are as close as possible to the preliminary information. We propose to solve this kind of problems using a simultaneous least-squares procedure based on the proportional first difference (PFD) criterion, a movement preservation principle proposed by Denton (1971). However, we suggest that a pure proportional adjustment is adopted for series with breaks and high volatility that deteriorate the meaningfulness of growth rates. We apply this procedure for reconciling the 1998-2002 U.S. annual input-output accounts, GDP-by-industry accounts and expenditure-basedGDP, subject to the 1997 and 2002 quinquennial benchmarks and all contemporaneous constraints of the input-output accounts for the in-between years.