|
Activity Number:
|
108
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 9, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Business and Economics Statistics Section
|
| Abstract - #300902 |
|
Title:
|
Time-varying Trading-day Effects in Seasonal Adjustment of Time Series
|
|
Author(s):
|
Donald E.K. Martin*+ and William R. Bell
|
|
Companies:
|
U.S. Census Bureau and U.S. Census Bureau
|
|
Address:
|
4700 Silver Hill Road, Washington, DC, 20233-9100,
|
|
Keywords:
|
trading day ; time-varying ; RegComponent
|
|
Abstract:
|
Trading-day effects reflect variations in a monthly time series due to the changing composition of months with respect to the numbers of times each day of the week occurs in the month. A relevant question regarding trading-day effects is whether they remain constant over time. This is especially pertinent for retail sales time series in which trading-day effects presumably depend on consumers' shopping patterns and on hours that retail stores are open, two things that have changed over time in the U.S. Seasonal adjustment practitioners sometimes deal with this issue by restricting the length of the series to which the trading-day model is fit. However, in series where the trading day varies through time, information is lost in so doing. We investigate possible time variation in trading-day effects in a large set of Census Bureau time series. We fit a model to the data that allows for stochastic trading-day coefficients that follow random walk models, and with residuals that follow an ARIMA model. As the models are a special case of the general RegComponent model, they are fit using the REGCMPNT program.
|
