The End Point of Time Series: Reporting Informatively on Trends

• Activity Number: 24
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): John Crequer and Sonya McGlone and Giles Reid and Richard Penny*+
• Companies: Statistics New Zealand and Statistics New Zealand and Statistics New Zealand and Statistics New Zealand
• Keywords: time series ; trends ; seasonal adjustment

Abstract

We will discuss the constraints and difficulties faced by National Statistics Offices (NSOs) in the reporting of trend movements and compare the sometimes conflicting approaches they use. We will also demonstrate the sensitivity analysis tool we have developed to help our subject experts to make reliable and informative statements about trend movements. Statistics NZ uses graphs and commentary about the seasonally adjusted and trend components of published time series to highlight stories in our figures. Trend movements have some advantages in comparison to seasonally adjusted movements or year-on-year comparisons. However, the sometimes significant revisions that can be made to the end point of a trend series as new data become available could undermine published interpretations and pose a risk to NSOs' reputations as trusted and unbiased sources. The trade-off between the risks posed by revisions and the early detection of shifts and turning points is made particularly difficult because the subject experts responsible for the releases are not necessarily technical experts in time series, and we rarely have the time to extensively model or analyse every published series.