Abstract

Seasonal Adjustment is a statistical tool that provides better understanding of the underlying trends, business cycle, and short-run movements in a time series data by estimating the unobserved components. It provides comprehensive information and timely interpretation of time series data which is vital for economic policy making and evidence-based decisions. Since the Gross Domestic Product (GDP) is one of the most important indicators to measure economic performance, it is imperative to provide short-term trend analyses to prepare economic managers and other users of official statistics on possible implications of economic shocks.
The Philippine Statistics Authority (PSA) releases the Seasonally Adjusted National Accounts (SANA) of the Philippines on a quarterly basis. The PSA is currently using the X-13 ARIMA-SEATS developed by the US Bureau of Census in generating the SANA and applying an indirect approach methodology up to the detailed level of 78 industries under production approach and 114 expenditure items for the demand side. Currently, the SANA is implementing the full review or concurrent adjustment methodology wherein each series is re-evaluated every addition in the data points or observations.
Concurrent adjustment provides the most accurate generated seasonally adjusted data due to the updated models, options, and parameters every new data observation is added. However, this may possibly lead to more frequent revisions in levels and growth rates for its time series due to possible changes in the models and parameters. Generating seasonally adjusted time series may require developing a coherent updated strategy and revision policy that should aim both at the frequency of the revisions to avoid unnecessary or abrupt changes and the quality of the estimates of seasonally adjusted series. To address this, one of the recommendations of the International Monetary Fund (IMF) is to use partial concurrent.
This research study aims to approximate the assumptions of the models, filters, and parameters for at least one year in accordance with the internationally agreed framework. Assumptions will be fixed for at least one year to prevent occurrence of large revisions due to changes in the behavior of a series or unstable seasonality. Optimal models and configurations for the seasonal adjustment will be used for possible best results while minimizing the revisions. However, the limitation of the study is that Interventions Analysis on structural changes in the series and exploration of other methods such as the TRAMO-SEATS were not applied in this paper.