Seasonal Adjustment of Chinese Macroeconomic Data for Investment Analysis

Summary

This research summary explains why monthly and year-over-year macroeconomic readings can mislead when economic series contain seasonal and calendar effects. It describes X-13-ARIMA-SEATS as a widely used adjustment approach and emphasizes that seasonal adjustment removes seasonal and calendar components while retaining trend and irregular variation; it does not simply smooth noise. For rolling historical strategy analysis, repeatedly allowing parameters to be selected automatically can make estimates unstable, so the report discusses fixed parameter choices.

The report notes seasonal effects across many commonly used Chinese indicators, with the moving Lunar New Year effect especially relevant for several series. Fiscal revenue and expenditure adjustments are described as unstable, prompting a recommendation to use their raw values in historical strategy tests. An application to small-cap equity premiums found substantial month-to-month differences in forecasts from adjusted and raw data, but no statistically significant overall difference. A separate bond risk-premium exercise found no significant information beyond the contemporaneous yield curve. These findings are bounded to the tested indicators, samples, and modeling choices.

Key ideas

• Month-over-month data can be distorted by seasonality, while year-over-year comparisons may remove only part of the effect and respond slowly to trend changes.
• Seasonal adjustment removes seasonal and calendar components but retains irregular variation, including noise and outliers.
• Repeated automatic parameter selection in rolling strategy analysis may make adjusted series unstable.
• The reported small-cap premium forecasts differed by month between adjusted and raw inputs, but not significantly overall.
• The report found unstable seasonal adjustment for fiscal revenue and expenditure and recommends raw values for those historical tests.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.