There are massive seasonal patterns in employment data. For example, in July, it is typical for the United States economy to lose over a million jobs. Adjusting for this normal seasonal variation is essential to interpreting month-to-month changes in employment. The approach for this seasonal adjustment that is presently used by the Bureau of Labor Statistics (BLS) puts very heavy weight on the current and last two years of data in assessing what are the typical patterns for each month.
Jonathan Wright of Johns Hopkins University argues in “Unseasonal Seasonals?” that a longer window should be used to estimate seasonal effects. He finds that using a different seasonal filter, known as the 3×9 filter, produces better results and more accurate forecasts. The key difference in the 3×9 filter is that it spreads weight over the most recent six years in estimating seasonal patterns.This makes the seasonal patterns more stable over time than in the current BLS seasonal adjustment method.
We calculate the month-over-month change in total nonfarm payrolls, seasonally adjusted by the 3×9 filter, for the most recent month. The corresponding data as published by the BLS are shown for comparison purposes. According to the alternative seasonal adjustments, the economy gained 289 thousand jobs last month. The Bureau of Labor Statistics reported that the economy gained 288 thousand jobs last month. The discrepancies between the two series are explained in Wright’s BPEA paper.