![]() ![]() The BLS made manual adjustments first by switching many series from having “multiplicative” to “additive” factors, but also by hardcoding that a particular month for a particular disaggregate was to be treated as an “additive outlier.” Within the X-13 statistical filter, which is used by U.S. Seasonal adjustment in the CES is done at the disaggregate sectoral level. The BLS also posts extensive documentation of its seasonal adjustment procedure. This measure is perhaps the most widely watched monthly economic indicator. The COVID-19 recession was so extreme that such interventions were necessary as discussed by the BLS commissioner.ĭo these adjustments mean that we will not see seasonal echoes of COVID-19 in the economic series in the future? We take as a case study, total nonfarm payrolls in the Current Employment Statistics (CES), produced by the BLS. Agencies dislike making these ad hoc adjustments because they want the data process to be transparent. Agencies doing seasonal adjustment were well aware of the problem and applied manual adjustments. If the seasonal filter were left to run without any special adjustment, the estimated seasonal factors would be completely dominated by the within-year patterns in 2020. Last year’s recession was an order of magnitude bigger than the Great Recession. Seasonal Echoes after COVID-19: Payroll Employment In fact, the Federal Reserve Board made such an adjustment in the 2010 annual revision in the official industrial production statistics. The problem could be mitigated by the user making manual adjustments. This contributed to a pattern where economic growth seemed to be strong in the spring only to fade later on in the year, as shown by Wright (2013). As a result, for the subsequent few years, an “echo” of the Great Recession took place as economic data kept exceeding the artificially low expectations for that time of year. Seasonal filters concluded that the normal employment for this time of year was lower. For example, the worst of the 2007-09 Great Recession was in early 2009. Since the seasonal filter determines the normal pattern for, say, January, by a weighted average of the last few Januaries, an unusual observation will have a big impact on estimated seasonal factors. It is easy then to miss just how large seasonal swings in the unadjusted economic data can be (for GDP, they are on average as big as a typical business cycle peak-to-trough fluctuation) and that the seasonal statistical filter itself can create spurious variation in the adjusted series.Ī pernicious problem with seasonal adjustment comes after a big shock that is not seasonal in its origin. ![]() Most analysts focus on seasonally adjusted data, without paying much attention to the unadjusted series or the adjustment process itself. Statistical agencies apply statistical filters to remove these seasonal effects so that the underlying economic trends can be easily compared over time. Many economic series present periodic patterns within each calendar year, generally referred to as seasonal effects. Seasonal Echoes after the Great Recession ![]() We note that seasonal echoes may lead the official numbers to overstate actual changes in payroll employment modestly between March and July of this year after which distortions flip the other way. In this Liberty Street Economics post, we discuss the prospects for these echo effects after last year’s sharp economic contraction by focusing on the payroll employment series published by the U.S. Large economic shocks, such as the 2007-09 downturn, can generate lasting seasonal echoes in subsequent data. Seasonal adjustment is a key statistical procedure underlying the creation of many economic series. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |