Consider the following three examples of seasonally adjusted versus unseasonally adjusted data.
figure 1: Nonfarm Payrolls, 000’s, seasonally adjusted (blue), nonseasonally adjusted (tan). Dates of peak-to-trough recessions as defined by NBER are shaded in gray. Sources: BLS via FRED, NBER.
Deep troughs associated with the 2020 recession could distort the process of estimating seasonality. Statistical agencies are not unaware of the challenges of extreme events.This is BLS to start employment in 2022and For CPI/PPI.
Here is a picture of GDP.
figure 2: 2012 GDP in billions, annualized, seasonally adjusted (blue), unseasonally adjusted (tan). Dates of peak-to-trough recessions as defined by NBER are shaded in gray. Source: BEA from FRED, NBER.
This is the export of goods.
image 3: Merchandise exports, million USD/month, seasonally adjusted (blue), unseasonally adjusted (tan). Dates of peak-to-trough recessions as defined by NBER are shaded in gray. Source: Census via FRED, NBER.
This is a Introduction to the Census On post-pandemic adjustments.
What should we do after a major event such as an epidemic occurs? Jonathan Wright wrote about the seasonal “echo” effect after Lehman (BPEA 2016), and issued Covid (the fed). Abern and Jacobs (JBCR 2022) Discuss Covid and typical filters.
The previous discussion on seasonal adjustment in the Covid era, here.
What about abnormal weather (although it is getting harder to define abnormal weather these days)? SF Fed There is an adjustment to rely on county-level data. Blue is the standard series on the change in nonfarm payrolls.
Figure 4: Monthly percent change in nonfarm payrolls, seasonally adjusted (blue), adjusted for regional heterogeneity (tan), and without regional heterogeneity (green). source: SF FedVisit 3/1.
It’s hard to see the significance of this weather adjustment with this kind of data span.this is a detail
Figure 5: Monthly percent change in nonfarm payrolls, seasonally adjusted (blue), adjusted for regional heterogeneity (tan), and without regional heterogeneity (green). source: SF FedVisit 3/1.
Adjusted for 2015, the San Francisco Fed calculated a change in NFP of 3.89 to 393,000 in January, compared with the official figure of 517,000.
Boulding and Wright (BPEA 2015) Suggest several ways to adjust the weather.
If you want to complain about seasonally adjusted data, by all means do so. But you should determine which aspect of seasonal adjustment is bothering you. You should do this after reading what the relevant statistical agency has done. In general, it’s usually possible to do better, but at the cost of using some kind of program that looks special from the outside. In other words, in turbulent times there is often a trade-off between better procedures and transparency of approach.
