Seasonal Adjustment in the Wake of Big Shocks, Economic and Otherwise

Consider the following three examples of seasonally adjusted vs. not seasonally adjusted data.

Figure 1: Nonfarm payroll employment, 000’s, seasonally adjusted (blue), not seasonally adjusted (tan). NBER defined peak-to-trough recession dates shaded gray. Source: BLS via FRED, NBER.

The deep trough associated with the 2020 recession can cause distortions in the process of estimating seasonal factors. It’s not as if the statistical agencies are unaware of the challenges of extreme events. Here’s the BLS take from 2022 on employment, and for CPI/PPI.

Here’s the picture for GDP.

Figure 2: GDP in bn Ch.2012$, annual rates, seasonally adjusted (blue), not seasonally adjusted (tan). NBER defined peak-to-trough recession dates shaded gray. Source: BEA via FRED, NBER.

And here is goods exports.

Figure 3: Goods exports, millions$/month, seasonally adjusted (blue), not seasonally adjusted (tan). NBER defined peak-to-trough recession dates shaded gray. Source: Census via FRED, NBER.

Here’s a Census presentation on adjustment post-pandemic.

What to do in the wake of big occurrences like the pandemic?┬áJonathan Wright writes on “echo” effect in seasonals, post Lehman (BPEA 2016), and post Covid (NY Fed). Abeln and Jacobs (JBCR 2022) discuss Covid and typical filters.

Previous discussion of seasonal adjustment in the Covid era, here.

What about anomalous weather (although what anomalous weather is in these times is getting harder and harder to define)? SF Fed has an adjustment, relying on county level data. Blue is the standard series on change in nonfarm payroll employment.

Figure 4: Month on month percent change in nonfarm payroll employment, seasonally adjusted (blue), adjusted with regional heterogeneity (tan) and without regional heterogeneity (green). Source: SF Fed, accessed 3/1.

It’s hard to see the importance of this weather adjustment using this span of data. here’s a detail

Figure 5: Month on month percent change in nonfarm payroll employment, seasonally adjusted (blue), adjusted with regional heterogeneity (tan) and without regional heterogeneity (green). Source: SF Fed, accessed 3/1.

Basing the adjustment to 2015, the SF Fed calculates January change in NFP at 389 to 393 thousand, versus official 517 thousand.

Boldin and Wright (BPEA 2015) present several approaches of adjusting for weather.

If you want to complain about the seasonally adjusted data, by all means, do so. But you should identify what aspect of the seasonal adjustment is bohering you. And you should do so after reading what the statistical agency in question has done. In general, it’s usually possible to do a better job, but at the cost of using some sort of procedure that looks ad hoc from the outside. In other words, there’s often a tradeoff between better procedures and transparency of methods, in turbulent times.

 

 

 

 

4 thoughts on “Seasonal Adjustment in the Wake of Big Shocks, Economic and Otherwise

  1. Macroduck

    Many, many thanks. In this environment, we need to be more cautious than usual about our conclusions.

    All these links and the Wolf book. Gonna need wine.

    Moses, I’ve got pinot grigio. What you got?

    1. Macroduck

      By the way, what weather giveth (in terms of data distortions), weather taketh away. Look for weather-related job gains in January to be handed back in February and perhaps March. Same with retail sales, and so on.

      1. Moses Herzog

        We had a tornado warning in the town I live, about 3 days ago. I’m glad the weather didn’t taketh me away. But in all seriousness there’s an above ground shelter I can use that could probably withstand anything short of a high-end F-4, or F-5 variety. The tornado 3 days ago wasn’t really much to worry about, it just kind of created what weathermen call a “signature” of a tornado on the radar, but not much in reality. When you’re not exactly sure and the town siren goes off it can be a rush. If you’ve ever seen “velocity” radar that’s where you see the green and the red colors juxtaposed to each other, and if you see a kind of circular “swirl” (best word I can think of) along the edge where the red and green colors meet up, you have something approaching to becoming or that is a tornado. There’s a suburb not far from me, that’s been pretty much destroyed TWICE in the last 15 years, but we always seem to luck out (knock on wood).
        https://images.app.goo.gl/R5hPAdYw2PpopiEA7

        https://images.app.goo.gl/zi67V2xPFMk4YehRA

        https://images.app.goo.gl/ZBreykepSgqUuF546

        If you’re looking at more the “standard radar” (doppler, or nexrad) then you’re looking more for something called a “hook echo” or possibly a “screaming eagle”, which “kinda” looks like this:
        https://images.app.goo.gl/b51yMLztVwAdto6NA

        You see either of those approaching where you live, get the hell underground or the center of your home with the most interior walls, and get the hell out of mobile homes. It’s like donald trump and a Big Mac. Tornados go like a magnet to mobile home parks every time, and demolish them every time.

    2. Moses Herzog

      Heh, I hope to have something in the house Friday/Saturday. In fact I will look for that wine (Grigio) in the store (please tell me it has at least 10%) I have a basic pattern~~10 days dry, get sauced, 10 days dry, get sauced, 10 days dry, get sauced etc. It’s the way I convince myself I have nothing more than a psychological need for alcohol. I should read more offline when I’m drinking. I know if I read offline while drinking Menzie’s blog comment garbage bin would thank me for all the stress relief. I got one of those plastic molded chairs in the backyard, I could bring my drinks out there. I’m just always afraid my “Okie” (in every derogatory sense of the word) backyard neighbors will start a conversation with me, and that it will turn to MAGA, and then I will tell them how dumb I think MAGA and Republicans are, and we’ll be off to the races with them jumping the fence and doing their best WWE cage match impression, which I’m really not interested in engaging in, but when I’m sauced I might let things progress and “go there”. My fuse, normally quite long, gets short about the 6th “tallboy can” in.

      I’m kind of like that Adam Sandler character in “Anger Management”. When I’m not drinking I am as anti-violent as a person can be. But after I drink a certain amount, even then, I don’t “start” things…. but I am not willing to listen to ANY of people’s crap at that point. So two sentences into someone else’s BS and I lose it.

      I do think I will request the Wolf book from the library, looks like they don’t have it yet—they tend to be slower on British authors for some reason unknown to me.

Comments are closed.