Yesterday we looked at the trend of unemployment based on the Obama years (but after the Great Recession), and based on the Trump years (but before the Covid shock). I wanted to show a third graph, with the trend based on the Biden years after the Covid shock, when unemployment appears to have returned to its normal behavior. But that graph didn't show anything useful. I don't know why.
Maybe there was not enough data between July 2022 and August 2024 to make a trend line that looked reasonable in the pre-Covid years. Or maybe the Fed's high interest rate started pushing unemployment up in early 2023 (as I said before) and put a "kink" in the trend line. And maybe the kink shows up as a massive mismatch between trend and data in those pre-Covid years. I dunno. But I couldn't use that graph.
So I went back to FRED's unemployment data. In red, I underlined two years of data before the Covid shock, and two years after it. I want to compare the two underlined periods:
This Graph at FRED: https://fred.stlouisfed.org/graph/?g=1uyts |
By eye -- my eye, at least -- the post-Covid increase seems to be accelerating upward noticeably faster than the pre-Covid decrease was slowing. As I said a couple of weeks ago:
To my eye, unemployment started going up around January 2023 - more than a year and a half ago, now -- and started accelerating around January 2024. They waited too long before bringing rates down.
It still looks like that, to me. But I don't trust my eye more than I trust arithmetic. So in Excel, I sorted the March-2018-to-Feb-2020 data in reverse-date order, turning the decrease into an increase. Then I showed the two periods starting together on a graph. Here is the result:
Comparison of Pre-Covid data (reversed) and Post-Covid data |
There's not much difference between the two lines. They start at 3.5 percent unemployment, both of them, and after two years they end up only 0.2 percent apart. On this graph, I don't see the recent accelerating increase of unemployment that I thought I saw on the underlined FRED graph.
But after sleeping on it, I notice that the left half of the graph shows the blue line mostly at-or-below the orange line. And the last seven or eight months show the blue line mostly at-or-above the orange. So the difference between the two lines may be more than the 0.2 percent difference we see between the August 2024 and March 2018 data.
I could take the blue line and move it up by 0.1, more or less centering the blue line on the orange in that first year. This might even be a reasonable manipulation of the data, given that I want to see the differences that arise in the second year.
Here's how the graph looks with the blue line values increased by 0.1:
By
eye, at least, the blue line is roughly centered on the orange through
2022 and 2023. In December 2023 and January 2024, the two lines are
identical at 3.8 percent unemployment. For the rest of 2024, blue gains
on orange: The post-Covid increase in unemployment outpaces the
(chronologically reversed) pre-Covid decline. This is the acceleration I
was looking for! The arithmetic confirms the eye.
Is it
reasonable to take a two-year decline in unemployment, put it in reverse
chronological order, and compare it to a two-year increase? I dunno. It seemed reasonable, at the time.
Are we gonna have a recession? I dunno. I expect one, yeah, but I don't predict.
Does it matter if we get a recession when we could have avoided one? Yes, definitely.
Does this post show that the Fed should have started lowering the interest rate during or before January 2024? Yes, definitely.