Consumer sentiment in the age of politics

Is the University of Michigan survey fit for purpose anymore?

The University of Michigan survey of inflation expectations and consumer sentiment has shown quite the divergence in measurement based on political affiliation:

Which raised the question in my head — is this useful? How do you even correct for such a thing? There are several steps between a respondent’s belief in what the political situation entails and actual changes in prices. There are fewer in the case of sentiment — if I think the economy is going to suck because the person in office put an unelected ketamine freak in charge of firing everyone that is in fact my sentiment!

Regardless, it made me realize I haven’t updated the sentiment model in public since December of 2019. Looking at the files, I apparently did some interim work in March and April of 2022 where I added two shock “estimates” for what seemed to be 1) the effect of stimulus or pandemic re-opening, and 2) the onset of the post-pandemic inflation. The inflation shock estimate did a decent job at estimating the bottom in 2022 — whereupon sentiment started to rise again:

Dynamic information equilibrium model of consumer sentiment as it was when it was last published on the blog in 2019 (blue), the subsequent stimulus (green dashed) and inflation (red dashed) shocks, and finally a “post inflation” shock (yellow). Note error bands are a constant ~0.1 in log units so look smaller at lower levels. Gray dashed lines are estimated “equilibrium paths” for the periods a) prior to 2001, b) 2003-2008, c) 2009-2013, d) 2015-2020, and e) ca. 2022. Data since 2022 is shown as black points.

I thought about trying two shocks for the most recent data — a rise post-inflation and a recent post-election drop — but given the size of the error bands it does not seem to add information. One interesting detail is that the model estimate says sentiment started to rebound from the inflation shock around Q3 of 2022 — which is exactly when inflation returned to normal¹:

DIEM inflation forecast using data vintage March of 2022, prior to the Fed raising rates for the first time since the pandemic, with post-forecast data in black.

So far, the sentiment data appears consistent with the end of an “end of inflation” shock (yellow dashed + band in the graph above). We can’t really answer the question in the subhed right now. However, since there are limited post-election data in the time series so far, we can use this new baseline to check if sentiment has gone MAGA or gone woke over the next few years assuming at least some of the infrastructure than enables me to write this blog continues to exist (FRED, the University of Michigan, my job, the internet, the United States, etc).

1. Despite it continuing to be a news story and a campaign issue through November of 2024 — two years later.