I haven't been blogging much the past two years — due to a) work taking nearly all of my energy, and b) the near daily update of data around COVID-19 being more conducive to Twitter than blogging. However, I thought it was time for a longer-term assessment of the economic time series in the context of Dynamic Information Equilibrium Models (DIEMs).
First, let's look at the consumer spending data from tracktherecovery.org (Raj Chetty and John Friedman's project using with proprietary credit card data and only bulk credit to the "OI team" of undergrad RAs) — beginning with a little history (and links to Twitter). I originally put together several models back in early June of 2020 to describe the shock to consumer spending data. About two months later (end of July 2020) I added in long run growth because it would start to become an important factor as the recovery dragged on. Towards the end of October, I decided to rank the performance of the four different models. Basically, all of them performed about equally — except for the "entropic shock" with a complete return to equilibrium. This means that there was some persistent gap in spending that wasn't made up until after the most recent round of stimulus checks in January of 2021.
Here are the four models (click to enlarge).
|Positive + negative shock|
|Step response (see here)|
|Negative "entropic" shock and return to equilibrium|
|Negative "entropic" shock and return to a different equilibrium|
I illustrate both the "entropic" shock models in this post on evaporating information. The basic idea is that there's a shock to the time series and it either evaporates entirely or leaves some residual: