top of page

Covid Unemployment modeling

Our data analysis and interpretation


JUMP to:

What is the short term unemployment model?

The ExtendPUA Short term model uses newly available data during the SARS-CoV-2 (Covid 19) pandemic. Google has been releasing mobility to assist those who are modeling Covid-19 outcomes. Google’s data is free and available to the public, and after running analysis on the data I believe that it may be useful as an early indicator of the coming months unemployment rate. 


This recession is different than most, in the recessions of 1981, 1990, 2001, and the great recession (recessions where data is easily accessible for analytics) they all follow similar trends of a slow and gradual rise to an eventual peak unemployment, and then as the economy recovers there is a slow and gradual return to recovery. Due to the nature of this recession with a full on economic shut down it is hard to model our trajectory through the early bits of the unemployment crisis compared to any recent recession. The Short Term Model can help with that.

Figure 1 shows what a typical recession rise to peak unemployment might look like:


Interpreting the graph

For reference: The heavy red line is the actual US Unemployment Rate, the Heavier Blue line is where the forecast of unemployment will be based on Google's data, the grey lines indicate where 66% certainty high and low unemployment could be, and the lighter blue lines are 95% certainty of where unemployment could be. Think of the blue and grey lines as a hurricane tracker making an estimate of where a hurricane is going to make landfall. The central blue line is where it is likely to strike, but the upper and lower light blue lines are also possible.


The current unemployment levels fall well outside the boundaries with which we can say with 95% confidence (margin of error) will be accurate. This leads to the creation of the short term unemployment model.

Figure 2 shows the short-term model:


As mentioned above the model uses previously unavailable data to make forecasts on the upcoming month. The margin of error is wider on these data, but that is due to the volatility across the states of this unemployment crisis, and the fact that there are only 51 states and 4 months of complete data to review and build a model from. 


Based on Google’s mobility Workplace Data (those going to work), and residential data (those spending time at home, or more specifically, in residential areas in general) the short term model currently suggests that July’s unemployment will go up from a national average of 11.1% to 12.0%, and that the 95% confidence margin of error lies between 6% and 18%. 


What is the Long term unemployment model?

Using the short term model as a guide, there is a long term model as well which utilizes data from previous recessions. It is our theory that after the volatile unemployment rates early in the COVID-19 recession, it may likely return to some sort of recession mean increase. 


Data from the Great Depression to today suggest that from the beginning of a recession (economists say ours officially began February 2020) to the peak of unemployment takes on average between 11 and 20 months.


The long term model assumes:

  • The volatility in the COVID-19 opening of the recession will normalize at a sort of point.

  • That unemployment will increase to a peak level per a usual recession, and then decline per a usual recession. 


Using the short term model for the next month's signaling, the long term model currently looks 3 months ahead, but as things normalize it may look further.

Figure 3 graphs the current model outcome results.


The current model assumes: May was our peak month for unemployment volatility, and that until we hit a peak unemployment level, the rise the unemployment will rise more slowly as per previous recessions; however, as more more data become available, the model will adjust to our circumstances.


Based on those assumptions, National unemployment in September is forecasted to be 11.9% and could be as high as 18.% and as low as 5.7% based on the margin of error. As we become more sure the data has normalized to typical recession growth (or recovery), the margin of error will shrink.


What does it all mean?

There has been lots of talk of a V-shaped recovery, that the economy would come roaring back, by easter, then summer, and now fall. But it seems that some people don’t want to face the possibility that this could be a long term situation, and that there isn’t necessarily a speedy recovery here. We very likely could be looking at unemployment at or above 10% for the remainder of the year, and that’s just by typical U3 unemployment standards. This doesn’t encompass the entirety of those without a job.


With all the talk of encouraging people back to work by cutting unemployment benefits, what work? Many of us want to go back to work, but the jobs just simply aren’t there and won't be for some time.

Why unemployment data modeling courtesy Andrew Griffin. research and data interpretation courtesy Andrew Griffin and Joe Beumer.


Our 100% Volunteer Based effort was created by two members of the entertainment industry who are currently unemployed due to COVID-19 and are relying on the FPUC $600 to pay their bills. We noticed despair, but little individual momentum, around the FPUC expiration and wanted to offer a resource for people to easily take clear and informed actions. In our research around the $600, we learned more about the overwhelming inequity of the system and we aim to change that too.


We are joined by a team of incredible collaborators, most of whom are also members of the entertainment industry.


This doesn’t happen without them.


We stand in solidarity with the over 30 million people who are unemployed.


Oh.  And we want everyone to wear a mask

Thank you for being a part of the movement.

Want more?

Read our talking points

bottom of page