Measuring State by State Responses to COVID through Podcast Downloads.

May 05, 2020

State

The following is outside of regular programming. We are a podcast attribution company and like data, but we aren’t COVID experts.

We are going to claim that you can roughly judge a state’s response to COVID by looking at podcast downloads between 7 am and 8 am local time. If downloads drop and stay down, people are staying at home and not commuting; therefore, new COVID cases fall. If they don’t, new cases rise.

Explore the data via our interactive tool.

The rest of this post outlines the methodology and why we even bothered to pull this data.

How We Got Here

There is much ado about podcast download numbers. We lead with it here, and many other people have written about it.

I’m going to switch sides and say that download numbers don’t matter in the short term.

Podcast downloads are always going to be trees in the COVID hurricane. Publishers only need high download numbers if they sell out of advertising inventory, and no one in the current market is selling out of inventory. Your loyal listeners are going to stay, and you have plenty of room for makegoods.

What is more interesting to me is how downloads are changing.

Dayparts

We’ve talked about dayparts a bit in a previous post. A daypart graph looks like this:

January 26
February 02
February 09
February 16
February 23
March 01
March 08
March 15
March 22
March 29
April 05
April 12
April 19

The Y-Axis is the percent change compared to the average number of downloads per hour. If there are 240 downloads in a day, 0% will represent ten downloads, 100% would represent twenty downloads, and -50% would be five downloads. Using percentages allows us to compare weeks, even though download numbers vary wildly.

People are reasonably predictable; we get up, we download a podcast, we go to work. We download something on the way home, and we go to bed. Over a large enough sample size, these are smooth, predictable lines till seven months after a bat bites a pangolin on the other side of the world.

What we have seen is these lines flatten, and we aren’t the only ones.

It’s clear from our data that morning routines have changed significantly. Every day now looks like the weekend. This trend was seen more significantly in Podcasts than in Music, likely due to the fact that Car and Commute use cases have changed quite dramatically - Spotify

This flattening is not universal in the United States though, that’s the interesting part. It is entirely dependent on the state you live in.

Let’s take New York.

January 26
February 02
February 09
February 16
February 23
March 01
March 08
March 15
March 22
March 29
April 05
April 12
April 19

Each line represents the average day part graph between Monday and Friday for a given week. We see the largest changes in morning listening. In early March, downloads in this period dropped by half.

If we compare that to say Iowa, it tells a different story. Downloads did drop in early March but have quickly recovered.

January 26
February 02
February 09
February 16
February 23
March 01
March 08
March 15
March 22
March 29
April 05
April 12
April 19

That got us interested; if fewer people are commuting in some states vs. others, what were the effects on new COVID cases? We pulled COVID cases data from the excellent Covid Tracking site and grouped new cases by week. The following graph shows the drop in downloads at 7 am and the percent of new cases per state for New York.

Covid cases
Impressions drop

Iowa presents a different picture.

Covid cases
Impressions drop

The above is an oversimplification of a massive pandemic, but if downloads at 7 am drop and stay down, people are staying at home, commuting less, and the world is healing.

We have all 50 states here.

The next time someone complains about download numbers dropping, let them know it’s a good thing.