How big is radio in the U.S. today? Truthfully, we don’t really know!
Largely because satellite radio (i.e., SiriusXM), cable TV radio channels (e.g., Music Choice), and Internet-only radio (e.g., Pandora), for different reasons in each case, have not been subscribers to Arbitron (a/k/a Nielsen Audio) in the past, we’ve never known exactly how much listening happens to the full spectrum of radio — i.e., via all four of radio’s delivery systems.
Last week, however, the research firm Edison Research, well-known in our industry for its annual “Infinite Dial” studies, released the topline findings of its first-ever “Share of Ear” study. The study is based on a sample of about 2,100 Americans age 13+ who were paid to keep a 24-hour diary documenting their audio listening behavior on a day in late May. (More details in RAIN story here.)
Although Edison CEO Larry Rosin announced that we can expect to see more findings released in coming weeks from clients who buy rights to this syndicated study, there are already some very interesting findings we can see and/or calculate:
Finding #1: Radio’s 80/20 Ratio Confirmed
The first key finding, I believe, is one that confirms a statistic which has been produced by other, less rigorous, research studies in the past: Of consumers’ total time listening to audio — i.e., both radio and their personal music collections (e.g., LPs, CDs, MP3s, and so forth) — the vast majority of that time is spent with radio and only about 20% of that time is spent listening to their personal music collections.
Specifically, the “Share of Ear” study found that 20.3% of the time consumers spend listening to a category that Edison Research called “Owned music.”
As previously reported in RAIN and elsewhere, here are the shares by category:
20.3% Owned music
9.6% Internet radio/music (“Pandora, Spotify, etc.”)
7.7% Satellite radio
5.2% TV music channels
100.0% Total audio
Truthfully, however, I believe that “20.3%” just slightly understates the number for owned music.
As shown above, Edison in its press release characterizes Spotify and its brethren (Rdio, Rhapsody, Beats Music, and so forth) as part of a category called “Internet Radio/Music” that comprises 11.6% of audio listening time.
But my belief is that that the on-demand music services (Spotify, Rhapsody, Rdio, Beats Music, and so forth) are primarily used by consumers in a way that they should not be grouped in with “Internet radio” as a form of what Edison’s press release called “preprogrammed music.”
Rather, I believe that consumers use on-demand services like Spotify as a replacement for CD or MP3 ownership.
For example, “I am going to the Richard Thompson show next week, so I would like to listen to his latest album a few times,” or “I am hosting a dinner party on Sunday night, so I am going to build a playlist of my favorite Frank Sinatra and Ella Fitzgerald songs,” or “I live in Colorado, it’s midnight, and I’m in the mood to listen to Pink Floyd’s Dark Side of the Moon.”
Depending on your preferred choice of technology, in those use cases you could put on a CD, you could call up the tracks in iTunes, or you could play them on Spotify. Those all fall in the category of “owned music” to me. (Albeit in the latter case the consumer owns rights rather than a physical item.)
I certainly understand that in doing this study it would have been challenging for Edison Research to ask diarykeepers to keep track of which feature of Spotify they were using — e.g., were they picking specific tracks or were they using the service’s “radio” feature (which admittedly they all have.)
But until I learn otherwise, I would guess that perhaps 10% of “owned music” listening nowadays is Spotify, et al., which would move two percentage points from the “Internet radio” category to the “owned music” category.
So that with that revised definition, the actual ratio is about 78/22. But that’s a minor quibble. The basic point is confirmed.
Finding #2: Radio’s Shares by Delivery Platform
Now let’s look specifically at the 78% or so of audio listening that I believe would be reasonably defined as “radio.”
How does that break out by delivery system?
First, let’s create a table of the data, expressed as %’s of all audio listening. (Note that I am also here ignoring the study’s 1.5% of “Other.”):
9.6 Internet radio (subtracting the 2 points I would guess are Spotify and its brethren)
7.7 Satellite radio
5.2 TV music channels
76.3 Total radio (all delivery systems)
Now let’s calculate those against our new denominator of 76.3:
12.6% Internet radio
10.1% Satellite radio
6.8% TV music channels
100.0% Total radio
Well, that’s very interesting: Radio, across all its various delivery platforms, is more popular than most of us in the industry imagined, because radio listening as measured by Arbitron (a/k/a Nielsen Audio today) has been picking up only abiut 2/3 of it!
Since we’ve already had some knowledge about the Internet radio numbers (from the monthly Webcast Metrics Top 20 press releases), the biggest insight here is the size of SiriusXM’s audience, which no one (maybe including SiriusXM!) has ever known before. SiriusXM reveals subscriber numbers, but not how much listening the subscribers (and their family members and fellow carpool members) do.
We have also had no idea how big the audiences for Music Choice, etc., are, and I think this study reveals that they’re bigger than most of us imagined.
Another way to look at this data is to express each delivery mechanism’s AQH as a percentage of AM/FM’s AQH:
18.4% Internet radio
14.8% Satellite radio
10.0% TV music channels
Here’s how to read those percentages: In the average city, the size of Internet radio’s AQH audience is about equal to 18.4 share points’ worth of AM/FM stations. The size of satellite radio’s audience is about as big as AM/FM stations with a combined share of 14.8. MusicChoice, etc., in your market is probably about as as big as a set of AM/FM stations with a combined 10.0 share. And podcasting’s AQH, at the average moment in the average market, is about as big as an AM/FM station with a 3.2 share.
Those are all bigger numbers than most of us would have probably imagined. But it’s hard to see how the Edison Research study could have had a methodological bias that would have gotten those numbers wrong!
Finding #3: Radio’s Hours per Week per Consumer
One question that’s on a lot of people’s minds nowadays is the number of hours per week that the average American spends with AM/FM radio. Analyses from diary markets give a bigger number than analyses from PPM markets, but overall, on average, it seems like that figure is about 9 to 10 hours per week (and declining about 5-6% per year lately).
However, Edison’s “Share of Ear” study says the average respondent spends about 28.5 hours per week listening to audio, with 52% of that to AM/FM radio, which would work out to 14.8 hours per week.
Why are these two findings so different?
Well, one adjustment we can make is that the “9 to 10 hours” estimate I quoted above is for the 18-hour broadcast day (6a-12m), whereas I believe the Edison number includes overnights. Roughly, I believe overnights comprise about 10% of radio listening hours, so let’s guess that Edison’s 6a-12m number about is 13.3 hours per week per person. That gets the two numbers closer.
The other factor I see that we could take into account is one called the “observer effect.” Think about it this way: If a research firm recruited you to keep a diary for 24 hours of your gum-chewing habits (or fast food consumption, or whisky drinking, or exercise), mightn’t that — at least subconsciously — encourage you to chew gum that day (or eat a fast-food meal, or drink a glass of whisky, or exercise)?
A variation on this point is called the “Heisenberg uncertainty principle,” developed in 1927 by Werner Heisenberg in the very building where I took my freshman year physics class, which says that you can never measure both the speed and position of a particle, because the very act of measuring one characteristic changes the other. (Clever RAIN readers will realize that this implies, of course, that the transporter on “Star Trek” could never actually work, as it requires knowing those characteristics of the atoms it is beaming from one place to the other. (However, the “Star Trek: The Next Generation Technical Manual” introduced the concept that Starfleet transporters have a “Heisenberg Compensator” to get around that problem. (When asked by Time magazine how that works, book co-author Michael Okuda answered, “Very well, thank you.”)))
So should we apply a Heisenberg compensator to adjust for the fact that the very act of keeping the diary for a day might have increased their audio consumption that day?
Well, we may have one data point soon to know the answer: Assuming Edison or one of its clients eventually releases Pandora’s share of the Internet radio listening in this study, we can calculate its apparent national AQH per “Share of Ear” and compare it to the server-based (and thus presumably reasonably accurate) AQH in Triton Digital’s Webcast Metric reports. (However, since Triton doesn’t report Pandora One (commercial-free) AQH, we may have to add in an estimate of that.) If the resulting AQH from Triton is higher than the implied AQH in the “Share of Ear” study, then a Heisenberg compensation to the TSLs for all forms of audio in the “Share of Ear” study may be appropriate.
Overall, this is a fascinating study with lots to recommend itself. As more details are released, either by Edison or by its clients, we’ll explore them further here.