Lies, Damn Lies, and the 45% Statistic

In May of 2013, “Trichordist Editor” wrote an article with a staggering headline: 45% Fewer Professional Working Musicians Since 2002. The lead sentence lays all the blame at the feet of the internet: “The numbers are simple and staggering. The internet has not empowered musicians, it has exploited them.”

Indeed, if that figure is right, then the numbers are staggering. It would mean that nearly half of the jobs for working musicians have evaporated in a single decade. And if there was even a shred of evidence that it was “the internet” that destroyed those jobs, then that would be a powerful indictment.

Fortunately for artists – but unfortunately for “Trichordist Editor” – that figure is a complete and utter fabrication.

The author claims that he got his data from the BLS (the Bureau of Labor Statistics), which he accurately calls “an agnostic government agency, not the RIAA.” But conveniently, he does not provide any links to the actual data that he used. He did, however, include URL’s in the simplistic graph that he posted on the page:


There is nowhere to click on a link to those URL’s, so in order to follow them, you have to type them in, by hand, from the low-resolution image. I’m betting he was counting on the notion that nobody would do that. I did. And if you do it too, you’ll see just exactly how bogus the 45% figure actually is.

Comparing Apples to Oranges

Following the links will reveal that the author compared two unrelated employment figures. The two inks are to BLS data for:

  • “Performing Arts, Spectator Sports, and Related Industries: NAICS 711”
    This is an industry group, not an occupation. The majority of workers in this industry are not musicians (it includes actors, dancers, and so forth), and the majority of working musicians are not employed in this industry. (It is, however, a very large employer – I’ll get to that later in the article.)
  • “Occupational Employment and Wages, May 2003: 27-2042 Musicians and Singers”
    This is an occupation, not an industry, so it includes all musicians and singers employed in all industries – not just those who work in the performing arts.

The two numbers count entirely different things, and comparing one to the other will tell us absolutely nothing about the status of working musicians (nor about the status of the performing arts industry).

As an aside: the “Musicians and Singers” category (27-2042) does not include such occupations as “Music Directors and Composers” (27-2041), “Art, Drama, and Music Teachers, Postsecondary” (25-1121), or “Sound Engineering Technicians” (27-4014). Many workers in these occupations are musicians, but they are not separated as such, so I will not deal with them here. This does mean that we cannot get a complete picture of all working musicians from BLS data alone.

Cherry picking

The fact that the author starts his decade on the year 2002 is also no coincidence. This is because 2002 was the year with the highest number of working musicians on record.

For example, in 1999 (the year Napster started operating), there were 46,440 working musicians. The number of employed musicians did not drop below this level until 2010. If we were to look at the percent change for the decade ending in 2009, we would say that 2% more musicians are “working today.”

This is not the only time that copyright maximalists have cherry-picked the data to compare. A classic example is comparing the decline in recorded music revenue, but using 1999 or 2000 as the starting date. What goes unmentioned is that these years had the highest income from recorded music sales in history.

Unfortunately, 1999 is the first year that BLS data is publicly available on the Web. For prior years, it is only available for download in Excel format, and the occupations are not categorized the same way. In particular, there is no “musicians and singers” category. If we truly wanted to see how “the internet” has affected musicians, we should go back to around 1996, the time when the World Wide Web became ubiquitous. Unfortunately, this is impossible using BLS data.

What The BLS Data Actually Shows

To actually compare apples to apples, I looked at only occupational category 27-2042, “Musicians and Singers.” The formula I used to calculate the percent decline is the standard one that is taught in elementary math courses:

% decline = (T0 – Tf) / T0 * 100
where T0 is the initial value, and Tf is the final value

In 2002, there were 53,900 musicians and singers. In 2012, that number was 42,100. This means that the decline was roughly 22% – less than half of what the article claimed.


To put this in perspective: a 45% drop would be a loss of 29,667 jobs. The actual job loss is 11,840. That is a difference 17,827 jobs. The difference alone is nearly 6,000 more than the number of jobs that were actually lost.

As for “Performing Arts, Spectator Sports, and Related Industries?” In 2002, there were 128,800 workers in that industry. In 2012, there were 116,580 workers. That is less than a 10% decline.

Now, don’t get me wrong. Even though we’re using the highest employment number in recorded history as our baseline, a 22% drop in employment is still very significant. If in fact, this was caused by “the internet,” then that would still be a big criticism.

But is it? By now, it’s a given that recorded music revenues have declined immensely. Most of that is due to lower payouts from legal services, and it’s extremely likely that some of it is due to piracy. Intentionally or not, the widespread use of the internet as a distribution mechanism has facilitated both.

But in theory, that should only result in employment losses if musicians and singers were employed in industries who make their money from recorded music. And, as it turns out, that’s not true.

Most working musicians and singers are not employed in the recorded music industry

Before we dig into the data, I should note that the BLS didn’t break down employment by employer category until 2002. But from the data that was gathered since then, one thing becomes absolutely clear: the industries that employed the most musicians and singers were never industries whose income derived primarily from the sale of recordings.

Across all the years covered by BLS data, these are the industries with the highest levels of employment for musicians and singers:

  • Performing arts companies
  • Religious organizations
  • Promoters of performing arts and sports
  • Amusement Parks and Arcades


Performing arts companies have always been the largest employer of musicians and singers (though it’s never hired more than half the employees in that occupation). Here is the U.S. Census description for that industry: “This industry group comprises establishments primarily engaged in producing live presentations involving the performances of actors and actresses, singers, dancers, musical groups and artists, and other performing artists.”

In other words, that industry earned its income, not from the sale of recordings (or videos), but from live performance income. In theory, this industry should be among the least affected by internet distribution, whether legal or illicit. On the other hand, there is no question that employment in this industry is declining – so the question is why it is declining. It is an especially tricky question, since employment in other industries – such as religious organizations – are growing. I don’t have the answer to that, but it’s clearly not anything as simplistic as “the internet.”

In contrast, the “Sound Recording Industries” (NAICS 5122) have employed less than 200 singers and musicians in every year since 2003. In fact, for most years (2004-2008 and 2010), the numbers are so small that the BLS does not list any data at all. The “Motion Picture and Video Industries” (NAICS 5121) actually employ more musicians and singers, but they have never employed more than 470 musicians and singers (in 2005). Like the sound recording industries, for many years (2004, 2007, 2012, 2013), the employment numbers are so low that the BLS does not list any data.

And there is some good news for “Independent Artists, Writers, and Performers” (NAICS 711500). Though their numbers are still small, their relative size is exploding.

In 2002, there were a mere 430 independent performers working as musicians and singers. By 2012, that number had grown to 1,830. In the same decade that the Trichordist is examining, the number of independent musicians and singers has grown by 326%. (The highest number in the BLS data was 1,910 in 2010).

Things may not look so good for those musicians and singers who have relied on the traditional industry for their income. But if you’re an independent musician, things are much, much better than they were.

Insults On Top Of Lies

The author hedges his bets by saying: “Of course there will always be people to nit pick the numbers, to argue and quibble about the Bureau Of Labor Statistics (BLS) methodology.” Right before he says “Those who debate the exact numbers are using that to delay action. Their job is similar to the commentators and ‘scientists’ funded by oil companies’ to deny global warming or say it needs ‘more study.'”

This is simply an ad hominem attack against anyone who would question his utterly bogus statistic. Those who demand that people not lie with statistics (including, I hope, myself) are not “nit picking.” We are demanding that we be told the truth.

And The Trichordist is not telling the truth.

So that you know I am, here are direct links to the BLS figures.

Data for 27-2042, Musicians and Singers

(Update: The 2013 data has been superceded by 2014 data, which is now the “current” data. Rather than updating everything each year, I put the “current” link on its own line.


4 thoughts on “Lies, Damn Lies, and the 45% Statistic

  1. Another area of stats that I would like to see explored is Streaming vs. Traditional CD sales.

    Most of what I have seen regrading this seems to put 1 month of streaming returns against 1 month of CD sales which is an apples to oranges comparison to me. Without taking into consideration that a CD sales is a 1 time transaction with unlimited plays for the life of the CD as opposed to per play streaming over the same timespan. I’ve tired to find some info on average song plays for the average lifespan of a CD, but wasn’t successful.

    My thinking is that the return on a CD sale compared to the streaming return on all the songs on a CD over 10 years isn’t all that far apart, but I haven’t found any stats to back this up yet.


    • Well, if you’re asking for stats over a 10-year period, then of course there wouldn’t be any stats to back it up… streaming hasn’t been around that long.

      I’m personally ambivalent about streaming. On the one hand, it completely does away with the concept of owning a copy of the music, and that’s not terribly good for consumers in the long run. At the moment, the money isn’t there for rights holders either, and there’s a horrible tendency towards major-label lock-in within the streaming companies.

      On the other hand, it definitely does seem to be what people want, and music companies have to go with the market.

      Really, though, the streaming revolution hasn’t even begun yet – at the moment, it’s too much like radio. Things won’t get truly interesting until streaming becomes a new form of communication, where direct artist-to-fan communications happen over the same channels. Something closer to what YouTube creators have now, in terms of social networking.

      Of course, it remains to be seen whether rights holders will even allow that to happen. If the curmudgeons among them have their way (no revenue-sharing schemes, massive increases in per-stream payouts, “windowing” of music towards downloads and away from streaming), then legal streaming sites will simply die. So the point might be moot anyway.


      • “Well, if you’re asking for stats over a 10-year period, then of course there wouldn’t be any stats to back it up… streaming hasn’t been around that long.”

        No, I realize that. The streaming part would have to be projected using current data.

        I ‘m more interested in what the actual “per play” return of a song on a CD is over the lifespan of the CD. I would guess that marketing studies on consumer’s CD listening habits exist somewhere, but I doubt that they would ever become public.

        Comparing the the profits from the sale of a CD (sold once, listened to until the CD fails) to streaming sales (pays on every play from now until potentially forever) seem very skewed to me. If CD sales could be broken down to a “per play” levels, the comparisons would have more meaning to me.


  2. Comparing the the profits from the sale of a CD (sold once, listened to until the CD fails) to streaming sales (pays on every play from now until potentially forever) seem very skewed to me. If CD sales could be broken down to a “per play” levels, the comparisons would have more meaning to me.

    Well, one possible option would be to look at how much consumers have paid for music downloads, and compare it to how much consumers pay for a subscription to streaming services in the same period.

    This is something similar to what Spotify has done: (scroll down a bit)

    They only compare their listeners with the general public, though. And, of course, that option doesn’t take into account the fact that users will (ideally) only pay for a downloaded track once in their lifetimes.

    To truly do an apples-to-apples comparison, as you want, is almost certainly impossible, even with all of the data available over ten years. You could look at how much a person pays for a download, and compare it to what they paid throughout ten years (or whatever) of streaming. But even this is misleading, because it assumes that if a person listens to a track – whether once or twice, or a million times – then they would have bought the download. Not only is this impossible to determine in general, its especially impossible given the fact that streaming services offer entire collections of works.

    So, say that someone listens to a track on Spotify once or twice in a ten-year period. If Spotify were not around, they would not have bought a download – they’re only listening to that song because it’s already “bundled” with the Spotify service. That’s easy to determine for songs that are only listened to once or twice, but what’s the cutoff point? Ten listens? One hundred listens?


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s