# of recording artists per given area

December 7, 2012 11:21 PM

I'm making a chart showing the number of musicians/recording artists per capita in different countries and US states. What I need is a dataset with the number of musicians in each area.

Clearly, this isn't totally easy because not all jurisdictions require musicians to register with the government, but I would be OK with slightly nonscientific data instead.

I tried to pull the information off last.fm, but the site only shows how many users used a tag, and how many times it has been used. If I could find out how many artists on last.fm have been tagged as "hungarian", that would be good. Population data I can find myself.

I'm aware of the Million Song Dataset, but I'm not a coder and don't think I could wrangle the Python code to get what I want (and besides, would rather not download a 200+GB file!)

posted by dunkadunc to media & arts (6 answers total)

There is no such generic database. Where DO musicians "register with the government?". Cuba?

And how narrowly do you define "musician?". Recording artists only? Million song data set only covers a fraction of musical genres, and only recorded music, so not representative of all musicians.

Not really a useful exercise as described.

You can probably find reports on the size and distribution of national music industries for many developed countries.
posted by spitbull at 4:48 AM on December 8, 2012

If you can live with people who reported their occupation as musician you might be able to find some useful numbers in the last census data.
posted by COD at 5:58 AM on December 8, 2012

Obviously musicians don't have to register with the government- not even saxophonists. I was being tongue-in-cheek. I am not an idiot.

I don't want demographic data, I want music data. I know such databases exist, which is why I mentioned last.fm. I would be happy with their database, except that their frontend doesn't tell you how many artists have been tagged with something, such as "japanese" or "hungarian". These tags aren't completely scientific, but they do give you a general idea, which would be enough for my purposes.

I am certain there are services out there other than last.fm where music has been tagged with locations, and there may even be an alternate interface that taps into last.fm's tag data.

If anyone knows of a dataset I can easily query that will give me a number of bands/artists that have been tagged with tag X so I get a rough idea of bands per region, that would be good. The Million Song Dataset would probably solve my problems, but I don't want to download a 200GB file and have to query it with Python.
posted by dunkadunc at 12:01 PM on December 8, 2012

I don't actually think the Million Song Data Set would give you what you're looking for, though.

If you look at the FAQ (under "Statistics of the dataset"), they've got 44,745 unique artists, which right off the bat sounds awfully low to me if you're talking "worldwide number of musicians." A quick & dirty search in the Economic Analysis section of the US Census gives a total of 189,510 "Musicians, singers, and related workers."

Then, while a million songs sounds like a lot, a quick bit of Googling suggests that iTunes has about 20 million songs, Amazon mp3 about 15 million, and last.fm something like 12 million. Undoubtedly there are a lot fewer artists than songs, and a lot of the artists are no longer producing, but that's still a significant difference in number of songs.

Finally, and I think most importantly, the MSDS is (if I'm reading the website correctly) a tool for end-users to R&D various analysis & matching algorithms. So it doesn't have to be anywhere near complete or comprehensive regarding the actual real-world number of artists or songs or whatever else - it just has to have enough different datapoints to allow an algorithm to function.

Which is to say, Amazon wants to know "If person likes Artist X, what other artist might they like?", and so it creates a program that looks at Artist X and classifies them as "(country of origin)+(BPM)+(genre)+(subgenre)+(gender of singer)" and then searches for matches. The MSDS gives them a set of data with which to test this program before actually applying it to the actual Amazon song collection. But I don't think you need all the countries in the world in order to test whether an algorithm can sort by country - for all I know you only need 2 or 5 or 10, and that could be all that's in the MSDS. Do you see what I mean? It's a testing tool, not a comprehensive database.

I think what you would be looking for is the databases of the end users of the MSDS, and I'd bet a lot of them are secret. The source for a significant number of the MSDS datapoints seems to be a company called The Echo Nest, but they're obviously a commercial enterprise - I'm sure their database is not cheap.

FWIW, I found this List of online music databases on Wikipedia.
posted by soundguy99 at 7:20 AM on December 9, 2012

But I think your project has bigger problems.

You say you're looking to create "a chart showing the number of musicians/recording artists per capita in different countries and US states", but you say you don't want demographic data. But "musicians per capita" kind of is the very definition of demographic data, y'know? So now I'm not sure what your end goal is.

As spitbull points out, a lot depends on what your definition of "musician" is for the purpose of this project. If you're trying to create a per capita number using only data you can pull off last.fm or AllMusic or such, you're going to be missing a significant number of musicians who have little or no recorded output. Pretty much all the musicians I know who are actually making a living in music are doing it through some combination of private lessons/teaching in one or more school systems/playing in cover bands. None of them would show up in a last.fm database. Or what about orchestras? A major orchestra can have over 100 members, but you'll only find "Cleveland Orchestra" on iTunes, so would that count as "1" in your chart? I think this would become even more of an issue as you move away from Western, first-world countries - do you think the number of musicians in Namibia would be accurately reflected at all in the number of artists tagged as such on last.fm? (I don't, since there are only 22 artists tagged "namibia.")

On the other hand, I doubt that many of the musicians I know who are making original music would show up in any sort of Census/IRS/Department of Labor/other governmental database - they'd show up as "barista" or "auto mechanic" or whatever else it is that they do for a day job. So COD's suggestion of looking at census data is good, but I suspect nowhere near complete.

If you'd like to go poking around in the US census data yourself, it's at American FactFinder.

Then there's the simple fact that the distribution of music has become so decentralized; there are so many routes for artists to connect with an audience that I just don't see how there could be only 1 database that has all musicians on it. I mean, you're a musician and you're on last.fm, but you're not on AllMusic or Pandora. And there have got to be others in the opposite position, or who appear on none of the above.

Plus a user-generated sorting system like the tags on last.fm will be likely to have wild inaccuracies. I just searched for artists tagged "cleveland" there, and my results included Iron Maiden, Slayer, and The Presidents of The United States of America, which . . . . huh ????? I guarantee you none of these bands have anything to do with Cleveland besides rolling through on tour every so often.

My overall point being, not only do I think no such database as you're asking for even exists, but if you want to create even a roughly accurate chart of "musicians per capita" you need to get data on musicians from several different sources and figure out some way of weeding out duplicates & mistakes before comparing your totals with total population data. There's just too many ways the music industry operates both below ground and above ground to rely on a single source.

P.S. It's ACCORDIONISTS who have to be registered with the government, I'm pretty sure . . . . . ;)
posted by soundguy99 at 9:06 AM on December 9, 2012

As I said before, I am not trying to be totally scientific. I am not trying to count Grandma, who plays guitar for the kids at Christmas, or some guy who plays drums with his friends in central Africa. I am trying to get a very rough idea, based on how many have been tagged as "finnish" or "scottish" or whatever in a music database such as last.fm.

I am only counting quasi-above-ground artists- artists or bands people have bothered to add to databases. As far as I'm concerned, those are the only ones that are relevant. As far as demographic data goes, bands per capita will simply be (#artists with tag)/(country population)

So to reiterate, so there's no confusion:

I don't care if it's scientific. I'm looking for very rough data. How can I poll a database like last.fm so I can find out how many artists have been tagged with "Icelandic" or "New Jersey", for example? Is there a tool I can download that will hook into an API?
posted by dunkadunc at 12:02 PM on December 9, 2012