The A la carte Composer Polls

Between October of last year and February of this year, I followed up my "A la carte" polls of most-liked classical works with a series of 43 A la carte Composer Polls. The same principles applied: in each poll, participants selected whichever of 14 composers they liked, with an option to say "I don't know or don't like any of them". 43 polls allowed me to slightly exceed 600 composers, which sounds like a lot - well, it is a lot! - but nonetheless required me to be more selective than I wished. The primary motivation for this set of polls was a thread started by TalkClassical regular mmsbls, which asked "Do you enjoy all the "top" composers?". I had (as my first contribution to the thread stated) already thought about doing composer polls, but I was spurred into action by the discovery that the 3 composers that seemed to be most responsible for people saying they didn't enjoy all of mmsbls's "top 10" were Wagner, Schumann and Handel - which was matched by my analysis of the popularity of these composers' works on the A la carte polls.

I can't remember how many composers I was going to include at first, or at what point I knew I'd have to have a "don't know/don't like" option (with a max of 15 options, having that option means I end up with multiples of 14 composers rather than 15, which is annoying if I want a total number ending in 0!) but eventually I settled on 504. 

But which composers should I include? There's a couple of hundred obvious ones that had to be included, but who then? Fortunately I have my painstakingly constructed spreadsheet that gives number of Arkiv recordings (now several years out of date but still useful) and length of 1980 New Grove entry for many many composers, so it was reasonably easy to select most of them. The difficulty came in deciding where to draw the line, or rather multiple lines because I wanted to be sure that the whole history of "classical" was adequately covered. One (Johannes Eccard) got no votes whatsoever, so it's reasonable to imagine that there are hundreds of other composers more popular than him.

For the nerd record, here's exactly how I settled on the final list. First I sorted the composers by birth decade and then by Arkiv recordings, and gave 1 point to the top 5 in each decade (or fewer than 5 if  it was a "weak" decade) and 0.5 points to any others that had a lot of recordings (this being a somewhat arbitrary decision). Then I did the same for New Grove entry length, but not giving out 0.5 scores in this case. Seeing as my New Grove is the 1980 edition, this put modern composers at a severe disadvantage, but the online version lets you know how long (in words) a composer's entry is, so I got those figures for all composers born in the 1920s to 1950s. 

Totting up the scores gave me 514 composers but then it got tricky again because some of them still seemed unlikely candidates. Anyone with 2 points was definitely included (oddly, there was no one with 1.5 points). But going through earlier composers that scored high with New Grove I found many that had relatively few recordings and thus were unlikely to be known, certainly compared with other less-early composers. For instance, Leonhard Lechner's New Grove entry is about 6 times longer than Franz Schmidt's, but the latter has 5 times as many recordings, so although Schmidt failed on both original criteria he's a much better candidate (in the end, 26% of people said they liked him). All told, 66 composers were parachuted in instead of others, and 10 more were removed altogether, giving a final list of 504, filling out exactly 36 polls. Not as ambitious as the works polls, but that was an open-ended project that evolved over time, whereas this had to be planned in advance. One important aspect was to ensure the composers were well balanced; I needed to have popular ones in each poll to draw people in. Cue some sort of ordering system I can't quite remember that also took into account my need to spread the composers around chronologically too.

And then when I'd sorted that out and actually posted a few polls on TC, I decided I wanted to increase the number of composers to 602, i.e. 7 more polls! In particular, I wanted more living composers but also there were several composers about whose popularity I was curious or that I thought would prove more popular than their position in the selection process might indicate. Selecting those 98 composers was basically my own judgement.

It's interesting now to compare the results of the polls to the way in which the composers were selected. Although Bernardo Pasquini was selected in the top 5 composers of the 1630s by both Arkiv and New Grove, he scored under 2% (2 out of 119 voters liked him). Whereas Kurt Atterberg, who didn't even make the original list of 504 composers, scored 29% and finished in the top 200 composers. But then again, David Gillingham was parachuted into the list of 504 and he scored less than 1%! So I'm not necessarily the best judge. But as I always say, a data point is a data point, and now we know Pasquini and Gillingham aren't popular composers.

I was pleased with the response to the polls but was well aware of the ineradicable problems with trying to do a poll series on TC: you need enough voters for the results to have some sort of statistical reliability, but you also need enough regular voters for you to be able to compare one poll with another. After a bit of an advertising push once the polls were completed, I was delighted that all but one poll ended up with at least 100 voters (poll 40 had 99). All told, 235 people participated, and I'd so love it if they'd all voted in all of the polls. But 88 people did vote in all of them, which was enough in itself to make some generalisations.

One rather ungratifying aspect of the polls was the appearance in my discussion of the results of a trollish individual named MartinD, who really didn't approve of the poll methodology despite my constant protestation that he was taking things too seriously. Inevitably the fact that my top 3 wasn't Bach-Mozart-Beethoven (in some order) drew ire, but the whole thing was a disgrace as far as he was concerned. This MartinD character was active on TC for about 3 months, and the majority of his posts seem to have been direct or oblique attacks on me. Weird. One other person named Genoveva also took me to task during poll 8, for similar reasons. I know that with 50 or 100 people you certainly can't extrapolate any result to the wider world, but it's very frustrating when serious people go on the offensive and try to, well, ruin other people's fun. The criticisms helped me better understand what I was trying to do, and not trying to do, but they left a bad taste in the mouth and were a contributor to my slowness to revisit the results. Side note: by a remarkable coincidence, Genoveva's last activity on TC was a couple of days before MartinD joined the forum.

The "leaderboard" is on Nereffid Supplemental. Unfortunately some people put too much store in the rankings - I fall back on them myself sometimes! - but it's the most convenient shorthand. The % score achieved by each composer is the most important thing, but it's handier to think "that composer's in the top 100" than "that composer got more than about 37%".

The A La Carte polls

If you go to the Google Sites page Nereffid supplemental, you can see the results of several months' worth of polls I conducted on TalkClassical last year and this.

TC users (some of them, at any rate) like to produce "recommended lists" that rank favourite works; just recently I hosted the Pre-1700 Works list. But the way these are done requires that participants have to choose which work to favour over another, and there's often going to be a clash between what you personally prefer and what you think should be on a list (for example, there's the desire to spread things round, and not let one composer dominate). When the Post-1950 Works list started, I was somewhat troubled - first because the voting system was terrible and second because some participants stated clearly that post-1950 music of a certain sort was not welcome (ie, Shostakovich). Round about this time I also started to become more interested in the notion of the classical "canon" and how one might go about quantifying it.

So I hit on the notion of a basic poll, where the voter is faced with a list of works and must simply pick the ones they like; putting all the results together would effectively produce a "recommended list" but without the complication of any individual having to rank anything. One interesting aspect of this is that a work that gets plenty of votes is not only well-liked but also well-known. There would no doubt be many works that might be seen as canonical by a handful of people but to the wider listening public they would be unfamiliar - and post-1950 works are a prime example. My original notion was that the polls would each focus on a particular year, so that bit-by-bit I could produce a list to reveal which post-1950 works people really liked. After a couple of dozen polls I got the feeling that some voters just didn't bother participating if it was an era they had no interest in (it's a curious thing; I added the "I don't know any of these works" option to overcome this, but it doesn't seem to encourage everyone), so I switched to themed polls and then, not wishing to put so much intellectual effort in if it wasn't going to pay off, just randomized lists of works. I got past 200 polls but they appeared to be running out of steam so I called a halt; but that was over 2,600 works, with at least 30 people (usually quite a few more) offering an opinion on each one.

There are the inevitable quirks you get when the polling population is different for each poll, and certainly the TC population overall has a few quirks too, but the collated results are largely what one might expect. Back in February, when I passed 2,000 works, I started a discussion thread on the topic.

It turns out that, yes, Shostakovich does seem to dominate post-1950 music!

One of the most fascinating discoveries was that - bearing in mind that TalkClassical's active members are by definition people who are enthusiastic about classical music - only 20% of the works were liked by at least half of the voters. Bear in mind too that the polls covered practically all of the "canonical" works (I didn't get all Beethoven's piano sonatas in, for instance, but that's the highest level of omission I'm talking about). And only 5% of the works (130 of them) were liked by at least two-thirds of voters. And of those 130, 43 are by Beethoven and Mozart! So what on earth is the canon? Surely, for instance, La Traviata and Parsifal are part of it? Well, they were liked by about 42% of voters in the A la carte polls, putting them well below Mendelssohn's 1st string quartet, Prokofiev's Scythian Suite, Chopin's op.41 Mazurkas, Bach's Toccata in C minor BWV 911... Clearly, the canon can't be defined by votes alone, and there are numerous other factors involved (vocal works generally are less popular than orchestral music in the polls; and some composers are so popular that even when I invented a fake Bach cantata it got enough votes to put it in the top three-quarters of the overall rankings!). But still, the undeniable implication of the poll results is that a work that is considered canon may well be unliked or unknown by a majority of self-professed classical music enthusiasts.

This post has gone on for a bit; if I remain in a posting mood there'll be plenty more analysis and vague thoughts to come...