I don't understand why they expected to see 11% back-to-back female songs instead of 11% * 11% = 1.2%
> If you listened to this station non-stop from midnight to 11:59pm today, you’d likely only hear 3 back-to-back songs by women, compared to 245 from men.
> Songs by women are already severely underplayed, making up only 11% of total plays in 2022. Even at this low rate of play, we’d expect to see 30-40 back-to-back songs by women for this same day, like in these 10 “coin flip” simulations. In reality, we see 3.
It would be nicer if they showed their work. Simulating this in Javascript with 300 random songs with 11% female artists results in:
let songs = new Array(300).fill(0).map(x => Math.random() > .11 ? 'm' : 'f');
let bb = {f: 0, m: 0};
for (let i = 1; i < songs.length; ++i) {
if (songs[i-1] === songs[i]) ++bb[songs[i]]
};
bb.pf = 100 * bb.f/(songs.length - 1);
bb.pm = 100 * bb.m/(songs.length - 1);
console.log(bb);
The output is 3 female back-to-back songs and 240 male. (Your random numbers might vary.)
> We looked at 19 dates throughout 2022 for this same radio station and found that out of 6,474 songs, only 64 (0.99%) were back-to-back songs by women plays. For back-to-back songs by men, it was 4,231 (65.53%). Back-to-back mixed-gender ensembles and collaborations account for 36 (0.56%) songs.
Y'all are 100% right – something wonky was going on! I’ve since made adjustments and updated the piece with a correction.
The visual and accompanying text were using the number of all women plays, not just the back-to-back ones.
It’s more likely that the simulation back-to-back numbers would fall between 7—17 a day NOT the originally stated 30–40. For the particular station in the intro, San Antonio (KCYY-FM), the number of women’s back-to-back plays that they actually played isn’t completely outside of the realm of possibilities if left up to chance, but we would be more likely to see higher numbers. For 17 other stations, the observed numbers do fall outside the simulation distribution. (See the histograms in the methodology section. Note that they are for the full 19 days.)
Thank y'all again for raising this issue! The data is only good as the humans behind it, and I definitely bungled this a bit.
That's still about twice or so the number that the simulations of sxp and myself give, but I think that we both used a different definition of back-to-back than you. We both are not counting the first song in a run of consecutive same gender artist songs as being back-to-back.
If the first song in a run is included in the back-to-back count then my simulation seems to be matching your updated simulations. If 11% of the songs are from female artists for example I get that ~2.27% of the total plays should be part of back-to-back runs of female artists.
If the first song in a run does not count as back-to-back, that was ~1.22%.
Note that the percentages for "first doesn't count" B2B simulations come out pretty close to the B2B percentages given in the article for most stations. That could just be a coincidence, but it is probably worth double checking how you calculated the B2B percentages for the actual station data to make sure that you are using "first counts" B2B.
If the station stats are using "first counts", then it looks like most of the stations are playing about 1/2 as many female artist songs B2B as they would if scheduling were random. But if the station stats are using "first doesn't count", then it looks like most are consistent with random playlists.
Reporting back again… We’ve checked the overall simulations and they check out when you use the summary data here (https://github.com/the-pudding/country-radio-data/blob/main/...) for the total number of songs (total_COUNT) and the percentage of women’s songs (onlyWomenSongs_PERCENT) for each station. For San Antonio (KCYY-FM), that would mean that the total observed number of women’s back-to-back songs (64) is within the realm of possibilities across all 19 days if left up to chance. But, the observed number is on the low end of the overall distribution, with the majority of numbers landing between 70-100. I have been able to recreate this using sxp’s original Javascript. When we’re talking about a difference of say 20 songs (across 19 days) you can’t visually see it, which is why the single day example was a poor choice for communicating this, especially because San Antonio (KCYY-FM) is one of the “better” stations.
We’ve removed the simulation section from the intro, made it clear what data the simulations use further into the story, and added a correction. Thank you again for keeping us honest.
I agree. They seem to have greatly overestimated the expected number of back-to-back female songs. I too did a quick simulation and my numbers are close to yours, and close to the radio station data in the article. My code is appended below.
Usage is "./bb.pl num_songs percent_by_women". E.g., "./bb_pl 100000 11" would do a 100000 song playlist where 11% of the songs are by women. Here was the output of a run with those arguments:
79227 MM, 1228 FF
9772 MF, 9772 FM
79.23% MM, 1.23% FF
9.77% MM, 9.77% FF
The first pair of output lines is saying that 79227 songs were songs by males (M) played immediately after an M song, 1228 songs were songs by females (F) played immediately after an F song, 9772 were F songs played immediately after an M song, and 9772 were M songs played immediately after an F song.
The second pair of output lines gives those as percentages of total songs.
#!/usr/bin/env perl
use strict;
use List::Util;
my $T_plays = shift @ARGV || die "Args: num_songs percent_by_women\n";
my $F_percent = shift @ARGV || die "Args: num_songs percent_by_women\n";
my $F_plays = int($T_plays * $F_percent / 100 + 0.5);
my $M_plays = $T_plays - $F_plays;
my @slot;
push @slot, 'M' foreach 1..$M_plays;
push @slot, 'F' foreach 1..$F_plays;
@slot = List::Util::shuffle @slot;
my $prior = '';
my $MM = 0;
my $FF = 0;
my $MF = 0;
my $FM = 0;
foreach (@slot) {
if ($_ eq $prior) {
if ($_ eq 'M') { ++$MM; }
else { ++$FF; }
} elsif ($prior ne '') {
if ($_ eq 'M') { ++$FM; }
else { ++$MF; }
}
$prior = $_;
}
print "$MM MM, $FF FF\n";
print "$MF MF, $FM FM\n";
print "\n";
printf "%.2f%% MM, %.2f%% FF\n", 100*$MM/$T_plays, 100*$FF/$T_plays;
printf "%.2f%% MM, %.2f%% FF\n", 100*$MF/$T_plays, 100*$FM/$T_plays;
I think the article needs to describe the population of overall songs to support this claim. What is the percentage of men, women, mixed gender, transgender, sexual orientation, and race? Are there many top 40 hits for country music songs performed by women? Or by LGBTQIA’s?
This seems like a simple comparison before claiming inquity. The closest is when they talk about a peak of female sung songs of 33% and a valley of 11%. But what’s the overall and what is the breakdown of songs that are likely to be played?
There aren’t many black country music singers so I think a station playing few songs is actually quite equitable. Same for sexual orientation. I don’t think that it’s very smart to spread overall total population characteristics over a specific non-random selection.
Just because there are 0.0% songs played by lgbtqia country artists doesn’t mean it’s a lack of equity. Popularity is not equitable by definition since it tends to focus attention on a few acts (Matthew principle in effect [0]) so it seems odd to talk about such a thing without more context and detail.
Just because the population is 50% female doesn’t mean that 50% of country songs should be sung by women or that 50% of country music radio stations will be women. There is bound to be variation just because country music isn’t the same as the overall population.
I wouldn’t say that the Tony awards is inequitable because the gender and sexual orientation of nominees and viewers doesn’t match the national demographics.
I also think that, in context, this is not exactly surprising. Country has been political for a while, a fair amount of people have been blackballed from country like the Chicks (nee. the Dixie chicks), and performers who don’t align with conservative values in country get a lot of heat (e.g. Kacey Musgraves, Beyonce performing at the CMAs): https://www.vox.com/platform/amp/culture/2016/11/4/13521928/...
And the entire notion of ‘equity’ is the current notion of ‘woke’ for others.
I agree that country music is more conservative. But what does that have to do with gender of singers? Are conservative men more likely to sing country songs? Or not listen to women singers? Or are there more men that listen to country music.
I expect that conservative people are also split 50% women just like most other large groups.
The kind of macho patriotism present in country is pretty masculine, so it does somewhat limit the topics that women can sing about and be popular for.
It's not a 50% split, not in my contry at least. I just checked the internet and it doesn't seem to be a 50% split in the US either [0].
And all that Twitter/public partisanship seems (at least to me) even more skewed against women, whether it's right wing or left-wing. And more authoritative right, it's between 5 and 10% (but women in those groups seemed to have a lot of power a few years ago). Here the source is 'trust me bro', but I used to be in an antifa surveillance group. We documented privately people going to far right meeting/exercises, some of us infiltrated the groups, and if the information was interesting, contacted journalist with the information (the last big 'revelation' about neonazis in the French army was from a group similar to mine). The 90/10 split male/female friends is what we noticed in meetings (in bars it was 95/5). Which was weird because it roughly matched the street occupation groups like LJG (watch groups were close to 50/50).
I know we're not supposed to bitch about UX, but could anyone get past 1/2 way? My totally-not flagship Android crashes trying to read this in Firefox. I was hooked but never got to the conclusion.
This may be because I'm a tab hoarder (usually my CPU usage is under 3% total even with everything open), but about halfway thru, total usage jumped to about 33%. an 11x increase!
On a totally unrelated note... if anybody knows how to better cool an HP Elitedesk 705 G4 with a 2400GE inside, I'm all ears.
I remember a long time ago there was a website that you could input an FM radio station (in the US) and time and it would tell you what song was played. It was really cool, but as far as I know it shut down and never came back.
I liked the scrolling and formatting but I don’t think it was a great way to display data because I don’t think it supported their point. Although I guess it’s still better than if it was just a bad point with a static pdf.
> If you listened to this station non-stop from midnight to 11:59pm today, you’d likely only hear 3 back-to-back songs by women, compared to 245 from men.
> Songs by women are already severely underplayed, making up only 11% of total plays in 2022. Even at this low rate of play, we’d expect to see 30-40 back-to-back songs by women for this same day, like in these 10 “coin flip” simulations. In reality, we see 3.
It would be nicer if they showed their work. Simulating this in Javascript with 300 random songs with 11% female artists results in:
The output is 3 female back-to-back songs and 240 male. (Your random numbers might vary.) > We looked at 19 dates throughout 2022 for this same radio station and found that out of 6,474 songs, only 64 (0.99%) were back-to-back songs by women plays. For back-to-back songs by men, it was 4,231 (65.53%). Back-to-back mixed-gender ensembles and collaborations account for 36 (0.56%) songs.with `...new Array(6474)...`: