On Race and Strictly Come Dancing - Some Numbers

[danieldwilliam]

Further to some conversation on my earlier post about racism on Strictly I've done a little data analysis.

(EDIT: Which has been edited slightly because I'd mis-categorised Ricky Whittle. He finished on the podium so moving him from category to another doesn't significantly alter my findings.)

These are some provisional observations. I'd like to get my daughter involved. She's doing joint honours Psychology and Biology and knows more about statistics than I do.

I've been keen for comments and questions.

I've created a data table of Strictly contestants.

Name

Week Out

Series

BAME Yes / No

How many weeks before the final they were eliminated.

I'm pulling some data from that table.

Up to series 15 there have been 207 contestants in Strictly Come Dancing. 37 of them have been BAME  or 17.87%. This  compares with 12.9% Non-White in the population of the UK overall. The total number of weeks danced is 1,620. 299 weeks danced by BAME contestants, 1,334 by non-BAME contestants, or 18.46% plays 81.54%

(I'm basing my BAME assessement on Wikipedia and my own memory. I haven't done a double check on every single contestent. Therefore I may have missed BAME people from communities like the Romany. I'm not too concerned about that as I think the question I'm trying to address is whether people who are visually obviously non-white are disadvantaged - in Strictly. EDIT: I did miss Ricky Whittle but by good fortune checking Ricky Whittle reminded me that American Gods existed on Amazon Prime - so I started watching that.)

Podium finishes. 11 out of 46 podium finishers have been BAME - 23.9%.

 

I've been trying to see if there is a trend for BAME contestants to go out earlier than average.

I have divided the competition in to First Half and Second Half based on the number of weeks before the final. 13 out of 72 First Half leavers are BAME 18.1% vs 23 out of 135 Second Half leavers, 17.0%. (NB not all Strictly series are the same length. Series 1 was 7 weeks, making everyone in it a Second Half Leaver. This is where my stats ability becomes stretched.)

Looking at the population of cummulative eliminations (e.g. there are 6 people who have been eliminated in Week 1, 0 BAME and 6 non-BAME, 100 % non-BAMe, by Week 2 4 BAME, 21 non-BAME, for 19% of total eliminations being BAME - compared to a to total population of 18.46% BAME contestants.

In Week 3 the figures are 8/36, or 22.22%, in Week 4, 9/51 or 17.65% - slightly less than the proportion of BAME contestants overall, which is 18.46%. This continues all the way through, with BAME contestants being slightly less likely to have gone out once we get past Week 3. This fits with the over representation of BAME contestants in the final (23.9% vs 18.46% of contestants.)

Excluding Week 1 Which has no BAME eliminations (NB many series had no eliminations in Week 1 so Goldie, first elimination but eliminated in Week 2) the average variance from of cummulative BAME eliminations from 17.65% is 1%. 10 weeks out 13 have  cummulatively fewer BAME contestants to have been eliminated.

There is a blip in Week 3 but it settles down pretty quickly. Where my stats fail me is working out if the blip in Week 3 is significant or just a blip. My gut feel says it's a blip. By Week 3 the cummulative % of BAME eliminations is 22.22% compared to an expectation of 18.46%

So, to conclude, BAME contestants are slightly over represented on Strictly. They are more likely to win, and more likely to finish on the podium than non-BAME contestants. Given that they are over-represented in the population as a whole they don't seem particularly likely to go out early. There seems to be a slight Week 3 effect but I think it's noise rather than signal.

Very, very happy to take suggestions on how this data analysis might be improved.

(I will also caveat this by saying that I've produced the data tables pretty quickly, there may be errors. EDIT - there were errrors - I  mis-categorised Ricky Whittle, who finished on the podium with Natalie Lowe a few years ago. I think I'd confused him with Ricky from Eastenders. As he made it all the way to the last week he doesn't impact the numbers or the conclusions for early entry much and he increases the representation of BAME contestants in the overall population and on the podium.)

EDIT Things that I found out following some questions

I think you can say that there is some evidence that BAME males do better than expected and BAME women slightly less well than expected but this looks like it might be to do with slightly fewer contestants being BAME women than we might expect given the overall BAME population in Strictly.

Thinking about a method to test is people are going out "early than the ought to" proved difficult.

The Week 3 blip might random or it might be because voters, unsure who to back early on, disproportionately back non-BAME contestants out of some racially biased default.

Comments

[mountainkiss]

Are the patterns similar or different for men and women?

[miss_s_b]

This is really interesting. I wonder why the perception has taken hold so strongly that racism is a factor?

[danieldwilliam]

I can think of two reasons which together might explain the perception.

1) BAME contestants are slightly over represented per capital on the show. BAME people are probably under-represented on television generally so it looks like lots of BAME people are being eliminated even if that is not the case in proportion to the actual number of BAME contestants.

2) There is a Week 3 blip, which even if it is random chance, might look like racial motivated voting.

(I also think that when Goldie went out unexpectably that fuelled the narrative.)

[danieldwilliam]

Do you mean, men and women overall or for BAME men and BAME women?

[miss_s_b]

Also Chizzy last year.
And Charles being bottom two this week really shocked me, because he's been really good.

[danieldwilliam]

I don't think Chizzy was that unexpected.

But I also am surprised by Charles. He's a much, much better, more entertaining dancer than two consecutive dance-offs suggest. He seems charming, urbane even. I am given to understand that he is considered handsome.

[miss_s_b]

He certainly falls into my "well, I would" category.

[alithea]

Yeah, I think maybe it is the week 3 blip - what it feels like to me is that early on, if you're not sure who is going to be in the bottom two because the worst dancers are likely to get the sympathy or amusement votes, a BME solidly mid-table contestant is the most likely candidate. But that sort of hypothesis requires judgement calls that I'm not sure you can make scientifically robustly in order to properly test it with statistics.

[danieldwilliam]

I could well believe that sufficient voters in Week 3 are not voting for a BAME contestant because latent racism in the context of a crowded and unknown field puts them off and for this to be a stronge emough effect for weaker BAME contestants to be slightly disadvantaged.

But I think you are spot on about testing it with the data I have. I think you'd have to some study that was a bit closer to the voters.

I *think* what I've managed to establish is that whatever is going on is weak enough not to be obvious in the data and it might just be noise but I don't think I've established that there is *definately* nothing going on.

[danieldwilliam]

There may be a queue.

[mountainkiss]

The latter, sorry.

[danieldwilliam]

A bit difficult to tell.

Male contestants = 102 (BAME men 20 contestants, non-BAME 82, 19.61% of men are BAME.

Female constestants = 105 (BAME 16, non-BAME, 89, 15.24% of women are BAME.

Both compared to 17.65% BAME population in Strictly and 12.9% in the UK

Of the total population on Strictly

9.66% are BAME men
39.61% are non-BAME men
7.73% are BAME women
43.00% are non-BAME women.

Looking at weeks danced

9.01% are BAME men
37.78% are non-BAME men
8.64% are BAME women
44.57% are non-BAME women.

Men dance less than women, BAME men dance more than BAME women. NB the population is small enough that one or two people could make the difference. If Vick makes it to the final and Charles and Danny leave before Blackpool that might unwind the difference

Overall 9 male winners, 6 female winners but 26 podium finishes for women vs 20 for men.

For BAME contestants

3 male winners, 1 female, podium finishes 5 males, 5 females.

The Week 3 blip seems equally gendered.

So I think you can say that there is some evidence that BAME males do better than expected and BAME women slightly less well than expected but this looks like it might be to do with slightly fewer contestants being BAME women than we might expect given the overall BAME population in Strictly.

[mountainkiss]

The winner vs. podium thing is interesting because of course the judges only influence the latter but do so significantly. So one can theoretically argue that high representation on the podium is a sign of lack of bias but also that a low rate of transition from podium to winning is the reverse. I asked the question because off the top of my head I could think of three BAME male winners (Ore, Louis Smith, Mark Ramprakash) but could not think of a female one. Off the top of my head is not statistics, though.

[danieldwilliam]

Yes, I think one could argue along those lines.

Female BAME winner is the famous Alesha Dixon (poacher turned gamekeeper).

Other podium finishes

Denise Lewis
Chelsee Healey
Natalie Gumede
Alexandra Burke

BAME podium

Colin Jackson
Simon Webbe

Jade Johnson from 2009 retired injured. She seemed to be doing okay but that was the year that Chris Hollins and Ola Jordan's Charleston won Strictly single-handedly.

[mountainkiss]

Alesha, of course.

[mountainkiss]

What about ethnicity of pro dancers?

[alithea]

Yes, I think a reasonable interpretation of your results is that racism doesn't stop good dancers winning/ making the final but *may* be a factor in less good dancers going out earlier than we might expect.

The proportion of earlier than expected exits that are BME might be an interesting stat to look at but again there's that judgement call issue, especially as I suspect those of us who are very aware of this might well be based towards considering a BME exit too early. But maybe you could devise a methodology to take their judges score/ table position into account to reduce the unmeasurable judgement call aspect there.

[danieldwilliam]

That's a harder question I think.

I think Oti Mabuse is the only black pro dancer who has ever been on Strictly.

There are large numbers of non-British European ethnicities. Eastern European women and Mediterranian men. For example; Olo, Kristina, Nadiya, Camilla, Lillia, two Katyas and Gionvanni, Vincent, Jared, Gorka).

How much they are percieved as different by the voting audience I don't know. Are the audience judging by appearance or by accent? Are some nationalities favoured? Russian sex-bombs and Latin lovers to the front!

[danieldwilliam]

I think one could look at something like average score and some factor of score improvement and then cross reference that with week of exit.

That would help.

I fear we might get to the point where noise overwhelms signal.

(It doesn't solve the Chris Hollins or Jamelia problems,. Chris Hollins won as a result of Ola's choreography and his personality making him more entertaining and watchable. What F called SeeMoreAbility. Jamelia, I think, carried in a reputation from her earlier career clashes that perhaps made her unpopular.)

I'd be reluctant to rely too heavily on my judgement in this because a) it's exactly the judgement of white voters that is under suspicion here and b) I have my own peculiarities - I think Denise van Outen should win every series and c) what exactly do we mean by "going out to early"?

[alithea]

Certainly I think by the time you've got a methodology for looking at whether the correlation between scoring & improvement and week of exit is different between white and BME dancers, there's not going to be enough data for any difference to be statistically significant. And yes, it still leaves the personality issue, and of course the fact that judgements on personality are also subject to racial bias - black women being more likely to be judged as divas likely affected both Jamelia and Heather Small...

[alithea]

Oh and yes, choreography too when it comes to the final - Erin's bizarre show dance with mannequins shafted Colin Jackson.

Ricky Whittle

[danieldwilliam]

I have mis categorised Ricky Whittle.

[danieldwilliam]

That's a good point on the perception of personallity being racially driven.

[danieldwilliam]

There's a lot in the overall package of the contestant to consider.

I think you're point that second-order considerations (such as views on personality) might have a racially discriminatory component is very helpful.

[mountainkiss]

Is there any difference in how low they have to get in judges’ rankings before being in the dance off? Am thinking of Jamelia who was consistently there despite being highly scored by judges. Am I making any sense?

[danieldwilliam]

I don't think there is a large enough population to do any valid statistical analysis. There aren't many BAME constestants who have gone out about half way through the season.

Looking at Jamielia specifically. She finished in 8th place, going out in week 9. Top score 32 in week 4, lowest score 21 in weeks 1 and 2. She was below average in Week 1 and Week 2 by a few points. Her average score overall was 26.22, which is the 8th highest score. In line with her actual performance of 8th.

In Week 2 when she was in the dance off with a combined Week 1 & 2 score of 42there were 4 people in the dance off with combined scores lower than her. These include Carol and Jeremy.

Carol Kirkwood and Pasha bucked the trend, average of 18 points, 14th out ouf 15, but placed 10th in the competition. As did Jeremy Vine, average of 20.25 (12th) but finished 9th. Also, Katie Derham who averaged 28.40 (7th) but finished 4th.

The average of all dances up to her elimination was 29.59 (vs 26.22). In the weeks she was in the dance off she was an average of 7.8 points higher than the lowest scoring dance but an average of 2.38 points below the average score that week.

So, she has probably been in the dance off more than her scores indicated - but her week of eventual exit was about right. The dance off mechanism saw her progress against 4 weaker dancers.

One could lay this at the door of overt racism, people not voting for a BAME woman, you could lay it at the door of a covert racism, people taking against her based on her (BAME) personality. You could lay it at the door of Carol and Jeremy being fan favourites. Or it could be that people didn't like Jamelia much.

[mountainkiss]

I was thinking that there would be statistics around ratio of score to tenure for each season (mean, standard deviation) and we could then look at the BAME contestants and see if they are within margin of error. From what you write, I'm pretty sure they would be, though. There might be a similar but better way to do this.

[danieldwilliam]

Would a scatter plot of tenure and average score work?

[mountainkiss]

No because I think you need to know standard deviation to gauge whether a given ratio is outside expected range.

[danieldwilliam]

I was thinking of comparing the best fit lines for BAME and non-BAME populations.

[mountainkiss]

I’m not sure that would tell us whether the difference was within expected range. But open to correction from better statisticians than I (many).

[danieldwilliam]

Ah - I see the problem.