Dresses at the Oscars

March 3, 2014

While the Academy Awards are really there to celebrate excellence in filmmaking, they are also essentially the Superbowl of looking at pretty dresses. Or rather, maybe the Academy Awards is the Superbowl of movies, and looking at pretty dresses is the Superbowl commercials of the Oscars. No? Anyway.

If movies are more your thing, check out Irmak’s project Movie vs. Movie. Because today we’re here to talk about dresses.

Just for fun, we created a dataset of famous ladies and the dresses they wore on the Oscars red carpet this year and the ten years previous[1]. Then we created an interactive visualization of dresses (and sometimes suits) worn by the Academy Awards nominees for Best Actress and Best Supporting Actress. Below, we give you some of the insights we unearthed as we poked through this most elegant of datasets.

Red carpet dresses worn by Academy Award nominees, 2004-2014

Each pearl is a nominee (winners are gold). Hover or click to learn more about the actress and designer.


Playing with the viz

This interactive visualization has one “pearl” per nominee, arranged in columns for each year, with Lead Actress nominees in the top half and Supporting Actress nominees in the bottom half. Hover over the pearl to see the actress’ name and her dress designer. Click the pearl (or any designer name) to highlight all of the dresses by that designer.

As you hover and click on different pearls and designers, you’ll see an area graph at the bottom of the visualization, shifting around. This shows the popularity of that designer on the red carpet over time[2]. For instance, you can watch Carolina Herrera disappear from the red carpet after 2009, the same year that dresses by Elie Saab began to gain popularity.

Click on multiple designers to compare the nominees they have dressed. Each subsequent designer gets a new color. When you’ve had fun with that, you can click ‘clear all’ to … ahem… clear all.


Most popular designers

So if we’re going to talk about fashion, we want to know who is the most popular, and who is the biggest winner.

In our dataset of the 11 Oscars ceremonies from 2004 to 2014, we ended up with 284 dresses representing the work of 87 design houses. There were 110 nominees and 22 winners.

Here are the top ten designers on the Oscars red carpet over the last ten years (#dresses):

#OverallAmong NomineesAmong non-Nominees
1. Versace (20) Armani (6) Versace (14)
2. Armani (16) Lanvin (5) Armani (10)
3. Christian Dior (15) Christian Dior (5) Christian Dior (10)
4. Valentino (14) Vera Wang (4) Valentino (10)
5. Marchesa (11) Prada (4) Oscar de la Renta (8)
6. Oscar de la Renta (10) Versace (4) Marchesa (7)
7. Vera Wang (9) Marchesa (4) Vera Wang (5)
8. Lanvin (9) Badgely Mischka (4) Elie Saab (5)
9. Elie Saab (8) Valentino (4) Chanel (5)
10. Prada (8)

And who is the biggest winner? Actually, it’s hard to say. The 22 winners in our sample wore dresses by 18 different designers. The only designers to dress two winners in the last eleven Oscars ceremonies were Christian Dior, Prada, Lanvin, and Tadashi Shoji.

If we had to pick one “winner” for the category of Dressing Oscar Winners, we’d probably have to pick Tadashi Shoji. In five years of Oscars appearances, Tadashi Shoji has sent five dresses down the red carpet, winning two Oscar statues out of three nominees. Even if it didn’t work out for June Squibb this year, if I were looking for a little bit of good juju (you know, when I am nominated for my Academy Award), I might pick out a Tadashi Shoji dress.

Poor Versace

We couldn’t help but notice that for all its red carpet appearances, no one seems to win wearing Versace. Until this Sunday, no one had won wearing Armani, either. At the other end of the popularity spectrum, there were five designers who dressed a winner with their one and only red carpet gown (Guy Laroche 2005 on Hilary Swank, Narciso Rodriguez 2006 on Rachel Weisz, Christian Lacroix 2007 on Helen Mirren, Balmain 2009 on Penelope Cruz, and Yves Saint Laurent 2009 on Kate Winslet).  

This inspired us to dig a little deeper. If there are so few winners wearing the most popular designers, and a handful of winners who were the lone wolf wearing theirs, is this a pattern? Maybe winners tend to dress a little outside of the pack. Since we do like we do, we thought we’d check this out with a few graphs.


Dressing with the herd: a nerd-out series of exploratory charts

For all of the following graphs, the x-axis is the designer’s popularity on the red carpet. This is just the count of total red carpet dresses in our sample. Versace’s popularity is 20, because Versace dresses appeared 20 times in our full sample. Other designers have been less popular: for example, Miu Miu, Dolce & Gabbana, and Yves Saint Laurent have only been worn once, so on our x-axis their popularity is 1.

Always start with a histogram

So first, we decided to check out the number of winners and losers at each level of designer popularity.

absolute_wins_losses_bar.png

So, by now we might know that Versace had 20 gowns in our sample, and a popularity of index of 20 (I love it when we keep the math simple). Of these 20 gowns, 4 were nominees and zero were winners, so the bar for “designers with 20 gowns” has a single red bar with height 4. So we can tell that nominees wearing dresses by designers of popularity level 20 won 0 out of 4 nominations.

There were two designers with a popularity of 9: Lanvin and Vera Wang each had 9 gowns on the red carpet from 2004 to 2014. Nine of these happened to be nominees, and thus two out of nine nominees wearing dresses with popularity index of 9 were winners.

What does this graph show us? Anything? Well, for one thing we can see that the most winners for any popularity level came from the singletons. And, for some reason, popularity level 5 seems to be a bit more successful than average. But there don’t really seem to be any exciting trends. Maybe if we scaled the data to take into account the different numbers of dresses in each category, that would make things easier to see.


Does designer popularity hang with win ratio?

The next graph shows the same information in the form of a winning percentage instead of absolute counts. So Versace, the most popular designer, has a win percentage of 0%, and the designers at popularity of 9 (Lanvin and Vera Wang) had a combined win percentage of 22% (2 of 9).

percent_vs_total_clean.png

Ok, this is good. After transforming the bars from raw counts to win ratios, the level of 5 still stands out above its neighbors, but level #1 doesn’t look quite so special anymore. It doesn't look like there are any trends connecting designer popularity and the probability of winning. Let’s see if there is anything interesting to say about the differences between nominees and non-nominees.


Under the spotlight: do nominees dress more with the pack?

The next graph illustrates the absolute number of dresses again, this time including the designers of the non-nominees (presenters and performers) as well. Here you can see that the red carpet designers are rather segmented; of the 284 total number of dresses, the top 5 most popular designers only count for 76 dresses (27%) of the red carpet gowns. 102 dresses (36%) come from “unique” designers, here defined as designers that have been represented at the Oscars 3 times or less. 39 designers (14%) have only been represented at the Oscars once.

abs_wins_losses_presenters.png

So, even though it seems like we hear about the same designers over and over again, most red carpet dresses at the Oscars are from more unique designers. This could be due to each star wanting to do their best to stand out, or by conscientious consumers of fashion doing their part to provide exposure to new designers. Of course, it’s likely that arrangements are made from time to time that make the dress selection beneficial to designer and attendee both.


Designer popularity by group, normalized for bin size

The last thing to check out is whether the choice of dress popularity differs between nominees and non-nominees. It’s hard to see that on the graph above, since the raw numbers make it difficult to compare proportions. To check this out, our final graph is normalized by designer popularity (i.e by the height of the bars on the previous graph). Here you can see the proportion of wins, losses, and presenters by designer popularity.

norm_wins_losses_presenters.png

So, not much of a trend here, either. Since the division between red and blue stays pretty much at the same height at each level of popularity, it doesn’t look like there’s any strong trend for one group to dress more uniquely and one to stick more with the pack.

  1. Technically, we had an army of strangers create the dataset via Mechanical Turk.
  2. Our full sample included all nominees in the categories of Best Actress in a Leading Role and Best Actress in a Supporting Role, as well as the women who acted as Presenters and Performers in the Academy Awards ceremony themselves. We threw out any Presenter or Performer whose dress designer we could not find (there were just a few of these). We left out tons of other people who show up on the red carpet and get their photos taken, like women nominated in non-acting categories, wives and dates of male nominees and presenters, rando socialites, and red carpet meta-celebrities like Giuliana Rancic and Joan Rivers.

contributors to this post

headshot of Laurie Skelly
Laurie
headshot of Bo Peng
Bo