Where Are All The Women in Modern Art?

Art curation has been heavily biased towards supporting male representation in the most elite art institutions. In 1985, a group of anonymous American female artists, the Guerrilla Girls, plastered New York City with 30 different posters. In fact, the group came about when Museum of Modern Art (MoMA) held an exhibition where less than 10% of the artists featured were female. Their work sought to motivate “museums, dealers, curators, critics and artists who they felt were actively responsible for…the exclusion of women and non-white artists from mainstream exhibitions and publications”.

Where did they get that 5% statistic from? They created their own datasets from museum reports and magazines.That was nearly 30 years ago, and we can do better with the data available to us today. And, we can start our analysis with the very museum that sparked the Guerrilla Girls’ call to action.

Since 1929, the Museum of Modern Art has acquired more than 200,000 works of art, of which 138,567 has been catalogued in a dataset released on Github.

Unlike open datasets from other museums, such as the Tate (last updated 2014) and Cooper-Hewitt (2016), MoMA automatically updates its data on a monthly basis. MoMA is one of the largest and most influential museums of modern art in the world. Its collection focuses on modern and contemporary art, including categories such as drawings, prints, photography, architecture, design, film, painting, sculpture, media, and performance.

Surprisingly, although the dataset has been released since 2015, it has only been lightly explored by a handful of individuals and groups. FiveThirtyEight narrowed in on the 2229 paintings at MoMA; researcher Florian Kräutli published a high-level time-series analysis; and other art enthusiasts created one-off charts on specific questions they had (e.g. “How Modern is Modern?” and “MoMA keeps things fresh”). Others even created Twitter bots that use the data from MoMA to create hypothetical art descriptions.

So, have things improved since the 1980s? Or, as critics and academics continue to argue, has very little changed?

The data comes in two separate CSV files: the first with basic metadata for artworks, the second (and much smaller) with demographic information on artists. My analysis focuses on the artworks.csv file, but I had to merge the datasets to include artist gender information from artist.csv. A quick look at the combined datasets has 114,372 individual art pieces, with 13.64% of them created by female artists. Assuming that this proportion holds for the 90,000+ pieces that have yet to be catalogued in the open dataset, this does not speak well of gender equity progress at MoMA given the amount of time passed.

I wanted to get a sense of acquisition trends historically, specifically how long it takes for a new piece of work to be acquired by MoMA. Using a seaborn scatterplot, I was able to quickly group this by gender. The graph reinforces how overwhelmingly male the collection is. We see that the acquisition of art by female artists has increased over time.

There is also a discernible lag between Year Created and Year Acquired, but it is not clear whether this differs between genders. To answer this question, I created a new feature, Art Age, by subtracting Year Created from Year Acquired. I then created Art Age bins for the cross tabulation. We can see a slight trend: the bulk of female artworks (68%) are acquired within 20 years of creation, versus 53% for male artists. A barplot of the means illustrates this more clearly: male artists see an average of 25 years before their art is acquired, female artists see an average of 17.5 years.

Is there a difference in what categories of art are acquired by the museum? The dataset provides us with three increasingly granular options for artwork categories: Department, Classification, and Medium. I went with Department as it captures a high-level overview of the 7 broadest categories of art necessary for exploratory analysis. The most popular category, for both male and female artists, is in Drawings and Prints.

How do acquisition trends break down by art categories? We see in the below two graphs that Drawings & Prints, and Architecture & Design dominate in both subgroups, making it hard to see trends in the smaller categories. Among male artists, we see two years in the 1960s when MoMA likely acquired a large collection through a gift donation. We also see similar, but more numerous acquisition spikes among female artists, with acquisition increasingly steadily over the decades.

What about the creation dates of these artworks? Again we see the bulk of acquired Drawings & Prints being created after the 1980s in both genders with a similar overall trend line. Architecture & Design looks a little different by subgroup: acquired works by female artists spiked sharply during the 1920s and 1930s, whereas male artists see acquisition rates spike in the early 1900s and later in the 1970s.

The Guerrilla Girls and other activist groups of their time did not just focus their critiques on gender equality. They were also interested in racial equality. The dataset provides us with artist nationalities, of which there are 125. Nationality is not a good proxy for race, so the question I want to answer has to be modified slightly: how American is MoMA?

To do so, I created a new feature that split the American-created art pieces from non-Americans. Surprisingly, we see that art created by foreign-born artists dominate the museum’s collection.

In the first chart, we see a similar trend as in the gender segmented analysis. There is a lag of approximately 20 years before artworks are acquired.

The most sought after pieces for both groups are Drawings & Prints followed by Photography, and Architecture and Design.

The dataset is extremely rich and there are further questions we can ask:

• Donors: Who are they? Do they differ in size of donations, types of artists favoured (including by gender and nationality)? This is especially difficult as there are 6722 unique donors to re-classify into smaller bins.
• Dimensions: The dataset provides dimensions and other details for each artwork. Do artists differ in the physical scope of art projects, by gender or nationality or age?
• Age: We can calculate how old an artist was when they created a piece of art. Does art creation cluster at a certain period of life? Does this differ by gender?