1 of 49

Queering

the Data

Jan Diehm, @jadiehm

Sarah Serpas, @sarahserp

NLGJA 2020

2 of 49

Y’all excited to nerd out with us?

3 of 49

The rundown

1. The People

2. The Projects

Sarah Serpas (she/her)

Urban Planner/Demographer

Jan Diehm (she/her)

Data Journalist

4 of 49

First off, data about people is complicated.

5 of 49

How do you study a population?

So let’s break it down. How do you study a population? How do you measure characteristics of a group of people?

If we wanted to know about individuals, we might do profiles or interviews. But to measure overall trends, we need to know about a lot more people than we could probably interview - especially for a population of hundreds of millions of people like the U.S. To make this data comparable, we have to do some grouping. We would place people into groups based on shared characteristics, identity, or experiences. Then we could compare between the groups.

Generally, we start with a question.

Let’s say that for this group of pretty people, we wanted to find out how many of them are identify as queer, and if the ones who are queer are more likely to have short hair cuts. We would likely send them a survey. But we probably wouldn’t want to read through paragraphs about their thoughts on their identity, so we’d probably make most of the survey questions multiple choice. We could simply say “Do you identify as Queer - yes or no?” or ask a more nuanced question with a list of different identities for them to select from, which we would then classify as queer.

We could also ask the hair length question in many different ways. And the essential question is who defines how short is “short”? Would we ask them details about their hair, then we decide what is short hair or not, or would we ask “is your hair short - yes or no?” letting them define what short hair means to them?

Lastly, we’re only measuring one moment in time - what if their hair is long at the time of the survey and they got a buzz cut the next day? We’d understandably miss something in the data there.

This is a weird example, but gets at some of the nuances we have to consider in population research and how questions are asked based on the needs of the people asking.

6 of 49

Options don’t always reflect people...

REALITY

(Spectrum of identities, experiences)

SURVEY/CENSUS FORMS

(check boxes that don’t reflect full identity or experience)

There are multitudes of identities and experiences people have that they are often asked to check a box for. And oftentimes the categories people are given on survey forms don’t match their whole identity or experience. On the slide I have a simple example of a rainbow gradient (to keep things, yknow, gay). Imagine that a person could fall anywhere along that spectrum. But a survey arrives in the mail, or they have employment forms, or tax forms, or any other source of data collection and are given those five color box options. What would the people experiencing “orange” check? Pink? Turquoise?

This is something we have to keep in mind as we look into data about queer and trans people, who are a statistically small portion of the general population. Reality contains multitudes of nuances in terms of a person’s identity.

How someone responds to questions like these can be shaped by internal and external factors, especially in terms of how someone identifies their race or ethnicity. Many questions have lots of different options for race or ethnicity, but those categories aren’t capturing the full reality, and race is a social construct that’s made valuable by the way people are treated.

An example is that most of the people in Puerto Rico identify their race as “White” on the Census, but Puerto Ricans in the Bronx mostly identify their race as “Black.” Does this mean that all of the Black Puerto Ricans moved from the island to the Bronx? No, but in the social context of New York City, Puerto Ricans are told by those around them that they are Black, and treated as such. And this changes their responses on the Census.

7 of 49

… which leads to oversimplification of identities.

Race (societally important social construct)
Ethnicity (often defined by family’s country of�origin - borders)
Gender Identity/Expression (often just binary �“sex” is asked)
Sexual Orientation (if asked, usually very simplified or only coupled & cohabitating people count)

The categories assigned to people or claimed by them in some cases can be vastly oversimplified and lack an intersectional focus. Even though people are allowed to self-select their choices for these boxes, the boxes themselves are often oversimplifications of their full selves.

A person’s race shape’s their outcomes, but two people who checked the same race category box could have vastly different experiences based on their other identities.

Ethnicity is not the same as race, but is often discussed in the same context as race. This generally asks a person to describe their family’s country of origin. Many people who are “Hispanic/Latino/Latinx” would not self-identify with those terms - they might prefer to identify as Mexican American or Dominican - but societally they might be treated as “Hispanic.”

Gender is often asked along binary Man/Woman lines, and typically focused on biology or “sex.”

Sexual orientation isn’t asked about much at all. We’ll talk later about how sexual orientation is often only measured for coupled people, leaving out single folks. It is also rarely treated as fluid, and focusing on couples does lead to erasure of different identities.

8 of 49

Categories are societally important.

People are treated differently by society based on these imperfect categories.

If we didn’t ask about race, we wouldn’t measure segregation/discrimination.
If we didn’t ask about gender identity, we wouldn’t measure the gender pay gap.

9 of 49

How the government counts people

10 of 49

Everyone has not been counted equally...

FROM THE CONSTITUTION:

Art I, Sec 2: “Representatives and direct taxes shall be apportioned among the several states… according to their respective numbers, which shall be determined by adding to the whole number of free persons, including those bound to service for a term of years, and excluding Indians not taxed, three fifths of all other Persons. The actual Enumeration shall be made within three years after the first meeting of the Congress of the United States [end of 1792], and within every subsequent term of ten years, in such manner as they [Congress] shall by law direct…

First off, we have to do a little U.S. government review. The census is required by the constitution to apportion the number of congressional representatives to the states. But… the U.S. has a pattern of counting some people “more” than others. The 3/5ths compromise, which was eradicated with the 14th amendment after the civil war, explicitly counted enslaved people as 3/5ths of a person in order to give more weight to the southern states. However, even with the 14th amendment, enslaved people and women were counted but not allowed to vote - their numbers generating political weight to the people they did not elect.

Today we still have issues of representation. I used to work for a nonprofit in NYC, and we did quite a bit of work around cities in Long Island with large immigrant populations - many of whom couldn’t vote in local or federal elections. So you effectively wind up with populations that generate political weight for politicians they cannot vote for - and who don’t care to listen to them because of their lack of voting power. Trump also just passed an executive memo to not count undocumented immigrants would hurt republican states as much as democrat states in terms of representation.

I say all this just to stress that political power is a function of population, and we don’t always grant everyone the same political power. We have to think about all of this in who we are counting and why.

11 of 49

What does the decennial Census ask? (12 questions)

BASIC HOME CHARACTERISTICS

Type (house, apartment, mobile home, etc)
Tenure (own/rent)

BASIC PERSON CHARACTERISTICS

Age
Race/Ethnicity
Sex
Relationship to other people �in household

Other Census Bureau Products (like the American Community Survey) are where you’ll find details on income, occupation, education, etc.

12 of 49

Who counts? - race/ethnicity

13 of 49

Race/ethnicity categories shift over time:

This chart shows are all the ways the race/ethnicity question has been asked since 1790. The options available on this question have shifted dramatically based on what was societally important at the time. At the time of the 3/5ths compromise, the government needed to know free vs. enslaved to apportion power. Later this shifted to a focus on who was foreign born, and as many white immigrants like the Irish and Germans assimilated into white society, this was less important and the categories continued to shift.

I show you this to make the point that the way we ask the question and the categories we use are shaped by the cultural environment at the time. But even with all these changes, the categories are still imperfect and leave out...��LINKS�Pew Research Center US Census timeline: https://www.pewresearch.org/interactives/what-census-calls-us/

The Pudding article on census evolution: https://pudding.cool/2020/03/census-history/

14 of 49

But still have gaps:

15 of 49

Who counts? - gender/sex

Binary, focused on “sex” (biology)

Hasn’t changed since first census in 1790

In fact, the question on gender ignores quite a bit. I’m sure we could make the same cartoon we saw in the last slide for someone who is trans or gnc trying to pick between these two options.

Race used to be considered a biological trait, and today we understand that it is not the case. Maybe eventually our Census will allow us to better reflect the population in terms of gender identity instead of asking simply “male” or “female”. Demographers often say sex is biological while gender can be more fluid, but the way the question is asked implies biology is only important. A trans man could self-select as “male” here, but we wouldn’t capture his full identity. Census responses are not cross-checked with other data sources, so a person can list their preferred name and “sex” without ramifications.

16 of 49

Who counts? - sexual orientation

Only counts coupled people living in same household (pretty heteronormative…)

17 of 49

Straight couples cause some of the issues...

2010 CENSUS ERRORS:

1 in 2,000 different-sex couples checked the wrong “sex” box leading to…
1 in 3 same-sex couple households being misidentified (actually a different-sex couple)

Gary & Karen

Sticking with sexual orientation and the census, not only are we ignoring single queer folks and erasing the trans experience all together, there’s even error in how we can count the same-sex married population because of… straight couples.

In the 2010 census, there were initially over 900,000 same-sex couples reported. However, this was much higher than previously thought or measured in future surveys. What went wrong?

Well, it turns out that 1 out of every 3 couples classified as “same-sex” were actually different-sex couples who had incorrectly marked their “sex” on the census form. This was verified by some researchers going into the forms and comparing the names that people had written down. They used other data sources to classify what names are “male” or “female”, and found that people like Gary and Karen here who had listed themselves as a married lesbian couple over 65 in thes suburbs, were likely not that. Of course you never know, but statistically it was unlikely. The new explicit options about relationships in the 2020 census should help with this error.

This is an example of a “false positive” problem -- While only 0.05% of straight couples made these errors, their large number led to a big error (28%) in the small population of same-sex couples.

18 of 49

Where same-sex couples live

19 of 49

To recap: you might run into...

Proxies for queer people
Sampling & miscoding errors
Undercounts
Asking the wrong questions

So, I wanted to give y’all an overview of some considerations as you look for queer data and some of the common pitfalls you might find. Overall, the categories we use in most data collection oversimplify things, but they are still very useful.

Some things you might run into include having to make proxies for queer life, such as only being able to identify couples. You might also find sampling and miscoding errors, since the LGBTQ population is a statistically small population. There are just fewer queer people to count — researchers estimate 4.5—10% of the US population is LGBTQ. Small data errors in the larger population (non-LGBTQ) can create big errors in the smaller population.

You could wind up with undercounts in surveys that actually ask about the queer population because people might feel at risk if they answer honestly. Queer, trans, and gender non-conforming youth are also more likely to be homeless, and that can lead to undercounts as well.

And perhaps we aren’t asking the right questions - for example, relying on the “biological” male/female binary is not inclusive.

With all that, I’m going to pass it on to Jan who will tell you more about what you can do with data. If you have any questions, definitely pop them in the chat and we’ll get to them as we go.

20 of 49

Let’s look at how you can make it work

21 of 49

Government data

Government data is one of journalists’ go-tos — and often for good reason. For the most part, it’s relatively reliable and stable, covers a large geography, and has a long temporal tail. But like Sarah outlined before, it comes with its own shortcomings and biases, so it’s important to still handle with care. It’s easy to get into the mindset that data = fact, but just as you would pose questions to someone you’re interviewing, you’ve also go to interrogate the data. Let’s look at some projects that have used government data:��LINKS�US Census data: https://www.census.gov/topics/families/same-sex-couples.html�US Treasury Department data: https://www.treasury.gov/resource-center/tax-policy/tax-analysis/Documents/WP-108.pdf�Centers for Disease Control data: https://www.cdc.gov/lgbthealth/index.htm

22 of 49

The Most Detailed Map of Gay Marriage in America

The New York Times, September 2016

At the time that this map from the New York Times was published in 2016, it was the most complete picture of gay marriage that we’d seen since the Windsor and Obergefell Supreme Court decisions. Joint tax filing data still might be one of the best ways to track gay marriages, even after a small shift in this year’s Census questions. 97.5% of married couple file their taxes jointly.��The map revealed what we anecdotally expected, but it allowed us to confirm it with data: the northeast and the west coast — home to states that legalized gay marriage before most of the country — have porportionally more same-sex married couples. It also shows how important cities and metropolitian areas are to queer populations, especially in the South and Midwest, see Atlanta, St. Louis, and Austin.��But it also gave journalists new leads to look into: like what’s happening in the eastern mountains of North Carolina? That’s how data often works: sometimes you’re using it to answer one question and it opens up a pandora’s box of more.��LINKS�NYTimes map: https://www.nytimes.com/2016/09/13/upshot/the-most-detailed-map-of-gay-marriage-in-america.html

Brookings data: https://www.brookings.edu/research/gay-marriage-in-america-after-windsor-and-obergefell/

23 of 49

Why Police Struggle to Report One of The Fastest-Growing Hate Crimes

The Marshall Project, November 2019

This is a project from the Marshall Project that shows just how government data collection practices can result in large undercounts. Currently, the FBI doesn’t require police departments to submit hate crime data to its national database, and when department do submit its self-reported with little oversight. Often the figures that are reported do not include hate crimes motivated by more than one bias. For example, an assault against a transgender Black woman may be motivated by their gender identity as well as their race, but it’s only classified once. ��Unsurprisingly when you look at victims’ surveys instead of law enforcement reports, there’s a big gap. ��LINKS�Marshall Project article: https://www.themarshallproject.org/2019/11/26/why-police-struggle-to-report-one-of-the-fastest-growing-hate-crimes

24 of 49

Public opinion data

25 of 49

How to Forecast an American’s Vote

The Economist,

November 2018

26 of 49

The Global Divide on Homosexuality Persists

Pew Research Center, �June 2020

(Mark Gevisser)

27 of 49

Academic research data

The Williams Institute
Scholars including: Dr. Gary Gates, Dr. Amin Ghaziani, Dr. Janice Fanning Madden, Dr. Matthew Ruther, and Jack Gieseking

28 of 49

LGBT Demographic Data Interactive

The Williams Institute, UCLA School of Law, �January 2019

29 of 49

This Is How Fast America Changes Its Mind

Bloomberg, June 2015

30 of 49

Non-profit research data

31 of 49

Mapping Transgender Equality in the United States

Movement Advancement Project, February 2017

32 of 49

The Dramatic Rise in State Efforts to Limit LGBT Rights

The Washington Post, updated June 2017

33 of 49

News media, social media, & tech data

Social sites including: Facebook, Twitter, Instagram, TikTok, Reddit, Yelp, and YouTube�
Trends from Google & Twitter�
Media coverage data including: Television News Archive and News Homepage Archive

You can’t mention data sources, without talking about the necessary evil that is social media and technology. I’m going to date myself here, but today’s queers will never know the inner turmoil of having to drive 3 towns over to a Best Buy where no one would know you to purchase a physical copy of The L Word DVD. Much of our queer world is online now, so it makes sense that that’s where a lot of the queer data is too.��LINKS�Facebook: https://developers.facebook.com/

Twitter: https://developer.twitter.com/en/docs

Instagram: https://developers.facebook.com/products/instagram/

TikTok: https://developers.tiktok.com/

Reddit: https://www.reddit.com/dev/api/

Yelp: https://www.yelp.com/developers

YouTube: https://www.yelp.com/developers

Google Trends: https://trends.google.com/trends/story/GB_cu_ztSYsWMBAAA15M_en_en-US�Twitter Trends: https://storywrangling.org/

Television News Archive: https://archive.org/details/tv

News Homepage Archive: https://nrjones8.github.io/news-archive-explorer/#/

34 of 49

Media's LGBTQ Focus Has Dwindled Since Trump’s Election

Real Clear Politics, September 2019

When I was preparing this presentation, I realized that a lot of the big LGBTQ+ data projects I had included were 5-10 years old. I wanted to find some more recent interactives, but it was hard. Then I came across this piece from Real Clear Politics that helped explain some of my struggle — major television networks really weren’t mentioning queer issues as much. Again, the data sparks some questions: Did broadcast media turn away when the battle for marriage equality — one of the most mainstream and sanitized representation of queerness — was over? Was it that the “feel-good” queer stories were replaced with daily threats from the Trump administration — often things that are narrative and not data driven? Did queer stories get shoved to the backburner because there was now an ever swirling dumpster fire? Or all of the above? Whatever it was, the piece makes it clear that despite queer stories disappearing from mainstream broadcast media, Google search interest for “LGBTQ” continued to climb over the same period.

LINKS

Real Clear Politics article: https://www.realclearpolitics.com/articles/2019/09/18/medias_lgbtq_focus_has_dwindled_since_trumps_election_141277.html

35 of 49

LGBTQ+ Pride Month‬‬

Google Trends, June 1, 2020 - Now

SEE ALSO

StoryWrangler

(University of Vermont Computational Story Lab)

36 of 49

Combine data

Collect and organize data from sources including: original documents, surveys, sensors, photos/video, news reports �and observations�
Combine sources into one measurement

37 of 49

Gay rights in the US, state by state

Guardian US, May 2012

Build the Radial Map in Excel (PolicyViz)

For this, the OG of queer data visualizations, the Guardian US interactive team researched gay rights including marriage, hospital visitation rights, adoption, employment and housing, and education discrimination protections, and hate crime laws. Instead of using a map, where you can only really encoded one, maybe two variables, the wheel allows you to layer on seven to get a more complete picture of gay rights. The geographic component isn’t completely lost doing it in this format, because the Guardian keeps the wheel divided into regions. You can clearly see that the Northeast has more sweeping protections compared to the Southeast.��If you’re interested in this type of visualization, the Guardian created a companion piece about global rights and authored a behind the scenes making of post. If you’re looking for a way to recreate this in a low-tech way, Jonathan Schwabish of PolicyViz has a great post about doing it in Excel.��LINKS�Guardian US states interactive: https://www.theguardian.com/world/interactive/2012/may/08/gay-rights-united-states

Guardian world interactive: https://www.theguardian.com/world/ng-interactive/2014/may/-sp-gay-rights-world-lesbian-bisexual-transgender

Guardian US making of: https://www.theguardian.com/news/datablog/2012/may/10/data-visualisation-us-gay-rights

Radial map in Excel: https://policyviz.com/2019/12/04/the-radial-map/

38 of 49

Men are from Chelsea, Women are from Park Slope

The Pudding, June 2018

SEE ALSO

Gayborhoods Index Data (The Pudding)

This is a piece that I did for The Pudding a couple of years ago (with lots of help from Sarah) — it’s really the thing that got me thinking and talking about queer data. After struggling with the validity of Census data to try to capture a gender divide in gayborhoods, Sarah convinced me to look more broadly and combine several datasets into one index. For this piece, I combined same-sex unmarried partner households from the Census’ American Communties Survey with same-sex married joint tax filers from the Treasury, gay bar locations from Yelp, and local Pride march routes to create a index. It gave me a much more robust and certain metric. If you’re joining us from San Francisco’s the Castro (zip code 94114) and identify as a man, or are from Boston’s Jamaica Plain (zip code 02130) and identify as a women, congrats, you’re in the neighborhood that we identified as a gayborhood with the most certainty.

LINKS

The Pudding gayborhoods article: https://pudding.cool/2018/06/gayborhoods/�Gayborhoods index data: https://github.com/the-pudding/data/tree/master/gayborhoods

39 of 49

Make your own data

Crowdsource personal experiences�
Or document your own

40 of 49

Queering the Map

A community generated mapping project that geo-locates queer moments, memories and histories in relation to physical space

41 of 49

Plot Me Genderfluid

Kelsey Campbell, GaytaScience

42 of 49

Things to remember

43 of 49

QUANTITATIVE

What numbers

tell you

QUALITATIVE

What people

tell you

Where you want to be

44 of 49

The best stories don’t start with a dataset.��They start with a question.

The best stories don’t start with a dataset — they start with a question. Of course, that question has to be able to answered with data. One of the pieces that I’m most internet famous for is a piece that looks at the size differences between jeans for people who identify as men versus people who identify as women. When my coauthor Amber Thomas and I started out, we could have asked the question: “Why are women’s pockets so small?” It’s a very valid question, but one that’s harder to answer with data. Reframing the question as “How much smaller are women’s pockets than men’s?” gave us a path forward. The flow chart on the right here is from a blog post we published at The Pudding about how we go from idea to data story. It’s worth a look to if you want to see all the twists and turns we encounter when starting with a question.��LINKS�Pudding process flow chart: https://pudding.cool/process/pivot-continue-down/�Women’s Pockets are Inferior: https://pudding.cool/2018/08/pockets/

45 of 49

46 of 49

1

big number does NOT make a story

47 of 49

IRE LGBTQ+ survey

bit.ly/irelgbtsurvey

48 of 49

Questions?

Jan Diehm, @jadiehm

Sarah Serpas, @sarahserp

NLGJA 2020

49 of 49

Extra resources: