Urbanicity and LGBT Demographics

As more and more surveys are including questions attempting to capture the size of the LGBT identifying population in the U.S., I’ve been interested. I’m interested in the measures being used, what they’re able to capture, how those estimates vary when we compare surveys, and the specific wording of questions used.

In the past few years, Gallup’s estimates (after partnering with The Williams Institute and Gary Gates) have received an incredible amount of attention. Reporting on the most recent wave of data collected, Gallup reported that 4.5% of the U.S. population identifies as LGBT. It’s a bold claim. And, as with most estimates, it is most likely a conservative estimate.

Because of how Gallup asks this question, we cannot disaggregate lesbians, gay men, bisexuals, and transgender folks from one another. They’re just lumped together. And much of the reason for this is… wait for it… probably best blamed on straight folks. Why? Designing survey questions that can reliably and accurately assess sexual identities is challenging for lots of reasons I’ve discussed before. But one reason worth noting is that straight folks – heterosexual people – are among the biggest hurdles. Enough heterosexual people cannot make sense of questions inquiring about their sexual identities that we worry about “false positives” (straight people not knowing how to answer and responding that they are “bisexual” or “lesbian” or anything other than straight not because they identify that way, but because they don’t understand the question). Read (or listen to) how sexual demographer Gary Gates puts in when talking about the simple question Gallup has asked survey respondents in two waves now:

“It’s a simple yes or no answer. One of the challenges that we’ve observed in measuring sexual orientation, and this may sound humorous to people, but heterosexuals often don’t know what their sexual orientation is and don’t routinely call themselves either heterosexual or straight. And so when you have questions where you’re asking people what they are, that very big population sometimes makes mistakes and it creates what we call ‘measurement error’ or ‘false positives.’ And it basically puts people in the LGB category that really aren’t…. With the Gallup question, you’re not asking that group what they are, you’re asking what they aren’t. And they more or less know that. So they may not use terms like heterosexual or straight. But they know they’re not gay.”

-Gary Gates (HERE)

In the Gallup Poll, respondents are asked, “Do you personally identify as lesbian, gay, bisexual, or transgender?” The responses are a simple yes/no. And all because straight folks and cisgender folks are so likely to misunderstand the question and inaccurately report their sexuality (and possibly gender).

So, it’s a conservative estimate that doesn’t allow us to break the LGBT population apart as much as we might like if we’re interested in understanding where growth in the population is and isn’t happening. But, because Gallup collects such a large sample, they are able to report on state-level estimates of LGBT populations throughout the U.S. I’ve written about this before. I updated a figure I previously produced for a lecture I’m giving and thought I’d share in case it’s useful to others as well.

LGBT by urban pop.png

We know that states with larger shares of the population living in urban areas have higher proportions of LGBT identifying individuals. There is more than one hypothesis about why this is or might be the case. I’ve charted these data before, but I added a new element to the figure below. Now it charts proportion identifying as LGBT by state by proportion of the state population living in urban areas AND data points vary by size according to the size of state populations relative to one another.

Personally, I’m excited to see the 2017-2018 data (which I imagine might be released soon) because from the work I’ve been reading, a great deal of growth in LGBT-identifying population is happening in the South in the U.S. And on the figure above, few southern states are above one standard deviation above the trend line (with the exception of Georgia). I wonder what this figure will look like when we map the population with more recent data. I have a feeling things have changed and I’m interested to see how.

_______________________

Special thanks to Connor Gilroy for answering my RStats question on Twitter, enabling me to figure out the code.

Baby Names and #MeToo, Take 2

By: Tristan Bridges and Philip N. Cohen

Last year in January, we wrote about name contamination as a telling illustration of how gender inequality is perpetuated. While a great deal of research and popular graphics examine the social forces behind which baby names become popular, name contamination refers to the ways social and cultural forces operate in steering people away from certain names. Thus, while we’d both love to see data showing that Donald Trump has effectively contaminated the name “Donald” for children born in the U.S., the data don’t support that. Donald is a name that was once rapidly rising in popularity – in the first few decades of the 20th century. But in the middle of the 1930s, “Donald” stared a free fall in popularity. The name became dramatically less popular among Americans in fairly short order. This is what scholars are talking about when they examine names that seem to be “contaminated.”

1

The source of the contamination? It’s Donald Duck, the character created by Walt Disney Productions in 1934. After Donald Duck, people just have never selected the name “Donald” for their babies in the same numbers. In his book, A Matter of Taste: How Names, Fashions, and Culture Change, Stanley Lieberson used Donald Duck as an example of “The Symbolic Contamination of Names,” tracking the name in California data and concluding the decline was probably triggered by the success of the Disney character.  Trump did not have this effect. In fact, Trump’s election and presidency has yet to register any effect on the number of baby boys in the U.S. named “Donald” – it just keeps trending steadily downward. Maybe Trump’s long-time celebrity means any further contamination associated with his name was already baked into the trends.

On the other hand, Melania and Ivanka – the two highest-profile women of the Trump family – might be provoking some identifiable name contamination. Both are quite rare names, with less than 400 girls given either name in 2018. And yet they have trended together quite closely since 1988, and both increased in 2016 and 2017 before taking a steep downward turn in 2018. We’ll have to keep an eye on those.

2

But last year, we were particularly interested in the name “Harvey.” The year 2017 was a big one for Harveys. Between the sexual violence of Harvey Weinstein and Hurricane Harvey. In our initial post, we discussed it as a quasi-natural experiment in what it takes for a man’s name to be contaminated? And this all happened in the middle of a very recent trend of the name Harvey becoming much more popular among Americans in very recent history. Most names that were once popular but are no longer never return to their former popularity. Some bounce back, like “Emma,” “Eleanor,” or “Oliver” have in recent history. Harvey was bouncing back, part of a resurgence of popularity of names that were previously popular in the first half of the 20th century.

Harvey appears to have been contaminated in 2017. Like “Forrest” following Forrest Gump, “Monica” following the news of President Bill Clinton’s affair with Monica Lewinsky, “Ellen” following Ellen DeGeneres coming out publicly (see here), “Hillary” following Bill Clinton’s term in office as President, “Alexa” following Amazon’s release of their digital assistant, and many others, Harvey seems to have started to become contaminated. One way of looking at this is by examining changes in the rank of the name among boys born in the U.S. (below).

3

While we see that Harvey was quickly rising up the ranks of name in recent years, the ranking of Harvey seemed to change its trajectory of popularity quickly in 2017. But, we also know that popular names used to be a lot more popular than they are currently. So, name rank is not a great measure to consult when considering whether a name has been contaminated.

Another measure to use is to consider the number of children born each year given the specific name. This might seem silly. But, as Philip has shown before, the best predictor of how many children are given any specific name this year is how many children received that name last year. When names buck this trend, the number of children given the specific name either increases substantially or decreases. Below, we charted the actual numbers of boys in the U.S. given the name Harvey since 1940 (the end of the period in which Harvey was popular in the first half of the 20th century).

4

Here too, you see the sharp decrease in the number of Harveys born in 2018. But we can also look at the rate of Harveys born to consider name contamination – that is, the number of boys given the name Harvey per number of boys born in a given year – to see if that rate changed. That figure is below, charting the number of boys born and named Harvey each year, per 1,000 boys born that year.

5

And finally, in our initial post, we examined name contamination by visualizing change in the rates of babies given potentially contaminated names by reporting name rates associated with the name in question relative to the name rate of the year in which the contamination may have occurred. Below, you can see that figure. Anything above the “100” line was a higher rate relative to the rate of babies named Harvey in 2017. And anything below is a lower rate relative to the rate of babies named Harvey in 2017.

6

By all of these measures, Harvey appears to have been contaminated. Name contamination is an important indicator of society’s values. Shifts in the name “Forrest” tells us something about inequality associated with ability. Changing in naming babies “Ellen” tells us something important about sexual inequality. Shifts in the names “Monica” and “Hillary” offer new powerful illustrations of the double binds women face in societies shaped by institutionalized forms of gender inequality. “William” was not contaminated following Bill Clinton’s affair. The popularity of the name “Clarence” did not noticeably change course following Anita Hill’s public accusations of sexual harassment. Harvey did.

There are a lot of ways to study cultural patterns today, including analysis of big data sources, such as social media posts and archival text mining. Names have something in common with these methods, but we like to study names because, unlike Facebook posts or tweets, they represent decisions that people consider very consequential – very personal, symbolically representing weighty statements of identity – and they are measured on a very large scale over a very long period. On the other hand, it’s impossible to know from these data exactly why people use the names they do. Name contamination is a good case because sharp reversals of fortune, such as that shown by “Harvey,” have readily interpretable causes.

In this case, although people debated the merits of the #MeToo movement, and what its impact would or should be, it appears that fewer people wanted to be associated with its leading man. Or maybe it was just the hurricane.

___________________________

Note: Thanks to Charles Seguin for noting that Stanley Lieberson first discussed “Donald” as a good illustration of name contamination in A Matter of Taste.

2018 GSS Update on the U.S. LGB Population

By: D’Lane Compton and Tristan Bridges

The 2018 General Social Survey data was just recently publicly released. Along with collections of social scientists interested in shifts in the various opinions measured on the GSS, we were eager to see how things shifted. We were particularly interested in the GSS demographic questions on sexual identity. As of 2018, it has officially been one decade that GSS has been asking the question. Over the course of that period of time, GSS has conducted six waves of data collection. To characterize your sexual identity on the survey, you can self-classify as “heterosexual or straight,” “gay, lesbian, homosexual,” “bisexual,” or “don’t know.” (Note: measuring sexuality on surveys poses a series of notorious challenges. This is one way of thinking about asking people about sexuality, but one among many options – see here for a discussion.)  As we have both noted previously, different surveys ask this question in slightly different ways, which makes direct comparisons across instruments messy. While the proportions shift around a bit by survey, the trends discovered on each about growth in LGB identity and the populations among whom that growth appears to be happening have been generally comparable across instruments in the past.

Some data on shifts in the LGBT population don’t share the data publicly allowing people to assess where change is happening. Gallup reports on the data they have collected in two waves of surveys in this way. But, as Tristan has noted before, reporting on shifts in the LGBT population treats the group as homogenous and artificially presents growth in LGBT identity as though it might be equally distributed among the L’s, G’s, B’s, and T’s. But that’s not true. Bisexual women account for the lion’s share on the growth in LGBT identification. And as Tristan and Mignon Moore showed in 2016, young Black women account for a disproportionate amount of the growth in LGB identification.

Gallup reported that much of the change in LGBT identification between 2012 and 2016 could be accounted for by young people, women, college-educated people, people of color, and those who are not religious (see here for a summary of the highlights). Data from GSS too has shown an increase in LGB identification between 2008 and 2018. The GSS is a smaller survey than Gallup. But the trends in sexual identification appear to be on trajectories and reasonable if we consider increases in social tolerance rates and queer visibility in mainstream media. They are not extreme and lie within smaller or similar ranges of measures from prior “national surveys” of sexual behavior, including but not limited to Kinsey’s (1948, 1953) work and Laumann et al.’s (1994) findings from the early 90s.

Below, we charted shifts in those identifying as lesbian and gay alongside those identifying as bisexual. Consistent with what Tristan showed in 2016, bisexual identification continues to be increasing at a steeper rate.

1.png

In fact, when you look at the proportions identifying as lesbian and gay or bisexual in 2008 and compare those with the proportions identifying as lesbian and gay or bisexual in 2018, lesbian and gay identifications have not really moved much. But bisexual identities continue to increase every year.

This is consistent with other national survey work. For example, the National Survey of Family Growth (NSFG) conducted by the U.S. National Center for Health Statistics was initially designed to be the national fertility survey of the United States and sought to explain trends and group differences in birth rates, such as contraception, infertility, sexual activity, and marriage. Compton, Farris and Chang 2015 found that almost nine percent of the women in the sample and about four percent of the men in the sample in 2008 and 2002 reported behavior as bisexual; that is, having sex with at least one male and at least one female partner in their lifetime. With regard to self-identification as bisexual for women, in 2008 0.43% and in 2002 0.26 percent. For men, in 2008 0.38% and in 2002 0.69% identified as being sexual. From this data we see how behavior and identity do not always align (which other work has noted as well). But these data do support the trends identified in GSS data, and might allow us to speculate that perhaps with increases in social tolerance rates people are more likely to claim bisexual identities than used to be the case regardless of their participation in the behavior.

Previous work has also suggested that much of the growth in LGB identities is happening among women. The GSS data show that the shift appears to be primarily happening among bisexual women. Indeed, bisexuality continues to be a more popular sexual identity than lesbian among women, but a less popular sexual identity than gay among men, as others have shown. As of 2018, almost 6% of women responding to the survey identified as bisexual compared with 1.5% in 2008. Comparatively, shifts in lesbian identities among women and both gay and bisexual identities among men really haven’t shifted much.

2.png

And similar to previous analyses (here, here, and here), this shift is particularly pronounced among the young. The figure below shows changes in lesbian and gay identities alongside shifts in bisexual identities for four separate age cohorts. The real shift in among bisexual identification among 18-34 year-olds. Between 7 and 8% identified as  bisexual on the 2018 GSS survey. This is all the more interesting when you look at the 2008 data on these figures. Bisexuality did not stand out in these data in 2008. In fact, in 2008, more people identified as lesbian and gay than bisexual. This shift has emerged and grown in an incredibly short period of time.

3.png

As other data has shown as well (see here and here, for instance), people of color account for a disproportionate amount of this shift. Black bisexuals accounted for almost 7% of Black respondents on the 2018 GSS. That’s a big shift as well. And while bisexual and lesbian/gay identities were moving along similar trajectories for Black Americans through 2016, as of 2018, bisexuality was much more common (as has been true for White respondents and those of other races… yes “other race” is actually the category GSS uses… and no, it’s not a good idea).

4.png

Gallup’s survey found that, as of 2016, education had become less predictive of LGB identification than it was in 2012. On the chart below, we show that this is generally true, but those with less than a high school education stand out in interesting ways here. Lesbian and gay identifying people with less than a high school education account are the least likely of any educational group to identify this way. Conversely, respondents with less than a high school education are also the most likely to identify as bisexual.This is interesting and probably deserves more scholarship to explain what precisely this means. But the contrast is more stark looking at the 2018 data (below).

5.png

Similarly, the subjective class identification data for sexual identity are all over the place. Part of this is a product of the vast majority of GSS respondents self classifying as working and middle class (see below). The slight trend that does seem generally visible is that lower classes have higher rates of LGB identification. Upper class (identifying) bisexuals are an outlier here, but they have also shifted a great deal in the last three waves of data collection and this may also be a product of “upper class” having the lowest numbers of respondents subjectively identifying with that class status. So, small n’s sometimes make for spiky graphs.

6.png

Additionally, consistent with a growing body of scholarship on sexual minorities outside of major cities in the northeast and west, the steady increase in lesbian and gay identification in the South is notable (see below). Indeed, just looking at lesbian and gay identifications on the figure below, the south and west have the highest rates of LG identification in the GSS sample. Bisexuality adds another dimension. Indeed, bisexuality seems to have become more common in the midwest between 2016 and 2018, while lesbian and gay identifications seemed to fall out of favor. This same trend can be seen for LGB folks in the northeast here as well (see below). And, finally, while more respondents identify as bisexual than lesbian and gay (combined), that trend is not true in the south, where – as of 2018 – more respondents identified as lesbian and gay than as bisexual. These facts too suggest we need more scholarship on the emergence of bisexuality and on sexual minorities in different parts of the country than where they have been most frequently studied.

7.png

Finally, rates of change among LGB identifying people by political party offers some interesting insights as well. As was true in the 2016 data as well, Democrats continue to have the highest rates of LGB identification and Republicans have the the lowest rates. The biggest change is the drop in rates of lesbian and gay identification among Independents. Whether these changes are indicative of people changing party identification or sexual identification would be interesting to know, but is not answerable with these data.

8.png

So what do we do with all of this? What can we take from the shifts? One thing that we ought to take from this is to take scholarship on bisexuality more seriously. Both of us have commented on this before. As a sexual identity, bisexuality is less studied than it ought to be. But bisexuality has continued to grow and continues to represent a larger number of people’s sexual identities than lesbian and gay combined. This is interesting for a number of reasons, but one is that much of the growth in the LGBT community might actually be the result of changes in the population of bisexual identifying people (and this is a group that is disproportionately composed of women). Whether bisexual identifying people understand themselves as a part of a distinct sexual minority, though, is a question that deserves more scholarship.

We should care about bisexuality as a sexual identity, because as a sexual identity, it is one that continues to be on the rise. And, as both of us has written about previously, if we are going to continue to group bisexuals with lesbian women and gay men when we report on shifts in LGB populations, this might be something that deserves better understanding and more attention. Context matters in how we understand identities and how they change or evolve over time.

Race, Education, Running and Spatial Inequalities

A while ago, I wrote a post at Sociological Images looking at the ways Dustin A. Cable’s map visualizing racial segregation in the U.S. compared with Kyle Walker’s map examining educational segregation in the U.S. My interest was in was in examining spatial inequality. In a nutshell, where you live matters. It plays an important role in what kinds of resources you have access to (or don’t). It shapes your future earnings, how much education you’re likely to receive (in addition to the quality of that education), how long you live, and much more.

I was interested in putting the two maps into conversation because sociologists who study inequality are interested in a specific social process wherein privilege and inequality tend to accumulate. That is, some kinds and qualities of resources (economic, social, cultural) are found in abundance in some contexts, to a lesser extent in others, and are virtually absent in many places. And you can see these accumulations. Non-white populations (specifically Black and Hispanic) are in high concentrations on the maps in the same areas lower levels of education are present.

Last week, Runner’s World ran a story that looks at where people run according to Strava’s maps to talk about racial segregation and it reminded me of this post. I went through and looked up some of the big cities to see how where most popular routes are for runners using Strava. [Note: I run and do not use Strava, and many runners I know do not either. So, this is not a measure of exercise in a given area; it’s a measure of how many people exercise using Strava.]

strava logo.pngIf you’re unfamiliar, Strava is a mobile app and website on which runners and cyclists are able to post their exercise and connect with others doing the same. So, it’s basically a facebook page for people who exercise and want to log their miles to go online and hold themselves and each other accountable (and probably to brag a little too).

But look at how racial segregation, educational segregation and Strava use map onto one another! Below, is the map I created so that you can see San Francisco, Berkeley, and San Jose, California in the same frame using Walker’s map of educational attainment (top) over Cable’s racial dot map (bottom). [Note the legends on each map when deciphering the meanings.]

San-Franciso-San-Jose-Berkeley-1.png

Here’s where people run using Strava in the same areas:

San Fran.png

Here’s Chicago:

Chicago-2.png

And Chicago’s Strava runners:

Chicago.png

Los Angeles:

L.A.-2.png

And L.A.’s Strava enthusiasts:

L.A..png

New York City:

New-York-2.png

NYC Strava users:

NYC.png

Detroit, Michigan:

Detroit-2.png

Detroit Strava users:

Detroit.png

and, finally, Houston, Texas:

Houston-2.png

And the Strava scene in Houston:

Houston.png

This could be an indication that white people are exercising more than are other groups (or, more accurately, that white people who exercise are simply a lot more likely to turn to Strava to tell everyone how much they exercised and where than are other racial groups). But it is also a kind of social network that people living in predominantly white areas seem to be tapped into that other groups are less likely to use (and part of this is both knowing about the Strava community and believing it would be worthwhile to join). It’s a social network white people seem to be using more than other groups. And that makes me wonder what else they might be getting from logging their extracurricular exercise on social media.

But it also made me think of Rashawn Ray‘s research on how varying racial compositions of neighborhoods influences Black men and women to engage in more of less physical activity in their leisure time. And it could provide a different kind of evidence for spatially-based health inequalities.

Segregation is one powerful way that inequalities persist–its also a way that many are kept blithely unaware of the existence of stark social inequalities. It can be hard to notice inequality when we’re segregated from each other (by race, class, education). But mapping where we run offers a powerful illustration of some of these inequalities – and the use of Strava is one small piece of how inequalities are sustained.

Where do LGBT people in the U.S. Live?

Originally posted at Sociological Images

I love gender and sexual demography.  It’s incredibly important work.  Understanding the size and movements of gender and sexual minority populations can help assess what kinds of resources different groups might require and where those resources would be best spent, among others things.  Gary J. Gates and Frank Newport initially published results from a then-new Gallup question on gender/sexual identity in 2012-2013 (here).  At the time, 3.4% of Americans identified as either lesbian, gay, bisexual, or transgender.  It’s a big deal – particularly as “identity” is likely a conservative measure when it comes to assessing the size of the population of LGBT persons.  After I read the report, I was critical of one element of the reporting: Gates and Newport reported proportions of LGBT persons by state.  As data visualizations go, I felt the decision concealed more than it revealed.

From 2015-2016, Gallup collected a second round of data. These new data allowed Gates to make some really amazing observations about shifts in the proportion of the U.S. population identifying themselves as LGBT.  It’s a population that is, quite literally on the move.  I posted on this latter report here.  The shifts are astonishing – particularly given the short period of time between waves of data collection.  But, again, data on where LGBT people are living was reported by state.  I suspect that much of this has to do with sample size or perhaps an inability to tie respondents to counties or anything beyond state and time zone.  But, I still think displaying the information in this way is misleading.  Here’s the map Gallup produced associated with the most recent report:

During the 2012-2013 data collection, Hawaii led U.S. states with the highest proportions of LGBT identifying persons (with 5.1% identifying as LGBT)–if we exclude Washington D.C. (with 10% identifying as LGBT).  By 2016, Vermont led U.S. states with 5.3%; Hawaii dropped to 3.8%.  Regardless of state rank, however, in both reports, the states are all neatly arranged with small incremental increases in the proportions of LGBT identifying persons, with one anomaly–Washington D.C.  Of course, D.C. is not an anomaly; it’s just not a state. And comparing Washington D.C. with other states is about as meaningful as examining crime rate by European nation and including Vatican City.  In both examples, one of these things is not like the others in a meaningful sense.

In my initial post, I suggested that the data would be much more meaningfully displayed in a different way.  The reason D.C. is an outlier is that a good deal of research suggests that gender and sexual minorities are more populous in cities; they’re more likely to live in urban areas.  Look at the 2015-2016 state-level data on proportion of LGBT people by the percentage of the state population living in urban areas (using 2010 Census data).  The color coding reflects Census regions (click to enlarge).

Vermont is still a state worth mentioning in the report as it bucks the trend in an impressive way (as do Maine and New Hampshire).  But I’d bet you a pint of Cherry Garcia and a Magic Hat #9 that this has more to do with Burlington than with thriving communities of LGBT folks in the towns like Middlesex, Maidstone, or Sutton.

I recognize that the survey might not have a sufficient sample to enable them to say anything more specific (the 2015-2016 sample is just shy of 500,000).  But, sometimes data visualizations obscure more than they reveal.  And this feels like a case of that to me.  In my initial post, I compared using state-level data here with maps of the U.S. after a presidential election.  While the maps clearly delineate which candidate walked away with the electoral votes, they tell us nothing of the how close it was in each state, nor do they provide information about whether all parts of the state voted for the same candidates or were regionally divided.  In most recent elections traditional electoral maps might leave you wondering how a Democrat ever gets elected with the sea of red blanketing much of the nation’s interior.  But, if you’ve ever seen a map showing you data by county, you realize there’s a lot of blue in that red as well–those are the cities, the urban areas of the nation.  Look at the results of the 2016 election by county (produced by physicist Mark Newman – here).  On the left, you see county level voting data, rather that simply seeing whether a state “went red” or “went blue.”  On the right, Newman uses a cartogram to alter the size of each county relative to its population density.  It paints a bit of a different picture, and to some, it probably makes that state-level data seem a whole lot less meaningful.

Maps from Mark Newman’s website: http://www-personal.umich.edu/~mejn/election/2016/

The more recent report also uses that state-level data to examine shifts in LGBT identification within Census regions as well.  Perhaps not surprisingly, there are more people identifying as LGBT everywhere in the U.S. today than there were 5 years ago (at least when we ask them on surveys).  But rates of identification are growing faster in some regions (like the Pacific, Middle Atlantic, and West Central) than others (like New England).  Gates suggests that while this might cause some to suggest that LGBT people are migrating to different regions, data don’t suggest that LGBT people are necessarily doing that at higher rates than other groups.

The recent shifts are largely produced by young people, Millennials in the Gallup sample.  And those shifts are more pronounced in those same states most likely to go blue in elections.  As Gates put it, “State-level rankings by the portion of adults identifying as LGBT clearly relate to the regional differences in LGBT social acceptance, which tend to be higher in the East and West and lower in the South and Midwest. Nevada is the only state in the top 10 that doesn’t have a coastal border. States ranked in the bottom 10 are dominated by those in the Midwest and South” (here).

When we compare waves of data collection, we can see lots of shifts in the LGBT-identifying population by state (see below; click to enlarge).  While the general trend was for states to have increasing proportions of people claiming LGBT identities in 2015-2016, a collection of states do not follow that trend.  And this struck me as an issue that ought to provoke some level of concern.  Look at Hawaii, Rhode Island, and South Dakota, for example.  These are among the biggest shifts among any of the states and they are all against the liberalizing trend Gates describes.

us-lgbt-state-shiftsPresentation of data is important.  And while the report might help you realize, if you’re LGBT, that you might enjoy living in Vermont or Hawaii more than Idaho or Alabama if living around others who share your gender or sexual identity is important to you, that’s a fact that probably wouldn’t surprise many.  I’d rather see maps illustrating proportions of LGBT persons by population density rather than by state.  I don’t think we’d be shocked by those results either.  But it seems like it would be provide a much better picture of the shifts documented by the report than state-level data allow.

Joan Acker and the Shift from Patriarchy to Gender

by: Tristan Bridges and James W. Messerschmidt

We’ve read some of the tributes to the feminist sociological genius of Joan Acker.  And much of that work has celebrated one specific application of her work.  For instance, Tristan posted last week on Acker’s most cited article—“Hierarchies, Jobs, Bodies: A Theory of Gendered Organizations” (1990)—which examined the ways that gender is so embedded in the structure of organizations that we often fail to appreciate just how much it shapes our lives, experiences, and opportunities.  But, this specific piece of her scholarship was actually her applied work. It was an application of a theoretical turn she was suggesting all sociologists of gender follow.  And we did.  Acker was involved in an incredibly important theoretical debate that helped shape the feminist sociology we practice today.

“Patriarchy” is a concept that is less used today in feminist social science than it was in the late-1970s and 1980s.  The term has a slippery and imprecise feel, but this wasn’t always the case. There were incredibly nuanced debates about patriarchy as a social structure or as one part of “dual systems” (capitalism + patriarchy) and exactly what this meant and involved theoretically. Today, we examine “gender.”  Indeed, the chief sociological publication is entitled Gender & Society, not Patriarchy & SocietyAcker - The Problem with PatriarchyBut in the 1970s and 1980s, patriarchy was employed theoretically much more often.  Feminist scholars identified patriarchy to focus the critique of existing theoretical work that offered problematic explanations of the subordination of women.  As Acker put it in “The Problem with Patriarchy,” a short article published in Sociology in 1989: “Existing theory attributed women’s domination by men either to nature or social necessity rather than to social structural processes, unequal power, or exploitation” (1989a: 235). The concept of patriarchy offered a focus for this critique.

Joan Acker was among a group of scholars concerned about the limitations of this focus; in particular, patriarchy was criticized for being a universal, trans-historical, and trans-cultural phenomenon—“women were everywhere oppressed by men in more or less the same ways” (1989a: 235).  Concluding that patriarchy could not be turned into a generally useful analytical concept, Acker proposed that feminist social science move in a different direction—a route that was eventually largely accepted and taken up.  It’s no exaggeration to suggest that Acker was among a small group of feminist scholars who shifted the conversation in an entire field.  We’ve been relying on their suggestion ever since.

Acker’s short 6-page article was published in the same journal that had published Raewyn Connell’s article, “Theorizing Gender” (1985), which spelled out her initial delineation of the problems with sex role theory and what she labeled “categoricalism.” Connell was also concerned with how feminist theories of patriarchy failed to differentiate among the categories of “women” and “men”—that is, femininities and masculinities. Judith Stacey and Barrie Thorne’s “The Missing Feminist Revolution in Sociology” (in Social Problems) was published that year as well (1985), specifically criticizing sociology for solely including gender as a variable but not as a theoretical construct. Acker (1989a) explained why feminist social scientists ought to follow this trend and shift their focus from patriarchy to gender relations and the construction of gender in social life.  As Acker wrote, “From asking about how the subordination of women is produced, maintained, and changed we move to questions about how gender is involved in processes and structures that previously have been conceived as having nothing to do with gender” (1989a: 238).  And in another piece published in the same year—“Making Gender Visible” (1989b) in the anthology, Feminism and Sociological Theory—Acker argued for a paradigm shift that would place gender more centrally in understanding social relations as a whole. Acker suggested a feminist theoretical framework that was able to conceptualize how all social relations are gendered—how “gender shapes and is implicated in all kinds of social phenomena” (1989b: 77). Today, this might read as a subtle shift.  But it was monumental when Acker proposed it and it helped open the door too much of what we recognize as feminist sociology today.

Acker published what became her most well-known article—“Hierarchies, Jobs, Bodies”—in Gender & Society (1990) as an illustration of what the type of work she was proposing would look like.  She was concerned with attempts that simply tacked patriarchy onto existing theories which had been casually treated as though they were gender-neutral.  She explained in detail how this assumption is problematic and limits our ability to understand “how deeply patriarchal modes are embedded in our theorizing” (1989: 239).  And Acker illustrated this potential in her theorizing about gender in organizations.  But her suggestion went far beyond organizational life.

And by all measures, we took up Acker’s suggestion:  “Gender,” “gender relations,” and “gender inequality” are now the central foci of sociological theory and research on gender.  But Acker also concluded her short 1989 article with a warning.  She wrote,

[T]here is a danger in abandoning the project of patriarchy.  In the move to gender, the connections between urgent political issues and theoretical analysis, which made the development of feminist thought possible, may be weakened.  Gender lacks the critical-political sharpness of patriarchy and may be more easily assimilated and coopted than patriarchy. (1989a: 239-240)

Certainly, Acker’s concern leads us to honestly ask: Will shifting the theoretical conversation from patriarchy to gender eventually result in simply a cursory consideration of gendered structured inequality? Will the shift to gender actually loosen our connections with conceptualizations of gendered power? We don’t think so but one way to commemorate the legacy of Joan Acker is to both celebrate gender diversity while simultaneously visualizing and practicing gender equality.  This means continuing to recognize that inequality is perpetuated by the very organization of society, the structure of social institutions, and the historical contexts which give rise to each.

___________________________
References
Acker, Joan. 1989a. “The Problem with Patriarchy.” Sociology 23(2): 235-240.
Acker, Joan. 1989b. “Making Gender Visible.” Pp. 65-81 in Wallace, P.A., Ed., Sociological Theory and Feminism. Newbury Park, CA: Sage.
Acker, Joan. 1990. “Hierarchies, Jobs, Bodies: A Theory of Gendered Organizations.” Gender & Society 4(2): 139-158.
Connell, Raewyn. 1985. “Theorising Gender.” Sociology 19(2): 260-272.
Stacey, Judith and Barrie Thorne. 1985. “The Missing Feminist Revolution in Sociology.” Social Problems 32(4): 301-316.

Visualizing the Sociology of Liana Sayer and Time Use Research

Originally posted at Feminist Reflections.

People are often shockingly wrong about how much time they dedicate to various tasks.  In general, we tend to overestimate how much time it takes to do things we dislike and underestimate how much time we spend on tasks we enjoy.  So, people ritualistically overestimate how much time they spend on laundry, cleaning bathrooms, working out and underestimate how much time they spend watching television, napping, eating, or doing any number of tasks that provide them with joy.

Asking about how people use their time has been a mainstay on surveys dealing with households and family life.  We ask people to assess how much time they spend on all manner of mundane tasks in their lives–everything from shopping, sleeping, watching television, attending to their children, and household labor is divided up into an astonishing number of variables.  The assumption, of course, is that people can provide meaningful information or that their responses are an accurate (or approximately accurate) portrayal of the time they actually spend (see here).  This is why time diary studies came into being–they produce a more accurate picture of how people use their time.  People record their actual time use throughout the day in a diary, marking starting and stopping points of various activities.  And there are a number of different scholars who rely on this method and these data.  But Liana Sayer is among the leading scholars in the field.  When I’ve seen her present, or others present on time use data, the data are almost always visualized in the same way (as stacked column charts).  Personally, I love seeing the data this way.  The changes jump out at me and I feel like I instantly recognize trends and distinctions they discuss. But I have learned in my classroom that students do not always have the same reaction.

I’m interested because I use data visualizations in class a lot.  And in my (admittedly limited) experience, students have an easier time interpreting the story of time use data when it is visualized in some ways over others.

All of the examples here are pretty basic changes in data visualizations.  But, learning these basics are necessary to help students read the more complex data visualizations they may encounter.  Being able to interpret visualizations of temporal data is important; it’s part of what helps social scientists consider, measure, and critique the idea of “feminist change.”  Distinguishing between men’s and women’s time use is only one pocket of this field.  But, it’s the one I’ll focus on here, and on which Liana Sayer is among the foremost experts.  The data I’m visualizing below come from one of Sayer’s most cited articles: “Gender, Time and Inequality: Trends in Women’s and Men’s Paid Work, Unpaid Work, and Free Time” (here – behind a paywall).

It’s fairly common to present time use data with a series of stacked columns (the same way the Census often illustrates shifts in household types).  Below is a visualization of the differences between women’s and men’s minutes per day allocated to paid work, unpaid work, free time, and time dedicated to self care.  It’s all time diary data and we talk about why this is more reliable and a better measure – but also why it is more difficult to collect, etc.  Some students see the story of this graph immediately.  I do too.  Men’s time allocated to paid labor decreased while women’s increased.  And women perform more unpaid work than do men.  Lots of students, however, are stumped. stacked column

But when I present the data differently, students often have an easier time seeing the story the chart is produced to illustrate.  The Pew Research Center visualizes a lot of their data using stacked bar graphs.  And maybe it’s because these are more easily recognized by people with less experience with data visualization.  I have found that more of my students are able to read the chart below than the one above (at least for temporal time use data comparisons).stacked bar

Another way of presenting these data might be to use clustered columns.  I have also found that students are more quick to recognize the trend in these data with the graph below than they are with the initial stacked column chart.cluster column

But, I’ve found that students have the easiest time with line charts for temporal time use data.  On the chart below, I deleted the grid lines because The New York Times sometimes displays time trend data this way (see Philip Cohen’s piece on NYT Opinionater, “How Can We Jump-Start the Struggle for Gender Equality?” for an example).  Students that struggle to recognize the trend in the clustered column chart, are much faster to see the trend here.line

These aren’t an exhaustive set of examples, and all of them are basic visualizations.  For instance, we might use a stacked area chart to show these data (as trends in the racial composition of the U.S. are often depicted), a scatterplot (as data on GDP and fertility rates are often illustrated), a series of pie charts (as men’s and women’s various compositions in different economic sectors are sometimes visualized), or something else entirely. Screen Shot 2016-03-22 at 9.59.28 AMIn fact, The New York Times produced a really incredible interactive stream graph visualizing data from The American Time Use Survey that illustrates differences in time use between groups. I sometimes have students explore this graph in my course on the sociology of gender.  But many struggle interpreting it.  This is, I think, in part due to the fact that we often take data visualization literacy for granted.  It’s a skill, and it’s one we should be better at teaching.

I think the point I want to make is that we (or I, at least) need to think more carefully about how we visualize our data and findings to different audiences.  Liana Sayer has an incredible mastery of this (she presents data in all of the ways mentioned in this post and more throughout her work).  One thing that I’m thinking more about as I write and teach about research and findings amenable to data visualization is which visualizations are best suited to which kinds of data (something all scientists are concerned with), but also which visualizations work with which kinds of audiences.  This is new territory to me.

Visualizing feminist change in a single chart is difficult.  And it’s often accompanied by, “Well, this is true, but let me tell you about what these data don’t show…”  But, I’m interested in how we make choices about visualizing feminist research and whether we need to make different kinds of choices when we talk about the findings with different kinds of audiences.

Why Popular Boy Names are More Popular than Popular Girl Names

Originally posted at Feminist Reflections.

UPDATE (4/18/19): I discovered that Andrew Gelman posted on this same trend with top ten boy and girl names in 2013. His post and a link to a piece he published in NYT is HERE.

In my introduction to Sociology class, I use trends in baby names to introduce students to sociological research and inquiry. It’s a fun way to show students just how much we can learn from what might feel like idiosyncratic details of our lives. I start by showing students the top 10 boy and girl names from the most recent year of data available (along with their relative frequencies). After this, I show them the most popular names and their relative frequencies from 100 years earlier. There are some names on both lists; but for the most part, the names on the latter list sound “old” to students. Screen Shot 2016-02-19 at 12.37.09 PM

When I ask students to characterize the types of names they see on the older list of names, someone usually says the names sound more “traditional.” I tell them that in 100 years, someone will probably say that about the most recent list of names they’re looking at: these future students will have a different idea of what makes a name “traditional.”

eva4ptth553dlbgtfzjf

19fed3usy9vmzgif

If you’re interested, someone produced these two GIF files that depict the most popular names by state between 1960 and 2012. I like to show one of these while I’m talking with students about what what names can help us learn. I ask students to raise their hand when they see their own name or the name of their best friend. As we get into the 80’s and 90’s, lots of hands start going up. But the GIFs are also interesting because they are a powerful visualization of the spread of cultural norms. Popular names move through a population in a way that appears to be similar to infectious diseases.

This is a fun way to show students that deciding what to name a child might feel like a personal decision, it’s actually a decision that is shaped by social forces. Names and name trends are great examples of what sociology can reveal because, as Stanley Lieberson points out so simply, while taste in most elements of culture is not a requirement, everyone has tastes in names. And, as it turns out, we can learn a lot about a society just by looking at patterns in which names we select for our children (and equally important are the types of names different groups tend to avoid).

SIDENOTE: I like to highlight a great finding by Stanley Lieberson, Susan Dumais, and Shyon Baumann from their article on trends in androgynous names (here). Androgynous names are names that are given to both boys and girls–think Taylor, Cameron, or Casey for current examples. Lieberson, Dumais, and Baumann found that androgynous names follow an incredibly common pattern once they achieve a critical level of popularity: they become girl names and become dramatically less common names for boys–a powerful example of the stigma associated with femininity for boys.

When I first started using the exercise, I was fascinated with the relative frequencies much more than the names on each list. But it’s an amazing shift. More than 1 in 20 girls born in 1914 was named Mary (the most popular name that year – and many other years too if you’re interested). By 2014, just over 1 in 100 girls born were given the most popular name that year, “Emma.” This is part of a larger trend in naming practices–popular names just aren’t as popular as they used to be. Stanley Lieberson refers to this as the “modernization theory” of name trends. The theory suggests that as institutional pressures associated with names decline (e.g., extended family rituals, religious rules), we see the proliferation of more diverse names. But there’s a twist. The phenomenon is also gendered: popular boy names have always been more popular (in aggregate) that popular girl names. Below, I’ve charted the proportion of boys and girls born in the U.S. with top 10 names from 1880-2014. Boys given top ten names in 1880, for instance, accounted for more than 40% of all boys born. And the most popular boy names have always accounted for a larger share of all boys born than the most popular girl names for girls born. It’s not a new fact and I’m not the first to notice it. (Though, as you can see below, the lines have just recently met, and they could conceivably cross paths any year now. And that will be something that has never happened.)Baby Name FrequenciesIn 1965, Alice Rossi suggested that part of what accounts for the discrepancy is related to gender inequality. As she put it, “Men are the symbolic carriers of the temporal continuity of the family” (here). Lieberson and Eleanor Bell later discovered that girls are more likely to have unique names as well (here). It’s an interesting example of something that many people teach in courses on men and masculinities. While men are, as a group, systematically advantaged, they may be held accountable to a more narrow range of gender performances than are women. And while men’s rights groups might frame this as an illustration of women being the group to benefit from gender inequality, it’s much better understood as what Michael Messner refers to as a “cost of privilege.”

Yet, this appears to be one costs of privilege that has decreased. In 1880, the top 10 boy names accounted for 41.26% of all boys born that year; the top 10 girl names accounted for 22.98%. There was more than an 18% gap. While boys’ popular names are still more popular than girls’ popular names, the gap shrunk to 0.27% by 2014. That’s a monumental shift. And I’m sure the modernization theory of name trends accounts for the lion’s share of the more general shift toward more secular names and a general decrease in name continuity between fathers and sons. But there is more than one way to read this shift. We might also say that this is a really simple illustration of one way that patriarchal family traditions have been chipped away over the past 100 years. Lots of data would support this conclusion.  We might account for it alongside, for instance, data showing the prevalence of women taking men’s surnames after marriage as a percentage of all marriages in a given year or opinions about surname change.  But it’s also an illustration of the ways that this process has meant changes for boys and men as well.

Masculinity has, quite literally, opened up. It’s something that has happened more for some racial and class groups than others. And whether this transformation–this “opening up”–is a sign of gender inequality being successfully challenged or reproduced in new and less easily recognizable ways is the subject of my favorite corner of the field.

 

 

Thinking Sociologically about #OscarsSoWhite: Measuring Inequality in Hollywood

Originally posted at Feminist Reflections

1498787_10202083647508448_647008496_oThe 2016 Oscar nominations were just announced.  This is the second year in a row that all 20 acting nominees are white–prompting the hashtag #OscarsSoWhite.  Matthew Hughey wrote on this issue last year as well.  The announcement got me thinking about inequality in film.  The nominees are selected by just over 7,000 members of the Academy of Motion Picture Arts and Sciences–so they are elected by a panel of peers.  But members of the AMPAS are not automatically voting members.  You have to apply, and your application has to be sponsored by existing member of the branch of the Academy for which you would like to be considered (here).  So, while the Oscars are awarded by a panel of peers, who make up the list of people who qualify as “peers” in the first place is a political matter.  And just like anywhere else, knowing someone who knows someone likely plays a role in gaining access.

Sociologists who study networks are often interested in how social networks provide access to various things people might want to acquire (wealth, status, access, “success” more generally, etc.).  This is why we have a concept for just how networked you are: “social capital.”  And certainly lots of people are complaining about the fact that Hollywood is an old, white, boy’s club and attempting to change this.  Indeed, Genna Davis founded an institute to study gender in the media.  April Reign (an editor at Broadway Black and NU Tribe Magazine) founded the hashtag #OscarsSoWhite after the all-white slate of nominees were announced last year.  And Maureen Dowd of The New York Times wrote an extensive article last year on the entrenched sexism that keeps women from occupying central roles in Hollywood.  Jessica Piven, one of directors quoted in the article, said:

“I feel that there is something going on underneath all of this which is the idea that women aren’t quite as interesting as men. That men have heroic lives, do heroic things, are these kind of warriors in the world, and that women have a certain set of rooms that they have to operate in.”

This belief system results in a network saturated with men and with precious few opportunities for women–and even fewer for women of color.  And as Effie Brown’s interaction with Matt Damon in “Project Greenlight” brought up, conversations about challenging the lack of diversity in Hollywood (similar to challenging the lack of diversity elsewhere) are often met with the presumption that diversity means compromising on ability, talent and creativity.  Entrenched sexism and inequality is a struggle to challenge in any institution because… well, because it’s entrenched.  So, it’s easy to feel like the most qualified guy who just happens to also be white without fully appreciating the fact that being a white guy might have been a big part of what gave you a foot in the door in the first place.

To think about this empirically, consider the party game “Six Degrees of Kevin Bacon”. The idea plays on the theory of “six degrees of separation”—part of a sociological puzzle called the “small world problem” asking just how connected everyone in the world is to everyone else.  The theory suggests that we are no more than six connections away from anyone in the world. In the early 1990s, some students at Albright University came up with the idea for the game: pick any actor and see if you can connect that actor with Kevin Bacon through shared movie appearances with other actors as the connections. Take Angela Bassett for example. Angela Bassett was in Sunshine State (2002) with Charlayne Woodard who was in He Said, She Said (1991) with… Kevin Bacon. So, Angela Bassett has a Bacon number of 2.Screen Shot 2016-01-15 at 4.18.58 PM

Later, a group of computer science students at the University of Virginia produced the network of actors to see how “central” Kevin Bacon actually is using IMDB.com (you can play around with the network on their site, www.oracleofbacon.org). And, as it turns out, Kevin Bacon is a central actor—he’s been in films with over 3,000 other actors and more than 99% of all of the almost 2 million actors listed on IMDB.com can be connected with Kevin Bacon in 5 connections or less. But, he’s not the most central actor. He’s actually the 411th most centrally connected actor (you can see the top 1,000 most “central” actors here). But, Kevin Bacon does share some things in common with the most central actor (Eric Roberts): they’re both white, they’re both men, and they were both born within two years of each other.  Coincidence?

When I encountered the list, I noticed that there weren’t many women. There are only 3 in the top 100 most central actors.  And all three are white.  So, I wrote a script to data mine some basic information on the top 1,000 to see who they are using data from IMDB.com (birth year) as well as NNDB.com (which lists race and gender).*  The list, perhaps unsurprisingly, is dominated by men (81.75%) and by white people (87.8%). Below is the breakdown for proportions of actors among the top 1,000 most central actor by gender and race.IMDB - Gender and RaceIt’s a powerful way of saying that Hollywood continues to be a (white) boy’s club. But they’re also an old white boy’s club as well. I also collected data on birth year. And while the 50’s were the best decade to be born in if you want to be among the 1,000 most “central” actors today, the data for the men skews a bit older.** This lends support to the claim that men do not struggle to find roles as much as women do as they age–which may also support the claim that there are more complex roles available to men (as a group) than women.IMDB Birth Year - MenIMDB Birth Year - WomenThe other things I noticed quickly were that: (1) Hispanic and Asian men among the top 1,000 actors list are extremely likely to be typecast as racial stereotypes, and (2) there are more multiracial women among the top 1,000 actors than either Hispanic or Asian women.

Part of what this tells us is that we like to watch movies about white people and men… white men mostly.  But part of why we like these movies is that these are the movies in which people are investing and that get produced.  As a result of this, there are a critical mass of super-connected white men in Hollywood.  So, it shouldn’t surprise us that white actors dominate the Oscar nominations. They’ve been hoarding social capital in the industry since it began.  #OscarsSoWhite

_________________________________

* To get the data, I wrote a Python script using the Unofficial IMDb API and the NNDB.com’s API. The results were able to read data for gender for all 1,000 people on the list but only gender, birth year, and race for 959 of the 1,000 people in the dataset. The other 41 had incomplete information on both sites. And I didn’t bother to clean the data up any more.

**Part of becoming a more central actor in the network of all actors has to do with having been a part of a mass of filmed projects with a variety of different actors.  The most central actor – Eric Roberts – has worked on projects with more than 8,000 other actors over the course of his career.  And being alive longer (perhaps obviously) helps.  But, it’s not all older actors.  And you don’t have to be living to be on this list.  But, actors born in the 1920’s, 30’s, and 40’s aren’t as central (as decade-based groups).  So, some of this is also having been in your 20’s, 30’s and 40’s between 1970 and 1990 which was a big period of growth for Hollywood.

What Constitutes a Mass Shooting and Why You Should Care

What Constitutes a Mass Shooting and Why You Should Care

By: Tristan Bridges, Tara Leigh Tober, and Nicole Wheeler

Originally posted at Feminist Reflections.

How many mass shootings occurred in the United States in 2015? It seems like a relatively simple question; it sounds like just a matter of counting them. Yet, it is challenging to answer for two separate reasons: one is related to how we define mass shootings and the other to reliable sources of data on mass shootings.  And neither of these challenges have easy solutions.

As scholars and teachers, we need to think about the kinds of events we should and should not include when we make claims about mass shootings.  Earlier this year, we posted a gendered analysis of the rise of mass shootings in the U.S. relying the Mass Shootings in America database produced by the Stanford Geospatial Center. That dataset shows an incredible increase in mass shootings in 2015. Through June of 2015, we showed that there were 43 mass shootings in the U.S. The next closest year in terms of number of mass shootings was 2014, which had 16 (see graph below).  That particular dataset relies heavily on mass shootings that achieve a good deal of media attention.  So, it’s possible that the increase is due to an increase in reporting on mass shootings, rather than an increase in the actual number of mass shootings that occurred.  Though, if and which mass shootings are receiving more media attention are certainly valid questions as well.

Mass Shootings (Stanford) 1If you’ve been following the news on mass shootings, you may have noticed that the Washington Post has repeatedly reported that there have been more mass shootings than days in 2015. That claim relies on a different dataset produced by ShootingTracker.com. And both the Stanford Geospatial Center dataset and ShootingTracker.com data differ from the report on mass shootings regularly updated by Mother Jones.* For instance, below are the figures from ShootingTracker.com for the years 2013-2015.

Mass Shootings, 2013-2015 (ShootingTracker.com)1For a detailed day-by-day visualization of the mass shootings collected in the ShootingTracker.com dataset between 2013 and 2015, see below (click each graph to enlarge).

Mass Shootings 2013

Mass Shootings 2014

Mass Shootings 2015

 

The reason for this discrepancy has to do with definition in addition to data collection.  The dataset produced by the Stanford Geospatial Center is not necessarily exhaustive.  But they also rely on different definitions to decide what qualifies as a “mass shooting” in the first place.

The Stanford Geospatial Center’s Mass Shootings in America database defines mass shootings as shooting incidents that are not identifiably gang- or drug-related with 3 or more shooting victims (not necessarily fatalities) not including the shooter.  The dramatic spike apparent in this dataset in 2015 is likely exaggerated due to online media and increased reporting on mass shootings in recent years.  ShootingTracker.com claims to ensure a more exhaustive sample (if over a shorter period of time).  These data include any incidents in which four or more people are shot and/or killed at the same general time and location.  Thus, some data do not include drug and gang related shootings or cases of domestic violence, while others do.  What is important to note is that neither dataset requires that a certain number of people is actually killed.  And this differs in important ways from how the FBI has counted these events.

Neither ShootingTracker.com nor the Stanford Geospatial Center dataset rely on the definition of mass shootings used by the Federal Bureau of Investigation’s Supplementary Homicide Reporting (SHR) program which tracks the number of mass shooting incidents involving at least four fatalities (not including the shooter). The table below indicates how different types of gun-related homicides are labeled by the FBI.

Screen Shot 2015-12-14 at 2.05.18 PMOften, the media report on events that involve a lot of shooting, but fail to qualify as “mass murders” or “spree killings” by the FBI’s definition.  Some scholarship has suggested that we stick with the objective definition supplied by the Federal Bureau of Investigation.  And when we do that, whether mass shootings are on the rise or not becomes less easy to say.  Some scholars suggest that they are not on the rise, while others suggest that they are.  And both of these perspectives, in addition to others, influence the media.

One way of looking at this issue is asking, “Who’s right?”  Which of these various ways of measuring mass shootings, in other words, is the most accurate?  This is, we think, the wrong question to be asking.  What is more likely true is that we’ll gather different kinds of information with different definitions – and that is an important realization, and one that ought to be taken more seriously.  For instance, does the racial and ethnic breakdown of shooters look similar or different with different definitions?  No matter which definition you use, men between the ages of 20 and 40 are almost the entire dataset.  We also know less than we should about the profiles of the victims (those injured and killed).  And we know even less about how those profiles might change as we adopt different definitions of the problem we’re measuring.

There is some recognition of this fact as, in 2013, President Obama signed the Investigative Assistance for Violent Crimes Act into law, granting the attorney general authority to study mass killings and attempted mass killings.  The result was the production of an FBI study of “active shooting incidents” between 2000 and 2013 in the U.S.  The study defines active shooting incidents as:

“an individual actively engaged in killing or attempting to kill people in a confined and populated area.” Implicit in this definition is that the subject’s criminal actions involve the use of firearms. (here: 5)

The study discovered 160 incidents between 2000 and 2013.  And, unlike mass murders (events shown to be relatively stable over the past 40 years), this study showed active shooter incidents to be on the rise.  This study is important as it helps to illustrate that the ways we have operationalized mass shootings in the past are keeping us from understanding all that we might be able to about them.  The graph below charts the numbers of incidents documented by some of the different datasets used to study mass shootings.

Mass Shootings Comparison

Fox and DeLateur suggested that it is a myth that mass shootings are on the rise using data collected by the FBI Supplementary Homicide Report.  We added a trendline to that particular dataset on the graph to illustrate that even with what is likely the most narrow definition (in terms of deaths), the absolute number of mass shootings appears to be on the rise. We do not include the ShootingTracker.com data here as those rates are so much higher that it renders much of what we can see on this graph invisible.  What is also less known is what kind of overlap there is between these different sources of data.

All of this is to say that when you hear someone say that mass shootings are on the rise, they are probably right.  But just how right they are is a matter of data and definition.  And we need to be more transparent about the limits of both.

_____________________

*Mother Jones defines mass shootings as single incidents that take place in a public setting focusing on cases in which a lone shooter acted with the apparent goal of committing indiscriminate mass murder and in which at least four people were killed (other than the shooter).  Thus, the Mother Jones dataset does not include gang violence, armed robbery, drug violence or domestic violence cases.  Some have suggested that not all of shootings they include are consistent with their definition (like Columbine or San Bernardino, both of which had more than one shooter).