Monday, October 05, 2015

Mining Big Data for Transgender Information

It's not easy to uncover information about the nation's transgender population. The Census Bureau does not include questions about sexual orientation or gender identity in the decennial census or American Community Survey. But there are other ways to shed light on the transgender segment—such as mining big data.

An economist at the Census Bureau, Benjamin Cerf Harris, has done just that. Harris mined the Social Security Administration's big data to determine whether transgender information could be gleaned from SSA records. By sifting through millions of Social Security files from 1936 to 2010, Harris located those in which a first name had been legally changed from male to female or female to male and also located those in which the sex-coding had changed. He found tens of thousands of transgender-consistent changes: 106,550 cases (at a 95 percent confidence threshold) in which a person aged 16 or older had legally changed his or her first name from one gender to another, and 28,234 cases in which there had been a first-name change and a sex-coding change. Overall, about 0.2 percent of files are what Harris calls "transgender consistent." The average age of identity change was mid-thirties.

The mining effort didn't stop there. For the transgender-consistent individuals who were still alive in 2010, Harris matched them to their census record and examined their response to the 2010 census question, "What is your sex?" He finds the transgender-consistents were more likely than non-transgender respondents to report being male and female (0.13 percent vs. 0.02 percent) or to report no sex at all (1.85 percent versus 1.13 percent). Harris also used 2010 census records to determine the geographical distribution of the likely transgender population.

Harris cautions that his research is not an attempt to estimate the size of the transgender population. Rather, it is an effort to show how big data can be mined to reveal something about the transgender population.

Source: Census Bureau, Likely Transgender Individuals in U.S. Federal Administrative Records and the 2010 Census

No comments: