Similar to using sexual behaviors and attraction to capture elements of sexual orientation, questions may also be devised that consider gender expression and non-conformity regardless of the terms individuals ple of these types of questions would be consideration of the relationship between the sex that individuals are assigned at birth and the degree to which that assignment conforms with how they express their gender.

Another factor that can create variation among estimates of the LGBT community is survey methodology. Survey methods can affect the willingness of respondents to report stigmatizing identities and behaviors. Feelings of confidentiality and anonymity increase the likelihood that respondents will be more accurate in reporting sensitive information. Survey methods that include face-to-face interviews may underestimate the size of the LGBT community while those that include methods that allow respondents to complete questions on a computer or via the internet may increase the likelihood of LGBT respondents identifying themselves. Varied sample sizes of surveys can also increase variation. Population-based surveys with a larger sample can produce more precise estimates (see S for more information about survey methodology).

While the surveys show a fairly wide variation in the overall percentage of adults who identify as LGB, the proportion who identify as lesbian/gay versus bisexual is somewhat more consistent (see Figure 2)

A final challenge in making population-based estimates of the LGBT community is the lack of questions asked over time on a single large survey. One way of assessing the reliability of estimates is to repeat questions over time using a consistent method and sampling strategy. Adding questions to more large-scale surveys that are repeated over time would substantially improve our ability to make better estimates of the size of the LGBT population.

Findings shown in Figure 1 consider estimates of the percentage of adults who self-identify as lesbian, gay, or bisexual across nine surveys conducted within the past seven years. Five of those surveys were fielded in the United States and the others are from Canada, the United Kingdom, Australia, and Norway. All are population-based surveys of adults, though some have age restrictions as noted.

Like the counterpart of measuring sexual orientation through identity, behavior, and attraction measures, these varying approaches capture related dimensions of who might be classified as transgender but may not individually address all aspects of assessing gender identity and expression

The lowest overall percentage comes from the Norwegian Living Conditions Survey at 1.2%, with the National Survey of Sexual Health and Behavior, conducted in the United States, producing the highest estimate at 5.6%. In general, the non-US surveys, which vary from 1.2% to 2.1%, estimate lower percentages of LGB-identified individuals than the US surveys, which range from 1.7% to 5.6%.

In most cases, these surveys were roughly 60% lesbian/gay versus 40% bisexual. The UK Integrated Household Survey found the proportion to be two-thirds lesbian/gay versus one-third bisexual.

The National Survey of Family Growth found results that were essentially the opposite of the UK survey with only 38% identifying as lesbian or gay compared to 62% identifying as bisexual. The National Survey of Sexual Health and Behavior and the Australian Longitudinal Study of Health and Relationships both found a majority of respondents (55% and 59%, respectively) identifying as bisexual.

The surveys show even greater consistency in differences between men and women associated with lesbian/gay versus bisexual identity. Women are substantially more likely than men to identify as bisexual. Bisexuals comprise more than half of the lesbian and bisexual population among women in eight of the nine surveys considered (see Figure 3). Conversely, gay men comprise substantially more than half of gay and bisexual men in seven of the nine surveys.