Do Nonbinary and Transgender People Experience Systematic Earnings Penalties?

Carpenter, Feir, Pendakur, and Warman (2025) examine whether nonbinary and transgender people experience earnings penalties compared to cisgender workers, and what explains those gaps. They use restricted 2021 Canadian Census data linked to tax records, covering earnings from 2019–2020 for adults ages 25–59. They find large earnings disparities relative to cisgender men. Nonbinary people assigned female at birth earn about 44–49 percent less in basic models and about 27 percent less after extensive controls. Nonbinary people assigned male at birth earn roughly 19–33 percent less. Transgender men and women earn about 13–33 percent less. Occupational sorting and time worked explain a substantial share of these gaps.

Why This Article Was Selected for The Policy Scientist

This article addresses a policy issue of broad and growing importance: how labor markets treat gender minorities. The topic is timely given recent debates over whether public data systems should collect information on gender identity and how such data inform employment law, civil rights enforcement, and economic policy. Carpenter et al. (2025), who have written extensively in this area, make a major contribution by using a mandatory national census linked to tax records, substantially improving data quality over prior survey-based work. The findings are plausibly generalizable to other high-income countries with similar labor markets. The statistical methods are rigorous but future work using causal inference designs would further strengthen the evidence.

Full Citation and Link to Article

Carpenter, C. S., Feir, D. L., Pendakur, K., & Warman, C. (Forthcoming 2026). Nonbinary and transgender identities and earnings: Evidence from a national census. American Economic Review: Insights. https://doi.org/10.1257/aeri.20240571

Central Research Question

This article asks whether nonbinary and transgender individuals experience systematic earnings differences relative to cisgender workers, and how large those differences are once observable characteristics are taken into account. A central focus is whether earnings penalties vary by gender identity and sex assigned at birth, particularly for nonbinary individuals assigned female versus male at birth. The authors also examine the extent to which these earnings gaps reflect differences in job sorting, hours worked, and position within the earnings distribution rather than simple mean differences alone. More broadly, the study seeks to establish credible population-level evidence on labor market outcomes for gender minorities using high-quality administrative data, addressing long-standing data limitations in this area of research.

Previous Literature

The study builds on a growing economics literature examining labor market outcomes for transgender individuals, which has consistently found earnings penalties relative to cisgender men. Prior work has relied largely on administrative data capturing legal sex marker changes or on survey data with small samples of transgender respondents, such as the U.S. Household Pulse Survey. These studies generally document lower earnings for transgender women and, to a lesser extent, transgender men, but are limited in their ability to control for detailed job characteristics or to analyze distributional effects.

In contrast, research on nonbinary individuals is extremely sparse. Existing studies either pool nonbinary respondents into broader gender-diverse categories or rely on non-representative online samples. As a result, little is known about how nonbinary individuals fare in the labor market, particularly when disaggregated by sex assigned at birth. The authors position their work as filling this major gap while also extending earlier research by examining earnings across the distribution and assessing the role of occupational and industry sorting. Conceptually, the paper relates to broader theories of identity, discrimination, and labor market segmentation, while remaining primarily empirical in focus.

Data

The analysis uses restricted-access data from the 2021 Canadian Census long form, which was the first Canadian census to separately measure sex assigned at birth and current gender identity using a two-step approach. Because the census is mandatory and administered to 25 percent of the population, the data provide unusually large and representative samples of nonbinary and transgender individuals. The authors identify approximately 6,400 nonbinary individuals and 7,600 transgender individuals ages 25–59 in the analytic sample, corresponding to tens of thousands of individuals at the population level.

A key strength of the dataset is its linkage to administrative tax records, which provide high-quality annual earnings data for 2019 and 2020. The census also includes rich demographic, geographic, and labor market information, including education, immigration status, health indicators, occupation, industry, and measures of time worked. This combination allows the authors to examine earnings gaps while progressively controlling for a wide range of covariates and to assess mechanisms such as job sorting and hours worked that are typically unavailable in smaller surveys.

Methods

The authors estimate a series of weighted linear regressions with log earnings as the dependent variable, using cisgender men as the reference group. Separate indicator variables are included for nonbinary individuals assigned female at birth, nonbinary individuals assigned male at birth, transgender men, transgender women, and cisgender women. The models are estimated sequentially, adding controls for demographics, education, immigration status, religion, family structure, health, and finally detailed occupation and industry fixed effects.

To capture heterogeneity across the earnings distribution, the authors also estimate conditional quantile regressions at multiple deciles. This approach allows them to assess whether earnings gaps are larger at the bottom or top of the distribution. Additional analyses examine weekly earnings, full-time employment, and weeks worked to assess the contribution of labor supply differences. The empirical strategy is descriptive and associational rather than causal, but it is implemented with careful attention to weighting, robustness, and sensitivity to alternative samples and specifications.

Findings/Size Effects

The results show large and persistent earnings gaps between gender minority groups and cisgender men. In baseline models, nonbinary individuals assigned female at birth earn roughly 44–49 percent less than comparable cisgender men, while those assigned male at birth earn about 30–33 percent less. Transgender men and women also earn substantially less than cisgender men, with gaps generally in the range of 25–35 percent depending on specification.

After adding extensive controls, including health and job characteristics, the gaps shrink but remain economically meaningful. In the most saturated models, nonbinary individuals assigned female at birth earn about 27 percent less than cisgender men, and those assigned male at birth earn about 19 percent less. Transgender men earn roughly 13 percent less, and transgender women about 22 percent less. Importantly, nonbinary individuals assigned female at birth earn significantly less than cisgender women as well, even after accounting for observable characteristics.

Quantile regressions reveal that earnings penalties are substantially larger at the bottom of the earnings distribution than at the top, indicating a pronounced “sticky floor” effect for gender minorities. Differences in occupation, industry, weeks worked, and full-time status explain a meaningful share of these gaps, particularly for nonbinary individuals, but do not eliminate them.

Conclusion

This article provides the first population-representative evidence on earnings outcomes for nonbinary individuals and the most comprehensive census-based evidence to date for transgender individuals. By leveraging mandatory census data linked to tax records, the authors substantially improve on prior work that relied on small or selective samples. The findings document large and persistent earnings disparities that vary systematically by gender identity, sex assigned at birth, and position in the earnings distribution.

While the analysis does not establish causal effects, it clearly demonstrates that earnings gaps remain even after accounting for extensive demographic and job-related factors. The results highlight the importance of job sorting and labor supply differences, while also suggesting that these mechanisms do not fully explain observed disparities. Overall, the study represents a significant contribution to the literature on gender identity and labor markets and establishes a strong empirical foundation for future work using causal inference designs to better isolate underlying mechanisms.