Sexual harassment survey percentages are comparable only when the underlying definitions, behaviors, settings, questions, populations, geography, timing, methods, and sources align closely. Before interpreting a difference as social change, researchers should document those elements survey by survey and explain any remaining mismatch.
Start with definitions, behaviors, and settings
Begin by identifying the construct behind each reported percentage. A measure combining sexual harassment and assault is not automatically equivalent to one limited to harassment. Likewise, surveys about street harassment, workplace harassment, public-space experiences, or a single behavior such as non-consensual sexual comments may capture different events and contexts.
The available studies illustrate this problem. The 2018 and 2024 #MeToo surveys assessed harassment and assault across public spaces and workplaces. The separate 2014 study focused on street harassment and also incorporated focus groups. A secondary statistics compilation may place figures about different behaviors beside one another even though they do not share a measurement framework.
For every survey, record the stated definition, included behaviors, setting, reference period, and whether measures are reported separately or combined. Do not treat similarly worded labels as proof of equivalence. The evidence supplied here is predominantly U.S.-focused and does not establish that its findings or instruments apply to Portugal or other countries. Cross-national use requires a separate examination of definitions, context, language, population, and design.
Compare questions across survey waves
Build a question map before comparing survey waves. Place each item from one questionnaire beside its nearest counterpart in the others, then label it retained, added, removed, or changed. Compare the behavior described, setting, reference period, response options, introductory wording, and any routing or follow-up structure. A shared topic is not enough: a wording or response change can alter what respondents understand and what the resulting percentage represents.
The 2019 study repeated some questions from the 2018 survey but also introduced questions about false accusations. That change demonstrates why the studies cannot be treated as if every wave used one fixed instrument. The 2018 source page links to survey questions and methodology, while the supplied excerpts do not reproduce the complete questionnaires for all years.
The package also lacks the full definitions, response options, weighting procedures, and sampling documentation needed for a complete equivalence assessment. Mark such fields “Details being checked” until the linked primary materials have been reviewed. Only describe a time-series difference as change over time when the relevant measures and methods are sufficiently consistent; otherwise, report the figures separately and disclose the mismatch.
Check samples, geography, and source provenance
Next, compare who was surveyed, where, when, and through which design. The 2018 study was a nationally representative GfK survey of 2,000 people. The 2019 nationally representative study used the AmeriSpeak Panel, included 1,182 women and 1,037 men, and was conducted by NORC in February and March 2019. The 2024 study surveyed more than 3,300 U.S. adults over age 18 through NORC in spring 2024. Its project also included state surveys in California, Louisiana, Mississippi, and Texas. These descriptions do not by themselves establish identical sampling, recruitment, weighting, or population coverage.
The 2014 street-harassment project used a nationally representative survey of 2,000 people and 11 focus groups involving varied demographic groups. Its narrower subject and qualitative component make it distinct from the later survey series. National findings, state findings, student figures, and results from different countries should therefore remain separate unless their populations and designs support a justified comparison.
Finally, classify every source as primary or secondary. The WorldMetrics page aggregates figures from multiple source types and labels results as verified, directional, or single-source. Its figures are secondary summaries, not one consistently designed survey series. Trace each statistic to the underlying study and record its institution, year, population, geography, wording, and method. If the primary basis cannot be established from the available material, label it “Availability not confirmed” and do not use it for a direct comparison.
Conclusion
A defensible comparison table should contain one row per survey and columns for the measured construct, included behaviors, setting, exact question wording, reference period, response options, sample, geography, fieldwork dates, sampling method, weighting, and source type. Add columns showing whether each question was retained, added, removed, or changed across waves.
Use the table to separate three judgments. First, determine whether two measures address the same concept and behaviors. Second, determine whether their populations, places, timing, and methods are sufficiently aligned. Third, confirm that every statistic can be traced to a primary study rather than relying solely on a mixed secondary compilation. Record missing documentation explicitly instead of assuming that designs match.
Differences can be described together for context even when direct comparison is inappropriate, provided the limitations are clear. They should not be presented as evidence of an increase or decrease merely because both sources use the phrase “sexual harassment.” Create a survey-by-survey evidence table now, record every definition, question, population, geography, and method, and consult the linked primary materials before drawing conclusions.
Disclosures and limitations
- This article was prepared with AI assistance from the supplied research package. Factual statements are limited to the identified sources, and missing questionnaire or methodological details require confirmation in the linked primary materials.
Related reading
Sources
- Sexual Harrassment Statistics | Fact-Checked 2026 — worldmetrics.org
- National Studies | Stop Street Harassment — Stop Street Harassment
- MeToo Research — Newcomb Institute