Reading a salary survey before using its numbers
Salary and rate comparisons depend on who answered, how employment and income were defined, and what the analysis omitted. Response rates, subgroup sizes, missing answers and differences between survey rounds affect how far the figures can be used as benchmarks.
Employment and income definitions
The 2007 American Medical Writers Association (AMWA) salary survey defined employment and income as follows.
- Employed: working part-time or full-time for a company, institution or individual, hourly, salaried or by contract, with the employer paying Social Security taxes and deducting state and federal taxes.
- Freelance: doing work for hire for a client, paying all Social Security taxes personally, without state and federal taxes deducted from client payments.
- Full-time: working at least 32 hours per week.
- Part-time: working fewer than 32 hours per week.
- Total employee income: income before taxes.
- Gross freelance income: all money collected from clients.
- Net freelance income: income after expenses, such as insurance, subcontracting and equipment, but before taxes and retirement contributions.
The survey requested salaries to the nearest $1,000 and referred to income earned in 2006, although its survey year was 2007.
The 2015 AMWA survey retained the employer-versus-writer tax distinction, gross employee income, gross and net freelance income, and the 32-hour threshold; only actively working medical communicators counted as respondents.
In 2019, freelance status included independent contractors, consultants and writers paying themselves a salary through their freelance company.
Employees reported total income before taxes earned in 2018, while freelances reported gross and net income and standard hourly fees.
Freelance full-time status depended on reported hours rather than self-description: at least 32 total hours per week, including billable and non-billable time, as in 2015.
The AMWA-Canada 2005 survey calculated an hourly rate by dividing annual gross income by 52 and then by weekly billable hours; its effective rate used total billable and non-billable hours instead.
Definitions also differ between salary resources: a 2015 comparison noted that the US Occupational Outlook Handbook grouped medical writers with technical writers, while other resources appeared limited to pharmaceutical writers or used small or unstated respondent counts.
Sampling frames and response rates
AMWA's 2007 survey received 1,704 responses, approximately 32% of its roughly 5,400 members, compared with a 47% response rate in 1994.
The 2007 analysis found no evidence that the sample was unrepresentative because respondents' demographic characteristics resembled those of the 2004 respondents and the overall membership.
For example, 83% were women, nearly identical to the membership proportion, and about a third belonged to the Delaware Valley, New York or New England chapters, matching the membership distribution and the concentration of pharmaceutical employers in the northeastern United States.
The 2011 AMWA survey invited 5,215 members and received 1,193 responses, a 23% response rate.
The 2012 European Medical Writers Association (EMWA) survey received 320 responses from approximately 1,000 members, compared with 145 respondents in 2006.
The 2015 AMWA survey invited its 4,421 members, including retirees, students and recruiters, and also recruited lapsed members and non-members through sister organisations and writing groups.
No response rate could be calculated because the total number of solicited medical communicators was unknown.
Participation increased by approximately 100 compared with 2011, but the effect of including non-members on results and comparisons with previous surveys was unclear, despite a demographic profile still resembling the membership.
The 2019 AMWA survey was sent to 7,456 people: 3,835 members and 3,621 non-members, including lapsed members and people who had never joined.
After 271 removal requests and 55 returned emails, the revised panel was 7,130; eligibility required paid work as a medical communicator during 2018.
It received 1,418 completed responses, a reported 20% response rate, including 938 employees and 480 freelances.
The self-reported counts of 1,141 members and 263 non-members came from an optional question and did not sum to the total; 81.3% of respondents answering about membership reported being members.
Among respondents, 54.6% said they had not completed the 2015 survey, 22.3% had and 23.1% did not remember.
The 2005 AMWA-Canada survey received 33 responses from 135 chapter members, a 24.4% response rate; participation was voluntary, and the response period was extended to February 2006 after an initial 19% response.
Because respondents were not randomly selected, its average rates should not be used as standards and its data require caution.
Participation and response bias
Low response rates can reflect several different influences: survey fatigue was a possible contributor in the 2011 AMWA survey, and some pharmaceutical companies prohibited employee participation in that and earlier rounds.
A participant at the 2011 survey presentation suggested that unemployed or underemployed members might be less likely to respond, although the survey asked about the previous year's income.
EMWA's 2017 survey identified possible bias in both directions: high earners might be less willing to disclose financial information, while new writers might not yet be members or feel qualified to participate.
A 2018 survey of 175 freelances identified four limitations.
- Recruitment partly through fee-charging forums may have favoured respondents who had already achieved a certain income.
- Previous research suggested that struggling and extremely successful freelances might be less likely to participate than moderately successful ones, particularly when income was requested.
- Social desirability bias might lead to overreporting of income and behaviours such as networking and volunteering.
- Experience categories overlapped at five years, allowing some respondents to choose either two-to-five or five-to-ten years.
AMWA's 2015 survey also cautioned that respondents might differ in important ways from medical communicators as a whole, making generalisability unclear.
EMWA's 2021 survey was open only to members, so its findings could not be generalised to the global or even European medical-writing community; voluntary participation could introduce selection bias.
Missing answers and small subgroups
A subgroup's reported number of respondents depends on who answered all questions used in that analysis, so similar groups in different tables can have different denominators and parts may not sum to an inclusive total.
The 2012 EMWA analysis calculated demographics and job characteristics for all respondents, but income statistics only for those supplying income data, and did not replace missing values.
It excluded two salaries judged implausibly low: €270 per year from Pakistan and €6,916 per year from Denmark.
The 2021 EMWA survey accepted incomplete responses and could not query missing or inconsistent data because responses were anonymous.
It allowed only one completion per device but could not detect one person responding from multiple devices; some questions also permitted multiple options where this was illogical.
In AMWA's 2011 survey, men made up 16% of respondents, consistent with membership, but their small numbers made conclusive gender comparisons in subgroup analyses difficult.
EMWA's 2017 analysis included 266 evaluable respondents, roughly a quarter of membership, yet some country, employer and job-title groups contained ten or fewer people and might not represent their populations.
The Swiss and Scandinavian averages might not reflect true regional averages because few writers responded; German and British figures, representing 58% of employee respondents, were suggested as a possible benchmark for a large proportion of European employed writers.
Spread, benefits and local context
AMWA's 2011 salary distributions often had large variation and skew, particularly in small samples; means were generally accompanied by respondent counts, standard deviations and medians, with ranges where space permitted, allowing medians to be considered instead of means where appropriate.
A 2020 critique of the 2019 AMWA survey recommended minimums and maximums, box plots or scatter plots alongside its percentiles to provide finer context.
Gross employee salaries in the 2011 AMWA survey omitted the value of health benefits, retirement plans, paid vacation, flexible spending accounts, bonuses and stock options.
The 2019 salary and fee figures likewise assigned no cash value to benefits such as health insurance or tuition reimbursement, precluding direct comparison of employee and freelance earnings.
The 2012 and 2021 EMWA analyses did not account for cost of living, and the 2012 analysis also omitted purchasing power.
AMWA's 2019 survey used census areas because most locations had too few respondents for more granular analysis, but pharmaceutical and biotechnology hubs such as Boston, San Diego and San Francisco can differ economically from regional and overall averages.
Locally matched data could support benchmarking against peers, expose niche markets and local disparities, and place broader averages in context.
Cross-country comparisons also require context about currency conversion, repatriation, taxation, socialised benefits, local markets and cost of living; within the United States, ease of changing jobs and distance from family can matter alongside living costs.
The 2019 analysis included 65 respondents outside the United States, reported as 7.2% of the total; their small share was not thought to meaningfully confound overall results.
Comparisons across surveys and over time
The 2007 AMWA regression left nearly half of salary variance unexplained, indicating that factors outside its model remained related to salaries.
The 2019 and 2015 AMWA full-time employee samples differed, preventing a direct comparison of their salary growth.
The 2012 EMWA analysis converted non-euro incomes using exchange rates for a single date, 28 January 2013.
Its comparisons with 2006 were sometimes indirect because questionnaire changes limited comparability, and most factors had no statistical comparisons beyond analyses of income and satisfaction; several country groups had only three or four respondents.
AMWA's 2011 and 2015 surveys kept the substance of key questions but changed wording and order, potentially affecting answers, completion and comparisons between rounds.
The 2019 AMWA comparison of experience between waves also noted a slight change in question wording.
The AMWA-Canada 2005 and national AMWA surveys asked different questions, preventing comparison of freelance rates for writing a clinical study report or editing a journal advertisement between Canada and the United States.
The 2011 AMWA session summary urged caution when comparing salary increases with government inflation figures, which it characterised as notoriously low.
A salary survey can show association rather than causality; an executive summary's condensed presentation also limits the information and conclusions available.