The self-employed pay gap is twice the salaried one

Because the number was set differently. Among employees in Switzerland working full time, median gross annual earned income in 2025 was CHF 92,000 for men and CHF 80,900 for women — CHF 11,100 apart, about 12%. Among the self-employed the medians were CHF 85,700 and CHF 67,500 — CHF 18,200 apart, about 21%. Twice the gap.

The first of those is a salary gap. Somebody else wrote the number down: a hiring manager, a pay band, a function-level grid, in some sectors a collective agreement. The second is a price-list gap. It was written by the person being paid.

Read quickly, that sounds like an accusation. It is not one, and the rest of the evidence is the reason why. The gap is not everywhere. In the closest comparable data it is 3.2% among full-time self-employed people and 20% among part-time ones, and in two of the measured fields it runs the other way. What looks like a fact about a population turns out to be a fact about where the work sits.

Two gaps, and only one of them was set by somebody else

The Swiss figures come from the Bundesamt für Statistik, reference year 2025, published on 25 June 2026 and drawn from the Schweizerische Arbeitskräfteerhebung — over 100,000 respondents, which is a serious sample by any standard.

Median gross annual earned income, Switzerland, reference year 2025. BFS, from the Schweizerische Arbeitskräfteerhebung.

Group	Men	Women	Difference
Employees, full time (90% workload and above)	CHF 92,000	CHF 80,900	CHF 11,100 (12%)
Self-employed	CHF 85,700	CHF 67,500	CHF 18,200 (21%)

The overall full-time median across both sexes is CHF 87,000, which puts the two employee figures either side of a familiar middle.

One point of precision, because it does real work later. The 90% workload threshold — the BFS definition of full time — is stated for the employee rows. The self-employed rows are the median gross annual earned income of the self-employed, without that restriction. So part of the difference in the second row is hours rather than rates. Hold on to that. It becomes most of the article.

At the median, self-employment costs a woman more than it costs a man

There is a second finding sitting in the same table that almost nobody reports, because it requires reading the rows against each other rather than across them.

A man's median falls from CHF 92,000 as an employee to CHF 85,700 self-employed. That is CHF 6,300 lower, or 6.8% of where he started. A woman's falls from CHF 80,900 to CHF 67,500 — CHF 13,400, or 16.6%. Both go down. Hers goes down by more than twice as much, in francs and in percentage terms.

The same caveat applies, and I would rather repeat it than bury it: the employee rows are restricted to full time and the self-employed rows are not, so some of that difference is working hours. But the hours difference would have to be both large and strikingly one-sided to produce a fall of 16.6% against 6.8%.

CHF 18,200 separates the median self-employed man from the median self-employed woman in Switzerland. Among employees working full time, the same difference is CHF 11,100.

48,019 people set their own hourly rate. This is what they set.

The cleanest look at what happens when the seller writes the number comes from Foong, Vincent, Hecht and Gerber, who analysed 48,019 freelancer profiles on a US online labour marketplace, 48.8% of them women. On that platform the advertised hourly bill rate is chosen by the freelancer. Nobody else is in the room.

Median advertised rate: women $26.00, men $35.00. Women's median is 74.3% of men's. The means are $35.61 and $46.35. After controlling for job category, experience and education, the average treatment effect was −$6.28 an hour, a 13.6% reduction in the mean advertised rate, significant at p < 0.001.

Then the finding that changes the meaning of all of the above. Median hours worked: women 48.83, men 32.50. Total revenue came out roughly equal.

Multiply the medians and you get $26.00 × 48.83 = $1,269.58 against $35.00 × 32.50 = $1,137.50. That is not a rigorous operation — medians do not multiply, and the two medians are not drawn from the same people — but it shows the shape the authors report. Roughly the same money, reached by working half as much again.

The gap was not paid in money. It was paid in time.

The employee gap is a number somebody else wrote down. The self-employed gap is a number she wrote down herself, and it is twice the size. The variable that changed is not the woman. It is who was holding the pen.

The eBay result that rules out the obvious explanation

The obvious counter-argument to everything above is that price differences usually encode quality differences, and buyers are not stupid. There is one study that tests exactly that, and it is the reason the counter-argument fails.

Kricheli-Katz and Regev analysed all eBay auction transactions of the most popular products by private sellers between 2009 and 2012. Women received about 80 cents for every dollar a man received when selling the identical new product, and 97 cents when selling the same used product. They also received fewer bids.

Look at which way round that is. For a used item the buyer has to assess condition, wear and honesty, so the seller's identity genuinely carries information about the good. That is where you would expect an identity effect to be largest. It is smallest there: 97 cents. For a new item — identical, sealed, nothing left to find out — it is 80 cents.

The discount was largest where there was nothing to learn.

That places part of the price on the buyer's side of the table, not the seller's. Not all of it. Part of it. And a part that no amount of repricing by the seller addresses, because the seller was not the one who set it. Where that portion of the price does get settled — in the conversation itself — the experimental evidence says what moves it is information about the bargaining range, not disposition.

The German data says the gap is small, and that is the argument

freelance.de surveyed over 3,300 freelancers in Germany between 27 January and 2 March 2026 on their average hourly rates. The headline gap is 6.2%. Almost every article that cites this study cites only that number, and it is the least interesting one in the file.

Average hourly rate, German freelancers. freelance.de, Freelancer-Studie 2026, over 3,300 participants.

Segment	Women	Men	Gap
All respondents	€96.69	€103.06	6.2%
Full-time self-employed	€101.04	€104.39	3.2%
Part-time self-employed	€75.49	€94.45	20%
Administration	€66.57	€104.97	37%
Coaching	€118.48	€115.18	women higher
IT	€103.82	€101.74	women higher

A gap that closes to 3.2% in full-time work, reverses in coaching and IT, and reaches 37% in administration is not a property of the people in the sample. It is a property of the segments they are working in.

The behavioural rows in the same study are worth reading twice. In the period covered, 26% of women raised their rates against 24% of men, and 14% of women lowered them against 17% of men. More women moved the number up, fewer moved it down. Whatever is producing the average, a reluctance to change the price is not it.

A gap that lives in part-time work, in an economy that is largely part-time

Now put the German finding next to the Swiss labour market. In SAKE in Kürze 2024, 58.7% of employed women in Switzerland work under a 90% workload, against 20.5% of men. Women are 71.8% of all part-time workers — 1.345 million against 528,000. The survey runs on roughly 120,000 interviews a year, across 4.876 million employed persons, of whom 14.1% are self-employed.

3.2% is the rate gap between full-time self-employed women and men in the German survey. Among part-time ones in the same survey it is 20%.

These are two countries and two different populations, and I am joining them deliberately rather than carelessly. The German survey tells you where the gap sits. The Swiss survey tells you how many people sit there. A gap concentrated in part-time work, in an economy where most working women are part time, produces a large national average without anything having gone wrong at the level of the individual price list.

That is a composition, not a character trait. It also happens to be the most useful sentence in the piece, because it does two jobs at once: it tells a woman running a full-time firm that the 21% headline is not a statement about her rate card, and it tells a woman working at 60% precisely where to look.

What this means in practice

Three things follow from the evidence, and none of them is about how anyone feels about their price.

The first is that a national median is an average over a population that contains you and a great many people whose commercial situation has nothing in common with yours — different fields, different capacity, different decades of experience, no controls of any kind. The 21% is a real difference between two distributions. It is not a diagnosis of any firm inside them.

The second is that part-time is a pricing structure and not a lifestyle detail. Someone working at 60% is rarely selling 60% of a full-time offer. They are usually selling the same offer, with a fraction of the capacity to absorb the unpaid hour, the extra round of revisions, the call that was not in scope. The fixed cost of doing the work well lands on fewer billable hours. That is where a 20% part-time gap comes from, and it is visible in the scoping of a proposal long before it is visible in an income statement.

The third is that the marketplace and eBay results describe two different mechanisms and only one of them is on the seller's side. Advertised rates are set by the seller. Fewer bids are not. A discount produced on the demand side does not respond to the seller lowering the price — it responds to a channel where the buyer arrives already convinced, or to fewer and better-qualified buyers, which is a distribution question rather than a pricing one. I have written separately about the mechanics of underpricing expert work, and about why capability and income come apart.

There is a fourth, quieter one. In the marketplace data the cost showed up in the hourly rate and not in the annual total. If the same thing is happening in a Swiss consultancy or studio, the annual revenue line looks fine and the realised rate per hour does not. Annual revenue is the figure everyone looks at. Realised rate per hour is the figure that would have told them. One level up, at company rather than person level, the pattern repeats: firms with identical survival rates and a third less revenue.

What this cannot tell you from the outside

None of this is an answer about a particular business, and it is worth being exact about why. The BFS medians are medians: no controls for industry, age, tenure or hours, and the self-employed row is not restricted to full time, which I have flagged twice because it is the single biggest limitation on the headline comparison. The marketplace study is one US platform, and it measures advertised rates, not what was finally invoiced. The eBay work is a consumer goods market between 2009 and 2012, not professional services. The German survey is self-selected: people who chose to answer a freelancer platform's questionnaire. The Swiss part-time figures cover all employed persons, not only the self-employed.

What would actually settle the question for one firm is a fortnight inside its own records. Realised price by client and by project, not the rate card. Win rate read against price, so you can see the level at which buyers start saying no rather than the level at which you start feeling uncomfortable. Discounts, written and verbal. Unbilled hours by project. Price achieved by channel — referral, inbound, tender, intermediary — because the eBay result says the channel is part of the price.

Every one of those facts already exists, in the invoices, the time records, the proposals sent and the proposals lost. What does not exist is anyone whose job it is to assemble them. The Treuhänder reports the past correctly. The tax adviser minimises what is owed on it. The bank reads the result. None of them has been engaged to work out what the work was worth, and none of them has been asked.

The person who would benefit most from that answer is usually the only one at the table who has commissioned no analysis at all.