Skip to content
Go back

Freelancing: You Earn Less Than as an Employee, and Do It Anyway

Illustration of a freelance professional with a laptop and clients

The most cited piece of research on self-employment is from 2000, and it says something no “how to become a freelancer” course mentions: most self-employed people earn less than they would as employees, and carry on being self-employed anyway.

That isn’t a demoralising fact, it is the correct starting point. If the median economic return is lower, the decision has to be made for other reasons. And those reasons have been measured too.

Hamilton’s result

Barton Hamilton (Journal of Political Economy, 2000) compared self-employed earnings with employee earnings using US longitudinal data, controlling for individual characteristics.

The finding has held up well: most self-employed people have lower initial earnings and lower earnings growth than they would have obtained as employees. The distribution, however, is highly skewed: a minority earns far more, pulling the mean up while the median stays below.

Averages for self-employment income are almost always misleading. The distribution has a long right tail: if you read an “average income”, you are reading the success of a few divided among everyone.

The gig-economy boom the authors themselves walked back

This episode is worth telling because it is rare and instructive.

Katz and Krueger (2016, later ILR Review, 2019) had estimated that the share of US workers in alternative work arrangements had risen from 10.7% to 15.8% in a decade. The figure went around the world as proof of the gig economy boom.

When the Bureau of Labor Statistics published its 2017 Contingent Worker Survey, the picture did not confirm that rise: the share of independent contractors in their main job came in around 6.9%, slightly down on earlier readings. Katz and Krueger published a downward revision of their own estimates.

The practical point for you isn’t statistical. It is that the “everyone is going freelance, jump in” narrative rests on numbers the authors themselves corrected.

The start-up problem, measured

The real obstacle at the start of freelancing isn’t competence: it is that nobody can verify it. Amanda Pallais (American Economic Review, 2014) measured exactly this with a field experiment.

On an online labour platform she hired 952 workers at random for a data entry task, and gave a subset of them a detailed public evaluation. Results:

Stanton and Thomas (Review of Economic Studies, 2016) documented the complementary mechanism: intermediary agencies help new workers land the first job, and the advantage is concentrated exactly in that phase.

PhaseThe real obstacleWhat the data says solves it
First jobNo verifiable informationAny public evaluation
First monthsIncome volatilityDiversifying clients
ConsolidationPrice anchored too low at the startExplicit re-anchoring
MaturityConcentration on few clientsRotation and reserves

Why you do it anyway: Benz and Frey’s answer

If median income is lower, why does independence keep attracting people?

Benz and Frey (Economica, 2008) analysed cross-country survey data and found that the self-employed report systematically higher job satisfaction than employees, controlling for income and other characteristics. Their explanation is what they call procedural utility: the value comes not from the outcome but from deciding how the work is done.

It is a result that makes the choice rational without pretending it is more lucrative. You are buying autonomy, and it has a measurable price.

Price: the one negotiation finding you need

On rates there is a robust and directly applicable experimental result. Galinsky and Mussweiler (Journal of Personality and Social Psychology, 2001) showed across a series of negotiation experiments that whoever makes the first offer systematically obtains a better outcome, through anchoring.

It is the same mechanism described by Tversky and Kahneman (Science, 1974): an initial number conditions all subsequent estimates, even when the recipient knows it is arbitrary.

The operational translation for a freelancer is precise and counterintuitive: the “how much does it cost?” question shouldn’t be deferred, it should be pre-empted. If you wait for the client to propose a budget, the anchor is theirs.

The first steps, in the right order

1. Buy public information, not clients. Pallais’s result says the initial constraint is verifiability. The first job exists to produce an evaluation, and should be judged on that, not on the fee.

2. Set your rate before the conversation. Write it on paper, with the floor you won’t go below. People who improvise the price during a call absorb the other side’s anchor, which is exactly what the experiments measure.

3. Compute the rate from capacity, not from a salary. Billable hours in a year are far fewer than hours worked: selling, admin and empty periods aren’t billed. Dividing a desired salary by theoretical working hours is the mistake that produces underpricing.

4. Don’t exceed a concentration threshold. If one client is worth more than half your revenue, you’re not independent: you’re an employee without protections. It is also the criterion many national rules use to reclassify the relationship.

5. Treat tax as a constraint to verify, not to improvise. Rules vary by country, revenue threshold and activity, and they change often: this is the one point in this guide where the only correct answer is to consult an up-to-date professional in your country.

6. Build the channel before you need it. The causal evidence on which contacts bring work is in professional networking; the “exist in a consultable form” part is in optimising your LinkedIn profile.

7. If you’re weighing the jump, weigh all of it. The comparison with alternatives is in how to change jobs, and negotiation techniques applied to fees are in asking for a raise.

My take: the low price that sticks is the real mistake

I think the single most expensive early mistake isn’t the low price but the low price agreed with someone who stays. A low price on a closed job costs you that job; a low price with a recurring client becomes your anchor for years.

I suspect that is Galinsky and Mussweiler’s dynamic observed over long timescales: the first agreed figure keeps working as a reference even when conditions change. I haven’t measured it, so I label it for what it is: my reading, not a result of the studies.

Frequently asked questions

Do freelancers earn more than employees? At the median, no, according to Hamilton’s work: initial earnings and growth both come out lower. The distribution is highly skewed, though, which makes averages useless as a personal benchmark.

Do I need a portfolio, or is experience enough? The experimental data says what counts is publicly verifiable information, not experience as such. A portfolio is one way to produce it; a public rating on a platform is another.

Are freelance platforms worth it? In the early phase they have a documented advantage: they solve the no-track- record problem. The cost is rate compression, so it makes sense to treat them as a launch tool rather than a permanent channel.

How do I handle income volatility? By diversifying clients and treating a fixed share of receipts as unavailable. It is the aspect self-employed people report as most stressful, and it is independent of income level.

Should I quit my job immediately? Nothing in the research cited suggests it has to be done in one move. The start-up finding says the constraint is producing verifiable evaluations, and you can start doing that without resigning.

Sources

Founder of UpSkillLab, with a Master's in Digital Marketing from the University of Florence. Every article starts from public sources cited in the text and is drafted with the help of AI tools, under my review and editorial responsibility: no purchased content, nothing published without human checking. Read the editorial method or report an error.


Share this post:

See all articles in Career & Work

Previous Post
Digital Skills: The 5% Who Can Solve Problems, and the Native Myth
Next Post
Growth Mindset: What the Research Actually Shows (and What It Doesn't)