Finland 2017-2018: what the experiment actually showed

Search for «basic income» in any public discussion and sooner or later Finland arrives. It is cited by supporters as proof that it works, and by opponents as proof that it achieves nothing. Both readings are wrong, and it is worth understanding why.

What was actually done

The experiment ran from 1 January 2017 to 31 December 2018. It was decided by the government then led by Juha Sipilä and carried out by Kela, the Finnish social insurance institution.

Two thousand unemployed people were selected at random from those who, in November 2016, were receiving unemployment benefit from Kela. They were paid a basic income of €560 a month, tax free, for two years, with two decisive features: there was no obligation to seek work, and the amount was not reduced if the recipient found a job or earned money elsewhere.

The control group consisted of people who, in that same November 2016, were receiving the same benefit but had not been selected. Participation was compulsory for those chosen.

This structure — random assignment, control group, national scale, public administration — is what makes the experiment exceptional. Basic income trials are almost always small, local and voluntary: people who volunteer differ from people who do not, and that ruins the comparison. Here it does not.

The employment results

Employment effects were measured over the period from November 2017 to October 2018.

The result: basic income recipients worked an average of 78 days in the period, six days more than the control group. A positive effect, but a small one.

There is a serious complication, however, and anyone citing the experiment without mentioning it is not reporting it properly. In 2018, halfway through, the Finnish government introduced an «activation model» that changed the rules for all unemployed people, control group included. This altered the context in which the experiment was running and makes the employment results considerably less easy to interpret than they would otherwise have been.

The evaluators say so explicitly: employment effects were small and their interpretation is complicated by the introduction of the activation model.

The wellbeing results

Here the picture is much clearer, and comes from a telephone survey conducted at the end of the period.

People who had received basic income described their condition systematically better than the control group: greater life satisfaction, less mental strain, less depression, less sadness, less loneliness.

They also reported a more positive view of their own cognitive capacities — memory, learning, ability to concentrate — and a more favourable assessment of their financial situation, which felt more manageable.

These are differences across many separate measures, all pointing the same way, which makes them more credible than a single isolated finding.

The four limits that have to be stated

Now the part supporters of basic income tend to skip.

First: it was not a universal basic income. Recipients were a narrow and specific group — people already unemployed and already on Kela benefits. Not the population. Citing the experiment as a test of a universal income misuses the word.

Second: the amount was not enough to live on. €560 a month is not a floor on which to build an independent life in Finland. The experiment tested a supplement, not a replacement for earned income. Many of the effects expected from a full basic income — above all on bargaining power — could not appear.

Third: no system effects. Two thousand people in a population of five and a half million do not move wages, prices, rents or the organisation of work. The hardest questions about basic income concern exactly those aggregate effects, and no experiment of this size can answer them.

Fourth: the financing was not tested. Two thousand people are paid for out of a budget line. A whole population is paid for by a tax reform, whose effects on behaviour and incentives are an integral part of the proposal. The experiment measured what happens to recipients, not what happens to the people who pay.

This last limit is not specific to Finland: it is structural in almost every basic income trial conducted anywhere. You can test the effect on those who receive; you cannot test the effect on the system, because to do that you would have to apply it to the system.

What can be concluded, then

Honestly, this.

The most widespread fear — that unconditional money leads people to stop working — found no confirmation. In the available comparison, recipients worked slightly more, not less. It is the most robust finding of the experiment, and it is not trivial: it is the objection you hear in every conversation on the subject.

The effects on mental health and on the sense of security were clear and consistent. Even a sum insufficient to live on, if it is certain and unconditional, changes how people stand in the world.

The promise of a significant rise in employment did not materialise. Those who argued that basic income would restart the labour market were wrong.

And everything to do with national scale, prices, wages and financing remains untested.

Why the experiment was built this way

A little-known detail helps explain why the results are what they are: the Finnish experiment was not conceived as a test of universal income. It was set up to answer a narrower question — whether the social security system could be simplified and made compatible with accepting discontinuous work.

In Finland, as elsewhere, finding a job did not necessarily increase an unemployed person's income, because earnings reduced benefits. The stated aim was to see whether removing that penalty changed behaviour.

That explains the choice of recipients — unemployed people rather than the general population — and it explains the amount, calibrated on the existing benefit rather than on any estimate of the cost of living. Judging the experiment as though it had set out to test something else amounts to blaming it for not answering questions it never asked.

Why this strengthens rather than weakens the case

It may look paradoxical to close an article that downgrades the evidence by still supporting the proposal. It is not.

If the available evidence were conclusive, the reasonable demand would be immediate introduction. It is not, and the reasonable demand becomes a different one: a serious European trial, large enough to measure system effects, long enough to capture changes in behaviour, with an amount sufficient to live on and a real financing mechanism.

It is expensive and complicated. It is also the only way out of a debate in which, for fifty years, both sides have cited the same two or three experiments and pulled them in opposite directions.

Finland demonstrated one thing that is often forgotten, and that the coordinator of the evaluation underlined: organising large-scale social experiments is possible. The rest still has to be done.

How to read experiments in general

The Finnish case also yields a useful test for any basic income trial, including the ones still to come.

The first question is who received: a population or a selected group? If selected, the results say something about that group, not about everyone.

The second is how much: a sum sufficient to live on or a partial supplement? Below that threshold, effects on bargaining power cannot appear, and citing those results to discuss a full income is incorrect.

The third is for how long: long-horizon decisions — changing occupation, returning to study, having a child — are not taken in a programme known in advance to be ending in two years.

The fourth is who paid: an external fund or ordinary taxation? In the first case, half the proposal has not been tested.

Apply those four questions to most of the studies cited in the debate and you find that almost none scores well on more than two. That should make everyone cautious, supporters included.

What a European trial should measure

If the main limitation of the experiments run so far is the impossibility of observing system effects, the practical consequence is that a useful trial has to be built differently.

It should cover whole territories rather than individuals drawn at random, because only then do effects on wages, prices and rents become observable. It should run long enough for people to make long-horizon choices: enrolling on a course, changing occupation, starting a business. It should pay an amount sufficient to live on, because it is only above that threshold that bargaining power changes. And it should include a real financing mechanism rather than an extraordinary budget line, because the effects on the paying side are part of the proposal.

It is an expensive and politically awkward design, and no single country has much reason to take it on alone. It is exactly the kind of undertaking a Union is for: pooling the cost of an experiment whose results would serve all twenty-seven.

A note on sources

The figures in this article come from the evaluation published by the Finnish Ministry of Social Affairs and Health together with Kela, and from the assessment of employment effects carried out by VATT.

The precision matters, because figures attributed to this experiment circulate in the debate that do not appear in the original documents, or that refer to preliminary findings later revised.

Anyone wanting to check should go back to the primary reports. It is advice that applies generally on this subject, where the distance between what the studies say and what is reported is unusually wide in both directions.