Between 2017 and 2018, Finland conducted an experiment with universal income that gave rise to significant media coverage. 2,000 unemployed people receiving the basic unemployment benefit (560 euros per month) received the same amount in the form of unconditional income, which could be combined with income from work for the duration of the experiment (2 years, not renewable). On 6 May 2020, the final report evaluating the experiment was published (here is a summary of the results). The evaluators concluded that the experimental universal income had moderate positive effects on employment and positive effects on economic security and mental health. According to the final report, on average individuals in the treatment group worked approximately 6 additional working days (they worked 78 days). They experienced significantly less mental stress, depression and loneliness, and their cognitive functioning was perceived as better. Life satisfaction was also significantly higher. The results of the experiment therefore seem to argue in favour of a universal income. But is it really possible to draw lessons from the experiment with a view to generalizing the system? In 2018, I wrote that experimenting with universal income was “impossible“. Does the Finnish experience contradict this claim? It turns out that it is indeed difficult to draw lessons.
The principle of a universal income, as it is commonly defined, is to pay a sum of money to all members of a political community, on an individual basis, without means-testing or any obligation to work or take a job.
Such experiments generally concern a small number of people (in Finland, 2,000 individuals): the universal aspect of the measure is therefore lost, but a measure’s impact can differ depending on whether it affects everyone or only some of the population. How are the individuals chosen? Two options are favoured by practitioners: a totally random draw, which favours the representativeness of the experimental sample, or a saturation site, which consists of including in the experimental sample an entire community (for example a single labour market area), which helps to capture externalities and interactions (“do I stop working more easily when my neighbour stops or when my spouse receives assistance?”). In Kenya, villages are used as saturation sites. In the Finnish experiment, 2,000 long-term unemployed people receiving end-of-entitlement benefits (equivalent in France to ASS assistance) constituted the experimental group, with the control group being made up of recipients of end-of-entitlement benefits who had not been randomly selected. This poses two problems. First, the experimental group is not representative of the Finnish population. The long-term unemployed make up only a small part of the population. So we cannot really say how people with jobs would have reacted (would they have reduced their working hours?). Second, interaction effects are not taken into account: for example, consider a job taken up by an unemployed person in the experimental group, who thus increases his or her labour supply in the context of the experiment – might this job have been taken up by a member of the control group?
The definition of universal income tells us nothing about its level or what benefits it replaces. All options are on the table. Programmes with a more liberal, free-market orientation offer a relatively low universal income and replace most social benefits and sectoral subsidies (notably in agriculture) or can even substitute for regulations on the labour market (the abolition of the minimum wage is envisaged). In a more social-democratic logic, universal income would replace only the social minimum (France’s RSA income support benefit) and income support for the in-work poor (in France, the Prime d’activité). The amount envisaged is often equal to or slightly higher than the social minimum. Finally, in a degrowth logic, the universal income could be lifted to at least the poverty line in order to eradicate statistical poverty. The effects expected from the reform depend greatly on the amount envisaged and the benefits it replaces. In the framework of the Finnish experiment, the universal income was 560 euros, the amount of the basic unemployment benefit received by the members of the experimental group. Simply replacing this basic allowance meant that at first the income of the unemployed in the experimental group remained unchanged. But the universal income could at the same time be cumulated with job income. This means that returning to work could lead to an additional financial gain of as much as 560 euros.
The experimentation thus increased the financial gains from a return to work. This is not a result that one usually thinks of in relation to establishing a universal income. One question often asked is, “What happens when you get 1,000 euros a month without working?” It turns out that, for those on low incomes, the generalized roll-out of a universal income could have ambiguous effects on the incentive to work: it increases income without work but it also provides additional income for the working poor. On the other hand, for those earning the highest incomes, the monetary gain from increasing their income would be reduced.
The evaluation was complicated by the introduction of activation measures during the second year of the experiment (2018). Based on the “activation model” put in place, people on unemployment benefits had to work a certain number of hours or undergo training, otherwise their benefit was reduced by 5%. These measures affected the experimental groups asymmetrically: two-thirds of the control group were affected, compared with only half of the experimental group (Van Parijs, 2020). Theoretically, the incentive to return to work was therefore greater for the control group. Note that activation goes against the principles of the universality and unconditionality of universal income.
Notwithstanding the activation measure, the results of the Finnish experiment tell us that the hours worked are higher for the experimental group than for the control group. The financial incentives to work would therefore have worked! In fact, the evaluators stress the moderate degree of the impact on employment. In the interim report, which covered the first year (2017), the impact was not significant. In 2018, the impact was significant, since the people in the experimental group worked an average of 78 days, or 6 days (8.3%) more than the control group. The impact is, however, not very significant: with a 95% confidence interval, it is between 1.09 and 10.96 days (i.e. between 1.5% and 15%). Kari Hämäläinen concludes: “All in all, the employment effects were small. This indicates that for some persons who receive unemployment benefits from Kela [Finland’s agency handling benefits for those at end of entitlement] the problems related to finding employment are not related to bureaucracy or to financial incentives”. On the other hand, the experiment tells us nothing about the effects of possible disincentives for higher earners due to the financing of the measure: by construction, an experimental universal income is not financed. More seriously, gender analysis is virtually absent from the final report. All we know is, from reading a table, that women in the experimental group worked 5.85 additional days compared to 6.19 for men, but there is no discussion of the issue of gender equality. The issue of how choices are negotiated within a household is also not posed. The impact on the lone parent group is not significant “due to its small size”. In an Op-Ed published by the New York Times, Antti Jauhiainen and Joona-Hermanni Mäkinen criticize the sample size, which is five times smaller than initially planned: the small size makes it difficult to draw any conclusions about subgroups.
The final report highlights the beneficial effects on mental health and economic well-being. The impacts on people’s life satisfaction and on stress and depression are very significant. However, two comments can be made. First, we do not know what comes from the higher living standards of the individuals in the treatment group and what comes from the mechanism of a universal income (the certainty that people will have an income whatever happens). Given the way the experimental income was actually designed (it functions like an employment bonus), one can easily assume that it is the income effect that takes precedence. Likewise, since the individuals in the experimental group are in all cases better off financially, it is not surprising that their economic well-being increases. Second, there may also be a reporting bias due to a Hawthorne Effect: individuals in the experimental group know that they are part of an experiment and that they were chosen so that they have an advantage over the control group. This can lead them to be more optimistic in their statements.
In the end, the Finnish experiment offers few lessons about the effects of the establishment of a global universal income, i.e. one for all citizens. Only a small category of the population was involved, and funding was not tested. Yet funding is half the mechanism; Finnish trade unions are also opposed to a universal income because they fear that the necessary tax increases will reduce earnings from working. In addition, a family and gender approach has been completely ignored, whereas a universal income has been denounced by feminists as being liable to discourage women from taking up jobs (likening it to a mother’s wage). As with the RSA income supplement experiment in France [article in French], the failure of the Finnish experiment is explained in part by the contradictory objectives of the various scientific and political actors. The evaluators hoped for a sample of 10,000 people including individuals with different employment statuses. They were constrained by a combination of time, money and a ruling political coalition that was no longer enthusiastic about the idea of testing a universal income (“Why Basic Income Failed in Finland”). The Prime Minister’s Centre Party was in fact interested in the question of financial incentives for the long-term unemployed, which is a long way from the idea of reconsidering the central role of market labour or being able to say no to low-quality jobs, which is often associated with universal income. This was certainly a limitation of these costly experiments: subject to the inevitable supervision of politics, they risk becoming showcases promoting the agenda of the government in power.