Dernière mise à jour : 19 févr. 2022
Most OECD countries have experienced a strong rise in income inequality and decrease in social mobility in the last thirty years. Yet, demand for redistribution seems to have stagnated on average across OECD countries as discussed in the last OECD report.
Is this because people do not incorporate information on rising inequality and stalling mobility in their perceptions? Or are there other factors that mute support for redistribution?
To answer these questions, Emanuele Ciani, Thomas Manfredi and I did a meta-analysis of survey experiments providing people with information on inequality. Our study combines the results from 84 information treatments coming from 36 in-survey experiments where a randomly selected group of respondents receive either information about the overall extent of inequalities, or about their position in the income distribution. Those experiments were carried all across the world as you can see on this interactive map which displays the number of treatments by country:
Our results show that providing information on inequality has a sizeable impact on people’s perceptions and concern over inequality, but a rather small effect on their demand for redistribution. This result is consistent with observational evidence available in the OECD report.
Moreover, correcting respondents’ misperceptions about their own position in the income distribution increases the preferences for redistribution for those who previously overestimated their position and decreases it for those who underestimated. Yet the average effects are small.
Inspecting the heterogeneity across treatments and outcomes helps to explain the small average effect on demand for redistribution, but the evidence is not yet conclusive about the potential explanations. There are several competing hypotheses in the literature. Three explanations seem the most consistent with the data.
First, citizens may believe social policies are ineffective. In several studies, researchers have built experiments with several arms. One group was given information about inequalities, and another was presented with information about policies with an emphasis on whether they could actually reduce inequality. Consistently with this hypothesis, the group that received information about the effectiveness of social policies was the one with the highest support for public interventions against inequalities.
Second, one can agree that there is a problem, but not on how to fix it. In our context, citizens may agree that certain policies are needed to reduce inequality, but not on which specific policies to implement. Indeed, our meta-analysis shows that approval for interventions is lower when respondents are asked about specific policies- a tax on the wealthiest, for example. It is easier to convince subjects that something should be done about inequality than it is to convince them of what exactly should be done.
Third, citizens may also be concerned about the costs of redistributive policies. Several experiments show that providing information about the costs of policies to reduce inequality (such as the minimum wage) reduces the support these policies receive. However, it seems that this factor plays a less important role than the two other factors.
There is an impressive diversity in the outcomes using to measure inequalities, perceptions and preferences. This is why we end our paper on recommendations which could improve the comparability of future studies on the topic:
-International surveys, such as the Social Inequality module of the International Social Survey Programme or the World Values Survey, provide long-standing and agreed formulation of different questions on these issues. Although they are not necessarily the best measures, they can provide a benchmark and facilitate the comparison with observational evidence as well.
- Given the wide variety of outcomes, standardizing them by the control group standard deviation – or at least providing the necessary information to do so – is key to allow comparability. Similarly, as studies differ in terms of the number of outcomes collected, it is crucial to calculate indices that average many of them. In this respect, the literature seems to agree on standardizing – in the microdata – each outcome by the control group standard deviation and then take the simple average.
- More clarity in the meaning of different outcomes would facilitate comparisons. In particular, it would be useful to separate, even in the construction of indices, the following dimensions: perceived inequality (what people think the level of inequality is); preferences (what people think the level of inequality should be); concern (to which extent people think inequality is larger than it should be); and demand for policy intervention (what people think the government should do).