Heterogeneity in bunching at tax kinks: Mechanisms and interpretations

Author: Håkan Selin, And

Dnr: 281/2025

The purpose of the project is to examine how differences in bunching between groups at a kink point should be interpreted.

When income crosses a kink point, the tax becomes higher on the portion above the threshold. Some people then adjust their incomes to stay below the threshold, either by working less or by taking out income in other ways, and thereby pay less tax. When unusually many people cluster at a particular income level, this is called ‘bunching’. Researchers use the magnitude of such tax bunching to estimate how strongly people's reported incomes respond to changes in taxes. Large bunching is usually interpreted as meaning that the reported incomes of the group are sensitive to tax changes. But that interpretation is not always straightforward.

The theoretical part of the project compares two models. In the first (the standard model), individuals can adjust their income without any particular fixed costs. In the second model, individuals must bear a fixed cost to, for example, reduce their labor earnings, convert labor earnings into another type of income, or otherwise position themselves at the kink point. When there are fixed costs, the relationship between the size of the bunching and people's sensitivity to taxes becomes more complicated than in the standard model. Differences between groups may then depend on how many people are initially located above the kink point.

The project examines the importance of the groups' income distributions. The same kink point applies to everyone, but it is located at different places in different groups' income distributions. For example, highly educated people have higher incomes on average than less educated people. The kink point is therefore relatively far down among the highly educated's incomes, but higher up among the less educated's. In a model with fixed costs, this can affect how large the concentration of taxpayers at the kink point becomes.

In the final part of the project, the theoretical model is adapted to Swedish income distributions and tax rules. Group-specific estimates from actual income data are compared with results from simulated data. The object of study is the Swedish kink point for the central government income tax. People who self-report their income, such as owners of closely held corporations, can in some cases choose to take out labor income up to the kink point and then take out dividends, which are taxed at a lower rate. Using data from Statistics Sweden's FRIDA database, actual income patterns are compared with the patterns the model predicts.

Project leader

Håkan Selin, IFAU