Polarization and the Wisdom of Crowds  [Link]
Abstract: Opinions are formed in a social environment. Existing models of social learning tend to assume that the social environment is static. In reality, it is formed by the actions of individuals over time. This paper develops a model where echo chambers form because the probability that two agents communicate depends on their actions. Agents that act similarly have a higher probability of communicating compared to agents that do not. Agents act on the basis of their opinions, so communication is more common among agents with similar opinions. Agents also adapt to new information in each period, reflecting that real learning occurs in the presence of external influences, e.g., from news and public debates. I find that this model can exhibit: (1) Polarization of opinions; opinions cluster into different groups even when opinions are initially identical. (2) Failure of the wisdom of crowds; the average opinion does not approach the average information signal, even if all signals are drawn from the same distribution. (3) Sensitivity to initial conditions; the average opinion converges to a level that depends on the initial distribution of opinions. Unlike existing models of polarization, learning in my model is assimilative, opinions are more similar after communication compared to before. Unlike existing models with assimilative learning, my results support long-term differences in opinions without restricting communication between agents; agents have some probability of adapting to any agent in the network. My results suggest that if a society worries about high levels of polarization, it should encourage a culture of communication between individuals with opposing opinions.

Dual income taxation and wealth inequality [Link]
Joint with Charlotte Paulie and Kerstin Westergren
Abstract: Since several decades, wealth inequality has increased in many countries. Do changes in the tax system contribute to these trends? Using a quantitative model, we examine the effect on wealth inequality of changing from a comprehensive to a dual tax system. We start out from a standard open-economy incomplete-markets model, to which we add an entrepreneurial sector. The entrepreneurs in the model exploit the duality of the tax system after the reform by declaring income as capital (taxed at a flat rate) rather than labor income (taxed progressively). The model is parameterized to match the characteristics of the Swedish economy under dual taxation. In contrast to previous studies, we estimate the parameters of the stochastic process for entrepreneurial income using micro-data observations. We find that the introduction of a dual tax system increases wealth inequality and that the possibility of the entrepreneurs to declare income as capital plays an important role for this result. We also find that the level of aggregate capital and the share of entrepreneurs is higher in an economy with a dual tax system.

Learning the Value of Education [Link]
Abstract: This paper will use a model of social learning and educational choice to show that the process of expectation formation is important and can cause biases in expectations that correlate with socioeconomic status. This will result in the inter-generational persistence of socioeconomic factors even in a very simple educational choice framework. The results suggest a link between social segregation and inter-generational persistence that could be mitigated by targeting the expectations of individuals.

Human Capital Risk and Wealth Inequality
What is the role of human capital risk for wealth inequality and intergenerational mobility? This project introduces a model that formalizes an idea from sociology that family wealth acts as insurance against human capital risk and facilitates human capital investment.  This mechanism remains relatively unexplored in economic models of wealth dynamics.