The effect of wealth on behavior is of substantial interest to policymakers. For example, many important public policies -- such as changes to retirement systems, property taxes, or lump-sum components of welfare payments -- often involve transfers of wealth. Moreover, the costs and benefits to income redistribution depend on exactly how people respond to changes in their wealth. Therefore, understanding how (and by how much) people respond to changes in wealth is important for predicting and evaluating a wide range of public policies.
Despite a rich literature studying the effects of wealth on a large number of economic and non-economic outcomes, there remains little consensus on the importance and magnitude of wealth effects in a wide variety of settings. This lack of consensus is likely a result of the substantial practical challenges associated with credibly estimating wealth effects. This research tries to improve our understanding of the causal effects of wealth on a wide range of behaviors and outcomes by studying three large populations of Swedish lottery players. Because prizes are assigned at random, this makes it relatively straightforward to produce credible estimates using the well-accepted and uncontroversial framework of a randomized controlled design.
The projects described in this research outline are all based on the analysis of three recently collected samples of Swedish lottery players, comprising more than three million individuals. With the help of Statistics Sweden, the players have been matched to population-based administrative records containing annual information about employment, earnings, health, and a number of other variables (with most of the variables available annually beginning in the late 1970s). Besides information on the lottery players themselves, the data also contains information on the players' spouses, partners, and children. Overall, the combined data set has several unique advantages. First, there are minimal biases due to attrition or non-response since the players have been matched to population-wide administrative data. Second, because of the large sample size, the data can be used to investigate differences across many different demographic groups. Lastly, there is significant variation in the amount of prizes awarded, which means that it is possible to estimate non-linear effects of the prizes. In total, the data contain 13,000 individuals who were randomly assigned more than the average annual income in Sweden and another 1,500 who were randomly assigned more than ten times the average annual income.
The assembled data will be used for three separate research projects. The first project examines how wealth shocks impact labor earnings, spousal labor supply, retirement behavior, social insurance take-up, and occupational switching. These outcomes span a wide range of interesting labor market effects that one might expect to observe in response to unexpected increases in wealth. The information on spousal earnings will enable separate estimates of wealth effects by individual and by household, which may also shed light on economic models of decision-making within the household. The second project explores how wealth impacts a rich set of proxies for health, such as cause of death, drug prescriptions, and hospitalizations. The findings from this project will hopefully provide an important step toward resolving a long-running debate in social science about the extent to which the strong socioeconomic gradient in health reflects the causal effect of wealth on health.
The third project explores the effect of wealth on political participation (such as voter turnout and nomination for political office) as well as political attitudes. This project is highly interdisciplinary in nature and should be of considerable interest to both political scientists and economists. While existing research in political science and economics has documented a strong correlation between socioeconomic status and political participation, the interpretation of these results remains controversial. The rich data and credible research design in this project will hopefully result in uncontroversial estimates of the effect of wealth on a number of different measures of political participation and political attitudes.
One of the core questions in labor economics is what determines individuals' and households' labor supply decisions. The interest in labor supply in part reflects a desire to understand the consequences of economic policies on labor force participation, employment status, hours worked, and the age at which individuals choose to retire. These labor supply behaviors often have direct implications for tax policy, transfer programs, and a host of aggregate macroeconomic variables. For example, many policy proposals -- such as changes to retirement systems, property taxes, or the lump-sum components of welfare payments -- often involve implicit or explicit transfers of wealth. In these cases, knowledge about the causal impact of wealth on labor supply decisions is directly relevant for predicting the consequences of such policies. However, despite a large empirical literature, there remains little consensus on the magnitude of the effect of wealth on individual and household labor supply. In this project, we confront these challenges by studying the effect of wealth on the labor earnings of individuals and households using a large sample of lottery participants in Sweden. We exploit the random variation in wealth induced by the lottery prizes in order to estimate the causal effect of wealth on labor earnings. Our analysis uses three separate samples of Swedish lottery players -- comprising roughly 3 million individuals in total -- matched to administrative data on labor earnings of lottery participants, labor earnings of their spouses, and a large number of socioeconomic and demographic variables. The unusually rich and high-quality data set has minimal attrition and very long panels, allowing us to study the longer run effects of shocks to wealth. Moreover, the representativeness of the lottery participants allows us to estimate heterogeneous wealth effects across a wide range of demographics. Overall, we find that winning a lottery prize modestly reduces labor earnings, with the effects roughly constant over time and persisting more than 10 years. Approximately 10 percent of a lump-sum lottery prize is spent reducing labor earnings. We find that the earnings reductions are larger for those with higher earnings levels (in the years before winning the lottery), and we find that the results are fairly similar by age and gender. We also show how a simple economic model can quantitatively account for our estimates both over the life cycle and across the earnings distribution. We then use this model to estimate key labor supply elasticities that we hope will be useful for policymakers and researchers. These parameter estimates can be used to evaluate public policies which affect labor supply incentives, such as the Affordable Care Act in the U.S. Lastly, we estimate labor supply effects across individuals within a household, and we find much larger earnings responses for winners than for their spouses, regardless of the gender of the winner. In other words, it seems that the identity of the winner determines who gets to "slack off" within the household. This finding is useful for researchers studying appropriate household labor supply models, since it implies that unearned income (such as lottery wealth) is not "pooled" across individuals within a household, but instead it is somehow "tagged" to individual members. This finding should be useful in distinguishing among alternative theories of household labor supply. Though our project is still a work-in-progress, our research paper has already been used to teach graduate students in economics in the United States (including at institutions unrelated to any of the researchers who received this grant).
