Periodic Reporting for period 4 - HETEROPOLIS (The Design of Social Policy in a Heterogeneous World)
Reporting period: 2021-08-01 to 2022-07-31
Modern societies are characterized by tremendous heterogeneity in economic outcomes: from heterogeneity in wages and employment, to heterogeneity in capital income, wealth and health outcomes. It is unclear, however, how to map heterogeneity in these outcomes to heterogeneity in welfare. This mapping is crucial for the design of tax and benefit systems, providing insurance against individual risk and redistributing income between individuals, while maintaining proper incentives.
The main objectives of HETEROPOLIS were: 1) to provide new insights on the relation between inequality in earnings, wealth and consumption, 2) to develop a new consumption-based method to measure welfare inequality and heterogeneity in the marginal value of social transfers, 3) to provide and implement a simple, but general evidence-based framework to evaluate the differential design of social insurance based on observable heterogeneity, 4) to analyse selection effects due to unobservable heterogeneity and how they affect social insurance design, 5) to analyse heterogeneity in behavioural “biases” and their consequences for policy design.
The first part of HETEROPOLIS analyses the use of registry-based consumption measures to evaluate heterogeneity in welfare and exploits a newly developed data set based on administrative registers for the universe of Swedish households providing comprehensive and detailed information on income, wealth, labour market outcomes and other variables. The second part develops and implements a general evidence-based framework to evaluate the design of multi-faceted social insurance programs in a heterogeneous world. The final part of HETEROPOLIS analyses and estimates different sources of heterogeneity that affect market efficiency and justify further government interventions.
The main objectives of HETEROPOLIS were: 1) to provide new insights on the relation between inequality in earnings, wealth and consumption, 2) to develop a new consumption-based method to measure welfare inequality and heterogeneity in the marginal value of social transfers, 3) to provide and implement a simple, but general evidence-based framework to evaluate the differential design of social insurance based on observable heterogeneity, 4) to analyse selection effects due to unobservable heterogeneity and how they affect social insurance design, 5) to analyse heterogeneity in behavioural “biases” and their consequences for policy design.
The first part of HETEROPOLIS analyses the use of registry-based consumption measures to evaluate heterogeneity in welfare and exploits a newly developed data set based on administrative registers for the universe of Swedish households providing comprehensive and detailed information on income, wealth, labour market outcomes and other variables. The second part develops and implements a general evidence-based framework to evaluate the design of multi-faceted social insurance programs in a heterogeneous world. The final part of HETEROPOLIS analyses and estimates different sources of heterogeneity that affect market efficiency and justify further government interventions.
The original proposal was formulated around three major topics. For all three topics, the research objectives met at the outset of proposal have been met.
On the measurement of consumption (Part I), we have created a registry-based measure of consumption that maximizes the potential of the granular data on income, assets and other resources available from administrative registers. The research article "Studying Consumption Patterns using Registry Data: Lessons From Swedish Administrative Data" summarizes the construction of this measures, provides descriptive results how consumption relates to income and illustrates the power of this measure for applied research relative to other measures of consumption (from National Accounts or consumption surveys).
On the evaluation of differentiated social policies with heterogeneous agents (Part II), we have developed a framework (1) to use consumption, consumption responses and other choice behavior to uncover heterogeneity in the value of social insurance; (2) to evaluate differentiated social policies with heterogeneous agents, in particular on the optimal timing of UI benefits over the unemployment spell and on the design of retirement benefits depending on employment histories; (3) to evaluate social policies when allowing for choice in a context with moral hazard, in particular on the design of a minimum mandate with the option for more comprehensive coverage, and in a context with behavioral frictions.
On the drivers of heterogeneity (Part III), we have made important progress in understanding the drivers of heterogeneity underlying insurance choices, and the importance of risk and frictions in particular, and how that affects optimal policy design. We have been working on two major applications: the first is the option to buy comprehensive UI on top of a basic coverage plan in Sweden, the second is the deductible in health insurance in the Netherlands. We have also performed conceptual work on the underlying identification of risk perceptions and risk types and on the welfare analysis when both sources of heterogeneity underlie individuals' behaviour.
We have extended the analysis in unforeseen ways, but very much in the spirit of the objectives initially set out. First, while the main focus has been on the design of social insurance and the corresponding transfers in particular, the welfare state also interferes through other interventions. These interventions are often targeted and thus the heterogeneity in risks that individuals are subject to becomes even more central for policy design. We have been studying how predictable long-term unemployment is and how its determinants change during the unemployment spell and over the business cycle. Second, the need for social insurance, including the need for appropriate healthcare, has been very much confirmed during the Covid crisis. We have studied the heterogeneity in impact of Covid along socio-economic dimensions, in particular studying how the income gradient in mortality has changed.
My updated CV includes an updated list of the published papers and working papers. The link is here http://personal.lse.ac.uk/spinnewi/cv.pdf(opens in new window).
On the measurement of consumption (Part I), we have created a registry-based measure of consumption that maximizes the potential of the granular data on income, assets and other resources available from administrative registers. The research article "Studying Consumption Patterns using Registry Data: Lessons From Swedish Administrative Data" summarizes the construction of this measures, provides descriptive results how consumption relates to income and illustrates the power of this measure for applied research relative to other measures of consumption (from National Accounts or consumption surveys).
On the evaluation of differentiated social policies with heterogeneous agents (Part II), we have developed a framework (1) to use consumption, consumption responses and other choice behavior to uncover heterogeneity in the value of social insurance; (2) to evaluate differentiated social policies with heterogeneous agents, in particular on the optimal timing of UI benefits over the unemployment spell and on the design of retirement benefits depending on employment histories; (3) to evaluate social policies when allowing for choice in a context with moral hazard, in particular on the design of a minimum mandate with the option for more comprehensive coverage, and in a context with behavioral frictions.
On the drivers of heterogeneity (Part III), we have made important progress in understanding the drivers of heterogeneity underlying insurance choices, and the importance of risk and frictions in particular, and how that affects optimal policy design. We have been working on two major applications: the first is the option to buy comprehensive UI on top of a basic coverage plan in Sweden, the second is the deductible in health insurance in the Netherlands. We have also performed conceptual work on the underlying identification of risk perceptions and risk types and on the welfare analysis when both sources of heterogeneity underlie individuals' behaviour.
We have extended the analysis in unforeseen ways, but very much in the spirit of the objectives initially set out. First, while the main focus has been on the design of social insurance and the corresponding transfers in particular, the welfare state also interferes through other interventions. These interventions are often targeted and thus the heterogeneity in risks that individuals are subject to becomes even more central for policy design. We have been studying how predictable long-term unemployment is and how its determinants change during the unemployment spell and over the business cycle. Second, the need for social insurance, including the need for appropriate healthcare, has been very much confirmed during the Covid crisis. We have studied the heterogeneity in impact of Covid along socio-economic dimensions, in particular studying how the income gradient in mortality has changed.
My updated CV includes an updated list of the published papers and working papers. The link is here http://personal.lse.ac.uk/spinnewi/cv.pdf(opens in new window).
We have found that the registry-based consumption measure is extremely powerful relative to consumption survey measures and opens up the analysis of various questions relevant to welfare and policy design that couldn't be answered before due to data limitations.
We have developed a novel methodology using state-dependent marginal propensity to consume to evaluate social transfers that condition on these states. We expect this approach to be impactful as it can relax some limiting constraint of other methodologies and is widely implementable.
We have further developed the so-called Sufficient Statistics methodology (1) to account for heterogenous agents, (2) to account for differentiated policies, and (3) to account for moral hazard, adverse selection and/or choice frictions.
We also further developed the approach originated by Hendren (2013, 2017) to use elicited perceptions to study heterogeneity in risk types when beliefs and the revision of beliefs may be biased. We applied this methodology to estimate heterogeneity in employment prospects across unemployed job seekers and how that affects the selection into long-term unemployment.
We have extended the sufficient statistics approach to the timing of unemployment benefits and the design of retirement benefits. We have shown that the use of different consumption moments provides very clear guidance on the type of dimensions along which reforms in the unemployment or pension policy are desirable.
We have also reached out successfully to administrations and government in various ways. Our research on health insurance choices has been central in the Dutch policy debate and discussed with the Minister of Health in the Dutch parliament. Our research on retirement has been presented to the top administration in France and has been disseminated more widely in the policy debate on pension reforms in France and in Belgium. Our research on the prediction of LT unemployment risk has lead to ongoing conversations with public employment services in different countries on evaluating their own profiling models and the use for assignment of job search counselling and training. Finally, our Covid research related to the income gradient in mortality has opened the eyes of administrators and policy makers in Belgium, in particular with respect to the value of linking administrative data sources.
We have developed a novel methodology using state-dependent marginal propensity to consume to evaluate social transfers that condition on these states. We expect this approach to be impactful as it can relax some limiting constraint of other methodologies and is widely implementable.
We have further developed the so-called Sufficient Statistics methodology (1) to account for heterogenous agents, (2) to account for differentiated policies, and (3) to account for moral hazard, adverse selection and/or choice frictions.
We also further developed the approach originated by Hendren (2013, 2017) to use elicited perceptions to study heterogeneity in risk types when beliefs and the revision of beliefs may be biased. We applied this methodology to estimate heterogeneity in employment prospects across unemployed job seekers and how that affects the selection into long-term unemployment.
We have extended the sufficient statistics approach to the timing of unemployment benefits and the design of retirement benefits. We have shown that the use of different consumption moments provides very clear guidance on the type of dimensions along which reforms in the unemployment or pension policy are desirable.
We have also reached out successfully to administrations and government in various ways. Our research on health insurance choices has been central in the Dutch policy debate and discussed with the Minister of Health in the Dutch parliament. Our research on retirement has been presented to the top administration in France and has been disseminated more widely in the policy debate on pension reforms in France and in Belgium. Our research on the prediction of LT unemployment risk has lead to ongoing conversations with public employment services in different countries on evaluating their own profiling models and the use for assignment of job search counselling and training. Finally, our Covid research related to the income gradient in mortality has opened the eyes of administrators and policy makers in Belgium, in particular with respect to the value of linking administrative data sources.