Publications

Across the developed world, there are growing calls to streamline and improve ever more complex income tax codes. Executing reform has proven difficult. Even when the desired outcomes of a reform are clear, the tools to design fitting reforms are lacking. To remedy this, we developed TaxSolver: a methodology to help policymakers realize optimal income tax reform. TaxSolver allows policymakers to focus solely on what they aim to achieve with a reform -- like redistributing wealth, incentivizing labor market participation or reducing complexity -- and the guarantees within which reform is acceptable -- like limiting fluctuations in taxpayer incomes, protecting households from falling into poverty or shocks to overall tax revenue. Given these goals and fiscal guarantees, TaxSolver finds the optimal set of tax rules that satisfies all the criteria or shows that the set of demands are not mathematically feasible. We illustrate TaxSolver by reforming various simulated examples of tax codes, including some that reflect the complexity and size of a real-world tax system.

• Preprint PDF

For over a century, life course researchers have faced a choice between two dominant methodological approaches: qualitative methods that analyze rich data but are constrained to small samples, and quantitative survey-based methods that study larger populations but sacrifice data richness for scale. Two recent technological developments now enable us to imagine a hybrid approach that combines some of the depth of the qualitative approach with the scale of quantitative methods. The first development is the steady rise of "complex log data," behavioral data that is logged for purposes other than research but that can be repurposed to construct rich accounts of people's lives. The second is the emergence of large language models (LLMs) with exceptional pattern recognition capabilities on plain text. In this paper, we take a necessary step toward creating this hybrid approach by developing a flexible procedure to transform complex log data into a textual representation of an individual's life trajectory across multiple domains, over time, and in context. We call this data representation a "book of life." We illustrate the feasibility of our approach by writing over 100 million books of life covering many different facets of life, over time and placed in social context using Dutch population-scale registry data. We open source the book of life toolkit (BOLT), and invite the research community to explore the many potential applications of this approach.

• Preprint PDF

Merchants often use personalized pricing: they charge different consumers different prices for the same product. We assess the ethicality of personalized pricing by generalizing and extending an earlier model by Coker and Izaret (Journal of Business Ethics 173:387–398, 2021) who found that price personalization ethically outperforms unitary pricing. Using a simulation analysis, we show that these results crucially depend on the choice of parameters and do not hold universally. We further incorporate additional sources of marginal cost into the utility function that will likely arise from personalized pricing. These include the expectation that personalized pricing is widely considered unfair by consumers who prefer that all consumers are charged the same price (unitary pricing), and that firms often approximate the consumers’ willingness-to-pay in ways that may raise negative sentiments among consumers who feel that their privacy is breached. By extending our model with disutility from unfairness perception and disutility from surveillance aversion, we demonstrate that personalized pricing is quickly outperformed by unitary pricing under social welfare functions that tend to prioritize total utility (utilitarianism and prioritarianism), whereas personalized pricing can ethically outperform unitary pricing under social welfare functions that tend to prioritize equality (egalitarianism and leximin). Our findings illustrate various intricacies and dynamics regarding the circumstances under which personalized pricing can be considered ethical.

• Code repository
• Open access pdf file

The COVID-19 pandemic led to reductions in non-COVID related healthcare use, but little is known whether this burden is shared equally. This study investigates whether reductions in administered care disproportionately affected certain sociodemographic strata, in particular marginalised groups. Using detailed medical claims data from the Dutch universal health care system and rich full population registry data, we predict expected healthcare use based on pre-pandemic trends (2017 – Feb 2020) and compare these expectations with observed healthcare use in 2020 and 2021. Our findings reveal a 10% decline in the number of weekly treated patients in 2020 and a 3% decline in 2021 relative to prior years. These declines are unequally distributed and are more pronounced for individuals below the poverty line, females, older people, and individuals with a migrant background, particularly during the initial wave of COVID-19 hospitalisations and for middle and low urgency procedures. While reductions in non-COVID related healthcare decreased following the initial shock of the pandemic, inequalities persist throughout 2020 and 2021. Our results demonstrate that the pandemic has not only had an unequal toll in terms of the direct health burden of the pandemic, but has also had a differential impact on the use of non-COVID healthcare.

• Code repository
• Open access pdf file

Quantitative sociologists frequently use simple linear functional forms to estimate associations among variables. However, there is little guidance on whether such simple functional forms correctly reflect the underlying data-generating process. Incorrect model specification can lead to misspecification bias, and a lack of scrutiny of functional forms fosters interference of researcher degrees of freedom in sociological work. In this article, I propose a framework that uses flexible machine learning (ML) methods to provide an indication of the fit potential in a dataset containing the exact same covariates as a researcher's hypothesized model. When this ML-based fit potential strongly outperforms the researcher's selfhypothesized functional form, it implies a lack of complexity in the latter. Advances in the field of explainable AI, like the increasingly popular Shapley values, can be used to generate understanding into the ML model such that the researcher's original functional form can be improved accordingly. The proposed framework aims to use ML beyond solely predictive questions, helping sociologists exploit the potential of ML to identify intricate patterns in data to specify better-fitting, interpretable models. I illustrate the proposed framework using a simulation and real-world examples.

• Code repository
• Open access pdf file

Machine Learning (ML) is gradually revolutionizing the social sciences as it has done for subjects like genomics and medicine. The new millennium brought an ambition to fnd the ‘Signal and the Noise’, followed by funding initiatives such as the creation of a working group in Computational Social Science by the Russell Sage Foundation. All aim to capitalize on ML’s ability to fnd intricate patterns; patterns which might have otherwise been missed in the traditional approach to model building.

• Code repository
• Open access pdf file

In times of crisis, real-time data mapping population displacements are invaluable for targeted humanitarian response. The Russian invasion of Ukraine on February 24, 2022, forcibly displaced millions of people from their homes including nearly 6 million refugees flowing across the border in just a few weeks, but information was scarce regarding displaced and vulnerable populations who remained inside Ukraine. We leveraged social media data from Facebook's advertising platform in combination with pre-conflict population data to build a real-time monitoring system to estimate subnational population sizes every day disaggregated by age and sex. Using this approach, we estimated that 5.3 million people had been internally displaced away from their baseline administrative region in the first three weeks after the start of the conflict. Results revealed four distinct displacement patterns: large-scale evacuations, refugee staging areas, internal areas of refuge, and irregular dynamics.

• Supplementary material
• Open access pdf file

Prediction is an underused tool in the social sciences, often for the wrong reasons. Many social scientists confuse prediction with unnecessarily complicated methods or with narrowly predicting the future. This is unfortunate. When we view prediction as the simple process of evaluating a model’s ability to approximate an outcome of interest, it becomes a more generally applicable and disarmingly simple technique. For all its simplicity, the value of prediction should not be underestimated. Prediction can address enduring sources of criticism plaguing the social sciences, like a lack of assessing a model’s ability to reflect the real world, or the use of overly simplistic models to capture social life. The author illustrates these benefits with empirical examples that merely skim the surface of the many and varied ways in which prediction can be applied, staking the claim that prediction is a truly illustrious “free lunch” that can greatly benefit social scientists in their empirical work.

• Code repository
• Open access pdf file

School closures have been a common tool in the battle against COVID-19. Yet, their costs and benefits remain insufficiently known. We use a natural experiment that occurred as national examinations in The Netherlands took place before and after lockdown to evaluate the impact of school closures on students’ learning. The Netherlands is interesting as a “best-case” scenario, with a short lockdown, equitable school funding, and world-leading rates of broadband access. Despite favorable conditions, we find that students made little or no progress while learning from home. Learning loss was most pronounced among students from disadvantaged homes.

• Code repository
• Pre-registration
• Open access pdf file

This article examines a new dimension of similarity, namely education and workplace similarity between lawyers and judges, and its impact on judicial outcomes. It builds on the similarity literature in law and economics, and uses the case study of the Hong Kong Court of Final Appeal to explore whether judges tend to decide in favor of parties represented by lawyers who are “similar” to them in terms of shared educational backgrounds or workplaces. Our findings show that lawyers who are more similar to judges perform significantly better in terms of winning cases. This association remains when controlling for lawyer, judge, and panel effects. The results point to the importance of social interactions inside and outside the courtroom on judicial decision-making, and prompt reflection regarding court design.

COVID-19 poses one of the most profound public health crises for a hundred years. As of mid-May 2020, across the world, almost 300,000 deaths and over 4 million confirmed cases were registered. Reaching over 30,000 deaths by early May, the UK had the highest number of recorded deaths in Europe, second in the world only to the USA. Hospitalization and death from COVID-19 have been linked to demographic and socioeconomic variation. Since this varies strongly by location, there is an urgent need to analyse the mismatch between health care demand and supply at the local level. As lockdown measures ease, reinfection may vary by area, necessitating a real-time tool for local and regional authorities to anticipate demand.

• Code repository
• Open access pdf file