My Profile

• Emergence of Slavery 

CSL MPI

On-going project

Throughout our history, the phenomenon of enslaving other humans has prevailed. Studies from archaeology and anthropology show that enslaving behavior existed even in prehistoric times. However, unlike in the most classic example of slavery, enslaved and free statuses are not binary and static in most societies. It is difficult to draw a line, but it is better to span a spectrum to describe one's social status between enslaved and free. Several theories have attempted to explain the origin of slavery. However, most of these theories focused on early state and agricultural societies, while ignoring the tension between enslaved and free statuses. This study attempts to explain this tension in a non-agricultural prehistoric context to help us better understand the emergence of slavery. Using evolutionary game theory and ethnographic empirical data, we aim to understand the dynamics between slavery and freedom. Our approach considers conflict over population and resources, as well as status flexibility and differentiation within and between societies. By integrating intergroup interaction patterns with variation in resource endowments -- factors typically treated separately across previous studies -- our approach offers a formalized explanation for how societies come to differentiate between more and less enslaved individuals. In doing so, the study sheds light on one of the foundational tensions in human society: the evolving tension between enslavement and freedom. 

• Wisdom of the Crowd: Crowd Analysis Project ---- Contributed one set of algorithms 

(https://woccap.com/)

CSL MPI 

Till Vater and I contributed four algorithms in this project, aiming to aggregate the most predictive crowd's guess on four fields: climate, economics, politics, and sports. The algorithm's aggregated crowd guess will be validated by the real-world data. Our algorithm implemented machine learning techniques (random forest) and automated the key predictor selection process (SHAP). Our algorithm is ranked as the fourth-best-performing algorithm in the field of politics, and we won 100 EUR for accurately predicting the decile of the reviewed algorithm. 

• Modern Slavery and Mistrust: a conceptual replication of Nunn & Wantchekon (AER, 2011) 

(https://cris.maastrichtuniversity.nl/en/publications/modern-slavery-and-mistrust-a-conceptual-replication-of-nunn-amp-/)

CSL MPI 

Human trafficking is modern slave trade, causing substantial direct damages to millions of victims annually. However, the broader societal impacts of this crime are not well understood. Here, we build on Nunn & Wantchekon (2011, AER) who traced how historical slave trading destroyed interpersonal trust in Africa. We study whether human trafficking analogously damages trust in two severely affected countries for which all required information is available sub-national levels: Romania and India. We find a robust link between victimization rates and reduced interpersonal trust. Our conceptual replication corroborates and generalizes results from Nunn & Wantchekon and highlights their contemporary relevance.

• Tricked into trouble: Deception, threat, and coercion in exploitative labor relations

(https://cris.maastrichtuniversity.nl/en/publications/tricked-into-trouble-deception-threat-and-coercion-in-exploitativ/)

CSL MPI 

Exploitative labor conditions are a massive global challenge, generating substantial illicit gains for delinquent employers. However, their strategic logic remains poorly understood. Here, we study the three practically most relevant forms of exploitative employer behavior in a principal-agent setting: deception, threat, and coercion. We analyze principals’ incentives for using these means, their welfare consequences, and the effects of introducing licensing to mitigate prevalent deception. We find that exploiters’ use of deception harms not only agents but also legitimate employers who are forced to compensate agents for the risk of exploitation. Moreover, we observe that increasing the costs of exploitation does not necessarily improve social welfare, as it can incentivize more employers to use milder forms of exploitation. Overall, we improve the economic understanding of exploitative labor relations by separating threat and coercion, integrating deception, providing insights into resulting market distortions, and identifying crucial pitfalls for seemingly first-best policy interventions.

• Final Project of Computational Statistics

(https://github.com/yudingshechu/Computational_Statistics/tree/Final-Project)

Uni Bonn

Used simulated data to implement LASSO in instrumental regression

Compared the post-LASSO instrumental regression algorithm in different simulated cases

• Term Paper of Research Module in Econometrics and Statistics 

(https://github.com/yudingshechu/Research-Module-Yield-Curve-Estimation-RKHS)

Uni Bonn

Replicated the implementation of RKHS non-parametric estimation in yield curve estimation

Used simulated data to show the advantages of RKHS estimation over other traditional estimations

• Literature Review of machine learning implementation in food security 

ZEF

Systematically collect, review, and summarize relevant papers, and explain machine learning algorithms in detail. Later this ends up to my first publication: "How robust are machine learning approaches for improving food security amid crises? - Evidence from COVID-19 in Uganda" (with Lukas Kornher and Clara Brandi), World Development, 196(2025), doi:10.1016/j.worlddev.2025.107171 

• Final Project of Effective Programming Practices for Economists 

(https://github.com/yudingshechu/EPP-Final-Project)

Uni Bonn

Clean and manipulate large volumes of China's national survey data

Replicated a DID paper with an automatically generated project template (pytask)

• Master Thesis: Robust Machine Learning for Food Security Forecasting (1.0/1.0) 

(https://github.com/yudingshechu/Masterarbeit)

Uni Bonn

The food insecurity issue is serious in eastern Africa, facing the shocks such as the Covid pandemic and war in Ukraine, a robust machine learning is needed for food insecurity forecasting. This study combines the Uganda National Households Survey data and other open-source data to predict household food insecurity before and during Covid. It is shown that models based on decision trees behave robustly when facing the shock of Covid. These tree-based models provide flexibility for policymakers to trade off the cost and benefit of food insecurity aiding. We also find that demographic and asset features provide the most prediction power. Finally, tree-based methods are robust against limited features or un-updated training data when facing a shock, implying they are robust in practical scenarios. 

Later this ends up to my first publication: "How robust are machine learning approaches for improving food security amid crises? - Evidence from COVID-19 in Uganda" (with Lukas Kornher and Clara Brandi), World Development, 196(2025), doi:10.1016/j.worlddev.2025.107171