#  SUPER 

 



## **Semester Undergraduate Program for Economics Research (SUPER)**

Harvard Economics is pleased to launch a new round of SUPER. Peruse available projects below and enter your application [**here**](https://docs.google.com/forms/d/e/1FAIpQLSd_WFKQGvqy-ZQVxRE2F05cewqB5GOXNrq33W487dubyWPWwA/viewform?usp=publish-editor) **by 1/9/2026.** RA matches will be announced by midnight on 1/21/2026 and RAships will begin on/around the week of 2/2/2026.

Applications require a transcript, résumé, list of relevant qualifications, and a ranking of your preferred project(s). Successful applicants will be placed in a semester-long RA position on one of their preferred projects (pay is $15 per hour, and hours vary by project - see descriptions).

SUPER is very much open to all currently enrolled Harvard undergraduate students, who will be on campus this semester.

**SPRING 2026 SUPER PROJECT DESCRIPTIONS:**

**The Economics of AI in the Workplace (Shengwu Li; Zoe Cullen, HBS; Danielle Li, MIT)**: We will be exploring company practices, running experiments and doing analysis on how human expertise is getting actively codified in the workplace, and how employees are responding to the codification of their expertise. The SUPER fellow will be keeping abreast of company practices through tech journalism, as well as carrying out several aspects of a field experiment that involves data collection and analysis. *Skills/prereqs: an understanding and interest in genAI tools, programming experience, strong communication skills. Weekly hours: negotiable.*

**Deep Learning for Economics** **(Melissa Dell):** Our research group has developed a variety of open-source packages for research at the intersection of economics and deep learning. These include Layout Parser (for document layout analysis), LinkTransformer (for LLM-based record linkage), MAR-S (for unbiased and robust inference with neural network predictions), EffOCR (for optical character recognition), and homoglyphs (vision transformer-based record linkage). The SUPER fellow will help to update, expand, and improve these packages, according to their interests. Familiarity with Python is essential. We have (optional) weekly group meetings, as well as regular meetings to discuss individual tasks. *Skills/prereqs: Python fluency. Weekly hours: 8.*

**The Economic Determinants of Isolationism** **(Melissa Dell):** This project examines isolationism in the United States in the period leading up to the Second World War, using individual level data on membership in a prominent isolationist organization (linked to individual census records), as well as rich data from local newspapers. The project examines how and why economic downturns influence isolationist sentiment, as well as examining how the media - and specifically the infiltration of German propaganda into U.S. newspapers - influenced the normalization of Nazi atrocities and isolationist sentiment. The SUPER RA will work with relevant newspaper content from the late 1930s and early 1940s. The work will include both qualitative assessments and designing/evaluating the quantitative extraction of information from the newspaper articles, through off-the-shelf or customized deep learning pipelines. We have weekly group meetings, as well as individual meetings to discuss specific tasks. *Skills/prereqs: Python fluency. Weekly hours: 8.*

**Employment, R&D Output, and Collective Bargaining in Higher Education and Hospitals Facing Cuts in Federal Support** **(Richard Freeman):** This project is a data-based analysis of the effect of Trump policies against higher education and collective bargaining on higher education and hospitals where grad students, post-docs and MD residents have unionized. We have a data set of hundreds of collective bargaining contracts and financial records of universities and other information to determine what impact of Trump Administration policies are having on what had been US comparative advantage in higher education. We are using text-as-data LLM-style modeling of contracts and personnel policies of institutions/workers without union contracts to assess whether collective bargaining can preserve academic freedom in a period of distress. The team has a mixture of experts in advanced analytics and more qualitative tools of analysis. Changes in policies and practices among universities create potential for assessing benefits and costs of different policies. The ultimate goal is for an artificial agent model of bargaining between unions and universities/hospitals -- the only sectors in US that have expanded union representation. Specific tasks are adaptable to student’s interests. *Skills/prereqs: Stat 104/Ec 20/Stat 110, Stata fluency, Python fluency, USE of AI for coding would help. Weekly hours: 10.*

**The Lens of Ownership: Enterprises and News Media** **(Nien-hê Hsieh, HBS):** The Ownership Project (TOP) at Harvard Business School is looking to engage two new SUPER undergraduate students in Spring 2026 alongside our existing team of undergraduate researchers. TOP aims to explore how structuring ownership of key elements of the economy can help individuals, businesses, and communities further their interests and address pressing challenges. We envision one student working with the existing team on ways that enterprise ownership can be structured to involve employees and other stakeholders (e.g., private equity, cooperatives, mutuals) and the strengths of weaknesses of different models. We envision another student helping to build out an emerging workstream, which explores how ownership shapes the functioning of news media and the viability of different forms of news media ownership (e.g., philanthropic, journalist owned, private equity, publicly traded). TOP undergraduates will engage in literature reviews to help build out the ownership lens on how best to address these areas of inquiry. Our working hypothesis is that only by examining the ownership of enterprises and exploring the full range of alternatives can we fully address pressing societal challenges today. *Skills/prereqs: none. Weekly hours: 10.*

**Impact of AI Adoption and Innovation on African Firms** **(Ebehi Iyoha, HBS):** This project examines how firms in Nigeria, Kenya, South Africa, and Morocco are innovating with and adopting AI technologies, and how this shapes productivity and participation in global value chains. The research draws on several data sources, including a field survey over 1000 firms, and datasets such as Orbis, Revelio Labs, Pitchbook, and the World Bank Enterprise Survey. Undergraduate RAs will support multiple parts of the research project: conducting focused literature reviews, assisting with survey preparation and rollout in Qualtrics, merging and cleaning firm-level datasets, and preparing short memos, descriptive summaries, and presentation materials. The project team will include one or two SUPER RAs working alongside three master’s-level RAs and two co-PIs, with weekly check-ins. *Skills/prereqs: Ec 1123/1126, Stat 104/Ec 20/Stat 110, Python fluency. Weekly hours: 15-20.*

**The Technological Origins of Objective Reporting in American Journalism (Quan Le, HBS):** This project examines whether U.S. newspapers became more objective and more informative after the Civil War, and whether falling printing costs and cheaper news collection (telegraph, AP/UP wires) drove that shift. The data combine structured newspaper panels (prices, circulation, pages, political affiliation), article-level text identifying wire content and linguistic measures of partisanship, AP subscription formulas, and predicted delivered newsprint prices constructed from newsprint-mill openings/closings. The RA will assemble and harmonize these sources, construct telegraph-cost and wire-adoption variables, implement the delivered-price model using existing mill data, build content measures from preprocessed text, and contribute directly to empirical analysis—estimating how cost shocks affect adoption, reporting intensity, and the move toward objective news. *Skills/prereqs: Ec 1123/1126, Stat 104/Ec 20/Stat 110, Python fluency, General interesting coding and learning about historical datasets. Weekly hours: negotiable (~10-15).*

**Tracing the Evolution of Economic Ideas with Generative AI** **(Vincent Pons, HBS):** This project uses generative AI to track how economic ideas emerge, spread, transform, and fade from the 18th century to today. By analyzing historical book corpora (Internet Archive, Google Books), we will identify when ideas first appear and map their diffusion across time and contexts. AI models will help generate a comprehensive list of economic ideas and automatically detect their presence in texts, enabling us to build a large dataset linking ideas to the books that discuss them. This dataset will support research on the life cycle of ideas and their interaction with major historical events. The RA will contribute to the creation of ground-truth data by reading economics books and identifying which economic ideas appear in each text. These annotations will be essential for testing and improving our idea detection methods. *Skills/prereqs: Strong reading comprehension and ability to accurately label ideas in economics texts, background in economics or the history of economic thought, experience with NLP, Python coding, or working with large language models is a plus. Weekly hours: 10.*

**Platforms and Policies** **(Vincent Pons, HBS):** When politicians adjust their platforms during an electoral campaign, is it mere cheap talk, or do these shifts carry over into their behavior once in office? We address this question by linking candidate platforms from French and U.S. elections to complete records of parliamentary speeches covering the French National Assembly since 1958 and the U.S. Congress since 1994. Using a close-election RDD, we measure changes in ideology, topical emphasis, and policy stances as politicians move from campaigning to legislating. The RA will assist with the analysis of the data, generating descriptive statistics as well as figures and table. *Skills/prereqs: Either R or Stata fluency. Weekly hours: 5-10.*

**Analyzing New Models of Psychology and Economic Theory** **(Matthew Rabin):** I have been working out the implications of some recent formal models attempting to embed more realistic psychology into mainstream economic analysis. Such models attempt to rigorously capture various improvements to the psychological realism of economics (to include such things as errors in statistical reasoning and in interpreting other people's behavior, self-control problems, and the effects of expectations, self-image, and other beliefs have on our utility). I place special emphasis on working out the implications of new assumptions across contexts rather than selectively focusing on situations where the models happen to match the evidence. (See my 2013 papers in the Journal of Economic Literature and the AER for outlines of long-term agenda motivating this research.) Potential research assistants would have some flexibility in choosing the existing models (by me and others) that interest them, and could contribute by (a) evaluating how existing empirical and experimental evidence accords to the predictions of these models, (b) analyzing the predictions of these new models across scenarios, and write up such analyses with guidance by me and other project members, or (c) help create calculators to and programs to streamline analysis of these models. Depending on student interest and current needs, there may also be opportunities to assist with current specific research papers in psychologically grounded economic theory. A strong background in microeconomic theory, math, and statistics is necessary; a background in psychology or behavioral economics could also be useful. The position is most likely to be valuable for those with potential interest in pursuing PhD studies after college. If it works out, continuation into future terms and between terms is typically available. *Skills/prereqs: Ec 1123/1126, Stat 104/Ec 20/Stat 110, Ec 1011a, R fluency, Python fluency, LaTeX/overleaf. Weekly hours: minimum 8 average, prefer ~10.*

**Computing Belief Biases from Data** **(Matthew Rabin):** In this project, SUPER participants will help Professor Rabin on a joint project with Professor Ben Bushong (visiting Harvard in the Spring) in building a scalable research pipeline for measuring how people form and update beliefs about real-world population facts (e.g., health behaviors, voting, internet access) when they see evidence. We will use both existing and new data. In our existing online experiment, participants stated an initial probability and then update after each binary signal drawn from large national surveys. We translate each person’s “belief path” into a set of implied moments of their underlying prior and run sharp computational checks for whether the path could have come from Bayesian updating, and if not what initial beliefs and errors in updating might best fit observed beliefs. SUPER participants will work on (i) cleaning and organizing belief-sequence data across topics and participants, (ii) implementing novel computational tests for Bayesian updating, and (iii) building fast diagnostics and visual summaries that quantify patterns like excess uncertainty, slow learning, and recency effects. We further expect to collect new data; and the SUPER may work on coding and planning future experiments. The work is coding- and data-focused and is ideal for students who want hands-on experience with empirical workflows and computational methods at the intersection of behavioral and experimental economics. *Skills/prereqs: Ec 1123/1126, Stat 104/Ec 20/Stat 110, Ec 1011a, R fluency, Stata fluency, Python fluency. Applicants who do not have all of the prerequisites above will be considered, but the position requires sufficient background in basic statistics and micreoeconomics and programming skills to help with the tasks described. Weekly hours: 8-10 minimum.*

**Integrating Machine Learning into New Estimators for Policy Evaluation** **(Rahul Singh):** Depending on interest and timing, we will select a project for the semester. Some options include: (i) how to estimate the effects of cash transfers, using satellite images as data; (ii) how to estimate the effects of schools, using student preferences as data; (iii) how to construct confidence intervals around effect estimates, using machine learning. The RA will have the opportunity to conduct simulations and a real-world policy evaluation, in Python or R, using novel estimators that integrate machine learning into econometrics. Off-the-shelf statistical packages do not exist; the work will be to adapt code from different problems to our problem of interest. The RA should have a strong background in programming, some experience in data science, and a curiosity for causal inference. We will meet one-on-one every other week. *Skills/prereqs: Ec 1123/1126, Stat 104/Ec 20/Stat 110, R fluency, Python fluency. Weekly hours: 10.*

**Wage Inequality and Mobility in Imperfectly Competitive Labor Markets** **(Oscar Volpe):** In this project, we will explore the determinants of earnings inequality and worker sorting using a two-sided matching framework. We will examine how firms compensate workers in thin labor markets, how workers choose jobs by trading off wages against nonpecuniary factors, and how variation in these preferences and practices across workers and firms shapes employment and social welfare outcomes. The RA will work directly with me and my co-authors on tasks that may include: assisting with data analysis, economic modeling, conducting literature reviews, and drafting a research proposal for data access. This position is particularly well-suited to students interested in labor economics and considering graduate study in economics. *Skills/prereqs: R fluency, Stata fluency. Weekly hours: 8-10.*

**International Power Over Two Centuries (David Yang):** This project develops a systematic, long-run measurement of international power—the ability of one country to shape the behavior, choices, and constraints of another—over the past two centuries. Building on our recent empirical frameworks that quantify bilateral influence through trade, the project extends this approach deep into historical contexts to capture how global power dynamics have evolved through major structural shocks. These include the era of colonization and its asymmetric dependencies; the waves of decolonization and state formation; the post-World War II institutional order; the ideological and geopolitical contest of the Cold War; and the subsequent reconfiguration of global influence marked by the rise of Japan and, more recently, the emergence of China as a central node in international networks. By integrating diverse historical sources with modern quantitative tools, the project seeks to describe and explain how countries gain, lose, and wield power across time and space. The RA will work with PIs, and responsible for data curation, data visualization, and empirical analyses. *Skills/prereqs: Ec 1010a, Ec 1010b, Stata fluency, Python fluency. Weekly hours: 10.*