Publications and record.

Who Fails the Filter? Size, Verification and the Incidence of an Unchecked Minimum Capital Requirement

Zhijun He · Yale University · 44 pages

Europe dismantled minimum capital requirements for private companies on the ground that they fail creditors. The other defence — that a minimum filters out the unserious — was settled by assertion rather than tested: against a European income it excludes nobody. This paper separates two parameters that are usually collapsed, size and verification, and reads four enacted regimes on both. A floor large enough to exclude a recognisable class is cheap only where the registry does not check; checking means paying for the exclusion one refusal at a time. Where both were set high, it was the size that got given up.

The evidence is one such floor. Brazil’s single-owner limited-liability form required paid-in capital of a hundred minimum wages, unverified — a stock, not a flow, standing at 1.0 to 1.6 times the maximum annual revenue the law allowed those firms in the regime they were leaving. In a register of 69.1 million firms, two populations under no requirement, five to nine years apart, declare the same nominal median against floors of R$78,800 and R$121,200. The unchecked floor bound anyway, through the filing offices: declarations of a 2020 revision ran four to six times above background two days after it was announced and seventeen days before it became law. It rationed a stock that had to survive later examination — sorting the few who reached the rung rather than holding many off it.

The paper closes on three common, cheaply reversible features of registry design that decide whether a reform of this kind can be evaluated at all.

He, Zhijun. “Who Fails the Filter?” SSRN Working Paper, 8 August 2026. Submitted to NEUDC 2026, Cornell. ssrn.com/abstract=7255539 · doi:10.2139/ssrn.7255539 · JEL L51, O54, O17, L26, K20, K22, D22

Governing Culture Through Capacity: A Tiered Institutional Framework for AI-Integrated Cultural Diversity Protection

Zhijun He (Yale, first author), Yuyan Wang (Columbia), Luoxianer Peng (Macalester) and Ziyu Liu (Macalester)

The UNESCO cultural convention system is capacity-agnostic: it assumes every signatory has roughly the administrative, legal and financial machinery needed to turn a ratified obligation into policy. That assumption is false by orders of magnitude, which is why a single universal framework underperforms hardest exactly where cultural diversity is under the most pressure. The recent wave of AI heritage projects has amplified the flaw rather than fixed it, producing three structural paradoxes — data sovereignty, model-validation externality, and infrastructure dependency — that turn preservation into a new form of extraction.

Three cases span the range. China as a high-capacity state, where the protection architecture is a product of prior institutional capacity. Rwanda as a mid-capacity state, and the analytically decisive case: gacaca, Ndi Umunyarwanda and the Kinyarwanda digitisation programme were not built on institutional capacity — they generated it, inverting the usual causal direction. Somalia as the limit case, where the argument is that cultural wealth does not produce cultural policy capacity: the Xeer and gacaca were structurally the same kind of institution, and what separated their fates was institutional trajectory rather than cultural richness. Documentation detached from the restoration of community governance is an archive without an institution.

The proposal turns this into an operational instrument: a Tiered Digital Cultural Governance Protocol, adoptable by resolution under the existing 2003 Convention without new treaty negotiation. Four requirements — a pre-deployment capacity assessment that fixes the governance tier; community-level data sovereignty as the test of bidirectionality, with communities as contracting parties rather than governments as proxies; offline-first technical standards for the lower two tiers; and built-in community validation of AI outputs. The feasibility section names its own risks: who defines “capacity” in the assessment instrument, elite capture inside communities, a three-to-four-year timeline against languages that have fewer than a hundred speakers left, and the absence of any way for a community to enforce a data agreement against an international organisation.

He, Z., Wang, Y., Peng, L., & Liu, Z. (2026). Honorable Proposal, Global Development Public Policy Youth Innovation Contest Finals — Renmin University of China, UNITAR SDGs Singapore Center and China Daily. Submitted 10 April 2026.

Interactive supplements: framework visualisation · stakeholder conflict map · Vanuatu generalisability test · capacity-assessment prototype

Climate-Resilient Infrastructure and Food Security: A Case Study of Common Harvest Farm in the Upper Midwest

Zhijun He · in Alternative Agriculture Futures in the Upper Midwest, supervised by William G. Moseley, Macalester College

A case study of one community-supported farm in Osceola, Wisconsin, built from a field visit and a three-hour interview with the owner-operator, and read against the six-dimensional food security framework — availability, access, utilisation, stability, agency and sustainability. Three investments are traced through their effects: a greenhouse with a wood-fired boiler that moved the season a month earlier, a solar array that cut electricity costs and kept cold storage running through a three-day outage, and an electric tractor conversion.

The more interesting finding is governance rather than technology. The CSA membership financed the solar installation by pre-purchasing five years of vegetables — an interest-free loan in exchange for a discount — and members were surveyed on whether to cut potato production when beetle pressure rose. The paper reads that as decentralised governance with a real constraint attached: as the farm grew, keeping members genuinely involved in decisions got harder, and a tiered membership was the compromise.

He, Zhijun. “Climate-Resilient Infrastructure and Food Security,” in Alternative Agriculture Futures in the Upper Midwest: A Collaborative Research Exploration (Macalester College Department of Geography, December 2024), 53–68. Full report, PDF

Economical Samaritans: Assessing the Impact of the HIPC Initiative on Inclusive Development in Ghana and Zambia

Zhijun He

On whether the Heavily Indebted Poor Countries Initiative delivered development rather than just relief — and for whom. Two cases with different post-relief trajectories, read for effects on inclusive development rather than on debt ratios alone. The question behind it is the one that recurs across this list: what a programme is credited with, and what actually changed.

Awarded top prize for individual research at the Elliott School Student Research Showcase, George Washington University, 24 April 2024, and presented again at the university-wide GW Research Showcase on 1 May. Selected by competitive review for the Midwest Political Science Association conference, Chicago, 2025. Showcase report

Cultivating youth-led agriculture solutions

Apoorva Reddy Neelapu, Sarah Edwards and Zhijun He · Jobs and Development

Written for the World Bank’s Solutions for Youth Employment programme, on three young people running agricultural initiatives in Sierra Leone, Peru and Colombia — a poultry and training operation in a district long dominated by mining, a regenerative cattle project in the Gran Chaco that trained smallholders’ children as field technicians, and an insect-protein feed company built on circular-economy principles. The piece ends on the constraint rather than the inspiration: youth-led initiatives need an enabling environment, and governments, donors, agribusinesses and farmer organisations are the ones who supply it or fail to.

Published 23 May 2024 on the Jobs and Development blog, from the S4YE Youth Advisory Group’s Agriculture and Informal Livelihoods thematic group. Read it · author page

Policy Substitutes or Complements? Industrial Policy and Institutional Quality in Rwanda

Zhijun He (Globalization Studies, Macalester; World Bank Youth Advisory Group), Ziyu Yan (Economics, NYU), Zonglin Zheng and Tianqi Zhao (Economics, Macalester) · 18 pages

Uses Rwanda’s Vision 2020 reform as a natural experiment, with World Bank Enterprise Survey data from 2006 to 2023 and a regression discontinuity design exploiting institutional variation between Kigali and the rest of the country. Industrial policy produces large short-term gains in export revenue and employment, but those gains shrink where institutions are already stronger, and they fade; institutional improvements last longer.

Presented in the undergraduate applied microeconomics session, Midwest Economics Association Annual Meetings, 21 March 2025. ssrn.com/abstract=5295229 · doi:10.2139/ssrn.5295229

After the Catastrophe: Rwanda’s Transformation

Zhijun He · Honours Project, Department of International Studies, Macalester College · 174 pages · defended April 2026, Pass with Highest Distinction

Standard development theory holds that recovery requires institutions first — courts, fiscal systems, the rule of law. Rwanda in 1994 had none of them and rebuilt a state anyway. The thesis treats that trajectory not as an exception to the good-governance consensus but as a refutation of it, working through four post-genocide mechanisms: gacaca community courts, umuganda collective labour, imihigo performance contracts and ingando political education camps.

Five concepts come out of that analysis. Convergent impossibility names the trap of prerequisites collapsing at the same time, which sequential development theory has no answer for. Paradox-breaking mechanisms are the institutional innovations that generate capacity where none existed. Negotiation-based capacity redefines state capacity itself — away from Weberian coherence, toward the strategic navigation of contradictory demands. Constitutive ambiguity describes the grey zones where coercion and consent became structurally indistinguishable, not as design failures but as the medium governance actually ran through. Bidirectionality, drawn from the Rwanda–Somalia–Ethiopia comparison, sets the boundary between governance and predation: the grey zone has to deliver to both the state and its citizens.

It closes on a sustainability paradox. The features that made the achievement possible — personalised authority, emergency legitimacy, community knowledge substituted for institutional independence — may be what prevents the form from reproducing under ordinary conditions. Using the Qin–Han transition as a structural analogy, the question becomes not whether the institutions survive their architect but whether their legitimacy framework can be replaced while the administrative architecture holds. That answer, the thesis argues, belongs to political choices not yet made by people not yet in power.

He, Zhijun. “After the Catastrophe: Rwanda’s Transformation.” International Studies Honors Projects 47, Macalester College. Committee: Ahmed I. Samatar (advisor), David Chioni Moore, Kathryn Linn Geurts (Hamline). Record · PDF, 973 KB

Note on the date: the repository record and its metadata give the year as 2025. The thesis was defended in April 2026, and the repository’s own comments field says so. A correction has been requested; until it is made, cite the defence year rather than the repository year.

Who Keeps the Record of Negotiation? Custody, Capacity, and the Digital Afterlife of Rwanda’s Gacaca Archive

Zhijun He · MADE graduate cohort, MacMillan Center, Yale

Between 2002 and 2012 Rwanda’s gacaca courts recorded, across 12,103 community jurisdictions, almost every adult’s relationship to the genocide. That record — an estimated sixty million pages, one of the largest transitional-justice repositories anywhere — has been scanned and indexed file by file since 2015, with completion slated for 2026. Custody has been contested at every step: commissions against NGO partners, Rwanda against the UN residual mechanism, a preservation mandate against an access regime that keeps the corpus largely closed.

The project is a custody history of the archive from 2012 to 2026, built from public records: the organic and closure laws, the custody chain through CNLG and its absorption into MINUBUMWE, parliamentary budget records, the paper trail of the Rwanda–IRMCT dispute, published archive metadata, press coverage across the period, and the existing gacaca fieldwork corpus. Two accounts of the last fourteen years are committed in advance to different observations — whether custody is unidirectional and tightening, or whether access, standards and trained capability are still being bargained over — and the discriminating evidence is named before it is looked for. The citizen side is the core test, not a footnote: whether ordinary Rwandans can reach their own files, and through which tier.

The general question underneath the case: what does datafication do to records whose original function depended on staying ambiguous? Rwanda’s median age is about twenty. For most of the population the genocide is inherited narrative rather than memory, which makes custody of the record custody of what the emergency was.

Article in preparation; target venue Journal of Genocide Research or equivalent, submission summer 2027.

Attributive Ambiguity: How Anticipated Deniability Deforms State Disaster Capacity

Zhijun He · The Whitney and Betty MacMillan Center for International and Area Studies, Yale University

A single disaster can be described, truthfully, both as a climatic extreme and as a failure of governance. The paper treats that dual attributability as an ex ante force in how resources get allocated, rather than as rhetoric that starts after the water goes down. Critical disaster studies established long ago that disasters are never natural; the argument here is that states treat that constructedness as an asset to be drawn down in advance. Because losses can later be settled into a “natural disaster” frame, the political cost of suppressing risk assessment and protective redundancy falls close to zero.

States therefore build visible, credit-claimable response capacity and let prevention capacity lag, since prevention is invisible and becomes legible only when it fails. Climate change intensifies the mechanism twice over: it raises real risk, and it hands every governance failure an alibi — non-stationarity makes “once in decades” language rest on a baseline that no longer exists. The long-run result is not general decay of state capacity but deformed specialisation, and it is an equilibrium of the legitimation circuit rather than a malfunction.

Building on Stone’s causal stories and on blame-avoidance theory, and engaging Pielke’s critique of disaster attribution, the concept is assessed through “settlement traces”: attributive wording in official texts, the scope of accountability, reconstruction standards and compensation channels. Two Himalayan cases ground it — Chamoli (2021) and the South Lhonak glacial lake outburst (2023) — and Derna’s 2023 dam collapse, where the natural-disaster frame failed to hold, is used to probe where the mechanism stops working.

Accepted for presentation, 2027 Annual Meeting of the American Association of Geographers · abstract code 53271 · session assignment in progress.

Does AI Clinical Decision Support Build or Erode Physician Competence? A Randomized Evaluation with Standardized Patients in Rural Primary Care

Zhijun He · research design

A trial design for the question the deployment literature keeps skipping: whether decision-support tools raise a clinician’s own diagnostic capability over time, or quietly substitute for it. Standardised patients make the counterfactual observable — the same case at the same difficulty, with and without the tool, measured on the physician rather than on the output.

First place, Outstanding Abstract Award, AI and Planetary Health Summer Institute (Peking University and Yale), Beijing, 2026.

Industrial Investment, Socio-Environmental Change, and Livelihood Resilience in Rural Tanzania: a case of Chinese manufacturing investment in Lulanzi

Enezael [surname] (China Agricultural University), Li Li (China Agricultural University) and Zhijun He (MacMillan Center, Yale University)

Rural livelihood resilience through cross-border investment: evidence from Chinese transnational township and village enterprises in Tanzania. A single-site study of what a manufacturing investment does to the livelihoods around it — the questions of attribution and of who bears the socio-environmental cost, asked at village scale rather than at the level of aggregate FDI statistics.

全球南方国别经验与新发展主义的理论要素建构

With Xu Xiuli, China Agricultural University (CIDGA) · in Chinese · commissioned for 《全球南方研究》 (Global South Studies)

An attempt to distil theoretical elements of new developmentalism out of country experience in the Global South, rather than restating the position. The starting point is Wang Hui’s two universalisms: the tension is not between the universal and the particular but between two universals, one of which — Western modernity — is a particular that achieved universal form. Grounding knowledge in a specific country is therefore an epistemological claim, not a methodological preliminary.

The framework runs as one line rather than three separate tests — who leads, how development is actually found, and how the resulting institutions get judged: 发展主体建构, building the political and social forces capable of carrying a long-run development task, without presupposing a strong state; 发展路径调适, adaptation through trial, alteration and diffusion, including the localisation of externally designed programmes — formative rather than normative; and 制度绩效导向, judging institutions by governance output under local constraint rather than by the completeness of their written text, which means accepting the standing tension between formal rules and informal practice.

Two cases carry the empirical section — China and Rwanda — with weaker-state values discussed in the conclusion rather than added as cases.

In preparation; sixth draft. An English version for a post-aid development knowledge landscape is under discussion.

Adaptation Lab protocol v0.1

SEESALT Adaptation Lab · open source

The published method that separates response capacity from prevention capacity, together with the multi-hazard data layer the separation requires. Published before it works, so that someone can tell us it doesn’t. Not yet run on a country. Method · Overview

Multi-hazard overlay with child-vulnerability assessment

Zhijun He and Tiago Maluta · UN Tech Over Challenge 2

Processes satellite data from NASA and NOAA to identify climate hazards affecting children and issues alerts to families; interactive maps were built for Cambodia, Kenya and Tajikistan. Released open source. What it did, and what it did not establish