Yupeng Liu

PhD student in Strategy and Entrepreneurship
Leeds School of Business, University of Colorado Boulder

My research sits at the intersection of innovation, entrepreneurship, and the policies that shape them. I am curious about how knowledge moves across the boundaries of firms, communities, and countries, and about what turns a new idea into a high-growth venture. Alongside my research, I also build tools for researchers and the wider academic community.

Open and user innovation

How firms govern ideas from developers, users, and customers, and what changes when geopolitics or AI steps in.

Three papers

Entrepreneurship

How new firms form across places and over time, and what their founding registrations reveal about their growth potential, measured from millions of business registrations.

Three papers

Tools and Artifacts

Resources for other researchers that I built or helped build: a record of tested hypotheses, a public startup map, and the systems that ran a conference.

Four projects

Research

Open and user innovation

Under review at Organization Science

Dependence Reconfiguration: Open-Source Governance as a Response to Geopolitical Restrictions on Critical Technologies

When access to an upstream technology is restricted, how do firms organize innovation downstream? After the October 2022 U.S. export controls on advanced AI chips, Chinese AI firms released more open-source code under permissive licenses (about 7% more) and became 15.7 percentage points more likely to have outside developers hold most write access, which brought more forks and more outside pull requests. Licensing widens who can take part, and gatekeeping deepens how much outsiders can shape the project. AI firms in India and Japan, which the controls did not cover, show no such change. The firms moved from concentrated supply-side dependence on a few players to distributed demand-side dependence.

With Tony W. Tong and Yuping Li. SMS Berlin 2026 handout

Abstract

Firms in the Global South often depend on a small set of Global North incumbents for critical upstream technologies. How can they sustain innovation when access to these inputs is restricted not by any single partner, but by an across-the-board supply constraint? We develop the concept of dependence reconfiguration, arguing that constrained firms can redirect exchange relationships away from concentrated upstream suppliers toward distributed downstream users and developers. Open source software provides governance mechanisms that enable this shift: permissive licensing broadens downstream adoption, while open gatekeeping structures deepen external developer participation. We test this argument using a matched sample of Chinese software firms, comparing AI-related firms with non-AI firms before and after the October 2022 U.S. export controls on advanced computing hardware. We find that, after the shock, AI firms increased their release of permissively licensed open source projects and granted more external developers gatekeeping authority. These governance changes, in turn, increased community adoption and external contribution. The findings extend resource dependence theory by showing how firms manage upstream constraints through the cultivation of downstream relationships. We further identify open source innovation as a strategic pathway through which Global South firms sustain innovation under technological restrictions.

Working paper, 2026

The Rise of National AI Ecosystems amid Geopolitical Competition: Governance and Strategy Implications

Once AI became a question of national power, the U.S. and Chinese governments stepped in to decide who may join their AI ecosystems and who owns the key assets. We call the result a national AI ecosystem, run jointly by the state and a few hub firms. Both countries are drawing firmer borders while leaving some doors open to keep ideas coming in.

With Tony W. Tong and Shihao Zhou.

Abstract

Existing ecosystem research offers an incomplete account of the competitive dynamics surrounding critical technologies such as artificial intelligence (AI), where intensifying geopolitical competition has elevated the state from a background institution to an active orchestrator. We advance a new framework centered on the concept of national AI ecosystems, arguing that state activism is driving the consolidation of firm-centric ecosystems into broader configurations aligned with national strategic objectives. We propose that these national AI ecosystems are governed through a dual-governor model, in which states and hub firms jointly orchestrate activities by navigating trade-offs along two core dimensions: control (whether to open ecosystems to rivals or close them for security) and ownership (whether to centralize strategic assets within the state or distribute them among ecosystem participants). This framework theorizes how geopolitical competition transforms governance mechanisms, shifting from market-based coordination toward state-firm co-governance arrangements. Analyzing the United States and China, we show that both countries are converging toward increasingly bounded national AI ecosystems while strategically leveraging selective openness to stimulate innovation. This dynamic has profound implications for firm strategy, compelling hub firms to navigate nuanced identity duality and constraining complementors’ affiliation choices. More broadly, our framework advances ecosystem research and responds to calls to reestablish firm-state strategic interdependence as a central line of inquiry in strategic management.

Research in progress

AI Review Summaries and the Allocation of Owner Attention: Evidence from Google Maps

Customers tell firms a lot in their reviews, but no owner reads everything. Google Maps now condenses a restaurant’s reviews into a short AI summary. Our hypothesis is that owners and managers pay more attention to the points in that summary and address them more in their replies. We are testing this on 13,718 restaurants and 13.66 million reviews in seven metro areas, 2022 to 2026.

With Jay Park and Shaoqin Tang.

Research

Entrepreneurship

Research Policy, 2022

The Startup Cartography Project: Measuring and Mapping Entrepreneurial Ecosystems

The project estimates each new firm’s chance of an IPO or a major acquisition from what its founding registration already shows, then maps that chance for every state, metro area, county, and ZIP code from 1988 to 2016. Growth outcomes are concentrated: the top 5% of new firms by estimated quality account for more than half of the later IPOs and acquisitions. The places with the most startups are often not the places with the most promising ones (Silicon Valley has the highest average quality but below-average startup formation per person; Miami is the reverse), and it is the promising ones that predict later growth.

With R.J. Andrews, Catherine Fazio, Jorge Guzman, and Scott Stern. Download data Interactive map

Abstract

This paper introduces the Startup Cartography Project, which provides entrepreneurial ecosystem statistics for U.S. states, metropolitan areas, counties, and zip codes from 1988 to 2016. Using business registration records and predictive analytics, we estimate each new firm’s probability of extreme growth (an IPO or high-value acquisition) and build four measures of an ecosystem: the rate of startup formation, average entrepreneurial quality, quality-adjusted quantity of entrepreneurship, and the realized performance of each startup cohort. Regional entrepreneurial quality, rather than quantity, predicts later regional economic growth. The project releases a public dataset and the interactive U.S. Startup Map.

NBER Working Paper 28787, 2021

How Is COVID Changing the Geography of Entrepreneurship? Evidence from the Startup Cartography Project

In eight states, new business registrations fell to 57% of their 2019 level by late March 2020, then rebounded, and the year ended with 21% more new firms than 2019. The rebound came from LLCs and partnerships, while Delaware corporations, the usual form for venture-backed startups, stayed near their 2019 level. Growth was higher in neighborhoods with a larger share of Black residents, especially higher-income ones. In New York City, it centered on the Bronx, Harlem, and parts of Brooklyn and Queens, while Manhattan never stayed above its 2019 level.

With Catherine Fazio, Jorge Guzman, and Scott Stern.

Abstract

Using business registration data from eight U.S. states in the Startup Cartography Project, we track how new firm formation changed during the COVID-19 pandemic. Registrations fell sharply at first, then recovered and rose above prior levels in all eight states. Within states, the recovery varied with neighborhood income and racial composition: growth in business formation from 2019 to 2020 was higher in areas with more Black residents, especially higher-income Black neighborhoods. Startup formation rose after the passage of federal relief packages, most in higher-income and more Black neighborhoods, even though the CARES Act did not directly target new business creation.

Working paper, 2019

Do Entrepreneurs Respond to Everyday Changes in Financing Costs?

We built the first daily index of U.S. firm formation, about 5,000 new registrations on an average business day from 1988 to 2014. A one-percentage-point rise in the 3-month Treasury rate is associated with 6.4% fewer new firms seven days later, but only 3.2% less quality-adjusted entrepreneurship, so the founders who drop out tend to be the less growth-oriented ones. The drop lasts about six weeks, and the lost firms do not come back later.

With Jorge Guzman.

Abstract

We ask whether U.S. entrepreneurs respond to the small, day-to-day movements in financing costs that follow monetary policy. Combining a high-frequency index of U.S. business registrations with identification through heteroskedasticity, we find that short-term interest rate shocks have a stable and statistically significant effect on entrepreneurship: startup registrations drop roughly seven days after a rate increase. The effect fades after about six weeks with no offsetting rebound, which suggests the lost startups are not simply delayed.

Tools and Artifacts

Alongside my research, I build tools and resources that other researchers use.

Research infrastructure

Construct Laplace

Construct Laplace is a searchable record of the hypotheses that strategy and organization research has tested. For each paper it shows the hypotheses, how the ideas were measured, and what was found, so researchers and AI assistants can see what the field already knows.

Paper: Construct Explorer, with Tony W. Tong, Shaoqin Tang, and Sen Yan. construct-laplace.org Video on YouTube

Public map

U.S. Startup Map

Every new business registered in the U.S. from 1988 to 2016, placed on a map and shaded by its estimated growth potential. The map comes from the Startup Cartography Project, whose approach was first introduced in Guzman and Stern (2015). I worked on the project as a full-time research associate.

startupcartography.com

Peer review

Paper Review System

The system that ran peer review for the 2026 Strategy Science Conference. It gathers each reviewer’s background, matches reviewers to submissions by expertise, assigns papers, and tracks every review to a decision.

lypjim.github.io/StrategyScience2026

Conference website

Strategy Science Conference 2026

The website and program for the 9th Strategy Science Conference, held at the Leeds School of Business in Boulder on May 28–30, 2026. It lists 90 papers in 20 sessions across four rooms, searchable by author or keyword. I built it on GitHub and Cloudflare Pages.

strategyscience2026.org

Talks

Oct 2026Strategic Management Society Annual Conference Berlin
Aug 2026Academy of Management Annual Meeting, Professional Development Workshop Philadelphia
Jul 2026Open and User Innovation Conference Harvard Business School
May 2026Strategy Science Conference University of Colorado Boulder
Feb 2020Eastern Economic Association Annual Conference Boston

About

Before the PhD I was a research associate on the Startup Cartography Project at Columbia Business School, then spent two years in Singapore at a global consulting firm, building data tools for forensic investigations.

PhD, Strategy and EntrepreneurshipCU Boulder, 2025 to present
MA, Applied EconomicsGeorgetown University, 2018
BSc, MathematicsUniversity of Birmingham, 2017
BA, ManagementHuazhong University of Science and Technology, 2017