Working Paper · v2.0 · June 2026

Terrain: A National Labor Organizing
Strategic Model

A county-level quantitative model of strategic organizing terrain across all 3,143 U.S. counties. Grounded in Jane McAlevey's whole-worker organizing framework, the model scores each county on structural labor leverage and political terrain to identify where organizing has outsized political and economic impact.

Sam Kaplan-Pettus
UC Berkeley Political Science · laborterrain.org
↓ scroll to explore
A letter from the author

About the Project

This project has been a labor of love and an incredible learning opportunity. While labor organizing has not been the focus of my academic work, as our traditional electoral political methods have proven increasingly ineffective at meeting the moment, I became ever more convinced of the inescapable fact that the most powerful force of political resistance are organized workers. I was deeply inspired by Jane McAlevey's laying out of the factors required for an effective and urgent labor movement that was radically visionary and unshakably grounded. I sought to take her profound theoretical and historical arguments and operationalize that theory with data. I see this as a natural extension of her argument: that what matters is to win, and we should use whatever tools we can to make that happen. At the same time, I believe we need to approach the use of data in politics and organizing with real caution. Too often the lived experiences and wisdom of those on the ground and in the field are overlooked by those who claim data, polls, and algorithms are the only answer. My hope and commitment with this project is to add to, and not detract from, the deep real-world organizing that is currently happening.

We need to be honest about the state of the labor movement in this country and the urgency that is required of us. With fascism knocking at the door, we must be endlessly strategic in our resistance. This requires recognizing that some terrain and some sectors in the country will have more leverage than others, and if we can identify those pressure points and opportunities, we can force real change.

My theory of change diverges from many of those who seek data-backed answers to our social questions. I believe that change rests on deep interpersonal worker organizing that produces solidarity across our communities — and I believe that it is only from that that we can see larger-scale political and electoral goals accomplished. I believe that the wins of worker-focused organizing will not just trickle down, but direct the macro political change we demand in this country.

I hope to communicate two fundamental priorities in this project: 1) that there should be a national strategy to deploy resources to the highest-leverage campaigns — not just based on my research, but on the vast amount of knowledge and wisdom in the labor movement; and 2) that the mechanism of social and political change is real worker organizing, not performative activism or mobilization that does not fundamentally change the balance of power.

I will continue to update and grow this project with new features and layers of research. I look forward to feedback, criticism, and peer review. I am working to make this as collaborative, open-source, and replicable as possible. Feel free to reach out with questions at sam.kp@berkeley.edu.

Thank you to anyone who spends some time exploring this project.

In solidarity, Sam


A note on AI

I would not have been able to do this project without AI. I do not have the technical skills or the time required to write the scraping tools, data software, models, and front-end code to do what I have done in this project. This is a reality that makes me both very uncomfortable and incredibly inspired. I made every theoretical and methodological decision in this study, I double-checked every source and fact, and I spent dozens of hours a week painstakingly working through every component step and outcome — many that did not make it to the V2 that is currently published.

I believe that we need to see AI as both an existential risk to our economic and social system and an inspiring tool that can be directed for social good. It could give us a three-day work week or mass unemployment, but it does not change the fundamental balance of power in our world. The problem is still capitalism's distribution of power and ownership, and we must continue to fight to change that system in every way we can.


A note on the jobs board

The results of this project are in many ways about how national and state coalitions could or should focus their resources, but for individuals who want to get involved, I made the jobs board. Most of the data is drawn from the incredible resource of unionjobs.com, but I've made an updated interface that brings in contextual and personal information to help users find a good fit for them.

It is also an evolving part of this project. My goal is to showcase not only union staff roles but also job openings in key rank-and-file sectors — this is our real opportunity to bring the insights of this research to the jobs market. I want users to get an idea of how to have the most leverage as a worker in any sector, and understand how their power can fit into a larger movement.

Where the counties land

Explore the Results

Tier 1 — CapitalHigh capital leverage + decisive terrain
Tier 1 — CommunityHigh community leverage + decisive terrain
Tier 2 — CapitalCapital leverage — build electoral conditions
Tier 2 — CommunityCommunity leverage — build electoral conditions
Tier 3 — ElectoralDecisive terrain — build the organizing base
Tier 4 — Lower priorityBelow current strategic thresholds
Explore the Results
Map view
Viewing: National lens
Awakened — scroll zooms the map. Click “Freeze” to scroll the page again.
Senate competitiveness tiers based on NPR analysis by Domenico Montanaro (May 2, 2026). Source: NPR — 2026 Senate Races to Watch.
Distribution view
Viewing: National lens
labor leverage → electoral leverage →

The argument

Theory of change

All workers deserve a union — so we organize where it builds the most power. Two sector pathways run the same deep-organizing playbook but win different kinds of impact, and both feed back to grow the movement.

All workers deserve a union moral premise
Some sectors are more strategic than others strategic claim
Community-sector leverage
Deep labor organizing
Super-majority strike
Electoral impact*
Capital-sector leverage
Deep labor organizing
Super-majority strike
Material impact
Worker wins
Community wins
Grow the movement

*primary focus of this research project

Methodology

The model, end to end

How the model reads a county

Raw inputs feed three leverage factors; the factors are read through two lenses; the lenses place each county on a strategic map. Click any box to jump to its section.

Input what we measure
Model factors → lenses
Output strategic map
labor leverage → electoral leverage →
Theory why this model exists
01 — Theory

The Stakes

The early 20th-century labor movement built power through profound sacrifice, skill, and strategy — organizing against violent suppression and winning anyway. After World War II, labor established itself as a central force in American politics and the Democratic coalition for a generation.

Since the 1970s, a sustained attack and the rise of a neoliberal program pushed labor away from on-the-ground workplace organizing and toward reliance on labor law, party relationships, and campaign contributions. We are once again in a political predicament that demands a return to the fundamentals of organizing.

Theory of change. First: durable power is built through well-placed, interpersonal, organized solidarity — irrespective of institutional permission. Second: winning material gains for workers in the workplace is the fastest and most effective way to build a movement capable of addressing our broader social and political needs. When workers organize, they win. When workers win, we all win.

02 — Theory

Core Assumptions

1 — Organized workers have structural power

The right to withhold labor cannot be fully legislated away. Organized workers can act collectively even under hostile legal conditions — and win. That power is real, not dependent on institutional permission, and cannot be neutralized by money or messaging alone.

2 — Building that power requires deep organizing

McAlevey distinguishes three forms of work: advocacy directs resources at policymakers; mobilization activates an existing base; organizing is the long-term, relational work of bringing unengaged workers into a campaign, training them, building worker-led structures, and winning material gains. The movement has relied heavily on advocacy and mobilization while under-investing in deep organizing.

3 — Not all organizing produces equal political leverage

Every worker in every county deserves a union; that right is not conditional on predicted strategic value. But workers in key sectors, demographics, and locations have an outsized political impact. The model prioritizes the highest-impact campaigns first so we can build a winning coalition.

McAlevey, J. (2016). No Shortcuts: Organizing for Power in the New Gilded Age. Oxford University Press.
Silver, B. (2003). Forces of Labor: Workers' Movements and Globalization since 1870. Cambridge University Press.
03 — Theory

What the Model Does

The model scores every U.S. county on two independent dimensions. The Strategic Leverage Score (SLS) asks: how much structural power would organized workers here have? It is computed separately as SLS-Capital (crisis-creating potential against capital flows — ports, logistics, energy, concentrated manufacturing) and SLS-Community (relational crisis-creating potential within the community — hospitals, schools, transit, home health).

The Political Terrain Score (PTS) asks: how much does organizing here translate into political impact? It has two sub-scores — P1 Electoral Leverage (how decisive is this county's geography in consequential elections?) and P2 Incumbent Alignment (how aligned are current elected officials with labor's legislative agenda?).

Two lenses are available. The National lens uses federal electoral tipping weights and federal incumbent alignment. The State lens uses state legislative chamber control and state incumbent alignment. The model shows terrain. The user brings the goal.

Input what we measure
04 — Input

Inputs: Sectoral

05 — Input

Inputs: Electoral

06 — Input

Inputs: Alignment

Model factors → lenses
07 — Model · Factor

Sectoral Leverage

08 — Model · Factor

Electoral Leverage

09 — Model · Factor

Incumbent Alignment

10 — Model · Lens

Federal lens

11 — Model · Lens

State lens

Output strategic map
12 — Output

The 2×2

13 — Output

Tiers & the two pathways

14 — Output

Map

15 — Output

Scatter

16 — Limitations

Known Limitations

These are not disclaimers. They are specifications of where the data ends and organizer judgment must begin.

1Public sector employment is understated

SLS scores use Census Bureau County Business Patterns data, which excludes government employees. Counties where government employment is dominant — state capitals, large military installations, Washington D.C. — have understated SLS scores. QCEW government data has been ingested separately; full integration is in progress.

2Sector coding involves human judgment

The 42 sector scores were assigned by human researchers based on best judgment, case study review, and existing labor literature. NAICS classifications do not always map cleanly onto how workers experience their industries, and sector scores are static.

Most significantly, the model does not yet distinguish between latent structural leverage and current active organizing capacity. A utility worker has enormous latent leverage even where current organizing density is low. The model scores structural position but cannot yet see the gap between what a sector could do and what it is currently capable of doing. Future work will add cross-coding by multiple independent experts.

3Electoral sub-scores use presidential margins as the primary proxy

The current P1 formula uses presidential electoral margins as the primary proxy for congressional and state legislative district competitiveness. True district-level margins require separate ingestion of district-level results, which is in progress. Counties spanning multiple competitive districts may be under- or over-scored.

4Key vote data coverage is uneven

P2 depends on data sources with uneven coverage across states. Federal key votes and ideology cover federal legislators comprehensively, but about half the roster has no non-zero key-vote score and falls back to ideology alone; state legislative alignment uses party composition (Open States) where seat-level voting records aren't available. The model flags coverage level for each county — full, partial, or unknown.

5Associational and community power are absent

The model has no signal for existing community infrastructure — labor councils, interfaith coalitions, tenant organizations, mutual aid networks — or the organizational readiness of a community to support a campaign. No public dataset captures these reliably at national scale. This is the gap most likely to be filled by organizer judgment rather than future data work.

6Electoral proxies imperfectly measure labor political alignment

The electoral leverage calculations rely in part on the two-party competitive margin framework — historically strong but becoming a less reliable proxy for labor alignment. The model attempts to address this through P2 (demonstrated legislative behavior rather than party label), but the underlying P1 calculation still depends on competitive-margin data. Future versions will seek alignment measures less dependent on the two-party frame.

7P2 incumbent alignment coverage is uneven

Federal P2 is complete for all 533 current federal legislators across 3,142 counties. State P2 is a different measure — the partisan composition of each county's state-legislative seats (Σ Democratic seat-share / Σ all-party share, from Open States rosters), used because seat-level labor voting records aren't uniformly available at the state level; it carries no key votes and no CFscores, and is flagged as coverage_type = party_proxy (with party_proxy_state_uniform / party_proxy_unavailable fallbacks).

17 — Data Sources

Data Sources

SourceWhat it powersVintage
U.S. Census Bureau — County Business PatternsPrivate sector employment by NAICS. Primary input to SLS.Annual
BLS Quarterly Census of Employment and WagesPublic sector employment by NAICS. Supplements CBP.Quarterly
MIT Election Data and Science LabCounty-level 2024 presidential margins. Powers electoral leverage.2024
U.S. Census Bureau — District Relationship FilesCongressional + state legislative district to county mapping.Decennial
National Conference of State LegislaturesState legislative chamber seat counts. Powers chamber tipping weight.Post-election
NAICSSector definitions and industry codes.Current
Congress.gov APIFederal voting records on defined key votes, verified against House Clerk + Senate.gov XML.Per vote
GovTrack Bulk DataFederal legislator ideology from sponsorship patterns. 119th Congress.Per Congress
DIME (Stanford)Federal legislator ideology (CFscores), blended with GovTrack for the federal P2 ideology component. 850M+ contribution records, 1979–2024.2024
Open States APICurrent state-legislative party rosters. Powers state P2 (party composition of each county's seats). 50 states + DC.Ongoing
18 — Literature

Theoretical Foundations

The works below ground the model. Citations marked under review are part of an ongoing manual citation pass and are not yet finalized.

AuthorYearWorkStatus
McAlevey, J.2016No Shortcuts: Organizing for Power in the New Gilded Agecomplete
Silver, B.2003Forces of Labor: Workers' Movements and Globalization since 1870complete
Wright, E.O.2000Working-Class Power, Capitalist-Class Interests, and Class Compromiseunder review
Womack, J.2005Working the Machine: The Logic of Industrial Technology and Labor Powerunder review
Fox-Hodess, K.2023Comparative analysis of dockworker organizingunder review
Feigenbaum, Hertel-Fernandez & Williamson2018From the Bargaining Table to the Ballot Box (NBER WP 24259)complete
Kolerman2024Regression Discontinuity Aggregation — unionization & commuting-zone voteunder review
Leighley, J. & Nagler, J.2007Unions, Voter Turnout, and Class Bias in the U.S. Electoratecomplete
Rosenfeld, J.2014What Unions No Longer Docomplete
Banzhaf, J.F.1968One Man, 3.312 Votes (Villanova Law Review 13(2))complete
Lalisse, M.2022Measuring the Impact of Campaign Finance on Congressional Votingunder review
Harvard CLJE2024The Varied Voice of Laborcomplete
19 — Roadmap

Further Research and Future Updates

Seven near-term priorities shape the model's roadmap:

1. Regression validation against known campaigns. Test predictive validity against a sample of known organizing campaigns and use the results to update weights and thresholds. 2. Latent vs. active capacity. Add a dimension distinguishing latent structural potential from active organizing capacity, drawing on NLRB petitions, strike history, and density. 3. Intervention type classification. A third layer classifying counties by the kind of intervention the terrain calls for. 4. Full two-stage electoral leverage. Implement a full empirical Banzhaf power index. 5. Expanded political alignment coverage. Standardize key-vote and campaign-finance coverage across states. 6. Cross-coding of sector scores. Independent expert review to strengthen inter-rater reliability. 7. Non-partisan alignment measurement. Develop alignment measures less dependent on the two-party frame.

20 — About

About This Project

Terrain was created by Sam Kaplan-Pettus, an independent researcher and political organizer with a degree in Political Science from UC Berkeley, to surface where labor and political power-building intersect across every U.S. county.

The project is open to collaboration, peer review, and proposals. If you are a researcher, organizer, or data scientist interested in contributing, reach out via GitHub or email.

A note on AI

We believe in transparency about AI's role in research. This project was made possible by AI and would have been inaccurate and misleading without human guidance. Every methodological, research, theoretical, and strategic decision was made by a human (Sam). Anthropic's Claude did the data processing and cleaning, software engineering, and visual output that those decisions specified — the human decided, the AI executed and implemented.

The site's copy and prose are themselves a mix of AI and human writing.

sam.kp@berkeley.edu · LinkedIn ↗ · GitHub ↗