The phrase "cross-sector collaboration" is a euphemism for what should actually be the central operating model of U.S. workforce policy: a sectoral partnership, in a defined regional labor market, with named employers committing to hiring, named training providers committing to curricula tied to those hires, and a public anchor — usually a community college or a workforce board — committing to funding and intermediation. The euphemism is what kills it. "Collaboration" is what gets written on the grant application. The four specific elements above are what actually moves the wage data.
This essay argues that the U.S. has roughly twenty years of evidence on what cross-sector workforce partnerships look like when they actually work, that the working models share four structural features, and that the bottleneck on scaling them is governance, not money. The Strive Together collective-impact framework, made famous in John Kania and Mark Kramer's 2011 Stanford Social Innovation Review essay "Collective Impact," gave the rhetorical scaffolding. What it lacked was teeth. Two decades on, the working examples have teeth.
The evidence base: which partnerships have actually moved wages
A short list of cross-sector workforce partnerships has produced credible long-term wage effects in independent evaluation. They are worth naming because the field tends to talk in generalities.
Project Quest in San Antonio, evaluated by the Economic Mobility Corporation across nine years of post-program follow-up, produced annual earnings impacts of roughly $5,000 per participant — a result that grew, rather than faded, with time. The structure is a sectoral healthcare partnership between local hospitals, Alamo Colleges, and community organizations, with wraparound supports paid for by a braided funding stream.
Year Up, evaluated by MDRC in a randomized controlled trial with results published in 2019, produced earnings gains of roughly $7,000 a year (about 30%) three years after program entry. The partnership structure pairs employer commitments (named hiring partners at Bank of America, JPMorgan Chase, Salesforce, and others) with structured internships and a deliberate young-adult target demographic.
Per Scholas in IT training, evaluated multiple times by MDRC, has produced earnings gains in the $5,000–$7,000 per year range for low-income adults, with strong employer partnerships across the financial-services and tech sectors.
The Apprenticeship Carolina partnership in South Carolina — a state-led model that increased registered apprenticeships from roughly 800 in 2007 to over 47,000 by 2023, according to the South Carolina Technical College System — is the most cited state-level success in the apprenticeship literature. It combines tax credit design (the state $1,000-per-apprentice credit), employer engagement infrastructure, and community-college intermediation in a way that the federal Apprenticeship USA model has struggled to replicate at scale.
What these partnerships share is not "collaboration" in the abstract. They share four specific structural features.
The four features that distinguish working partnerships from grant theatre
Named employer commitments to hiring, not letters of support
Most workforce-grant applications include "letters of support" from employers. These are worth almost nothing. The partnerships that work require employers to commit, in writing, to interview a stated number of program graduates per year, with hiring conversion targets and reporting back. Year Up's named-employer model and Per Scholas's IT employer partnerships have this; the typical WIOA-funded training program does not.
Curriculum tied to a specific job, not a general occupation
"Train workers for healthcare jobs" is too vague to produce wage outcomes. "Train workers for the patient-care technician role at Methodist Hospital, with the specific competencies their unit-based clinical educators have validated" is the level of specificity that the working partnerships hit. The MIT Sloan researcher Paul Osterman has documented this in his work on sectoral programs, including his 2017 book Who Will Care for Us? on the direct-care workforce.
A backbone intermediary with paid staff
The collective-impact literature's central operational insight, often ignored, is that successful partnerships have a "backbone organization" with paid staff whose only job is coordination. JFF (formerly Jobs for the Future), Per Scholas's local affiliate structure, and the National Fund for Workforce Solutions' regional collaboratives have functioned this way. Volunteer steering committees do not produce hires.
Shared outcome metrics tied to UI wage records
The partnerships that survive over time tie themselves to outcome metrics drawn from state Unemployment Insurance wage records — actual employment and actual wages two and four years after program exit — rather than to self-reported survey data. That measurement discipline is what allowed Project Quest's effects to be credibly documented across nine years.
Why scale stays out of reach
The puzzle in U.S. workforce policy is not that successful sectoral partnerships don't exist. They do, and we know what makes them work. The puzzle is why so few of them scale.
Three structural reasons. First, the federal funding stack is fragmented across WIOA, Trade Adjustment Assistance (when active), Perkins (for career and technical education), Pell (for short-term programs), and a long tail of agency-specific grants. Each has different reporting, different eligibility, different timelines. A regional sectoral partnership has to braid five or six funding streams to operate, and the braiding cost itself consumes meaningful management bandwidth.
Second, governance is structurally weak. State Workforce Development Boards under WIOA are required to include employers, but their authority is mostly advisory and their planning cycles do not match employer hiring cycles. Brookings' Harry Holzer, in repeated policy analyses, has argued for stronger board authority and longer planning horizons; the federal statute has not followed.
Third, the evaluation infrastructure that distinguishes successful partnerships from unsuccessful ones is expensive. UI-wage-record matching, comparison-group construction, and multi-year follow-up cost real money. The Department of Labor's Chief Evaluation Office has funded some of this, but the budget is small relative to the program budget. Without rigorous evaluation, weak partnerships persist on autopilot because they pass the "people trained" metric.
The governance reform that would matter
The federal reform that would do the most to scale working partnerships is consolidation. A single competitive funding stream — call it a Sectoral Partnership Fund — with five-year grant durations, evaluation requirements tied to UI wage records, and authority for regional consortia to braid existing federal streams into a single performance plan, would change the operating environment substantially.
The state-level analog is to give Workforce Development Boards binding planning authority, with a defined regional sectoral-pipeline obligation and the ability to require employer participation in exchange for state procurement preferences or tax credit access. South Carolina's tax-credit model and Massachusetts' Workforce Skills Cabinet are working examples of what this looks like in practice.
For the broader frame on which training models actually move wages in the AI era — and how sectoral partnerships fit into the larger reskilling debate — see our flagship Reskilling for Real →.
"Cross-sector collaboration" is what gets written on the grant application. Named employer hiring commitments, curriculum tied to a specific job at a specific firm, a paid backbone staff, and UI-wage outcome tracking are what actually moves the wage data.
The collective-impact framework has been around long enough that the evaluation evidence is in. The good partnerships work. There are not enough of them, they are not large enough, and the funding stack actively discourages the structural features — multi-year grants, employer hiring commitments, paid backbones — that distinguish them. None of this is a mystery to the people running working programs. It is a mystery to the policy debate, which keeps re-litigating whether collaboration matters instead of operationalizing the specific institutional design that makes it work.
Updated May 21, 2026. This piece was substantively rewritten as part of NWLB's 2026 editorial refresh.
