Career-Connected Education Outcomes Research Review: Parts 2 to 4

How to read the evidence

Career-connected programs reliably raise high school completion by a small amount, shift some students toward two-year colleges, and work best when they are deep, well built and tied to real local demand. Part 1 of this series covered employment and earnings. Parts 2 to 4 cover high school completion and credentials, college access and completion, and the program and policy features that change results. Scope is United States K-12 programs, with a few international studies used only as design guardrails.

Every finding carries a tag in the form [Author(s) Year; design; sample/setting]. Where a study already appears in the S4IS evidence appendix (corrected Oct. 3, 2026), the tag is copied from it. Where this review confirmed a detail the appendix still marks as open, the tag gives the confirmed detail and the change is listed in the review status section.

Design classes. You will see three classes, from strongest to weakest causal warrant.

Lottery studies. Early college and P-TECH studies use admissions lotteries. The researchers, the What Works Clearinghouse (WWC) and Arnold Ventures often call these randomized. The S4IS appendix tags lottery-based admission studies as quasi-experimental (lottery-based), and this review follows that convention so client text stays consistent. They are among the strongest designs here, but they compare applicants only, so they tell you about students who chose to apply.

Matching is not randomization. Propensity score matching and regression with controls handle observed differences only. Several authors in this review say so directly (Giani 2022; Plasman 2026; Glennie, Ottem & Lauff 2020; Witzen 2019). Read those results as associations.

Effect sizes. Results are given as the study reports them: percentage points (pp), percent, odds ratios, or Hedges’ g. Nothing has been converted or rounded up.

The anchor synthesis. The CTE Research Network meta-analysis is the best single summary for Parts 2 and 3 [Lindsay, Hughes, Dougherty, Reese & Joshi 2024 (AIR, CTE Research Network); systematic review and meta-analysis of 28 secondary CTE studies meeting WWC standards (RCT and quasi-experimental); U.S. secondary students].

The authors found no statistically significant negative impacts and “insufficient evidence to distinguish between the impacts of different CTE types”; no moderator (program type, outcome domain, gender, disability) reached significance.

Winchester. The Shihadeh Innovation Center in Winchester, Virginia, is not cited as evidence anywhere in this review. No published outcome study of it exists.

This review is research, not legal advice.

Part 2: High school completion, persistence and credential attainment

CTE raises high school completion by a small, consistent amount (meta-analytic g = 0.137), stand-alone technical schools show the largest graduation effects, and early colleges show the largest credential effects, mostly in associate degrees.

CTE concentration

• Nebraska and South Dakota. Concentrators were 7 pp more likely than similar non-concentrators to graduate on time, in a sample of 112,764 students (2012-13 to 2016-17 cohorts). The study meets WWC standards with reservations. [Brodersen, Gagnon, Liu & Tedeschi 2021 (REL Central 2021-087); quasi-experimental (matched comparison of concentrators and similar non-concentrators); Nebraska and South Dakota high school students]
• Arkansas. Each additional CTE course raised on-time graduation by 3.2 pp, and concentrators were 21 pp more likely to graduate. [Dougherty 2016 (Thomas B. Fordham Institute); quasi-experimental (instrumental variables using within-school variation in course offerings, plus coarsened exact matching, alongside OLS with school and cohort fixed effects); three Arkansas cohorts entering 9th grade 2008 to 2010, more than 100,000 students]
• Washington State. Students with more than one CTE credit were 5.7 to 12.0 pp more likely to graduate than students with minimal CTE. Nearly all students (94% of the 2024 cohort) earn some CTE credit, but only 45% concentrate in a single pathway. [Riggs & Cigarroa Kennedy 2025 (Washington ERDC); descriptive/correlational (longitudinal administrative data); more than 750,000 Washington students, 2013-14 to 2023-24]
• Timing. CTE taken later in high school is more protective against dropout than CTE taken earlier. [Gottfried & Plasman 2018 (AERJ 55(2)); descriptive/correlational (regression on national longitudinal transcript data); U.S. national sample] The odds figures in circulation (1.6% vs. 1.2% per course) come from press coverage and have not been checked against the article.
• Dose. A middle-range mix of CTE and academic courses is associated with the lowest dropout risk, but only for students who entered 9th grade at age 14 or younger. [Plank, DeLuca & Estacion 2008; descriptive/correlational (discrete-time hazard/survival models); NLSY97 national sample]

The often-quoted national figure (concentrators graduate at about 96% vs. about 85% overall) is a raw comparison with no controls. [ACTE 2024 fact sheet; descriptive/correlational; national administrative rates] It was not re-checked. Confirm with ACTE before using it, and do not present it as an effect.

Stand-alone technical high schools

Two admissions-cutoff studies show the largest graduation effects in the literature, for students near the cutoff.

• Connecticut. Male admittees were 10 pp more likely to graduate (control mean 80%). There was little evidence of any effect for female students. More than 57,000 8th-grade applicants, 2006 to 2013. [Brunner, Dougherty & Ross 2023 (Review of Economics and Statistics 105(4); NBER w28790, 2021); quasi-experimental (regression discontinuity at technical high school admission cutoffs); Connecticut Technical High School System applicants, P20 WIN administrative data]
• Massachusetts. Attending a regional vocational and technical high school raised on-time graduation by 7 to 10 pp for higher-income students, with “suggestively larger” effects for lower-income students. [Dougherty 2018 (Education Finance and Policy 13(2)); quasi-experimental (regression discontinuity at admission cutoffs); applicants to three oversubscribed Massachusetts regional vocational and technical high schools]
• Baltimore. A CTE center admission offer raised pathway completion and consistent four-year enrollment, but had no significant effect on on-time graduation, and admitted students had lower GPAs. [Stein, Plasman, Durham & Smith-Onyewu 2023 (MLDS Center, CTE in Baltimore City; related paper Plasman et al. 2025, EEPA); quasi-experimental (regression discontinuity at CTE center admission cutoffs); Baltimore City Public Schools students]
• Philadelphia. Lottery winners for CTE high schools did better on graduation, credit accumulation and completing the Algebra 1, Algebra 2 and Geometry sequence, with mixed results on science, GPA and test scores. [Neild, Boccanfuso & Byrnes 2015 (Career and Technical Education Research 40(1)); quasi-experimental (lottery-based admission); Philadelphia CTE high school applicants, classes of 2003 to 2005]

Career academies

• MDRC randomized trial. Over eight years, more than 90% of both academy and control students earned a diploma or GED. WWC rates the completion effect “potentially positive” (96% vs. 94%, not statistically significant). About 1,400 students, roughly 85% Hispanic or African American. [Kemple with Willner 2008 (MDRC); RCT (applicants randomly assigned to academy or control); 9 urban high schools, 8-year follow-up]
• High-risk students. In the earlier follow-up, academies cut dropout among students at high risk of dropping out from 32% to 21%. For medium-risk students the effect depended on how strongly the academy was organized (see Part 4). [Kemple & Snipes 2000 (MDRC); RCT with subgroup analysis by baseline risk; same 9 high schools]
• A contemporary academy. Winning the lottery for a North Carolina information technology academy raised graduation by about 8 pp and sharply raised IT industry certification. Attendance and certification partly explain the graduation effect. [Hemelt, Lenard & Paeplow 2019 (CALDER; Economics of Education Review 68); quasi-experimental (lottery-based admission); one North Carolina career academy]

P-TECH 9-14

P-TECH pairs a high school with a community college and employer partners over six years. MDRC studied the first seven New York City schools using the city’s admissions lottery (3,161 students; 86% Black or Hispanic). WWC rates the study “Meets Evidence Standards without Reservations.”

[Rosen, Alterman, Treskon et al. 2023 (MDRC, P-TECH 9-14 Pathways to Success); quasi-experimental (lottery-based admission); 3,161 applicants to the first seven New York City P-TECH 9-14 schools] The seven-year results come from part of the sample (Arnold Ventures puts it at 16%) and, by Arnold Ventures’ reading, did not quite reach statistical significance. Secondary costs were about 17% higher per student, and cost-effectiveness was inconclusive. Part 1 called this a “lottery-based RD study”; the design is an admissions lottery, not a regression discontinuity.

Early college high schools

Early colleges have the strongest credential results of any model in this review. Most of the gain is in associate degrees.

• National study, first follow-up. Early college students graduated high school at 86% vs. 81%, enrolled in college at 80% vs. 71%, and had earned a college degree by one year after high school at 21% vs. 1%. Impacts on degrees were stronger for female, minority and lower-income students. 2,458 students, lotteries in 2005-06 to 2007-08. [Berger et al. 2013 (American Institutes for Research); quasi-experimental (lottery-based admission); 10 early college high schools with admission lotteries]
• Six years after high school. [Song & Zeiser 2019 (AIR, Early College, Continued Success); quasi-experimental (lottery-based admission); same 10 early colleges, 2,458 students]
• Ten years after high school. The overall degree gap was 22.5 pp in year 1 and narrowed to 8.5 pp by year 10 as the comparison group caught up. The bachelor’s effect was significantly larger for underrepresented minority students (12.2 pp vs. 2.1 pp). [Song, Neering, Zeiser, Schwarzhaupt & Mitchell 2024 (AERJ); quasi-experimental (lottery-based admission); 2,388 students in 10 early colleges in five states]
• North Carolina. High school graduation was 85.4% vs. 81.4% (not significant); any college within a year 89.9% vs. 74.3%. [Edmunds, Unlu, Furey, Glennie & Arshavsky 2020 (EEPA 42(2)); quasi-experimental (lottery-based admission); 1,687 North Carolina early college applicants, 2005 to 2011] A later follow-up found associate attainment 23.2 pp higher, a bachelor’s difference of +2.4 pp that was not significant overall, significant bachelor’s gains for economically disadvantaged and first-generation students, and time to degree shortened by about half a year. [Edmunds, Unlu, Phillips, Glennie & Mulhern 2024 (AERA Open); quasi-experimental (lottery-based admission); 4,073 students in 19 North Carolina early colleges]
• Cost. Early colleges cost about $3,819 more per student over four years and returned an estimated $57,682 in lifetime benefits per student (benefit-cost ratio 15.1, or 4.6 under conservative assumptions). This is a model, not a measured earnings result. [Atchison, Zeiser, Mohammed, Levin & Knight 2019 (AIR, The Costs and Benefits of Early College High Schools); descriptive/correlational (cost-benefit model built on lottery estimates); same 10 early colleges]

Dual enrollment

• WWC review. Five studies met standards (about 77,249 students). WWC rated the effects positive for college degree attainment, college access and enrollment, credit accumulation, high school completion and high school academic achievement. [What Works Clearinghouse 2017 (Dual Enrollment Programs intervention report); systematic review of 5 studies meeting WWC group-design standards; U.S. high school students]
• P-TECH and early colleges (above) show that dual enrollment inside a structured model moves associate degrees far more than high school graduation. Standard dual enrollment and college outcomes are in Part 3.

Industry-recognized credentials: value

The credential literature is almost entirely descriptive. Credentials are associated with better short-run earnings and two-year outcomes, with a fading wage premium and weaker four-year outcomes.

• Texas. About 5.9% of 2017-2019 graduates earned an industry-recognized credential (IRC). Earning one was associated with about 9% higher annual earnings (12.6% for students who did not go to college) and about 1 pp higher college enrollment. Returns varied widely by field. The author states the estimates “cannot be interpreted as causal” and may be biased upward. [Giani 2022 (Thomas B. Fordham Institute); descriptive/correlational (regression with school fixed effects plus matching); more than 1 million Texas graduates, 2017 to 2019]
• Ohio. Credential earners earned 21% more than non-earners the year after high school, falling to 5% by their mid-twenties. Manufacturing credentials showed the strongest return (37% at age 25). Men gained (23% by mid-twenties); women showed “no wage advantage at all.” Credential earners were about 9 pp less likely to enroll in college (40.6% vs. 49.3%). [Plasman 2026 (Thomas B. Fordham Institute, College or Career Readiness?); descriptive/correlational (fixed-effects regression with demographic and prior-achievement controls); about 1.3 million Ohio students, classes of 2016 to 2022]
• Florida. Certification earners had 2.9 pp higher community college enrollment and 4.5 pp higher associate attainment, 1.6 pp lower university enrollment, and a negative association with bachelor’s completion. “Gold Standard” certifications added nothing over regular certifications. The authors say unobserved aspirations limit causal inference. [Glennie, Ottem & Lauff 2020 (Journal of Career and Technical Education 35(1)); descriptive/correlational (propensity score matching with inverse probability weighting); 255,650 Florida first-time 9th graders, 2007-08]
• Florida, Indiana, Kentucky. Credentials were positively associated with on-time graduation in all three states and with higher wages at age 24 and older in Florida and Indiana. In Florida, only certifications (not software credentials or licenses) carried a wage gain, about 12%. [ExcelinEd & Burning Glass Technologies 2019 (Credentials Matter Report 2, Where Credentials Meet the Market); descriptive/correlational (regression with demographic and achievement controls); Florida, Indiana and Kentucky longitudinal data]
• North Carolina. The record 382,964 credentials in 2024-25 is a count, not an outcome. [NC DPI 2026 (press release, Feb 4, 2026); descriptive/correlational; NC administrative data] Confirm with NC DPI before citing.

Industry-recognized credentials: the credential-inflation evidence

The main warning about high school credentials is a supply-demand mismatch, not a finding that credentials harm students.

• National landscape. Across 24 states with usable data and over 780,000 credentials from 367 vendors, only 19% of credentials earned by K-12 students were demanded by employers in job postings. Ten of the 15 most-earned credentials were oversupplied. No state was highly aligned; 12 were moderately and 12 poorly aligned. [ExcelinEd & Burning Glass Technologies 2019 (Credentials Matter Report 1, A National Landscape of High School Student Credential Attainment Compared to Workforce Demand); descriptive/correlational (credential counts compared with more than 150 million online job postings); 24 states, 2018-19 data collection] The 19% figure is confirmed through Fordham’s summary of the report; the executive summary does not print it. The report also notes many credentials are required for jobs without being named in postings, which would understate demand.
• Ohio example. In 2017-18, Ohio students earned more than 49,000 credentials, fewer than 13,000 of them on the state’s in-demand list. [Poiner 2019 (Thomas B. Fordham Institute commentary on Credentials Matter); descriptive/correlational; Ohio 2017-18 credential counts]

What this means for you: a count of credentials is not a quality measure. The studies that find value find it for certifications tied to occupations with real local demand, and mostly for young men entering work directly.

Part 3: College access, enrollment and completion

CTE moves some students toward two-year colleges and does not raise four-year enrollment on average; whether it lowers four-year outcomes depends on the field, the design and the setting. Dual enrollment, by contrast, raises college outcomes, with the largest gains in structured models.

Does CTE divert students from college?

Tags for each row:

• [Ames 2021 (Journal of Career and Technical Education 36(1)); descriptive/correlational (logistic regression with background, GPA and curricular-intensity controls); HSLS:09, about 9,400 first-time 9th graders in public schools]
• [Dougherty 2016 (Thomas B. Fordham Institute); quasi-experimental (instrumental variables using within-school variation in course offerings, plus coarsened exact matching, alongside OLS with school and cohort fixed effects); three Arkansas cohorts entering 9th grade 2008 to 2010, more than 100,000 students]
• [Brodersen, Gagnon, Liu & Tedeschi 2021 (REL Central 2021-087); quasi-experimental (matched comparison); Nebraska and South Dakota high school students] The five-year results use 42,398 students from two cohorts. Read the bachelor’s figure as a net difference after controls, not a causal sentence to a lower credential.
• [Witzen 2019 (Maryland Longitudinal Data System Center); quasi-experimental (propensity score matching, with an instrumental-variables check using distance to CTE centers); Maryland public high school graduates of 2009-10 and 2010-11]
• [Brunner, Dougherty & Ross 2023 (Review of Economics and Statistics 105(4); NBER w28790, 2021); quasi-experimental (regression discontinuity at technical high school admission cutoffs); Connecticut Technical High School System applicants, P20 WIN administrative data]
• [Stein, Plasman, Durham & Smith-Onyewu 2023 (MLDS Center, CTE in Baltimore City); quasi-experimental (regression discontinuity at CTE center admission cutoffs); Baltimore City Public Schools students] The same study found higher immediate employment (b = 0.11).
• [Ecton & Dougherty 2023 (Educational Evaluation and Policy Analysis 45(1)); descriptive/correlational (regression on course-taking records linked to UI wages and National Student Clearinghouse data); state administrative data, state not confirmed] The working paper version (Ecton & Dougherty 2021, ERIC ED616805) identifies the state as Massachusetts (251,437 students, 2009 to 2011 cohorts).

Field decides the college pattern

The average hides opposite effects by field. In Massachusetts, health care and education concentrators were 15.1 and 13.0 pp more likely to attend college, while construction and transportation concentrators were 12.7 to 14.9 pp less likely (Ecton & Dougherty). Black and Latino concentrators showed an 8.9 pp increase in college attendance. In Nebraska and South Dakota, two-year enrollment ran 85 to 86% for finance, marketing, health science and government concentrators but 51% for transportation and 61% for manufacturing (Brodersen et al.).

When a parent asks whether a pathway affects college, the honest answer is that it depends on the pathway. Health, business and IT pathways look like college pathways. Trades, transportation and manufacturing pathways look like direct-to-work pathways, with higher early earnings (Part 1) and lower bachelor’s rates.

Career academies and pathway models

• MDRC Career Academies. No difference in educational attainment; about half of both groups completed a postsecondary credential within eight years. The earnings gains in Part 1 came without a college tradeoff. [Kemple with Willner 2008 (MDRC); RCT (applicants randomly assigned to academy or control); 9 urban high schools, 8-year follow-up]
• North Carolina IT academy. College enrollment within a year rose for males but not females. [Hemelt, Lenard & Paeplow 2019 (CALDER; Economics of Education Review 68); quasi-experimental (lottery-based admission); one North Carolina career academy]
• California Linked Learning. Students in certified pathways were 7.6 pp more likely to graduate and earned about 15.5 more credits, were as likely as similar peers to enroll in college, and when enrolled were equally or slightly more likely to choose a four-year college. The four-year effect was significant and positive for African American students and students with low prior achievement. [SRI International 2016, authors not confirmed (Taking Stock of the California Linked Learning District Initiative, seventh-year evaluation, for the James Irvine Foundation); descriptive/correlational (propensity score matching); classes of 2013 to 2015 in nine California districts]
• Massachusetts Innovation Career Pathways. Participants had 1.34 times the odds of enrolling in college within six months (about 5.6 to 6.7 pp) and 1.37 times the odds of persisting to a second year, with no significant differences by race or income. The authors describe their weighting as estimating causal effects; it adjusts for observed differences only. [Lucien, Lindorff & Strand 2024 (ERIC ED672362); descriptive/correlational (inverse probability of treatment weighting on propensity scores); 1,354 participants and 25,630 non-participants at 43 Massachusetts high schools, 2018-19 to 2021-22]
• P-TECH. No difference in college enrollment or degree at six years (60% vs. 60%), with a preliminary associate-degree gain at seven years (Part 2).

Selective admissions and college eligibility

Direct evidence on CTE and selective-college admission is thin. Three findings are usable.

• Baltimore selective CTE centers. An admission offer had no effect, in either direction, on enrollment at four-year, selective or more-selective colleges (Stein et al. 2023). The same offer lowered two-year and any-college enrollment and lowered high school GPA, which is the more important caution for college-bound students.
• Elite admission drivers sit outside CTE. At the same SAT or ACT score, applicants from top 1% families were 58% more likely to be admitted to Ivy-Plus colleges than middle-class applicants, driven by legacy preference, non-academic ratings and athletic recruitment. The appendix and July corpus gave this figure as 34%; the paper says 58%. [Chetty, Deming & Friedman 2023 (Opportunity Insights; NBER w31492); descriptive/correlational for the admissions gap (causal estimates of attending use waitlist variation); Ivy-Plus applicant and admissions data]
• Eligibility is a course-approval question. In states where public universities set a required course list, a CTE course protects eligibility only if the university system accepts it. California is the clearest case: CTE courses can be approved toward the UC “a-g” requirements, with more than 9,000 approved CTE courses by about 2010, and UC Curriculum Integration (UCCI) lists 76 integrated CTE-academic course models. [Achieve, about 2010 (Snapshot of University of California “a-g” Approval Process for CTE Courses); descriptive/correlational (program description); California] These counts are dated and Changing. Confirm current counts with UC High School Articulation. Last checked Oct. 3, 2026.

What this means for you: the risk to college eligibility comes from what a CTE course replaces, not from taking CTE. You protect it by keeping the full college-admission course core, or by getting the integrated course approved as meeting that core. See the Norway evidence in Part 4.

Dual enrollment and college outcomes

• WWC (2017) rated dual enrollment positive on degree attainment and college access (Part 2).
• Low-income students. Dual enrollment raised degree attainment more for low-income students than for affluent peers, and would “modestly reduce gaps.” WWC rates it as meeting standards with reservations. [An 2013 (EEPA 35(1)); descriptive/correlational (propensity score methods with sensitivity analysis); NELS:88, about 8,800 students]
• Maryland. Dual enrollment was associated with higher two- and four-year enrollment, degree attainment and earnings six years after high school, with stronger effects for Black students and students eligible for free and reduced-price meals. [Henneberger, Witzen & Preston 2022 (Emerging Adulthood); descriptive/correlational (propensity score matching); Maryland statewide administrative data]
• Texas. Dual credit “modestly improved” college enrollment and completion and shortened time to degree by about a semester, but effects were more positive for advantaged students and negative for low-income students with low prior test scores. [Miller, Kosiewicz, Tanenbaum, Atchison et al. 2018 (AIR for the Texas Higher Education Coordinating Board, Dual Credit Education Programs in Texas: Phase II); quasi-experimental (instrumental variables within a difference-in-differences framework); 16 cohorts of Texas high school juniors, 2001 to 2016]
• Tennessee randomized trial. A dual-credit math course, randomly assigned across 103 high schools (61,766 students), did not change overall college enrollment but shifted students from two-year to four-year colleges, most clearly for middle-achieving students. [Hemelt, Schwartz & Dynarski 2019 (EdWorkingPaper, Dual-Credit Courses and the Road to College); RCT (school-level random assignment); 103 Tennessee high schools]
• Meta-analysis. Across 162 effect sizes, dual enrollment was positively associated with GPA, credits, enrollment, persistence and degree attainment, and with shorter time to graduation. [Schaller, Routon, Partridge & Berry 2023 (Journal of College Student Retention 27(1)); meta-analysis of mostly non-experimental studies; U.S. dual enrollment research]
• CTE plus dual enrollment. In Texas, fewer than 5% of students combined a CTE focus with dual enrollment; those who did earned $9,746 per quarter at age 24 vs. $8,097 for CTE-only students. [CCRC 2025 (Promising Combinations of Dual Enrollment, AP/IB, and CTE); descriptive/correlational; Texas graduates of 2015-16 and 2016-17 followed six years]
• Who uses dual enrollment. Fifteen percent of fall 2010 community college entrants nationally were high school dual enrollment students; 46% of those who started at a community college after high school earned a credential within five years. [Fink et al. 2017 (CCRC, What Happens to Students Who Take Community College “Dual Enrollment” Courses in High School?); descriptive/correlational; more than 200,000 dual enrollment students, National Student Clearinghouse data]

Part 4: Policy moderators

Depth, structured support, demand alignment, academic integration and implementation fidelity each change results, but most moderator evidence comes from single studies, and the meta-analysis found no significant moderators at all, mainly because there are too few studies. Read each item for its design.

Concentration versus dabbling

Depth matters more than exposure, for completion and for wages.

• Upper-level vocational courses carry about a 2% wage premium per year; introductory courses carry none. Part 1 labeled this study quasi-experimental; it is descriptive/correlational. [Kreisman & Stange 2017 (NBER w23851; published in Education Finance and Policy 15(1)); descriptive/correlational (regression with rich controls for ability and background within a curriculum-choice framework); NLSY97 national sample]
• Concentrators, not course samplers, drive the earnings advantage in Massachusetts (Ecton & Dougherty 2023, Part 3).
• Later CTE is more protective than earlier CTE (Gottfried & Plasman 2018), and a middle-range CTE share is associated with the lowest dropout risk for on-age students (Plank, DeLuca & Estacion 2008).
• In Washington, almost every student takes some CTE but fewer than half concentrate, and underserved groups concentrate in a single pathway at 6 to 10 pp lower rates (Riggs & Cigarroa Kennedy 2025, Part 2).
• Credentials combined with coursework beat credentials alone: Ohio students who paired coursework with credentials earned 18% more by their mid-twenties vs. 4% for credential-only earners (Plasman 2026, Part 2).

Work-based learning quality

Offering work-based learning (WBL) is not enough; structure and support around it matter, and causal evidence on WBL alone is mixed.

• Urban Alliance internships. In the first evaluation, males offered the program were 12 pp more likely to enroll in college. [Theodos, Pergamit, Hanson, Edelstein, Daniels & Srini 2017 (Urban Institute); RCT; Urban Alliance applicants in Washington, D.C., and Baltimore] In a later trial with 1,435 seniors from 143 schools, there was no difference in college attendance, four-year attendance, persistence or employment at two years, and lower earnings for the treatment group in Baltimore wage records. [Theodos, Pergamit, Hanson, Teles et al. 2025 (Urban Institute); RCT; 1,435 Urban Alliance applicants in Baltimore, Chicago, Northern Virginia and Washington, D.C.]
• P-TECH. Internship participation rose 38 pp, yet high school graduation and six-year college outcomes did not change (Part 2).
• Linked Learning. WBL was underdeveloped throughout the initiative; pathway students did report more applied learning and better job quality (paid leave, health insurance) one year out (SRI International 2016, Part 3).
• Kentucky. CTE students with WBL outperformed other CTE students on enrollment, completion, credits and GPA, but only 0.14% of CTE students had a recorded WBL experience, so this is likely a selection story until shown otherwise. [Mulolli, Liu, Koehler & Dickinson 2025 (HumRRO for the Kentucky Department of Education); descriptive/correlational (propensity score matching and hierarchical models); 93,006 Kentucky graduates, 2021-22 to 2023-24]
• Registered apprenticeship, the most structured WBL, has no randomized or matched U.S. earnings study in the S4IS appendix; the authors of the main study call their findings “suggestive, but not conclusive.” [Reed, Liu, Kleinman, Mastri, Reed, Sattar & Ziegler 2012 (Mathematica Policy Research for U.S. DOL); descriptive/correlational (dosage model: earnings regressed on share of the program completed, controlling for pre-enrollment earnings, employment and demographics; no matched or randomized comparison group); RA registrants in 10 states (FL, GA, IA, KY, MD, MO, NJ, OH, PA, TX), UI wage records 2000-2010]

Regional demand alignment

Student choices only partly track local jobs, programs respond slowly, and returns are highest where training meets real demand.

• A 10 pp higher local share of jobs in a field corresponded to about a 3 pp higher share of students concentrating in it. When a field’s employment grew 10%, aligned concentrators rose only about 0.9 to 1.4% after two to three years, and only after growth in jobs needing college. [Carruthers, Dougherty, Goldring, Kreisman, Theobald, Urban & Villero 2024 (AERA Open); descriptive/correlational (student course records merged with BLS employment data); Massachusetts, Michigan, Montana, Tennessee and Washington]
• Only 19% of credentials earned by K-12 students matched employer demand (Credentials Matter Report 1, Part 2).
• Returns follow demand-heavy fields: manufacturing credentials in Ohio (37% at age 25) and trades and transportation in Part 1.
• Internationally, oversupplied narrow tracks scar early careers. Use as a guardrail only. [TREE panel studies (Switzerland), multiple authors; descriptive/correlational (20-year longitudinal panel); Swiss PISA 2000 cohort]

Credit integration

Integrating academics into CTE can raise academic results without costing technical skill, and keeping the academic path open protects long-run outcomes.

• Math-in-CTE. CTE teachers randomly assigned to co-design math-enhanced lessons with math teachers produced significantly higher student scores on TerraNova and ACCUPLACER after one year, with no loss on technical skill tests. The study covered 136 classrooms and about 1,600 students in five occupational areas, with 10 days of professional development. WWC rates it as meeting standards with reservations. [Stone, Alfeld & Pearson 2008 (AERJ 45(3)); RCT (teachers randomly assigned within occupational area); 136 CTE classrooms, five occupational areas]
• Meta-analysis. CTE had a small positive effect on academic achievement (g = 0.093) and college readiness (g = 0.077) (Lindsay et al. 2024).
• Norway. Adding general academics to vocational tracks and keeping a route to college captured the vocational entry benefit, raised earnings and lowered crime among disadvantaged men, and somewhat widened the gender earnings gap. [Bertrand, Mogstad & Mountjoy 2021 (Journal of Labor Economics); quasi-experimental (difference-in-discontinuities around the reform cohort cutoff); Norwegian population registers]
• Lifecycle tradeoff. Narrow vocational training’s employment advantage in youth reverses later in life, most strongly in apprenticeship countries. [Hanushek, Schwerdt, Woessmann & Zhang 2017 (Journal of Human Resources); quasi-experimental (difference-in-differences across age groups); IALS adult samples, mostly European countries]

Tracking risks for equity groups

CTE benefits many underserved students, but access, field sorting and diversion risks fall unevenly.

• Access is between schools. Differences in CTE credits by race, income and disability are generally larger across schools than within them. [Georgia Policy Labs, year not confirmed (Multi-State Study of Equity in CTE; Carruthers, Dougherty, Kreisman & Theobald); descriptive/correlational; Massachusetts, Tennessee, Washington and the Atlanta metro area]
• Gender. Graduation and earnings effects in Connecticut, the MDRC academies and Ohio credentials accrue mainly to young men; women sort into fields with smaller direct returns (Part 1). California’s career pathway grants cut dropout more for females than males (below), so the pattern is not universal.
• Who gets diverted. Texas dual credit helped advantaged students more and was negative for low-income, low-scoring students (Miller et al. 2018). Baltimore CTE center offers lowered GPA and two-year enrollment (Stein et al. 2023). Credential earners in Ohio and Florida were less likely to reach a bachelor’s degree.
• Who gains. Early college bachelor’s gains were concentrated among underrepresented minority, economically disadvantaged and first-generation students (Song et al. 2024; Edmunds et al. 2024). Dual enrollment helped low-income students more in national data (An 2013). Linked Learning’s four-year college effect was positive for African American and low-achieving students.
• Historical tracking. A policy brief documents vocational education’s history as a tracking mechanism and warns of “a tale of two CTEs,” with high-status STEM pathways and traditional pathways sorted by background. [Hodge, Dougherty & Burris 2020 (National Education Policy Center); descriptive/correlational (policy review); U.S. national data and literature]
• Who earns credentials. In Texas, Hispanic students earned IRCs at 6.9% and Black students at 3.4%, and the school a student attended explained most of the variation (Giani 2022).

Implementation fidelity

The same model works or fails depending on how fully it is built.

• Career academies. For medium-risk students, academies with a tight school-within-a-school structure and dedicated teachers reduced dropout, while loosely organized academies increased it. Personal support from teachers and peers was the strongest predictor of positive results. The comparison across academy types is non-experimental. [Kemple & Snipes 2000 (MDRC); RCT with subgroup analysis by baseline risk; same 9 high schools]
• Linked Learning. Certified pathways (meeting external quality standards) showed consistent graduation and credit effects; non-certified pathways showed almost none (SRI International 2016).
• California Career Pathways Trust. Districts that just won grants raised per-pupil CTE spending 21.7% and cut dropout by 23%, more for females than males. [Bonilla 2020 (Economics of Education Review 75); quasi-experimental (regression discontinuity at the grant-scoring threshold); California districts applying for Career Pathways Trust grants]
• Math-in-CTE worked with sustained teacher teaming and professional development, not with materials alone (Stone, Alfeld & Pearson 2008).

What the evidence does not show

Seven gaps limit what you can claim from Parts 2 to 4.

• No outcome study of any universal-participation model, including Winchester. All the positive evidence comes from applicants, lottery winners or opt-in students.
• No causal evidence that industry credentials themselves raise outcomes. Every credential study here is descriptive, and the wage premium fades.
• No reliable causal estimate for high school WBL on its own. Randomized internship results are mixed and partly null.
• No evidence that CTE raises four-year college enrollment on average. Some designs (Maryland, Baltimore, the credential studies) show modest movement away from four-year and bachelor’s outcomes.
• No direct study of CTE and admission to selective colleges beyond the Baltimore result.
• No significant moderators in the meta-analysis, so the moderator findings in Part 4 rest on single studies.
• Little post-2018 (Perkins V) cohort evidence. Most causal studies follow students who entered high school before 2015.

Review status and sources

Status: New Parts 2 to 4 written and verified Oct. 3, 2026, against primary reports, journal abstracts and official summaries. Research only; not legal advice. Last checked Oct. 3, 2026.

Corrections this review makes to Part 1 and the appendix

Open questions

Sources

• Lindsay et al. 2024, CTE Research Network systematic review (ERIC ED674303)
• AIR press release on the 2024 systematic review
• Brodersen et al. 2021, REL Central 2021-087 brief
• Dougherty 2016, Fordham Arkansas report (hosted copy)
• Riggs & Cigarroa Kennedy 2025, Washington ERDC CTE study
• Gottfried & Plasman 2018, AERJ abstract
• Plank, DeLuca & Estacion 2008, ERIC EJ889492
• Brunner, Dougherty & Ross 2023, REStat abstract
• Brunner, Dougherty & Ross, NBER w28790
• Dougherty 2018, ERIC EJ1175151
• Stein et al. 2023, CTE in Baltimore City (MLDS Center)
• Plasman, Stein, Durham & Smith-Onyewu 2025, EEPA
• Neild, Boccanfuso & Byrnes 2015, ERIC EJ1132478
• Neild & Byrnes 2014, postsecondary outcomes of Philadelphia CTE schools
• Kemple with Willner 2008, MDRC Career Academies long-term impacts
• WWC Career Academies intervention report 2015
• Kemple & Snipes 2000, MDRC (ERIC ED441075)
• Hemelt, Lenard & Paeplow 2019, Economics of Education Review
• Rosen et al. 2023, MDRC P-TECH 9-14 final report
• Arnold Ventures review of P-TECH
• Berger et al. 2013, AIR press release
• Song & Zeiser 2019, Early College, Continued Success (ERIC ED602451)
• Song et al. 2024, AERJ (ERIC EJ1452418)
• Edmunds et al. 2020, WWC study review
• Edmunds et al. 2024, AERA Open
• Atchison et al. 2019, Costs and Benefits of Early College High Schools
• WWC Dual Enrollment Programs 2017 (ERIC ED572842)
• Giani 2022, Fordham industry-recognized credentials in Texas
• Giani 2022 full report (ERIC ED625598)
• Plasman 2026, Fordham Ohio credentials report
• Fordham commentary on Plasman 2026
• Glennie, Ottem & Lauff 2020 (ERIC EJ1310506)
• Credentials Matter Report 1 executive summary
• Credentials Matter Report 2, Where Credentials Meet the Market
• Fordham summary of Credentials Matter (Poiner, May 15, 2019)
• Fordham Ohio credential commentary (Poiner, June 6, 2019)
• ExcelinEd coverage of Credentials Matter
• NC DPI credential press release, Feb. 4, 2026
• ACTE CTE Works fact sheet 2024
• Ames 2021 (NSF Public Access copy)
• Witzen 2019, Maryland CTE report
• Ecton & Dougherty 2021 working paper (ERIC ED616805)
• SRI International 2016, Linked Learning seventh-year report (ERIC ED664229)
• Lucien, Lindorff & Strand 2024 (ERIC ED672362)
• Chetty, Deming & Friedman 2023 paper
• Chetty, Deming & Friedman, NBER w31492
• Achieve, UC a-g approval for CTE courses
• UC Curriculum Integration (UCCI)
• An 2013, ERIC EJ1009522
• Henneberger, Witzen & Preston 2022, Emerging Adulthood
• Miller et al. 2018, Dual Credit in Texas Phase II
• Hemelt, Schwartz & Dynarski 2019 (ERIC ED670873)
• Schaller et al. 2023 meta-analysis
• CCRC 2025, Promising Combinations (press release)
• Fink et al. 2017, CCRC (ERIC ED578185)
• Kreisman & Stange, NBER w23851
• Theodos et al. 2017, Urban Alliance evaluation
• Theodos et al. 2025, Urban Alliance long-term evaluation
• Mulolli et al. 2025, HumRRO for Kentucky Department of Education
• Reed et al. 2012, Mathematica for DOL
• Carruthers et al. 2024, AERA Open
• Stone, Alfeld & Pearson 2008, ERIC EJ807311
• Stone et al. 2008, NRCCTE report
• Bertrand, Mogstad & Mountjoy, NBER w25679
• Hanushek et al. 2017, Journal of Human Resources
• Georgia Policy Labs, Multi-State Study of Equity in CTE
• Hodge, Dougherty & Burris 2020, NEPC brief
• Bonilla 2020, Economics of Education Review