Online only courses: What the Research Says
Purpose
The research does not say online instruction fails. It says that in community colleges, fully online course-taking has lowered course success for two decades, most for less-prepared students, and that a mix of online and in-person courses does better than either extreme.
These notes collect the studies behind that summary, including the ones that point the other way.
Methodology
The studies below came from two sources: targeted searches for well-established studies, and a systematic ERIC search to check and extend them.
Core studies. 18 studies selected from the established literature, located through ERIC and journal and publisher pages, and checked against publisher and index listings. Summaries are based on abstracts; two full texts (Bernard et al., 2004; Wladis et al., 2025) were read.
ERIC search. Run November 4, 2026, through the ERIC API. Criteria: peer-reviewed, published 2018鈥2026, postsecondary students, comparing online with face-to-face delivery on an objective outcome (grades, exam scores, pass or completion, withdrawal, retention, or graduation).
| Screening stage | Records |
|---|---|
| Retrieved from ERIC | 808 |
| Passed automated relevance check | 580 |
| Passed title screen | 171 |
| Included after abstract screen | 127 |
Of the 127 included studies, 5 are randomized, 19 quasi-experimental, and 103 descriptive comparisons. All nine cited peer-reviewed studies that ERIC indexes for 2018 onward were retrieved.
| Modality coverage of the 127 (categories overlap) | Studies |
|---|---|
| Planned online courses | 70 |
| Pandemic emergency remote instruction | 57 |
| Includes a hybrid, blended, or HyFlex group | 26 |
| Includes synchronous (live video) online instruction | 24 |
| Set in two-year colleges | 12 |
Modality counts are approximate, coded from abstract keywords. One reviewer screened titles and abstracts. Comparisons of hybrid with face-to-face only were out of scope; a separate study will cover all three formats.
Trends that hold up
Five findings recur across states, decades and research designs.
| Trend | Evidence | Published |
|---|---|---|
| Average differences are near zero, but the spread is wide. Mode alone predicts little; design and pedagogy predict more. | Bernard et al. (2004): 232 studies, effect sizes 鈥渆ssentially zero鈥 with wide variability; pedagogy explained more variation than media. Means et al. (2010): online modestly better overall, driven by blended conditions that also added time and instruction. | 2004, 2010 |
| In community colleges, course success is lower in fully online sections. | Xu & Jaggars (2014), Washington. Johnson & Mejia (2014), all 112 California colleges: 60.4% online vs 70.6% in-person success in 2012; 11鈥14 points lower after controls. Hart, Friedmann & Hill (2018), California. Xu, Ran & Zhou (2023): online developmental courses cut completion by 13 points. Worsham, Swanson & Avery (2025), Los Angeles: slightly lower GPA and moderately lower persistence. | 2014鈥2025 |
| The gap is largest for less-prepared and underrepresented students. | Xu & Jaggars (2014). Johnson & Mejia (2014): the white鈥揃lack success gap widened from 12.9 to 17.5 points online. Bettinger et al. (2017). Kofoed et al. (2024). Hart et al. (2026): trends most pronounced for Hispanic and Black students. | 2014鈥2026 |
| The gap has narrowed since the pandemic, and is smaller for synchronous online. | Hart, Baker, Hill, Alonso & Xu (2026), California community colleges 2015鈥16 to 2021鈥22: gaps remain but are smaller than before 2020; by 2021鈥22, passing gaps were smaller in synchronous than asynchronous sections. | 2026 |
| Some online coursework helps completion; a heavy online load does not. | Johnson & Mejia (2014): students taking some online courses were more likely to earn a degree, certificate or transfer. Shea & Bidjerano (2018), 30 SUNY community colleges, n = 45,557: students taking more than 40% of courses online 鈥渂egin to lose the benefits.鈥 Worsham et al. (2025): negative effects driven by dosage. | 2014鈥2025 |
Online formats hurt “least prepared and at-risk” students most
All five studies find lower outcomes online, and four of them report the largest losses among lower-achieving or at-risk students.
| Study | Setting and design | Key finding |
|---|---|---|
| Xu & Jaggars (2014), , Journal of Higher Education 85(5) 路 | ~500,000 course enrollments, 40,000+ Washington State community and technical college students; observational with controls | Every student group did worse online. Gaps were largest for male, younger, and Black students and those with lower GPAs; strongest in social sciences and applied professions. |
| Bettinger, Fox, Loeb & Taylor (2017), , American Economic Review 107(9) 路 | 230,000+ students, large for-profit university; instrumental-variables design | Taking a course online lowered the course grade by 0.44 grade points (about 0.33 SD), lowered grades in later courses, and made students less likely to stay enrolled. Effects were larger for students with lower prior GPAs. |
| Figlio, Rush & Yin (2013), , Journal of Labor Economics 31(4) 路 | Randomized: live vs. online-video lectures in intro microeconomics at a large research university | Modest advantage for live lectures, largest for Hispanic, male, and lower-achieving students. |
| Alpert, Couch & Harmon (2016), , AER Papers & Proceedings 106(5) | Randomized into face-to-face, blended, or fully online microeconomics principles | Online scored about 5鈥10 points lower on the final exam; blended was not significantly different from face-to-face. |
| Kofoed, Gebhart, Gilmore & Moschitto (2024), , AER: Insights 6(3); first released as IZA DP 14356 (2021) | Randomized at West Point, fall 2020, ~550 students in intro economics | Online lowered final grades by 0.215 SD, most for academically at-risk students. Online students reported weaker concentration and less connection to instructors and peers. |
Publisher links (DOIs) are mostly paywalled; free copies are linked where available.
More recent studies (2018 onward)
Newer work mostly confirms lower course completion online. Two studies question whether online delivery itself causes it: Wladis et al. attribute the gap to student self-selection, and Fischer et al. find faster graduation. Kofoed et al. (2024, above) is also in this group.
| Study | Setting and design | Key finding |
|---|---|---|
| Wladis, Hachey, Manly & Conway (2025), , Online Learning 29(3) 路 | All CUNY two- and four-year college courses, fall 2019 and spring 2020; within-student fixed effects, using the forced spring 2020 shift to separate chosen mode from actual mode | Fully online courses had lower completion, but the gap followed the mode students originally chose, not how the course was taught. Students also put electives online more often than major requirements, and electives have lower completion. |
| Bird, Castleman & Lohner (2022), , AERA Open 8 路 | Virginia community college system, spring 2020 shift to online; two estimation approaches | Course completion fell about 3鈥6%, more for lower-performing and less experienced students. Instructors’ prior online teaching experience did not offset the drop. |
| Fischer, Baker, Li, Orona & Warschauer (2022), , Educational Evaluation and Policy Analysis 44(3) | 10,572 students in 13 majors at a public research university; online course offerings as an instrumental variable | Taking major-required courses online was linked to a higher chance of four-year graduation and slightly shorter time-to-degree. The benefit was smaller for lower-SES students. |
| Altindag, Filiz & Tekin (2021), , NBER Working Paper 29113 | Pre- and post-pandemic course records comparing instruction formats | In-person instruction was associated with higher grades and lower withdrawal rates. |
| Hart, Friedmann & Hill (2018), , Education Finance and Policy 13(1) | ~217,000 first-time entrants, 2008鈥12; student and instructor fixed effects | Online students were less likely to complete, pass, or earn an A or B. Gaps were largest in summer and intersession terms, and controlling for instructors shrank them without eliminating them. |
Where the evidence is mixed
The direction above is consistent; the size and the cause are less settled.
- Selection or cause? Most large studies are observational. Wladis et al. (2025) used the spring 2020 forced shift at CUNY and found the completion gap followed the mode students had chosen, not how the course was taught. Some of the gap may be who enrolls online, not online itself.
- Different institutions, different results. At a public research university, Fischer et al. (2022) linked online major courses to a higher chance of four-year graduation, with a smaller benefit for lower-SES students. Community college results are less favorable.
- Blended does at least as well as in-person. Alpert, Couch & Harmon (2016) found blended sections no different from face-to-face, while fully online scored lower. Means et al. (2010) caution that blended conditions often added time and instruction, so the medium alone gets no credit.
- The randomized trials are narrow. Figlio et al. (2013), Alpert et al. (2016) and Kofoed et al. (2024) all study introductory economics at selective or research institutions.
- Older course designs. Much of the evidence predates 2020. Hart et al. (2026) show the gap shrinking since, so pre-pandemic numbers likely overstate today鈥檚 gap.
- Students鈥 own view. In Jaggars (2014), community college students said they had to 鈥渢each themselves鈥 online and preferred 鈥渄ifficult鈥 or 鈥渋mportant鈥 subjects face-to-face. The author鈥檚 conclusion: build more instructor presence online, or keep ample face-to-face sections.
About this note
Prepared by Matthew MacNaughton with the help of an AI research assistant (Claude). I asked it to survey the research on how fully online courses compare with in-person and blended courses in college, to identify trends that hold up across studies and over time, and to include evidence that cuts the other way. I told it not to cite anything it could not verify. Sources were located through ERIC and journal and publisher pages. Findings are summarized mostly from abstracts; two studies (Bernard et al., 2004; Wladis et al., 2025) were read in full. Please consult the original sources before relying on any single figure.
Full APA references: Online Learning Bibliography
漏 2026 Matthew MacNaughton. This work is licensed under a .