The Evidence on Teaching

Failure rates in introductory programming revisited

Watson, C., & Li, F. W. B. · 2014

grade Creviewindependentreplicated
Sample
161 CS1 courses across 51 institutions in 15 countries
Population
Introductory programming courses worldwide.
Design
A systematic review harvesting reported pass rates from published articles, so the sample is courses whose instructors chose to publish about them — a selected sample. Descriptive; no causal claim. Read at abstract depth.
Key findings
The worldwide mean CS1 pass rate is 67.7%, almost identical to earlier estimates. Pass rates differed by grade level, country and class size, but NOT over time and NOT by programming language taught. The premise that CS1 is uniquely brutal — which motivates most instructional-innovation papers in the field — is only weakly true.
Genetic confound
Not applicable — descriptive base rates, no treatment.
Replication notes
Directly replicates and extends Bennedsen & Caspersen (2007); Bennedsen & Caspersen (2019) then re-replicated with a larger sample and obtained a 28% failure rate, consistent within noise.
DOI / URL
10.1145/2591708.2591749

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Worldwide CS1 pass ratemean proportion67.7% across 161 coursesadministrativeend of coursenonenot-applicableattainment
Change in pass rate over time and variation by language taughtmoderator testno significant difference on eitheradministrativeacross cohorts from the 1970s to the 2010snonenot-applicableattainment

Cited by