ESL placement tests misplace 30% of English learners into wrong proficiency levels, wasting semesters of instruction

English as a Second Language (ESL) placement tests at U.S. colleges and adult education programs — most commonly CASAS, TABE, and Accuplacer — assess reading and grammar but poorly measure speaking and listening proficiency, resulting in roughly 30% of students being placed into levels that are too high or too low for their actual communicative ability. A student who reads well but speaks poorly is placed into an advanced class they cannot participate in; a student who speaks fluently but reads at a basic level is placed into a beginner class that wastes their time. So what? A misplaced student spends an entire semester (12-16 weeks) in a class that does not match their needs, neither improving their weakest skills nor being challenged appropriately. So what? In adult ESL programs where funding is tied to level gains (as measured by the National Reporting System), students who do not show measurable improvement because they were misplaced cause the program to lose federal funding, creating a perverse cycle where bad placement leads to worse program resources. So what? For immigrant workers, each wasted semester in the wrong ESL level delays their ability to pass citizenship tests, obtain professional licenses that require English proficiency, or qualify for jobs that require English fluency — directly extending their period of underemployment. So what? Underemployment of skilled immigrants represents a massive economic loss: immigrants with professional credentials from their home countries (doctors, engineers, accountants) are stuck in low-wage work while their skills atrophy. So what? Their children, observing the futility, are less likely to invest in their own education, creating intergenerational effects from a single bad placement test. It persists because ESL placement tests are optimized for cost and scalability (multiple-choice, machine-gradable) rather than accuracy, and assessing speaking/listening at scale requires trained human raters, which programs cannot afford.