On this page
In 1985, Tennessee tried something rare in education: a large randomised experiment on class size. Children and teachers were assigned to classes by lottery. Economist Alan Krueger re-analysed the data in the Quarterly Journal of Economics in 1999, and his paper remains a standard reference in debates about smaller classes.
What was the Tennessee STAR experiment?
STAR stands for Student/Teacher Achievement Ratio. According to Krueger's working paper version (NBER Working Paper 6051), the design had three arms, starting in the 1985-86 school year:
Small classes of 13 to 17 students per teacher.
Regular classes of 22 to 25 students.
Regular classes of 22 to 25 students with a full-time teacher's aide.
Students were meant to stay in the same class type for four years, from kindergarten to third grade (roughly ages 5 to 9). Over those four years the sample included 11,600 students from 80 schools, and random assignment happened within each school, Krueger reports. The Tennessee legislature paid about $12 million over the four years.
The first analyses came from the state's own research team. Covering kindergarten and first grade, Finn and Achilles (1990) reported that students in reduced-size classes did better in both reading and mathematics, with evidence that minority students benefited in particular.
What did Krueger's 1999 study find?
Krueger's main findings, from the abstract of the 1997 NBER working paper version:
Test scores rose by about 4 percentile points in a student's first year in a small class, whatever the grade.
After that first year, small-class students gained only about 1 more percentile point per year over regular classes.
Teacher aides had little effect on achievement.
Smaller classes helped minority students and students eligible for free lunch more.
The gains did not look like a Hawthorne effect, where people improve simply because they know they are being studied.
His conclusion, in the same paper, was that adjusting for attrition, re-assignment and drifting class sizes did not overturn the main result: small-class students scored higher on standardised tests.
What are the limits and criticisms of the STAR study?
The experiment was not clean, and Krueger listed the problems himself. Regular-class students were re-randomised between aide and no-aide classes after kindergarten. About 10 percent of students switched between small and regular classes between grades, mainly because of behaviour problems or parent complaints. Actual class sizes ranged from 11 to 20 in small classes and 15 to 30 in regular ones. Half of the kindergarten students were missing in at least one later year, the paper notes.
Economist Eric Hanushek made a prominent critique in Educational Evaluation and Policy Analysis (1999). He argued that design and implementation problems leave considerable uncertainty about the size of the effect, and that the commonly cited results are likely biased upwards. Even taken at face value, he wrote, STAR shows effects only for very large, expensive cuts in kindergarten or possibly first grade, and offers no support for smaller cuts that leave classes above 13 to 17 students or for reductions in later grades.
Later follow-ups added a twist. A 2011 paper by Chetty and colleagues in the Quarterly Journal of Economics, which linked 11,571 STAR students to adult administrative records, found that those in small classes were significantly more likely to attend college, but found no significant effect of class size on earnings at age 27, an estimate the authors call imprecise.
What can Indian students and families take from STAR?
Nothing in STAR was tested in India, in Class 9 to 12, or in exam coaching. What follows is our inference, not a finding.
Size is a proxy, not the goal. STAR did not test why small classes helped; Finn and Achilles said the question remained largely unanswered. More teacher attention per child is one plausible reason. A big batch with real doubt-clearing may beat a small one.
Early years matter most. STAR covered only kindergarten to third grade, and Hanushek finds effects only for cuts in kindergarten or possibly first grade. If you are choosing a primary school, class size is a fair question to ask.
Do not pay a premium for "small batch" alone in Class 11-12. STAR gives no evidence for later grades.
Teaching at the child's level has been tested in India. The Indian trials in our explainer on Teaching at the Right Level and the Kenyan ability-tracking trial are closer to the question most families face.
If the one-to-one attention is what you are missing, Fermi's AI tutor works out what a student already knows and builds a plan for the session, then checks understanding as it goes.



