Investigate the institution
Examine an actual class-size experiment and audit what happens when assignment, attendance and missing outcomes diverge. Does assigning a resource identify how that resource works?
Education can distribute skills, credentials, peer contact and institutional recognition. Class size is one resource dimension; it is not the entire institution.
Random assignment creates a comparison strategy. Noncompliance and attrition can change what an analysis estimates, so preserve assignment and actual exposure separately.
An achievement effect is consistent with several mechanisms, including time for feedback or classroom interaction. A design that changes class size does not automatically isolate each mechanism.
Distinguish authentic evidence from the practice ledger
Project STAR analysis
The Tennessee experiment assigned pupils and teachers to small classes, regular classes or regular classes with an aide. Krueger reports about a four-percentile-point initial achievement advantage from small-class assignment, while examining attrition and movement between class types.
Primary research summary. The study began in 1985–86. A percentile rank is not the percentage of test items answered correctly. Implementation departures require analysis; assignment alone does not guarantee complete follow-up.
Alan B. Krueger — NBER working paper 6051, June 1997; QJE publication 1999. This is an attributed summary, not reproduced article text.
Later outcomes require later measures
A follow-up examines later test performance and whether participants took college entrance examinations after the classroom intervention ended.
This is follow-up evidence from the original experiment, not a fresh randomized trial. Test taking is an access/participation outcome and is different from admission or completion.
Alan Krueger and Diane Whitmore — NBER working paper 7656, April 2000. This is an attributed summary, not reproduced article text.
Invented assignment and outcome ledger
A fictional school assigns 100 pupils to each of two classroom formats. In each group, 80 end-of-year tests are observed. Among observed tests, 48 in A and 40 in B pass. The other 20 outcomes in each group are missing.
A separate exercise, not STAR microdata. You cannot silently count a missing test as a failure or assume missing pupils resemble observed pupils.
Inspect the fictional ledger
Fictional practice data, not the study results. Scroll horizontally on narrow screens.| Fictional assigned group | Assigned pupils | Observed tests | Known passes | Missing outcomes |
|---|
| A | 100 | 80 | 48 | 20 |
|---|
| B | 100 | 80 | 40 | 20 |
|---|
Draw two 100-cell grids. Use labeled patterns for known passes, known non-passes and missing outcomes, not color alone. Fill every missing cell once as a pass and once as a non-pass. Does the ordering of the groups survive both scenarios? Explain why comparing only observed pupils asks a different question.
Work the comparison
For a separate invented group with 10 assigned pupils, 8 observed tests and 6 observed passes, the observed pass rate is 6/8=75%. Across all assigned pupils the pass fraction lies between 6/10=60% and 8/10=80%, depending on the two missing outcomes. Those are logical bounds, not a confidence interval.
Make an inspectable artifact
Draw assignment → actual classroom experience → measured outcome, with a separate branch for missing observations. Calculate observed pass rates and worst/best missing-outcome bounds in the invented ledger. Explain what an institutional resource effect does and does not identify.
Record your diagram or table as a written trace in the claim field. Number each link, identify the relevant source ID and distinguish measured evidence from a proposed mechanism. Save a capture before revising your interpretation.
Test a stronger claim
Connect a sociological mechanism to evidence it would require. Distinguish the reported STAR estimate, fictional missing-outcome bounds, original assignment and actual classroom exposure. State a defensible limit on policy transfer.
Independent extension
Read the original methods discussion of reassignment and attrition. Propose assignment-based and exposure-based analyses, identify the added assumptions of each, and discuss classroom clustering, staffing costs and scale-up constraints.
Use the supplied records and public research descriptions. Do not conduct deceptive hiring or housing applications, collect classmates’ sensitive information, or treat the fictional ledger as a real effect estimate.
