For teachers and instructors
The slide deck is meant to be cut apart and adapted — take the exhibits you need, drop the rest, change the framing to suit your course. The LaTeX source is provided for exactly that reason.
Download the deck
The deck and its LaTeX source are published alongside the paper on the corresponding author's site. Sixteen frames covering the puzzle, the design, the three main figures and the national benchmark, what the evidence does and does not support, and discussion questions — each block marked so it can be dropped on its own.
Discussion questions
The deck's own discussion questions will be listed here verbatim once the deck is posted. The prompts below are drawn only from claims stated on this site, and are marked as such rather than attributed to the deck.
- 01
The raw gap is −0.32σ and the reweighted net gap is −0.16σ. What changed between those two numbers, and why is reweighting to a common-support ability distribution necessary before the difference can be interpreted at all?
- 02
Ability contributes +0.25σ and inexperience −0.30σ. What would have to be true about the sample for those two to be separately identified, and what does the additive form assume away?
- 03
The route remainder is −0.10σ with a 95% interval of [−0.37, +0.17]. Write down two substantively different conclusions that are both consistent with that interval.
- 04
Enseña Chile recruits sit at the 97th percentile of ability but the 42nd percentile of model-predicted value-added. Why does the second number carry no residual dispersion, and how does that change how you read the percentile?
- 05
The comparison group was chosen to overlap on ability rather than to represent the country. Which questions can this sample answer, and which questions has that choice put out of reach?
- 06
Why does reporting unshrunk value-added make the reported gradients conservative rather than overstated?