Read a record as a claim with a creator, purpose and scope. A count, interpretation and value judgment are different kinds of evidence. Percent means per hundred; percentage points compare two percentages.
operational definition
An explicit rule for observing a concept
teaching record
An authored fictional example designed to make reasoning inspectable; not evidence that an event occurred.
01 · TRY
What is the difference in gains?
Compute both gains and draw the counterfactual claim. Identify one pre-program observation that would test its plausibility.
Use this as optional interdisciplinary evidence practice; there is no standalone AP Social research methods course. Compare two explanations using the supplied records and state a limitation.
Make a provisional claim
Loading saved evidence…
Inspect evidence and test an explanation
The investigation controls appear when your local evidence is ready.
04 · UNDERSTAND
Connect evidence and mechanism
Separate a comparison from a causal claim
A causal question asks what would change under a different action, holding relevant conditions comparable. Random assignment helps balance confounders on average; observational designs need explicit assumptions. Before-after change alone can reflect other events.
Inspect the case
Teaching program group rises from 50 to 65 points. A comparison group rises from 40 to 50 during the same period.
Teaching notes say both groups faced a common curriculum change, but the program group volunteered and had different prior trends.
Both records are original fictional teaching materials. They are not historical primary sources and do not report real participants.
Work through the mechanism
Program gain is 15; comparison gain is 10; the difference in gains is 5. Calling that a causal effect requires assumptions about what would have happened without the program.
Your investigation
Compute both gains and draw the counterfactual claim. Identify one pre-program observation that would test its plausibility.
Make a two-column evidence table: what the record directly supports, and what you infer. Use record IDs when connecting observations to claims. Preserve your first interpretation before revising it.
Practice and transfer
What is the difference in gains?
Try the practice check in the activity, explain your choice, then apply the idea to this unfamiliar case: Earlier records show the program group was already improving faster. What happens?
Limits and responsible interpretation
A difference-in-differences calculation does not prove its own assumptions. No observational checklist removes every confounder.
A successful answer identifies a mechanism and its boundaries. A saved response records practice; it does not automatically establish independent understanding.
Further reading
OpenStax: approaches to research provides disciplinary context. The teaching records above were authored for this lesson and are not quotations from that source.
Choose your depth
High school: Use the supplied records to complete this task: Compute both gains and draw the counterfactual claim. Identify one pre-program observation that would test its plausibility. Explain your reasoning in a short paragraph or an annotated diagram with a text equivalent.
AP connection: Use this as optional interdisciplinary evidence practice; there is no standalone AP Social research methods course. Compare two explanations using the supplied records and state a limitation.
Introductory college: Frame a researchable version of the case, document definitions and source selection, and compare the supplied mechanism with a rival explanation. Use the linked disciplinary reading to find one relevant scholarly or institutional source; distinguish that real source from these fictional records. Submit a revision log and methods note.
Test the missing counterfactual, not just the subtraction
These original fictional records extend the earlier program/comparison example. Each number is a site-group mean on a fixed 0–100 point rubric, measured at equal intervals. They are aggregate teaching records, not real learners or a matched-person panel. Three periods precede the program, which starts between 0 and +1. No sample sizes or uncertainty estimates are supplied.
Before calculating: state what must be true for the comparison group's change to represent the program group's change without the program. Specify a record that could contradict your claim. You may preserve one draft for comparison with your revision.
No draft captured. Capturing does not prove the work was unseen or independently assessed.
Context record: No concurrent intervention is recorded. Absence of a record is not evidence that none occurred.
Fixed time positions −2, −1, 0, +1 and score 0–80 on the plot; the rubric's full range is 0–100. Solid navy program, dashed red comparison. Dotted gold segment is an assumed counterfactual, not an observation.
Period
Program mean
Comparison mean
Program minus comparison
-2
40
30
10
-1
45
35
10
0
50
40
10
+1
65
50
15
From 0 to +1: program gain=15; comparison gain=10; difference in gains=5 points. Earlier gap changes (“placebo” differences before the actual program) are 0 and 0 points. These are arithmetic comparisons, not statistical tests; there are no standard errors or p-values.
Make the identifying assumption visible
δ says how much more the program group would have gained without the program than the comparison group. It is unobserved. Counterfactual post mean=50+10+δ=60. Observed 65 minus this assumed value gives a sensitivity residual=5 points. δ=0 encodes equal untreated changes; it is not verified by clicking an option. δ=5 makes the residual zero; δ=10 reverses its sign. None is a confidence interval or measured causal effect.
Similar pre-period changes do not prove equal untreated changes after the program. Different earlier changes challenge a simple parallel-trends story, but extrapolating them exactly is also a substantive assumption. Changing composition, measurement, anticipation, spillovers into the comparison group and different concurrent events can undermine the interpretation.
In the concurrent case, the arithmetic and pretrends look the same as the similar-pretrend case. Explain why these means cannot separate the original program from the equipment change. Propose a comparison or additional timing record that could help, including why it could still fail. Do not subtract an invented equipment effect.
Reasoning guide
All three cases have post difference-in-differences 5. Similar and concurrent cases have pre-placebos 0/0; divergent pretrends have 5/5. Assuming the prior five-point differential continues implies δ=5 and residual 0, but the observations alone do not establish that continuation. A comparison with no second intervention, staggered event timing or independent documentation might sharpen the question; none automatically fixes selection or measurement.
The concurrent case cannot identify the named program's effect separately from the second change with these records. Parallel pretrends are evidence to consider, not a causal certificate. Report the observed differences, the assumed counterfactual and the remaining alternatives separately.
Method source and original-record scope
David McKenzie, World Bank, 3 February 2020: sensitivity to parallel-trend violations explains why exact continuation of earlier trend differences is a strong assumption. This activity uses original small panels, not that article's empirical examples or an implementation of its cited estimation methods. There are no sampling intervals, modern staggered-adoption estimators or authentic microdata here.
Copy the design-audit report
Draft, one captured draft and revision are temporary, each limited to 3,000 characters. Reset clears them and restores similar/δ=0, retaining known guide/report exposure this mount; reload clears exposure. No independent mastery is awarded.
Worked example
Program gain is 15; comparison gain is 10; the difference in gains is 5. Calling that a causal effect requires assumptions about what would have happened without the program.
