demog-theory-building

Theory & Argument Building (demog-theory-building)

At Demography a result is not a contribution until it is attached to a **claim population science can

use** — a mechanism that explains a demographic process, a sharper estimate that revises the record,

or a formal result that unifies or clarifies. This skill turns findings into argument: explicit

mechanisms, scope conditions, and observable implications, in the idiom appropriate to your kind of

work.

When to trigger

• The empirics are strong but the "so what / why" is thin
• A reviewer said the paper is "descriptive," "atheoretical," or "just a correlation"
• You need to state mechanisms, identifying assumptions, or scope conditions explicitly
• Formal demography: deciding what to model and what the model buys you

Build the argument (by mode of work)

Explanatory population study

1. Concept & measure — define the demographic construct (e.g., parity progression, lifespan

inequality, net migration) precisely; distinguish it from neighbors and from its measure.

1. Mechanism — the population story: which behaviors, exposures, or compositional shifts move the

rate, for whom, and why (incentives, constraints, selection, cohort experience).

1. Observable implications — what we should see if the mechanism operates (age pattern, cohort

signature, subgroup contrast) and what we should not see. These become the tests in

demog-research-design.

1. Scope conditions — which populations, periods, and regimes the argument covers.

Formal / mathematical demography

• State the substantive population puzzle the model addresses before the setup.
• Keep assumptions (stability, stationarity, Markov transitions, independence) **transparent and

motivated**; flag which results depend on which assumptions.

• Translate results into interpretable demographic quantities (e.g., contributions to life

expectancy, sensitivities/elasticities, equilibrium structure) a reader can recognize.

• Say what the model buys: a non-obvious decomposition, a unifying identity, a corrected intuition.

Measurement / decomposition contribution

• Make explicit what the new measure or decomposition separates that prior work conflated (e.g.,

tempo vs. quantum, composition vs. rate, age vs. cohort).

• Show the substantive payoff: the trend now attributes to a different component than was assumed.

The "portability" test (Demography-specific)

Ask: *Could a demographer studying a different component or population import this mechanism, measure,

or decomposition?* If yes, you have a population-science contribution. If it only works for your exact

case, generalize the logic or reframe (back to demog-topic-selection).

Anti-patterns

• "Hypothesizing after results are known" — state the argument before the tests
• A formal model with opaque assumptions chosen to produce the desired identity
• Mechanisms named but never made observable in age/cohort/subgroup patterns
• Treating a regression coefficient as a mechanism with no demographic story
• Universal claims with no scope conditions on population, period, or regime

Worked micro-example: from finding to population-science claim (illustrative)

A hypothetical study observes that completed cohort fertility fell across successive birth cohorts. The

argument is built in the idiom Demography — the Population Association of America flagship at Duke

University Press — rewards (numbers invented to illustrate):

• Bare finding: "Cohort TFR fell from ~2.1 to ~1.7 across the 1955-1975 birth cohorts."
• Mechanism: Postponement of first births raised the mean age at first birth, and recuperation at

older ages was incomplete — a quantum decline operating through tempo, not a uniform shift.

• Observable implication: Parity-progression ratios from parity 0 to 1 should fall most at younger

ages and only partly rebound later; a pure quantum story would show uniform decline across ages.

• Scope condition: The claim covers low-fertility settings with delayed childbearing.
• Portability: A mortality scholar can import the tempo-vs-quantum logic to lifespan compression;

that import makes it a population-science contribution, not a single-country fact.

Referee-pushback patterns and the theory-side fix

• "This is descriptive — where is the mechanism?" -> Name the behavior/exposure/compositional shift

that moves the rate, for whom, and translate it into an age or cohort signature the design can test.

• "You assert a mechanism but a compositional shift would produce the same trend." -> State the rival

(composition) explicitly and the observable that separates it from your account before the results.

• "The formal model's assumptions are chosen to deliver the identity." -> Flag which results depend on

stability/Markov/independence assumptions and motivate each substantively.

• "This only works for your one case." -> Generalize the mechanism so another demographer studying a

different component could reuse it; otherwise reframe via demog-topic-selection.

Output format

【Core claim】one sentence
【Mechanism / identity】the population story or formal result
【Assumptions】(formal) the load-bearing ones
【Observable implications】testable demographic signatures -> research-design
【Scope conditions】which populations / periods it covers
【Portability】who else in population science can use this
【Next】demog-research-design

Supplementary resources

• [../../resources/external_tools.md](../../resources/external_tools.md) — formal-demography and decomposition tooling
• [../../resources/official-source-map.md](../../resources/official-source-map.md) — Demography scope and contribution expectations