No Plan, No Aid?

“No Plan, No Aid?” asks whether a domestic policy reform can change how donors allocate scarce aid. We situate the question within the classic need–merit dilemma of aid allocation and within the IPCC AR6 risk framework, which separates physical hazard and exposure (recipient need) from adaptive capacity (recipient merit). We model the adoption of a National Adaptation Plan (NAP) as an information shock that updates donors' perception of a country's adaptive capacity while leaving its physical hazard unchanged — providing the first causal identification of the allocative role of this institutional component.

Empirically, we build a country–year panel linking the staggered timing of NAP submission to the UNFCCC to project-level adaptation finance from the OECD Creditor Reporting System (CRS) (Rio adaptation markers), over 2009–2024. Using the Callaway & Sant'Anna (2021) staggered difference-in-differences estimator — and corroborating with de Chaisemartin & d'Haultfœuille (2024) — we find that NAP adoption raises adaptation commitments by about 34% and a country's share of global adaptation finance by about 0.3 percentage points, with total and non-adaptation aid statistically flat. The pattern points to a composition channel: finance is reallocated toward adaptation rather than the overall envelope expanding. The response is concentrated among DAC bilateral and multilateral donors, consistent with NAPs functioning as capacity (readiness) signals.

On the theory side, the paper embeds NAP adoption in a recipient–donor allocation model. A donor splits a fixed global budget across countries and, within each country, between adaptation aid and other aid using nested CES aggregators. NAP adoption enters as a policy-driven rise in a country's effective adaptive capacity V_i, generating two empirically separable channels:

• Composition channel: higher perceived capacity raises the adaptation share of a country's envelope.
• Envelope (scale) channel: higher perceived returns may expand the total envelope allocated to adopters — but need not.
• Global reallocation: adopters should capture a larger share of a broadly fixed global adaptation pool.

The model maps the IPCC AR6 risk components onto aid-allocation motives: time-invariant hazard is a primitive absorbed by country fixed effects; slow-moving exposure/sensitivity enter as controls; and adaptive capacity — the merit margin — is the only component NAP adoption shifts. This is what makes NAP submission an information shock to capacity rather than a shock to climate need.

On the empirical side, the paper combines:

• Adaptation finance: OECD Creditor Reporting System project-level commitments tagged with the Rio adaptation marker (principal or significant), aggregated to the recipient–year level over 2009–2024.
• NAP adoption: the year of first official NAP submission to the UNFCCC (NAP Central tracking tool), an absorbing treatment.
• Controls: pre-treatment WGI Government Effectiveness and log population, both evaluated at g − 1, plus country and year fixed effects.
• Estimation: the Callaway & Sant'Anna (2021) staggered difference-in-differences estimator (doubly robust, never-treated comparison, multiplier-bootstrap inference), corroborated by de Chaisemartin & d'Haultfœuille (2024) and a Goodman–Bacon decomposition.

Average effects. The ATT on log adaptation commitments is 0.294 (SE = 0.125, p < 0.05), a ≈ 34% increase. Total commitments (0.101) and non-adaptation commitments (0.071) are statistically indistinguishable from zero. The effect is reallocation within a roughly constant envelope — the model's composition channel — not an expansion of overall finance. The country's share of global adaptation finance rises by 0.32 percentage points (SE = 0.140, p < 0.05). Disbursements move in the same direction but are imprecise (0.142), consistent with implementation lags.

Dynamics and mechanism. Adaptation commitments rise progressively after adoption rather than jumping on impact, consistent with donor programming cycles. A placebo test shifting treatment two years earlier yields a null (ATT −0.019), and a mitigation falsification — an outcome NAPs carry no mandate over — shows no significant effect, sharpening the reading that donors respond to the adaptation-capacity content of a NAP rather than to a generic climate-policy posture.

Who responds. By donor type, the effect is driven by DAC bilaterals (≈ 36%, p < 0.05) and multilaterals (≈ 74%, p < 0.05), while non-traditional donors show no response — the pattern predicted for capacity-screening donors. The effect is concentrated in below-median-governance adopters and among low- and lower-middle-income countries (≈ 38–52%), consistent with a credentialing interpretation. LDC/non-LDC splits are imprecise, most plausibly because the LDC estimation sample contains only 17 countries.

Robustness & caveats. The headline effect is stable across alternative comparison groups (not-yet-treated), outcome units and the treatment of zeros (the unit-invariant intensive-margin and PPML estimates give ≈ 33% and ≈ 45%), and two heterogeneity-robust estimators. We are transparent about limits: formal pre-trend tests reject for the total and non-adaptation outcomes (so the "flat envelope" reading is suggestive, not established), and HonestDiD sensitivity shows the positive adaptation and share effects are guaranteed only for parallel-trend violations below 25% of the largest observed pre-trend — warranted caution given the short post-treatment window.

Bottom line. NAPs pay — not by equally expanding every envelope, but by signaling readiness to the donors who respond to it, shifting each adopter's envelope toward adaptation and reallocating global adaptation finance toward adopters within a broadly constant total. Because the mechanism rewards the capacity (merit) dimension of climate risk, it also risks steering finance away from high-need, low-capacity non-adopters — the very need–merit trade-off the framework makes visible.