---
marp: true
theme: default
paginate: true
---

# Decision and Authorization Latency in NASA Programs

A cliometric analysis of program cost, schedule, and mission cadence, 1958 to 2026

**Candidate PHD-08**
COLLEGIUM 1st Battalion
Category: Mission Program Execution Management
Date: 2026-06-15

The stewardship of a national space program rests on the care with which public resources are committed and the speed with which sound decisions are reached. This work measures that speed.

Dissertation defense brief. Design stage. No estimate is executed; expected signs are labeled as such.

---

## The answer, stated first

- **Contribution:** a consistent, documentary-rule-based series of NASA decision-and-authorization latency can be built across 1958 to 2026, and this dissertation pre-registers a falsifiable panel test of whether longer latency worsens program execution.
- **The measurement stands even if the null is retained.** No such latency series exists today.
- **H0 (null):** authorization latency has no association with cost growth, schedule slip, or mission cadence, after program and era fixed effects and technical controls.
- **H1 (alternative):** longer latency is associated with greater cost growth, more schedule slip, and lower mission cadence, after the same accounting.
- One coefficient, beta, estimated for three outcomes under one pre-registered decision rule. Either result is informative.

---

## The problem

- NASA programs frequently exceed the cost and schedule baselines set at commitment; the cost-estimating community documents this and sets reserves by it [1, 2, 3, 6].
- One GAO accounting: 15 development-phase projects, roughly 12 billion dollars in overruns and 28 cumulative years of delay [34].
- Practitioners attribute part of the growth to slow internal decision and authorization, not to engineering difficulty alone.
- Cited examples: Constellation spent about nine billion dollars over roughly five years with no operational missions [102]; the first-to-second crewed SLS interval spanned about 41 months (program-record-derived).
- These are anecdotes, not measurements. The attribution has never been tested as a quantitative hypothesis across the agency's history.

---

## The gap: two literatures never joined

- **NASA cost-and-schedule literature:** sophisticated on technical and contractual drivers (mass, power, TRL, contract type); treats schedule as an outcome; never isolates administrative latency as a measured regressor over the full history [1, 4, 6, 19].
- **Public-administration literature:** validated constructs for red tape and administrative burden, related to performance, but never applied to NASA as a long-run program panel [7, 8, 10, 11].
- Neither closes the question. The right outcomes sit in one literature; the right explanatory construct sits in the other.
- The empty cell: a within-NASA, long-run, latency-specific, identified panel test. That is the contribution space.

---

## Theoretical framework: the cliometric lens

- **Maddison.** Consistent constant-unit, single-rule measurement is the precondition of any cross-era comparison. This fixes the latency-construction rule and the constant-dollar deflation [16, 26].
- **Fogel.** The counterfactual ("would have cost less under faster authorization") must be named and reported as a bounded range, not a point. Within-program and within-era comparison is the feasible counterfactual [17].
- **North (with Williamson).** Authorization latency is an internal transaction cost: the time to measure an action against the rules and obtain authorization [18, 27].
- **North path dependence.** Authorization regimes generate increasing returns and lock in, so latency is large and durable. This is why era fixed effects are required, not optional [18].

---

## The mechanism (named, not a correlation)

**Driver:** fragmented decision authority and sequential, multi-layer authorization gates paced by an annual appropriations cycle.

**Mechanism:** each pending action accrues elapsed time (latency) during which standing program costs continue and requirements drift.

**Observable effect:** measured latency per program-phase rises.

**Operational consequence:** higher cost growth and schedule slip; fewer flight or delivery events per period (lower cadence).

**Strategic implication:** a portion of NASA cost and schedule performance is a controllable process variable, distinct from irreducible engineering difficulty, a lever for program execution management, including JPL's long-cycle programs.

---

## Key variable definitions

- **Authorization latency (explanatory):** median elapsed months between a documented trigger event (authorization becomes due) and the documented resolution event, per program-phase. Two resolutions: coarse (milestone-to-milestone, full span) and fine (key-decision-point, modern subperiod), reported separately.
- **Cost growth (outcome 1):** (actual minus baseline phase cost) / baseline, in constant FY dollars via the NASA New Start Inflation Index.
- **Schedule slip (outcome 2):** (actual minus baseline phase duration) / baseline duration.
- **Mission cadence (outcome 3):** events per period within a program family; reported under multiple family definitions.
- **Unit of analysis:** the program-phase observation. Panel = program i; time = phase p and era t.

---

## Data: three named sources

- **NASA historical budgets and program records:** baselines, actuals, and the dated decision and authorization events (KDPs, PCRs, confirmation reviews, budget actions).
- **NASA Technical Reports Server (NTRS):** technical parameters for controls and process documentation (Standing Review Board Handbook, NPR 7120.5 practice records, JCL policy) [80, 86, 94, 95].
- **GAO major-project assessments:** an independent, consistently formatted measure of cost and schedule performance for the modern era; anchors and cross-checks the panel.
- Joined on the program-phase key; triangulated and validated against published figures before any estimate.

---

## Data: coverage and honest gaps

- Coverage spans 1958 to 2026. Dense and fine in the modern era; sparse and coarse in the early decades. The panel is **unbalanced by construction**.
- The two-resolution latency design is the response to changing documentary density, not an afterthought.
- **Acknowledged corpus gaps:** the GAO assessment series carries report numbers, not DOIs, and must be entered as discrete records; the NASA New Start Inflation Index source document must be added. Both are named, not papered over.
- Ethics: documentary records, no human subjects, public or archivally accessible, reproducible from source.

---

## Research design: the estimator

For program i, phase p, era t:

**Outcome(i,p,t) = beta * Latency(i,p,t) + gamma * X(i,p,t) + alpha_i + delta_t + epsilon(i,p,t)**

- alpha_i: program fixed effects (absorb time-invariant program difficulty).
- delta_t: era fixed effects (absorb common shocks and durable rule regimes).
- X: technical and programmatic controls (mass, power, TRL, mission class, contract type, partners, funding-instability index).
- epsilon: clustered by program; wild-cluster bootstrap given few clusters.
- Identification: within-program and within-era variation in latency. Under H1, beta > 0 for cost and schedule, beta < 0 for cadence.

---

## Identification: the endogeneity threat

- **Central threat:** latency is endogenous to phase-specific difficulty. A hard phase may both take longer to authorize and overrun more, manufacturing a spurious positive coefficient.
- Program fixed effects remove fixed difficulty but not difficulty that varies by phase.
- **Response (instrumental-variable strategy, after Decarolis [12]):** instrument latency with sources that move authorization timing for reasons unrelated to phase difficulty:
  - authorizing-office workload (queue congestion);
  - timing relative to the appropriations calendar (budget-paced, exogenous to engineering).
- Relevance and exclusion are argued and tested; first-stage strength reported; imperfect-instrument bounds reported regardless of the over-identification test.

---

## Identification: the staggered-regime concern

- Administrative regimes switch on at staggered times. The naive two-way fixed-effects difference-in-differences estimand can carry negative weights and wrong-sign bias under heterogeneous, staggered effects [21, 29].
- **First defense:** latency enters as a continuous within-program variable, not a binary staggered treatment.
- **For any discrete-regime analysis:** heterogeneity-robust estimators replace two-way fixed effects:
  - Callaway and Sant'Anna group-time [22];
  - de Chaisemartin and D'Haultfoeuille intertemporal [23];
  - Goodman-Bacon decomposition reported as a diagnostic [21, 29]; Imai-Kim governs interpretation [24].

---

## Threats to validity

- **Internal:** endogeneity (fixed effects + IV); baseline gaming (reserves-to-baseline conservatism test); reverse causation (early-in-phase latency + instruments).
- **External:** NASA-specific. The federal-procurement evidence [12, 53] is corroboration of the mechanism, not a basis for transferring magnitude.
- **Construct:** latency operationalized from documentary records, defended at two resolutions; cadence is most fragile and reported under multiple family definitions.
- **Statistical-conclusion:** unbalanced panel, modest cluster count. Wild-cluster bootstrap, alternative clustering, alternative fixed-effects structures.
- Each rival has a pre-specified observable signature and a check that could vindicate it [33].

---

## Analysis plan: five pre-registered steps

1. **Assemble and validate** the panel; reproduce published GAO and Aerospace Conference figures for overlapping programs.
2. **Describe before estimating:** univariate latency, cost, schedule, cadence distributions by era, constant-dollar. The latency series is the standalone first product.
3. **Fixed-effects baseline** for all three outcomes, program-clustered.
4. **Address endogeneity:** IV; report first-stage strength; compare FE vs IV.
5. **Robustness and heterogeneity:** both latency resolutions, alternative cadence definitions, heterogeneity-robust estimators, Goodman-Bacon diagnostic; report bounded ranges, in the Fogel manner.

---

## The binding decision rule

- **Reject H0 only if** the latency coefficient is statistically distinguishable from zero, carries the predicted sign in the same direction across the fixed-effects and instrumented specifications, and survives the robustness battery.
- **Retain H0 (contribution fails) if** the coefficient is indistinguishable from zero, reverses sign between specifications, vanishes under the heterogeneity-robust estimators or the alternative latency resolution, or is fully explained by baseline gaming or reverse causation.
- The rule is conjunctive and fixed before estimation. No single specification can carry a rejection. The null is a genuine possibility, not a straw figure.

---

## Expected results (design-stage, NOT findings)

Directional expectations under H1, with mechanisms. These are not estimates.

| Outcome | Expected sign | Mechanism |
|---|---|---|
| Cost growth | + | standing costs accrue while waiting; delay invites requirements change |
| Schedule slip | + (largest, most robust) | latency is itself a component of elapsed schedule |
| Mission cadence | - | higher latency per decision lengthens the interval between flight events |

The illustrative coefficient table is **left unpopulated by design**. Reporting fabricated coefficients as if real would violate the falsifiability standard.

---

## Confidence and uncertainty

- **Measurement contribution (the latency series):** *high* confidence, largely independent of the regression. Lowered only if the early-decade record is too sparse for even the coarse measure.
- **Causal claim (latency moves the outcomes):** *moderate* at design stage. Raised by a strong first stage, imperfect-instrument bounds excluding zero, and stability across resolutions and estimators.
- **Cadence inference:** *low to moderate*, the most construct-fragile outcome; requires concordance across family definitions.
- Statistical significance is necessary but not sufficient; magnitude is interpreted against the historical cost-growth distribution, fixed before estimation.

---

## How the argument holds together

The case is built in five steps, each tied to evidence already on the table:

- **The problem is real:** cost-growth literature [1, 2, 3, 5, 6]; contractor analysis [34].
- **The problem is material:** fat-tailed overrun [13, 30, 59]; Constellation [102].
- **The design addresses the mechanism:** documentary latency entered net of fixed effects, with an instrument [12, 18, 80, 86].
- **The design improves on the alternatives:** heterogeneity-robust estimators with a Goodman-Bacon diagnostic, not naive two-way fixed effects [21, 22, 23, 24].
- **The residual risk is managed:** the instrument, the baseline-conservatism test, early-in-phase measurement, and the two latency resolutions [12, 33].

**Residual risk:** endogeneity bounded not eliminated; small cluster count; cadence fragility. Each is paired with a pre-committed check, and confidence is downgraded where a risk cannot be fully defeated.

---

## Scope: architecture traceability out of scope

- This is a cliometric econometric study. Its objects are programs, phases, documentary events, and panel estimates.
- No capability, system, data or service exchange, or enterprise architecture is being designed or fielded.
- No DoDAF or BEA architecture-traceability table; forcing that vocabulary would be a category error.
- The single permitted architecture-adjacent statement: faster, lower-layer authorization as a program-execution-management lever, stated in management terms, conditional on H1.

---

## Implications under each outcome

- **If H1 holds:** a measurable share of cost growth and schedule slip is a controllable process variable, independent of technology investment and contract structure. Largest return on long-cycle programs (JPL). A dollar-and-month instantiation of the red-tape construct.
- **If H0 holds:** authorization latency is not the lever; redirect reform to technical and estimating factors [1, 4, 6]. Disciplines the practitioner narrative by showing it does not survive measurement.
- The design is symmetric: either result is a usable finding. That symmetry is the mark of a well-posed test.

---

## Limitations, stated without softening

- The test has not been run; every coefficient is expected or illustrative.
- Baseline-definition drift across eras (response: two resolutions, era fixed effects).
- Undated events force bounded latency (response: Fogel-style ranges).
- Optimistic baselines / baseline gaming (response: reserves-to-baseline conservatism test).
- Endogeneity to phase-specific difficulty (response: IV, both estimates reported).
- External validity is NASA-specific; cadence is family-definition-fragile.
- Corpus gaps named: GAO series and New Start Inflation Index must be entered as discrete records.

---

## The threefold contribution, restated

1. **A consistent long-run NASA latency series**, built to a single documentary rule, 1958 to 2026. Stands regardless of the regression result.
2. **Joining two literatures** for one agency over the long run: NASA cost-and-schedule meets public-administration administrative process.
3. **A pre-registered falsifiable test** with a fixed decision rule, a bounded counterfactual, and a refusal to report invented estimates as real.

The first stands now. The second is a design and framing contribution. The third awaits execution. The document closes on the claim it opened with.

---

## Future work

- **Near horizon:** execute the five-step procedure; populate the illustrative table with estimated coefficients and bounded ranges; deliver the descriptive latency series. Prerequisite: enter the GAO records and the New Start Inflation Index.
- **Middle horizon:** extend to comparator agencies sharing the appropriations-calendar instrument; a multi-agency panel with agency fixed effects tests whether the mechanism travels [12].
- **Far horizon (conditional on rejecting H0):** turn the latency series into a forward-looking management instrument that flags accruing authorization delay before it converts into measured growth.

---

## Selected references

- [1] Emmons, Bitten, Freaner (2007). Historical NASA Cost and Schedule Growth. IEEE Aero. doi:10.1109/aero.2007.353027
- [6] National Research Council (2010). Controlling Cost Growth of NASA Earth and Space Science Missions. doi:10.17226/12946
- [12] Decarolis et al. (2018). Bureaucratic Competence and Procurement Outcomes. NBER 24201. doi:10.3386/w24201
- [16] Bolt, van Zanden (2017). The Maddison Project. doi:10.5195/jwhi.2017.46
- [17] Herranz-Loncan (2006). Railroad Impact in Backward Economies. doi:10.1017/s0022050706000350
- [18] North (1990). Institutions, Institutional Change and Economic Performance. doi:10.1017/CBO9780511808678
- [21] Goodman-Bacon (2018). DiD with Variation in Treatment Timing. NBER 25018. doi:10.3386/w25018
- [22] Callaway, Sant'Anna (2020). DiD with multiple time periods. doi:10.1016/j.jeconom.2020.12.001

Full 139-entry reference list appears in the dissertation. All entries carry a clickable DOI or resolvable URL.

---

## Closing

State the proposition quantitatively. Build the measures from primary records to a single rule. Bound the estimate. Let the data decide, under a rule fixed in advance.

The contribution is the design and the measurement. The verdict belongs to the data, when they are made to speak.

**Thank you. Questions welcome.**