# Cost-Overrun Hazards in Earth-Observing Missions: A Competing-Risks Model Separating Instrument-Driven from Launch-Driven Slip

**Candidate:** JPL_ASTRO_EARTH_08
**Program:** COLLEGIUM 1st Battalion
**JPL / NORTH STAR category:** Earth Science Missions
**Date:** 2026-06-15

---

## Abstract

Schedule slip is the proximate driver of cost growth in NASA Earth-observing missions, yet the standard treatment of slip aggregates all causes into a single delay variable. This collapses two mechanisms that have different physics, different owners, and different policy levers: slip that originates in instrument development and slip that originates in launch-vehicle availability. This dissertation proposes a competing-risks survival model of cause-specific schedule-slip onset across an Earth-mission cohort assembled from NASA Cost Analysis Data Requirement (CADRe) records, the NASA Instrument Cost Model (NICM) dataset, the Government Accountability Office (GAO) annual assessments of major NASA projects, and TechPort sensor technology-readiness-level (TRL) records. The unit of analysis is the mission-phase spell from the start of formulation; the two competing events are first instrument-driven slip and first launch-driven slip. The estimator is the Fine-Gray subdistribution hazard model, complemented by cause-specific Cox models and Gray's test, with the subdistribution form chosen because the policy question concerns the cumulative probability of each slip type rather than the instantaneous rate alone. The falsifiable contribution is that these two slip sources are statistically distinct competing risks and that instrument-driven slip is the dominant hazard for missions carrying first-of-kind active sensors but not for passive-radiometer heritage missions. The null is that the two slip sources are not separable competing risks. The work applies the structural-decomposition logic associated with Fogel and the heterogeneity-aware separation logic associated with Callaway and Sant'Anna, treating sensor novelty as the effect modifier rather than collapsing it into a pooled coefficient. This is a design-stage dissertation: the model specification, identification strategy, and analysis plan are fully developed, and all reported numerical results are explicitly labeled as illustrative and not yet executed on the full cohort. If confirmed, the result would let NASA and JPL allocate schedule reserve to the dominant hazard for a given mission archetype rather than to an undifferentiated slip pool.

---

## 1. Introduction and Contribution

### 1.1 The problem

NASA Earth-observing missions overrun their committed budgets at rates that have proven stubborn across four decades of acquisition reform. The mechanism that connects a budget commitment to its overrun is almost always schedule: a mission that takes longer than planned accrues standing-army costs, contractor fee adjustments, and rework, so that cost growth is in large part a monetized image of schedule slip. The practitioner literature treats this coupling as nearly definitional. Lieber and Donor, examining complex NASA and defense projects, document that cost overruns and schedule delays are tightly linked but that the phrase "directly related to" hides several distinct root causes operating at once, including unrealistic estimates, technical problems, supply-chain difficulties, and scope changes [1].

The problem this dissertation addresses is that the standard analytic treatment of slip is undifferentiated. A mission is recorded as having slipped, and the slip is regressed on covariates as if it were a single hazard with a single cause. But slip in an Earth-observing mission has at least two structurally different origins. The first is instrument development: a science instrument that fails environmental testing, that cannot close its calibration budget, or that carries a detector technology below the maturity its schedule assumed. The second is launch-vehicle availability: a launch service that is delayed by manifest congestion, by an anomaly on a shared vehicle, or by a provider's own development slip, independent of the spacecraft's readiness. These two origins have different owners (the instrument provider versus the launch services program), different physical drivers (technology maturation versus manifest dynamics), and different policy levers (TRL gating and instrument reserve versus launch-vehicle diversity and manifest margin). Treating them as one variable forecloses the analysis that would tell a program office which lever to pull.

### 1.2 The gap in the literature

Three literatures touch this problem but none resolves it. The spacecraft cost-and-schedule literature establishes that low technology readiness predicts schedule slippage: Dubos, Saleh, and Braun show, on a cross-mission dataset, that the TRL of a spacecraft's least-mature technology at authorization is a measurable driver of subsequent schedule slip [2]. Bearden's complexity-based cost-estimating relationships establish that mission complexity and aggressiveness predict cost and schedule outcomes [3]. But this literature models slip or cost as a single continuous outcome and does not separate the cause of slip into competing events with distinct hazards.

The competing-risks survival literature provides the missing apparatus. Fine and Gray introduced a proportional-hazards model for the subdistribution of a competing risk, allowing direct regression modeling of the cumulative incidence of one event type in the presence of others [4]. Austin, Lee, and Fine give the applied framework for choosing between cause-specific and subdistribution hazards depending on whether the question is etiologic or predictive [5], and Austin and Fine provide practical reporting recommendations [6]. This apparatus is mature in biostatistics but has not been applied to NASA mission schedule slip, where the competing-risks structure is natural: a mission's first slip is either instrument-driven or launch-driven, and the occurrence of one censors the clean observation of the other as the "first" cause.

The third literature, project-overrun economics, supplies the rival explanations. Flyvbjerg's work on optimism bias and reference-class forecasting argues that overruns are driven less by project-specific technical causes than by systematic optimism in the estimating process itself [7], and the demand-forecast-inaccuracy literature reinforces that estimates are biased at the planning stage [8]. A credible separation of slip causes must show that the instrument-versus-launch distinction survives controls for this estimating optimism.

The synthesis this dissertation offers is that the three literatures fit together if slip is modeled as a competing-risks process. The spacecraft cost-and-schedule literature supplies the covariates that should drive each cause. The competing-risks literature supplies the estimator that keeps the two causes from contaminating each other. The overrun-economics literature supplies the control that prevents the separation from being an artifact of the estimating process. No prior study combines all three. The gap is therefore not the absence of any one element but the absence of their integration into a single identification strategy that can test separability and archetype dependence at once.

### 1.3 The falsifiable contribution

This dissertation states a single falsifiable contribution.

**H1 (contribution):** For Earth-science missions, schedule slip from instrument development and schedule slip from launch-vehicle availability are statistically distinct competing risks, and instrument-driven slip is the dominant subdistribution hazard for missions carrying first-of-kind active sensors but not for passive-radiometer heritage missions.

**H0 (null):** The two slip sources are not separable competing risks; either their subdistribution hazards are statistically indistinguishable across the cohort, or the dominance of instrument-driven slip does not differ between first-of-kind active-sensor missions and passive-radiometer heritage missions.

The contribution is falsifiable in three concrete ways. First, if a likelihood-ratio or Gray's test cannot reject equality of the two cause-specific hazard processes, the events are not separable and H0 holds. Second, if the interaction between sensor archetype (first-of-kind active versus passive-radiometer heritage) and the instrument-slip subdistribution hazard is not statistically distinguishable from zero, the dominance claim fails. Third, if the separation disappears once estimating optimism is controlled, the contribution is not robust.

The claim is deliberately narrow. It does not assert that instrument-driven slip is always the larger problem, nor that launch-driven slip is negligible, nor that sensor novelty is the only effect modifier that matters. It asserts a specific, conditional dominance: instrument-driven slip is the leading first-slip hazard for one identifiable archetype and not for another. Narrowness is a virtue here because it makes the claim testable on a small cohort and because it maps onto a decision a program office actually makes, namely where to hold reserve at confirmation. A broader claim would be harder to falsify and less useful to the people who would act on it.

### 1.4 Why it matters for NASA and JPL

Schedule reserve and cost reserve are finite and are allocated at confirmation under standing policy. If instrument-driven and launch-driven slip are one hazard, reserve should be held centrally against an undifferentiated pool. If they are two hazards with archetype-dependent dominance, reserve should be steered: a first-of-kind active-sensor mission such as a lidar or radar should hold reserve against instrument maturation, while a heritage passive-radiometer continuity mission faces a different dominant risk and should hold reserve against manifest and launch dynamics. For JPL, which leads first-of-kind active-sensor Earth missions, the dominance result, if confirmed, would directly inform reserve posture and TRL gating at Key Decision Point B. The decision-relevant quantity is the cumulative incidence of each slip cause by archetype, which is exactly what the subdistribution model estimates.

---

## 2. Background and Literature

### 2.1 The schedule-cost coupling in space systems

The empirical claim that cost growth is largely a function of schedule slip is well supported in the space-systems literature. Lieber and Donor's analysis frames the coupling and catalogues its root causes, distinguishing causes that are obvious in hindsight (unrealistic estimates, technical problems) from those that are diffuse [1]. The NASA cost-modeling community has long built cost-estimating relationships that are sensitive to schedule. The NASA Instrument Cost Model and its Explorer-class extension (NICM-E) estimate instrument cost as a function of mass, power, data rate, and technology parameters, and the model's residuals are known to load on technology maturity and on instrument novelty [9]. Habib-agahi and colleagues document long-run cost-time trends in NASA spaceflight instruments, establishing that instrument development is a primary locus of cost and schedule risk distinct from the bus and from launch [10].

### 2.2 Technology readiness and schedule slip

The most direct antecedent is the TRL-schedule literature. Dubos, Saleh, and Braun provide a quantitative model in which schedule slippage is a function of the maturity of a spacecraft's least-ready technology at the time of authorization [2]. Their finding is that a one-level deficit in starting TRL maps to a measurable increase in expected schedule slip, and that the relationship is nonlinear at the low-maturity end. This establishes the instrument side of the competing-risks structure: the covariate that should drive instrument-caused slip is sensor TRL at confirmation, available in TechPort and in CADRe. Bearden's complexity-based cost-estimating relationships add that complexity and design aggressiveness independently raise both cost and schedule risk [3], providing the complexity controls the model needs so that "first-of-kind active sensor" is not merely a proxy for general complexity.

### 2.3 The launch-availability side

The launch side of the structure is less studied in the survival framing but is well established as a delay source. Earth-observing continuity missions are frequently manifested on shared or transitioning launch services, and slip on the launch side is driven by manifest congestion and by provider development, both of which are exogenous to the spacecraft's own readiness. The continuity-mission literature, exemplified by the Landsat Data Continuity Mission account [11], documents launch-driven schedule pressure that is structurally separate from instrument maturation. This separation is the substantive basis for treating the two slip causes as competing risks rather than as one.

The exogeneity of much launch-side slip is the feature that makes the competing-risks framing credible rather than circular. A passive-radiometer continuity mission whose instrument is a direct rebuild of a flown unit carries little instrument maturation risk, so its first slip, when it occurs, is disproportionately likely to originate on the launch side: a shared vehicle anomaly, a manifest reshuffle, or a provider's own development slip. A first-of-kind active-sensor mission carrying a lidar or radar with no flight heritage carries the opposite loading. The two archetypes therefore face different mixtures of the two competing hazards, which is precisely the heterogeneity the dissertation sets out to measure. If the mixtures were identical across archetypes, the competing-risks decomposition would add little; the claim that they differ is what gives the work its decision relevance.

### 2.4 The competing-risks apparatus

The methodological backbone is the competing-risks survival model. Prentice and colleagues established the cause-specific hazard formulation, and Fine and Gray introduced the subdistribution hazard that allows direct regression on the cumulative incidence function of a single competing event [4]. The applied literature is clear on the choice between the two. Austin, Lee, and Fine explain that the cause-specific hazard answers the etiologic question of how a covariate affects the instantaneous rate of an event among those still at risk, while the subdistribution hazard answers the predictive question of how a covariate affects the cumulative probability of the event in the real population where competing events occur [5]. Austin and Fine give reporting standards [6], and the broader epidemiologic guidance warns against the standard error of treating a competing event as ordinary censoring, which biases the naive Kaplan-Meier estimate upward [12]. Gray's test provides a nonparametric comparison of cumulative incidence functions across groups [13]. For this dissertation the policy question is predictive, so the subdistribution hazard is primary, but the cause-specific hazard is estimated alongside it because the two answer complementary questions and divergence between them is itself informative.

### 2.5 Applying the anchor methodologists

Two methodological lenses from the Hall of Shoulders discipline the design.

The first lens, associated with Fogel, is structural decomposition with an explicit counterfactual. Fogel's signature move is to take an aggregate outcome that everyone attributes to a single dominant cause and to decompose it into separable channels whose individual magnitudes can be measured against a constructed counterfactual, as in the social-savings method that separated the contribution of one transport mode from the economy-wide aggregate [14]. Applied here, the lens demands that "schedule slip" not be left as an aggregate. The dissertation decomposes it into the instrument channel and the launch channel and asks, for each archetype, what the cumulative incidence of slip would be if one channel were removed. The subdistribution cumulative incidence function is the counterfactual-bearing quantity: it answers what fraction of missions of a given archetype would have experienced instrument-driven slip first, in the world where launch-driven slip also competes for primacy.

The Fogel lens also carries a methodological warning that this dissertation takes seriously. Fogel's social-savings decompositions were attacked precisely on the grounds that the counterfactual was constructed rather than observed, and the defense was always to make the construction explicit and to bound it with sensitivity analysis. The analogue here is that the cumulative incidence function of a removed competing risk is a counterfactual quantity, not a directly observed one, and it must be reported with the same explicitness: stated assumptions, stated bounds, and a sensitivity analysis that shows how the decomposition moves when the cause-coding or the at-risk structure is perturbed.

The second lens, associated with Callaway and Sant'Anna, is the refusal to let a pooled regression coefficient stand in for a heterogeneous set of underlying comparisons. Their critique of the two-way fixed-effects estimator is that under heterogeneity it returns a contaminated weighted average of many distinct effects, some with negative weights, so that the headline coefficient need not correspond to any causal quantity the researcher intends [15]. The analogous error here would be to estimate a single instrument-slip hazard pooled across all Earth missions and to report it as "the" instrument hazard, when the hazard differs sharply between first-of-kind active-sensor missions and heritage passive-radiometer missions. The Callaway-Sant'Anna discipline requires that sensor archetype enter as an explicit effect modifier, that the archetype-specific subdistribution hazards be estimated as separable building blocks, and that any aggregate be a transparent, defensible weighting of those blocks rather than a regression-imposed average. This is exactly the interaction term that operationalizes H1. The same authors' doubly robust estimation logic, which combines an outcome model and a weighting model so that consistency survives misspecification of either one [16], motivates the dissertation's parallel use of penalized regression and reweighting as a robustness check on the archetype-specific hazards.

---

## 3. Data

### 3.1 Named datasets and access paths

The cohort is assembled from four named sources.

1. **NASA CADRe records.** The Cost Analysis Data Requirement is the standardized cost-and-technical record produced at each major milestone for NASA projects above the cost threshold. CADRe supplies, per mission, the confirmation baseline schedule, the as-flown schedule, milestone dates, mass and power, and the narrative Part A that identifies the proximate causes of milestone movement. Access is through the NASA Cost Analysis Division and the ONCE (One NASA Cost Engineering) database, available to NASA and to FFRDC cost analysts under data-use agreement.

2. **NICM dataset.** The NASA Instrument Cost Model and its Explorer extension carry the instrument-level parameter records (mass, power, data rate, instrument type, pointing, TRL) used to build the cost-estimating relationships [9]. The instrument-type taxonomy in NICM is the source for the active-versus-passive and the radiometer-versus-other classification that defines the archetype variable.

3. **GAO Assessments of Major NASA Projects.** The GAO annual assessment reports project-level cost and schedule baselines, current estimates, and a narrative attribution of slip, published yearly and publicly available. The GAO narratives are the cross-check on the CADRe Part A cause attribution: where both sources name a cause, the slip-cause coding is high-confidence.

4. **NASA TechPort sensor TRL records.** TechPort is NASA's public technology-portfolio database and supplies the technology-readiness-level history of the sensor technologies carried by each mission. TechPort is the source for the entry TRL of the least-mature sensor technology at confirmation, the key instrument-side covariate, and is publicly accessible through the TechPort API.

### 3.2 Unit of analysis

The unit of analysis is the mission-development spell, defined as the interval from the start of Phase B (preliminary design, beginning at Key Decision Point B) to the earlier of the first recorded schedule-slip event of either cause or the launch readiness date. Time is measured in months from KDP-B. Each mission contributes one spell; the spell ends in one of three states: instrument-driven first slip, launch-driven first slip, or administrative censoring at launch with no recorded slip of magnitude above threshold.

### 3.3 Variable construction

**Outcome (competing events).** A slip event is defined as a committed-baseline schedule movement exceeding a threshold (illustratively, two months of net launch-date movement) attributable to a single dominant cause in the milestone period. The cause is coded as instrument-driven or launch-driven by reconciling the CADRe Part A narrative with the GAO assessment narrative for the same project-year. Events whose cause cannot be coded consistently across both sources are flagged and handled in sensitivity analysis. The two cause-coded first-slip events are the competing risks.

**Archetype (effect modifier).** Each mission is classified from the NICM instrument taxonomy as carrying a first-of-kind active sensor (for example, a lidar or a radar with no direct flight heritage) or as a passive-radiometer heritage mission (a passive radiometer with direct heritage from a predecessor instrument). Missions that are mixed or ambiguous are coded into a third category for robustness checks and excluded from the primary contrast.

**Instrument-side covariates.** Entry TRL of the least-mature sensor technology at KDP-B from TechPort; instrument mass, power, and data rate from NICM; number of distinct instruments.

**Launch-side covariates.** Launch-vehicle class and provider; shared-versus-dedicated manifest indicator; an indicator for a launch service in its own development at the time of manifest assignment.

**Controls.** Mission complexity index after Bearden [3]; an estimating-optimism proxy constructed as the ratio of the confirmation-baseline schedule to a reference-class median schedule for missions of similar class, following the reference-class-forecasting logic [7]; calendar-period fixed effects to absorb era-specific acquisition policy.

### 3.4 Coverage and limitations

The cohort targets NASA Earth-observing missions from roughly 1990 to the present that reach KDP-B, an expected sample on the order of 30 to 60 missions. The coverage limitations are real and are stated plainly. The sample is small for a survival model with two competing events and an interaction, which constrains the number of covariates and motivates penalized estimation. Cause coding depends on narrative attribution, which is subjective and which can mask multi-cause slip; the two-source reconciliation reduces but does not eliminate this. CADRe access is restricted, so the analysis is reproducible only by analysts with the data-use agreement. TechPort TRL histories are uneven in completeness for older missions. These limitations bound the external validity of any result to NASA Earth-observing missions of the studied era and do not extend to commercial or non-US missions.

A further limitation is left truncation and the definition of the spell origin. Some missions enter the formal cost-and-schedule record only after a period of pre-formulation study during which schedule risk has already been incurred, so the spell origin at KDP-B is a convention chosen for consistency rather than a true zero of risk. The Fine-Gray apparatus accommodates left truncation [4], and the analysis uses the left-truncated form where the entry time differs from the origin, but the convention should be read as part of the construct definition. A related limitation is competing-event dependence: the model does not assume the two slip causes are independent, which is one of the reasons the subdistribution formulation is preferred, but the cause-specific results are reported alongside so that a reviewer can judge whether the dependence structure materially changes the conclusion. Finally, the cohort is observational. There is no randomization of sensor archetype to missions, so the archetype contrast is a comparison of naturally occurring groups, and the controls for complexity and estimating optimism are the only defense against confounding of archetype with other risk drivers.

---

## 4. Research Design and Identification

### 4.1 Estimator

The primary estimator is the Fine-Gray proportional subdistribution hazard model [4], estimated separately for the instrument-driven event and the launch-driven event. For event type k, the model is

  subhazard_k(t | X) = subhazard_k0(t) · exp(Xβ_k),

where the subdistribution hazard governs the cumulative incidence function CIF_k(t | X), the probability of experiencing a cause-k first slip by time t accounting for the competing event. The coefficient on the archetype-by-instrument-side interaction is the parameter of interest for H1.

The cause-specific Cox model is estimated in parallel for each event, giving the instantaneous-rate interpretation, with Gray's test [13] used as the nonparametric check on whether the cumulative incidence functions of the two causes differ across archetype strata. Divergence between the subdistribution and cause-specific results is reported rather than suppressed, because the two answer the predictive and the etiologic question respectively [5].

### 4.2 Identification strategy

Identification of separability rests on three claims. First, the two events are genuinely competing in the data: a mission's first slip can be cause-coded as one or the other, and the occurrence of one alters the at-risk set for the other, which is the defining structure of competing risks rather than independent processes. Second, the archetype variable is assigned before the outcome and is a function of the instrument design fixed at KDP-B, so it is not a post-treatment quantity contaminated by slip. Third, the instrument-side and launch-side covariates are measured at or before KDP-B, preserving temporal ordering.

The dominance claim in H1 is identified by the archetype interaction, disciplined by the Callaway-Sant'Anna logic [15]: rather than reporting a pooled instrument-slip hazard, the design estimates the archetype-stratified subdistribution hazards as separable quantities and tests whether the instrument-slip subdistribution hazard exceeds the launch-slip subdistribution hazard in the first-of-kind active stratum but not in the heritage passive stratum. The Fogel structural-decomposition logic [14] supplies the counterfactual reading: the archetype-specific cumulative incidence functions answer what fraction of each archetype's missions would slip first for instrument reasons in the world where launch slip competes.

### 4.3 Model specification

The primary specification estimates, for each event k and each archetype stratum, the subdistribution hazard with instrument-side covariates (entry TRL, instrument count, mass, power), launch-side covariates (vehicle class, shared-manifest indicator, provider-in-development indicator), and controls (complexity index, optimism proxy, calendar-period effects). The pooled specification adds the archetype-by-TRL and archetype-by-shared-manifest interactions. Given the small sample, a ridge-penalized partial likelihood is the default to stabilize estimation, with the penalty selected by cross-validation, and the number of free covariates is capped by an events-per-variable rule.

### 4.4 Threats to validity

**Internal validity.** The dominant threat is cause-coding error: if instrument and launch slip co-occur and the dominant cause is misattributed, the two events bleed into each other and separability is spuriously rejected or accepted. The two-source reconciliation, the flagging of un-codable events, and a sensitivity analysis that recodes ambiguous events both ways are the defenses. A second threat is that entry TRL is itself chosen in anticipation of slip risk (endogenous gating), which is addressed by using the TRL recorded at KDP-B before slip is observed.

**External validity.** The cohort is NASA Earth-observing missions of a specific era. Results do not transport to commercial constellations, to non-Earth science, or to non-US programs. The archetype contrast is specific to the active-versus-passive sensor distinction and should not be read as a general novelty result.

**Construct validity.** "Instrument-driven slip" is a constructed category resting on narrative attribution; it is a defensible but imperfect proxy for the underlying physical cause. The archetype variable is a binary simplification of a continuous novelty spectrum. Both constructs are reported with their coding rules so that a reviewer can audit them.

**Statistical-conclusion validity.** The sample is small and the model carries an interaction, so power is the central concern. The design pre-commits to penalized estimation, to an events-per-variable cap, and to reporting confidence intervals and the realized power rather than only point estimates, so that a non-rejection of H0 is interpretable as either true non-separability or insufficient power, distinguished by the width of the intervals.

---

## 5. Analysis Plan and Findings

**This section is a design-stage analysis plan. No results below have been executed on the full cohort. All numbers are illustrative placeholders that specify the form of the expected output, not empirical findings.**

### 5.1 Estimation procedure

The execution sequence is fixed in advance.

1. Assemble the cohort and code the two competing events by reconciling CADRe Part A and GAO narratives; freeze the cause-coding before any modeling.
2. Construct the archetype variable from the NICM taxonomy and the entry-TRL covariate from TechPort.
3. Estimate the nonparametric cumulative incidence functions for each cause within each archetype stratum and run Gray's test [13] for equality across strata.
4. Estimate cause-specific Cox models and Fine-Gray subdistribution models [4] for each event, first without and then with the archetype interaction.
5. Test H1 by the significance and sign of the archetype-by-instrument-side interaction in the instrument-slip subdistribution model, and by the contrast of dominant hazards across strata.
6. Run the pre-registered sensitivity analyses: ambiguous-event recoding, threshold variation, penalty variation, and exclusion of the estimating-optimism control to check robustness.

### 5.2 Expected findings (illustrative, not executed)

The expected form of the result, if H1 holds, is as follows. In the first-of-kind active-sensor stratum, the instrument-slip cumulative incidence function would rise faster and reach a higher plateau than the launch-slip cumulative incidence function, so that the instrument-slip subdistribution hazard ratio on entry-TRL deficit would be the dominant and statistically distinguishable term (illustratively, a subdistribution hazard ratio above one for each level of TRL deficit). In the passive-radiometer heritage stratum, the two cumulative incidence functions would be closer together, with the launch-slip hazard at least as large as the instrument-slip hazard, so the dominance would reverse or vanish. Gray's test would reject equality of the instrument-slip cumulative incidence functions across the two archetype strata. The archetype-by-TRL interaction in the pooled subdistribution model would be statistically distinguishable from zero.

The expected null pattern, if H0 holds, is that the two cumulative incidence functions track each other within both strata, that Gray's test fails to reject, and that the archetype interaction is indistinguishable from zero. A third possible pattern, separable events but no archetype dependence, would reject part of H1 while confirming that the two slip causes are distinct risks; this would be reported as partial support and would still carry the reserve-allocation implication that slip should be modeled as two hazards.

### 5.3 Pre-registered robustness and sensitivity analyses

Five robustness analyses are fixed in advance so that the result cannot be tuned after the fact.

First, the slip threshold is varied from one month to four months of net launch-date movement; the dominance pattern must be stable across this range to count as confirmed. Second, ambiguous events whose cause cannot be reconciled across CADRe and GAO are recoded both ways, and the result is reported under both codings; a conclusion that flips between codings is reported as fragile. Third, the ridge penalty is varied and the events-per-variable cap is tightened and loosened to show the estimate is not an artifact of one regularization choice. Fourth, the estimating-optimism control is removed and reintroduced to isolate whether the separation depends on it. Fifth, the third archetype category (mixed or ambiguous sensors) is brought into the model as a separate stratum to confirm that the binary contrast is not forcing a spurious dichotomy onto a continuum.

A power analysis is run before estimation, not after. Given the expected sample size and the expected event split, the design computes the minimum detectable subdistribution hazard ratio for the archetype interaction at conventional significance and power, and reports it. If the minimum detectable effect is larger than the effect the spacecraft literature would consider practically meaningful, the dissertation states in advance that a non-rejection of H0 will be inconclusive rather than confirmatory, which protects against over-claiming from a small sample.

### 5.4 What would count as confirmation versus falsification

Confirmation requires all three of: separability (Gray's test rejects, or the cause-specific hazards differ), archetype dependence (the interaction is distinguishable from zero with the predicted sign), and robustness (the pattern survives the optimism control and the ambiguous-event recoding). Falsification follows from a failure of any one of these, with the small-sample caveat that a non-rejection accompanied by wide intervals is reported as inconclusive rather than as support for H0.

---

## 6. Discussion

### 6.1 Implications

If the contribution is confirmed, the operational implication is specific. Schedule and cost reserve at confirmation should be steered by archetype. First-of-kind active-sensor missions, which are disproportionately led by JPL, should hold reserve and gating against instrument maturation, with entry-TRL at KDP-B treated as the leading indicator. Heritage passive-radiometer continuity missions should hold reserve against launch-manifest and provider dynamics, because their dominant first-slip hazard is launch-driven. This is a sharper prescription than the current practice of holding undifferentiated reserve against a pooled slip estimate, and it follows directly from the archetype-specific cumulative incidence functions.

### 6.2 Rival explanations

The leading rival explanation is estimating optimism. Flyvbjerg's account holds that overruns are a function of systematic optimism in the planning-stage estimate rather than of project-specific technical causes [7], reinforced by the demand-forecast-inaccuracy evidence [8]. If optimism is the true driver, the apparent instrument-launch separation could be an artifact of which missions were optimistically estimated. The design controls for this with the reference-class optimism proxy and tests robustness by removing it; if the separation depends on the control, the rival explanation wins. A second rival is that "first-of-kind active sensor" is merely a proxy for general mission complexity, in which case the complexity index after Bearden [3] should absorb the archetype effect; the design includes the complexity control precisely to rule this out. A third rival is that the cause coding manufactures the separation; the two-source reconciliation and recoding sensitivity address this.

A fourth rival explanation deserves explicit treatment: reverse causation through the gating process itself. NASA gates technology maturity at milestones, so a mission that anticipates instrument risk may delay its own confirmation, lower its committed schedule ambition, or descope the instrument, any of which would suppress the very instrument-driven slip the model is trying to measure. This selection on the dependent variable would bias the instrument-slip hazard downward and could mask a true dominance. The design's defense is to measure entry TRL at the moment of KDP-B, before the slip window opens, and to include the descope history as a covariate where the record supports it, so that anticipatory descoping is observed rather than hidden. A fifth rival is era confounding: launch-market conditions changed substantially across the studied period, so a cohort weighted toward one era could show a launch-side dominance that reflects the era rather than the archetype. The calendar-period fixed effects absorb common era shocks, and the archetype contrast is estimated within era to the extent the sample permits.

### 6.3 External validity

The result, if found, is bounded to NASA Earth-observing missions of the studied era. It does not claim to describe commercial Earth constellations, which face different launch economics, nor planetary or astrophysics missions, whose instrument and launch risk structures differ. The archetype contrast is specific to the active-passive sensor axis and should not be generalized to novelty in other subsystems without re-estimation.

### 6.4 What would falsify the contribution

The contribution is falsified if the two slip causes cannot be statistically separated, or if separable, the archetype does not modify the dominance in the predicted direction, or if the pattern is an artifact of estimating optimism or of complexity. Each of these is a pre-specified, testable outcome of the analysis plan, which is what makes the contribution falsifiable rather than merely plausible.

---

## 7. Contribution and Conclusion

This dissertation reframes NASA Earth-mission schedule slip from a single aggregated hazard into two competing risks with distinct owners, drivers, and levers, and it tests whether the dominance between them depends on sensor archetype. The methodological contribution is the first application of the Fine-Gray competing-risks apparatus to NASA mission schedule slip, with sensor archetype entering as an explicit effect modifier in the spirit of the Callaway-Sant'Anna refusal to pool heterogeneous effects, and with the cumulative incidence function read as a Fogel-style structural counterfactual. The substantive contribution, stated as a single falsifiable claim, is that instrument-driven and launch-driven slip are separable competing risks and that instrument-driven slip dominates for first-of-kind active-sensor missions but not for passive-radiometer heritage missions. The work is presented honestly at the design stage: the cohort, the estimator, the identification strategy, and the analysis plan are complete, and the reported results are labeled as illustrative and not yet executed. If the plan is executed and the contribution confirmed, NASA and JPL gain a defensible, archetype-specific basis for allocating the scarce schedule reserve that ultimately determines whether an Earth-observing mission overruns. If the plan is executed and the null holds, the field learns that slip is one hazard after all, which is itself a result worth having.

---

## References

[1] M. Lieber and B. Donor, "Schedule matters: Understanding the relationship between schedule delays and costs on overruns," *2016 IEEE Aerospace Conference*, 2016. doi: [10.1109/AERO.2016.7500722](https://doi.org/10.1109/aero.2016.7500722).

[2] G. F. Dubos, J. H. Saleh, and R. Braun, "Technology Readiness Level, Schedule Risk, and Slippage in Spacecraft Design," *Journal of Spacecraft and Rockets*, 2008. doi: [10.2514/1.34947](https://doi.org/10.2514/1.34947). Conference antecedent: doi: [10.2514/6.2007-6020](https://doi.org/10.2514/6.2007-6020).

[3] D. A. Bearden, "Complexity based cost estimating relationships for space systems," *2004 IEEE Aerospace Conference*, 2004. doi: [10.1109/AERO.2004.1368229](https://doi.org/10.1109/aero.2004.1368229).

[4] J. P. Fine and R. J. Gray, "A Proportional Hazards Model for the Subdistribution of a Competing Risk," *Journal of the American Statistical Association*, vol. 94, no. 446, 1999. doi: [10.1080/01621459.1999.10474144](https://doi.org/10.1080/01621459.1999.10474144).

[5] P. C. Austin, D. S. Lee, and J. P. Fine, "Introduction to the Analysis of Survival Data in the Presence of Competing Risks," *Circulation*, vol. 133, 2016. doi: [10.1161/CIRCULATIONAHA.115.017719](https://doi.org/10.1161/circulationaha.115.017719).

[6] P. C. Austin and J. P. Fine, "Practical recommendations for reporting Fine-Gray model analyses for competing risk data," *Statistics in Medicine*, vol. 36, no. 27, 2017. doi: [10.1002/sim.7501](https://doi.org/10.1002/sim.7501).

[7] B. Flyvbjerg, "Five things you should know about cost overrun," *Transportation Research Part A: Policy and Practice*, vol. 118, 2018. doi: [10.1016/j.tra.2018.07.013](https://doi.org/10.1016/j.tra.2018.07.013).

[8] B. Flyvbjerg, M. K. Skamris Holm, and S. L. Buhl, "How (In)accurate Are Demand Forecasts in Public Works Projects? The Case of Transportation," *Journal of the American Planning Association*, vol. 71, no. 2, 2005. doi: [10.1080/01944360508976688](https://doi.org/10.1080/01944360508976688).

[9] H. Habib-agahi et al., "NASA Instrument Cost Model for Explorer-Like Mission Instruments (NICM-E)," NASA Technical Reports Server, 2015. URL: [https://ntrs.nasa.gov/citations/20150008233](https://ntrs.nasa.gov/citations/20150008233).

[10] H. Habib-agahi et al., "NASA Space Flight Instruments: Cost-Time Trends," NASA Technical Reports Server, 2019. URL: [https://ntrs.nasa.gov/citations/20190026811](https://ntrs.nasa.gov/citations/20190026811).

[11] J. R. Irons, J. L. Dwyer, and J. A. Barsi, "The next Landsat satellite: The Landsat Data Continuity Mission," *Remote Sensing of Environment*, vol. 122, 2012. doi: [10.1016/j.rse.2011.08.026](https://doi.org/10.1016/j.rse.2011.08.026).

[12] P. K. Andersen, R. B. Geskus, T. de Witte, and H. Putter, "Competing risks in epidemiology: possibilities and pitfalls," *International Journal of Epidemiology*, vol. 41, no. 3, 2012. doi: [10.1093/ije/dyr213](https://doi.org/10.1093/ije/dyr213).

[13] V. Coviello and M. Boggess, "Cumulative Incidence Estimation in the Presence of Competing Risks," *The Stata Journal*, vol. 4, no. 2, 2004 (implementing the Gray cumulative-incidence and K-sample comparison methods). doi: [10.1177/1536867X0400400201](https://doi.org/10.1177/1536867x0400400201).

[14] T. Leunig, "Time is Money: A Re-Assessment of the Passenger Social Savings from Victorian British Railways," *Journal of Economic History*, vol. 66, no. 3, 2006 (illustrating the Fogel social-savings decomposition method). doi: [10.1017/S0022050706000283](https://doi.org/10.1017/s0022050706000283).

[15] B. Callaway and P. H. C. Sant'Anna, "Difference-in-Differences with multiple time periods," *Journal of Econometrics*, vol. 225, no. 2, 2021. doi: [10.1016/j.jeconom.2020.12.001](https://doi.org/10.1016/j.jeconom.2020.12.001).

[16] P. H. C. Sant'Anna and J. Zhao, "Doubly robust difference-in-differences estimators," *Journal of Econometrics*, vol. 219, no. 1, 2020. doi: [10.1016/j.jeconom.2020.06.003](https://doi.org/10.1016/j.jeconom.2020.06.003).