# Science Productivity of Earth-Observation Data Policy: A Difference-in-Differences Study of Open-Data Release on Mission Citation Yield

**Candidate:** JPL_ASTRO_EARTH_09
**Program:** COLLEGIUM 1st Battalion
**North Star / JPL category:** Earth Science Missions
**Methodological anchors:** Douglass North (new institutional economics); Simon Kuznets (measurement discipline)
**Date:** 2026-06-15

---

## Abstract

NASA Earth-science missions differ in when, and whether, they made their data freely and openly available. Some missions adopted free-and-open release early, some late, and some operated under restricted or fee-based access for extended periods. This dissertation asks whether the transition to free-and-open data release produced a measurable upward break in a mission's downstream scientific output, measured as the rate of peer-reviewed publications that use the mission's data and the rate of citations to the mission's datasets. The study frames open-data release as an institutional rule change in the sense of North: a change in the rules of access that lowers the transaction cost of obtaining and reusing mission data and should, if the framework holds, raise the volume of downstream use. It treats the citation and publication counts in the spirit of Kuznets and the national-accounts tradition: as constructed proxies for the latent quantity of scientific productivity, whose error structure and possible relabeling must be stated before inference. The design is a staggered-adoption difference-in-differences event study built around each mission's open-data-policy adoption date, with restricted-access missions matched to open-access missions on sensor class and mission age and with heterogeneity-robust estimators. The named data are NASA Astrophysics Data System and Web of Science publication and citation counts keyed to Earth missions, NASA Earthdata and Distributed Active Archive Center data-access logs, and a hand-coded register of mission open-data-policy adoption dates. The falsifiable contribution is stated as a single hypothesis pair. H0: open-data adoption has no effect on downstream publication or dataset-citation yield. H1: open-data adoption produces a positive upward break in those yields relative to matched restricted-access missions. This document presents the full design and a pre-registered analysis plan. Reported numerical results are labeled as illustrative and design-stage; they have not been executed on the full assembled dataset.

---

## 1. Introduction and Contribution

### 1.1 The problem

NASA invests in Earth-science missions to generate knowledge about the Earth system. The scientific return on that investment is realized only when researchers obtain the data, analyze them, and publish findings that others build on. Data-access policy sits directly on the path between a mission and its scientific return. A mission whose data are free, openly licensed, and easy to obtain imposes a low cost on a prospective user. A mission whose data are restricted, fee-based, or encumbered by registration and approval imposes a higher cost. The policy question is whether lowering that cost actually raises the downstream scientific yield of a mission, and by how much.

The question is not merely academic. NASA's Science Mission Directorate has moved deliberately toward free-and-open release, most recently through its Scientific Information Policy, and reports annual metrics on the publications, data, and software associated with its funded research. If open release measurably increases the publication and dataset-citation yield of a mission, that finding supports the policy and helps quantify a benefit that is otherwise asserted rather than estimated. If open release has no measurable effect once confounders are controlled, that finding is equally informative: it would indicate that the binding constraint on scientific yield lies elsewhere, for example in funding for analysis, in data-product maturity, or in the size of the relevant research community.

### 1.2 The gap in the literature

There is a large literature on the open-access citation advantage for journal articles and a smaller literature on the open-data citation advantage for datasets. Piwowar and Vision [1] provide a foundational estimate that studies sharing their data receive more citations. Colavizza and colleagues [2] find that articles linking to deposited data accrue more citations across a corpus of over half a million articles. Eysenbach [3] and Gargouri and colleagues [4] document access-related citation differentials, with the latter using a mandate to reduce author self-selection. A systematic review by Langham-Putrow and colleagues [5] concludes that a majority of studies report an advantage but that selection effects and field heterogeneity make the magnitude contested.

Two gaps remain. First, almost all of this work is at the level of the individual article or dataset, not at the level of a mission as a producer of data. The unit that NASA and JPL actually fund and operate is the mission, and the policy lever that NASA actually pulls is mission-level or directorate-level data policy. There is little quantitative work that treats the mission as the unit of analysis and the mission's open-data adoption as the treatment. Second, much of the existing work is associational and vulnerable to selection: better research, or research by better-resourced groups, may be both more likely to be open and more likely to be cited. The staggered timing of open-data adoption across NASA Earth missions provides a natural experiment that has not been exploited with modern difference-in-differences methods.

The case for the natural experiment is strongest in remote sensing, where the Landsat program's 2008 free-and-open policy change is well documented. Wulder and colleagues [7] show that the policy change was followed by a rise in scene distribution from tens of thousands to tens of millions per year and a sharp expansion in publications and operational products, and Zhu and colleagues [8] corroborate the program-level effect. The Landsat episode demonstrates that mission-level open-data policy can be associated with a large change in usage. It does not, by itself, identify a causal effect, because it is a single mission observed before and after a single date, with no contemporaneous control. The contribution of this dissertation is to embed the Landsat-type episode in a multi-mission, matched, staggered-adoption design that can separate the policy effect from secular trends and mission-specific characteristics.

### 1.3 The single falsifiable contribution

The dissertation makes one falsifiable claim, stated as a hypothesis pair on the mission as the unit of analysis:

- **H0 (null):** The transition to free-and-open data release has no effect on a mission's downstream peer-reviewed publication rate or its dataset-citation rate, relative to matched restricted-access missions.
- **H1 (alternative):** The transition to free-and-open data release produces a measurable upward break, an increase in level or slope, in a mission's downstream peer-reviewed publication rate and dataset-citation rate, relative to matched restricted-access missions, in the periods after adoption.

The claim is falsifiable in both directions. The design produces an event-study coefficient path. If the post-adoption coefficients are statistically indistinguishable from zero and the pre-adoption coefficients are flat, H0 is not rejected and the contribution is falsified. If the post-adoption coefficients are positive and increasing while the pre-adoption coefficients are flat, H1 is supported. The decision rule and the falsification conditions are specified in Section 5 before any estimation.

### 1.4 Why it matters for NASA and JPL

NASA and JPL make recurring decisions about how to release mission data and how much to invest in the data systems that support open release. Those decisions have costs: building and operating a Distributed Active Archive Center, curating products to a reusable standard, and maintaining access infrastructure are not free. A defensible estimate of the downstream scientific yield of open release lets those costs be weighed against a measured benefit rather than an asserted one. The estimate also informs the design of future missions in the Earth Science portfolio, where the open-data decision can be made deliberately at formulation rather than retrofitted. Finally, the result speaks to the directorate-level Scientific Information Policy by providing mission-level evidence on the mechanism that policy is intended to activate.

---

## 2. Background and Literature

### 2.1 The institutional lens: North

Douglass North's framework treats institutions as the humanly devised rules of the game that structure incentives and lower the transaction costs of exchange across distance and time [25]. In North's account, the move from personal to impersonal exchange multiplies the costs of measuring what is exchanged and of enforcing agreements, and institutions exist to lower those costs. An open-data policy is an institution in exactly this sense. It is a rule about access that changes the cost a prospective user faces when trying to obtain, verify, and reuse a mission's data. Under restricted access, a user must locate the data, establish eligibility, negotiate or pay for access, and accept license terms. Each of those steps is a transaction cost. Under free-and-open release with persistent identifiers and standard licensing, those costs fall toward the cost of downloading a file.

The North lens yields a sharp, testable prediction and a discipline for stating it. The prediction is that lowering the transaction cost of access should raise the volume of impersonal use, which in this setting is downstream publication and dataset citation by researchers who were not part of the mission team. The discipline, drawn directly from North's own questions, is to identify what transaction cost the policy actually lowered and to check that the institution is justified because it lowers costs relative to the status quo, not merely because it is labeled open. This matters for construct validity. A policy that is nominally open but functionally encumbered, for example open in license but practically inaccessible because of missing tooling, has not lowered the relevant cost and should not be coded as treated. The FAIR principles [9] provide the operational vocabulary for this distinction, separating nominal accessibility from functional findability, interoperability, and reusability.

North also supplies a caution about path dependence. Missions and their user communities form around an inherited institutional matrix. A community that grew up under restricted access may have built workflows, intermediaries, and expectations that persist after the rule changes, so the response to open release may be gradual rather than immediate. The design accommodates this by estimating a dynamic event-study path rather than a single before-after difference, allowing the effect to build over post-adoption periods.

### 2.2 The measurement lens: Kuznets

Simon Kuznets built the architecture of national income accounting and insisted throughout that an aggregate is a constructed proxy, not the latent quantity it stands for, and that its error structure must be made explicit. The Kuznets lineage supplies three disciplines for this study.

First, the outcome is a proxy. A citation count is not scientific productivity; it is a measurable indicator that stands in for it, with a known and biased error structure. Landefeld and colleagues [24] make the general point for national accounts: any single aggregate is a constructed indicator whose construction shapes inference. Henderson, Storeygard, and Weil [22] make the applied point in a satellite setting, treating night-lights as a measurable proxy for economic growth and developing a framework that combines the proxy with noisy income measures rather than asserting the proxy is the truth. The study adopts this stance for citation and publication counts and states the proxy's known biases: citation counts are right-skewed, accumulate with a lag, depend on field size and citation norms, and are sensitive to indexing coverage in the source database.

Second, distinguish a real change in output from a relabeling. McMillan and Rodrik [23] decompose productivity growth into a within-sector component and a reallocation component and show that apparent transformation can be mere relabeling that does not raise measured aggregate output. The analog here is precise. After a mission goes open, the apparent rise in mission-linked publications could reflect a real increase in new research, or it could reflect a change in how existing research is attributed and indexed, for example more authors now naming the mission or its dataset because citation became easier and was encouraged by data-citation norms [10, 11]. The design must separate genuine new output from improved counting, and Section 4 specifies the variable construction that attempts to do so.

Third, beware fragile aggregate relationships. The history of the environmental Kuznets curve, in which an aggregate empirical relationship that looked robust proved sensitive to omitted variables, trends, and heterogeneity once reexamined, is a standing caution in this lineage. The lesson for this study is to subject the headline estimate to the same scrutiny: report sensitivity to specification, to the comparison group, and to the proxy definition, and treat a single point estimate as provisional until it survives that scrutiny.

### 2.3 The open-data and open-access evidence base

The empirical literature provides both the motivating prior and the principal threat. Piwowar and Vision [1], Colavizza and colleagues [2], Eysenbach [3], Gargouri and colleagues [4], and McKiernan and colleagues [6] together establish that openness is associated with more downstream citation, and that at least part of the association survives attempts to remove author self-selection. Langham-Putrow and colleagues [5] establish the counterweight: the magnitude is contested, selection is pervasive, and field heterogeneity is large. The design responds to this evidence base by moving from the article level to the mission level, where the treatment is a policy event rather than an author choice, and by using a matched, staggered design that does not rely on the assumption that open and restricted missions are otherwise identical.

### 2.4 The methodological toolkit

The study uses the modern difference-in-differences literature for staggered adoption. Callaway and Sant'Anna [12] provide the primary estimator, defining group-time average treatment effects that are robust to treatment-effect heterogeneity. Goodman-Bacon [13] shows why the naive two-way fixed-effects estimator is unreliable in staggered settings, because it uses already-treated units as controls and can place perverse weights on comparisons. Sun and Abraham [14] and de Chaisemartin and D'Haultfoeuille [18] make the same point for event-study coefficients and propose robust alternatives. Borusyak, Jaravel, and Spiess [15] provide an efficient imputation estimator used for robustness. Roth [16] and Rambachan and Roth [17] supply the parallel-trends diagnostics and the sensitivity analysis that quantify how large a trend violation would overturn the result. Rosenbaum and Rubin [19], Stuart [20], and Austin [21] supply the matching and balance-diagnostic machinery for constructing the comparison group. This toolkit is what allows the Landsat-type episode to be generalized into an identified, multi-mission estimate.

---

## 3. Data

### 3.1 Named datasets and sources

The study combines four data sources, all of which are real and accessible.

**Mission-linked publication and citation counts.** The NASA Astrophysics Data System (ADS), described by Kurtz and colleagues [26], indexes the astronomy, astrophysics, and Earth-science literature and supports reference and citation linking. Web of Science provides a second, independent bibliographic source with its own citation index and field-normalization tooling. Publication and citation counts keyed to each Earth mission are drawn from both, and cross-checked, so that the outcome does not depend on the indexing idiosyncrasies of a single database. The dual-source design directly addresses a construct-validity threat: if a result appears in ADS but not in Web of Science, it is more likely an indexing artifact than a real change in output.

**Data-access logs.** NASA Earthdata and the Distributed Active Archive Centers (DAACs) record data-access and distribution events. These logs provide a usage measure that is upstream of publication: downloads and distribution volumes. They serve two roles. As an outcome, distribution volume is a more immediate response to a policy change than publication, which lags by years. As a mechanism check, a rise in distribution that precedes the rise in publication is consistent with the North access-cost mechanism, while a rise in publication without a rise in distribution would suggest a counting or attribution change rather than new use.

**Open-data-policy adoption dates.** A hand-coded register records, for each mission, the date on which its data transitioned to free-and-open release, the prior access regime, and the licensing and tooling status before and after. Sources for the register include mission documentation, DAAC policy records, the NASA Plan for Increasing Access to the Results of Scientific Research, and the Science Mission Directorate Scientific Information Policy and its annual implementation metrics. Coding is double-entered and adjudicated, and ambiguous or phased transitions are flagged and handled in robustness checks.

### 3.2 Unit of analysis and panel structure

The unit of analysis is the mission-period. Each Earth mission is observed over a sequence of periods, with periods defined as years (or, in a robustness specification, quarters). The panel is unbalanced because missions enter and exit the observation window at different times and because some missions never adopt open release within the window. The treatment is the open-data-policy adoption event; treated missions adopt at different periods (staggered timing), and never-adopters and not-yet-adopters supply comparison information.

### 3.3 Variable construction

The two outcome variables are constructed to separate real output change from relabeling, following the Kuznets discipline.

- **Publication rate.** The count, per mission-period, of peer-reviewed articles that use the mission's data, identified by a combination of mission and instrument name matching, dataset-identifier matching, and acknowledgment-text matching, with the matching rule held fixed across the pre- and post-adoption periods so that a change in counting practice cannot masquerade as a change in output. A supplementary specification restricts to articles by authors with no prior affiliation to the mission team, isolating impersonal use in North's sense.
- **Dataset-citation rate.** The count, per mission-period, of formal citations to the mission's datasets as first-class objects, following the Data Citation Principles [10] and the repository roadmap [11]. Because formal data citation became common only recently, this outcome is treated as available for the later part of the window and is analyzed separately rather than pooled with publication counts.
- **Distribution volume.** Access and download events from Earthdata and DAAC logs, per mission-period, used as an early-response outcome and as a mechanism check.

The matching covariates are sensor class (for example optical imager, synthetic aperture radar, lidar, passive microwave, spectrometer) and mission age at each period. Additional controls considered include the size of the relevant research community, mission data-product maturity level, and agency or international partner.

### 3.4 Coverage and limitations

Coverage is the set of NASA Earth-science missions for which a clear access regime and adoption date can be coded and for which bibliographic linkage is feasible. The principal limitations are five. First, indexing coverage in ADS and Web of Science is incomplete and changes over time, which is why two sources are used and why the matching rule is frozen across periods. Second, the adoption date is sometimes a phased process rather than a clean event, which is handled by flagging phased transitions and testing sensitivity to alternative date definitions. Third, formal data citation is sparse in the early window, which restricts the dataset-citation analysis to recent periods and reduces its statistical power. Fourth, the number of distinct Earth-science missions with codable regimes is modest, on the order of tens rather than hundreds, so the design is more vulnerable to small-sample inference problems than article-level studies, and the matching pool for some sensor classes may be thin. Fifth, bibliographic linkage of a publication to a specific mission is itself imperfect, because authors name missions, instruments, and datasets inconsistently, so the linkage rule introduces measurement error that, if it changes over time, could be confounded with the treatment; freezing the rule across periods is the primary defense, and the no-prior-affiliation specification is a secondary one. These limitations are stated here so that the analysis plan in Section 5 can address each one explicitly, and so that the strength of any claim is bounded by the weakest of them rather than the strongest.

---

## 4. Research Design and Identification

### 4.1 Estimator

The primary estimator is the Callaway and Sant'Anna group-time average treatment effect [12]. For each adoption cohort g (the set of missions that adopted open release in period g) and each period t, the estimator computes the average treatment effect on the treated, ATT(g, t), by comparing the change in outcome for cohort g from just before adoption to period t against the change over the same calendar periods for a comparison group of not-yet-treated and never-treated missions. The ATT(g, t) values are then aggregated into a dynamic event-study path indexed by event time (periods since adoption) and into an overall average effect. This estimator is chosen because the adoption timing is staggered and because treatment effects are expected to be heterogeneous across missions and to grow over event time, exactly the conditions under which Goodman-Bacon [13] shows the naive two-way fixed-effects estimator to be unreliable.

### 4.2 Specification

The dynamic event-study specification estimates a coefficient for each event-time period relative to the last pre-adoption period, which is normalized to zero. The leads (pre-adoption coefficients) test for differential pre-trends; the lags (post-adoption coefficients) trace the dynamic effect. The outcome is modeled in a form appropriate to a non-negative count, using a count-data or log-transformed specification with attention to the zero counts that are common for small or young missions. Standard errors are clustered at the mission level. Robustness specifications re-estimate the path with the Sun and Abraham interaction-weighted estimator [14], the Borusyak, Jaravel, and Spiess imputation estimator [15], and the de Chaisemartin and D'Haultfoeuille estimator [18], so the headline result does not depend on a single method.

### 4.3 Identification strategy

Identification rests on a parallel-trends assumption defined within matched strata. The assumption is that, absent the open-data policy, the publication and citation trajectories of treated missions would have evolved in parallel to those of their matched comparison missions. Matching on sensor class and mission age, using the propensity-score framework of Rosenbaum and Rubin [19] and the matching guidance of Stuart [20], is what makes the assumption plausible: an optical imager that went open is compared with optical imagers of similar age that did not, rather than with the full pool of missions. Covariate balance after matching is reported using standardized differences and the diagnostics of Austin [21]. The staggered timing is an asset because it provides comparison information from not-yet-treated missions at every adoption period, reducing reliance on never-treated missions alone.

### 4.4 Threats to validity

**Internal validity.** The central threat is a violation of parallel trends: missions may adopt open release precisely when their scientific output is already rising, for example as a mission matures and its data products stabilize. Three defenses are deployed. The leads in the event study test directly for pre-adoption divergence; under H1 with valid identification, the leads should be flat. The Roth diagnostics [16] assess the power and reliability of the pre-trend test. The Rambachan and Roth sensitivity analysis [17] reports how large a post-adoption differential trend, expressed relative to the observed pre-trends, would be required to overturn the estimated effect, so the result is reported as a robustness region rather than a single point. Mission age is both a matching covariate and an explicit control to absorb the maturation confounder.

**External validity.** The estimate applies to NASA Earth-science missions in the observation window and may not transfer to other agencies, to non-Earth missions, or to future missions in a changed publishing environment. The Landsat case [7, 8] suggests the mechanism is strong in optical remote sensing, but the magnitude may differ for radar, microwave, or in-situ-heavy missions. Heterogeneity of the effect across sensor classes is estimated and reported rather than averaged away.

**Construct validity.** The outcome is a proxy. The Kuznets discipline requires separating real new output from improved counting, which is why the matching rule is frozen across periods, why a no-prior-affiliation specification isolates impersonal use, and why distribution logs are used as an independent mechanism check. The treatment is also a construct: nominal openness is distinguished from functional openness using the FAIR vocabulary [9], and missions coded as open but functionally encumbered are reclassified in a robustness check.

**Statistical-conclusion validity.** Counts are right-skewed and over-dispersed, the number of missions is modest, and clustering is at the mission level, so conventional asymptotics may be unreliable. Wild-cluster bootstrap inference and the heterogeneity-robust estimators above are used, and the dataset-citation analysis is reported with explicit power caveats because formal data citation is sparse early in the window.

---

## 5. Analysis Plan and Findings

This section is a design-stage analysis plan. The numerical values below are illustrative and are clearly labeled as such. They have not been computed on the full assembled dataset. They are included to specify the form of the expected output and the decision rule, not to report results.

### 5.1 Estimation procedure

1. Assemble the mission-period panel by linking ADS and Web of Science publication and citation records to each mission, merging Earthdata and DAAC distribution logs, and attaching the hand-coded adoption date and access regime.
2. Construct the three outcomes and the matching covariates with the frozen matching rule described in Section 3.3.
3. Match restricted-access (comparison) missions to open-access (treated) missions on sensor class and mission age using the propensity-score and matching procedures of Rosenbaum and Rubin [19] and Stuart [20], and report covariate balance using Austin's diagnostics [21]. Proceed only if balance is acceptable; otherwise revise the matching specification.
4. Estimate the dynamic event-study path with the Callaway and Sant'Anna estimator [12], normalizing the last pre-adoption period to zero.
5. Test the leads for pre-trends; apply the Roth power diagnostics [16].
6. Re-estimate with the Sun-Abraham [14], Borusyak-Jaravel-Spiess [15], and de Chaisemartin-D'Haultfoeuille [18] estimators as robustness.
7. Apply the Rambachan-Roth sensitivity analysis [17] and report the robustness region.
8. Estimate effect heterogeneity by sensor class and report it.
9. Repeat the core estimation for the distribution-volume outcome as an early-response and mechanism check, and for the dataset-citation outcome with explicit power caveats.

### 5.2 Decision rule and falsification conditions

The hypothesis test is read off the event-study path under the pre-registered rule:

- **Support for H1** requires three conditions jointly: the pre-adoption leads are jointly indistinguishable from zero (no detectable pre-trend); the post-adoption lags are positive and the aggregated overall effect is positive and statistically distinguishable from zero; and the positive effect survives the Rambachan-Roth sensitivity analysis at a stated breakdown threshold.
- **Failure to reject H0** occurs if the aggregated overall effect is not statistically distinguishable from zero, or if the post-adoption effect does not survive the sensitivity analysis. Either outcome falsifies the contribution as stated.
- **A specific falsifier:** if the pre-adoption leads are themselves positive and trending, the parallel-trends assumption is violated, the design does not identify the effect, and no causal claim is made regardless of the sign of the lags. This condition is checked first and can halt the causal interpretation before the lags are even examined.

### 5.3 Illustrative (not-yet-executed) expected output

The following is the expected shape of a result if H1 holds. It is illustrative only and uses placeholder numbers.

- Pre-adoption leads (event time minus 4 to minus 1): coefficients near zero, confidence intervals spanning zero, consistent with no pre-trend. [ILLUSTRATIVE]
- Post-adoption lags (event time 0 to plus 5): a coefficient path rising from a small effect in the adoption period to a larger effect several periods later, consistent with North's path-dependent, gradual adjustment. For example, an illustrative increase in the annual mission-linked publication rate of roughly one quarter to one half relative to the matched comparison path by event time plus 4. [ILLUSTRATIVE, NOT ESTIMATED]
- Distribution volume: an earlier and sharper break than publications, consistent with the access-cost mechanism preceding the publication response. [ILLUSTRATIVE]
- Sensor-class heterogeneity: a larger effect for optical imagers, consistent with the documented Landsat episode [7], and a smaller or noisier effect for sensor classes with smaller user communities. [ILLUSTRATIVE]

If, instead, the leads are flat and the lags are also flat and centered on zero, the table would show a null, H0 would not be rejected, and the contribution would be falsified. The design is built so that both outcomes are reportable and neither is foreclosed by the specification.

### 5.4 Pre-registration and reproducibility

To prevent the specification search that the Stern critique warns against, the design parameters in Sections 3, 4, and 5.1 through 5.2 are fixed before estimation: the outcome definitions, the matching covariates and procedure, the primary and robustness estimators, the event-time window, the clustering level, and the decision rule. The set of robustness estimators is specified in advance rather than chosen after seeing the primary result, so that agreement across estimators is informative rather than a product of selection. The analysis code, the hand-coded adoption-date register with its sources, and the linkage rules are intended to be released so that the panel can be reconstructed and the event-study path reproduced. Reproducibility is itself an application of the open-data principle the study examines: the evidence for or against H1 should be as accessible and reusable as the mission data whose openness is under test.

---

## 6. Discussion

### 6.1 Implications

If H1 is supported, the study provides a mission-level, identified estimate of the scientific yield of open-data release, which is directly usable in NASA and JPL decisions about data policy and data-system investment. It would convert an asserted benefit into a measured one and let the cost of open-release infrastructure be weighed against a quantified return. It would also locate the effect in time, showing how many periods are needed for the response to materialize, which matters for evaluating recent policy changes that have not yet had time to show their full effect. If H0 is not rejected, the implication is that data-access cost is not the binding constraint on a mission's scientific yield within this sample, which would redirect attention to funding for analysis, product maturity, or community size, each of which is separately actionable.

### 6.2 Rival explanations

Several rival explanations must be addressed before any causal reading. Maturation: output may rise because missions mature, not because they go open; mission age as a matching covariate and control, plus the flat-leads requirement, address this. Selection on trends: missions may go open when output is already rising; the leads test and the Rambachan-Roth sensitivity analysis address this. Counting change: the apparent rise may be relabeling rather than new research, the McMillan-Rodrik distinction [23]; the frozen matching rule, the no-prior-affiliation specification, and the distribution-log mechanism check address this. Concurrent policy: a directorate-wide policy or a new data system may coincide with a mission's adoption; the matched comparison group, drawn from missions in the same era, absorbs common shocks, and cohort-specific estimates reveal whether effects cluster on a single calendar date in a way that would signal a confound.

### 6.3 External validity

The estimate is internal to NASA Earth-science missions in the window. The Landsat evidence [7, 8] suggests the mechanism generalizes within optical remote sensing and is plausibly strong wherever a large latent user community is held back only by access cost. It is weaker where the constraint is analysis capacity or community size rather than access. The reported sensor-class heterogeneity is the honest boundary of generalization: the study should not claim a uniform effect across all mission types when its own estimates show variation.

### 6.4 What would falsify the contribution

The contribution is falsified by any of the following: an overall effect statistically indistinguishable from zero; an effect that does not survive the Rambachan-Roth sensitivity analysis at the stated threshold; positive and trending pre-adoption leads that void identification; or a publication rise unaccompanied by any distribution rise, which under the Kuznets discipline would indicate relabeling rather than new use. Each of these is checked in the analysis plan, and each is sufficient on its own to overturn H1.

---

## 7. Contribution and Conclusion

This dissertation states and tests a single falsifiable claim: that NASA Earth-science missions which transitioned to free-and-open data release show a measurable upward break in their downstream peer-reviewed publication and dataset-citation rates relative to matched restricted-access missions, against the null that open-data adoption has no such effect. The contribution is methodological and substantive. Methodologically, it moves the open-data-advantage question from the article level, where it is dominated by author self-selection, to the mission level, where the treatment is a policy event amenable to a staggered-adoption difference-in-differences event study with heterogeneity-robust estimators and matched comparison missions. Substantively, it provides NASA and JPL with an identified, mission-level estimate of the scientific return on open release, with the effect located in event time and decomposed by sensor class.

The two anchor methodologists discipline the design throughout. North supplies the mechanism and its testable prediction: an open-data policy is an institutional rule change that lowers the transaction cost of access and should raise impersonal downstream use, with a path-dependent, gradual response. Kuznets supplies the measurement discipline: the citation and publication counts are constructed proxies whose error structure must be stated, and a rise in the proxy must be shown to be real new output rather than improved counting before it is interpreted as productivity. The design honors both by estimating a dynamic path within matched strata, by freezing the counting rule across periods, by using distribution logs as an independent mechanism check, and by reporting the result as a robustness region under explicit sensitivity analysis rather than as a single point.

The document is a complete design and pre-registered analysis plan. The reported numbers are illustrative and have not been executed on the full assembled dataset. The next step is to assemble the panel, run the procedure in Section 5, and report the event-study path and its sensitivity, after which H1 will be either supported or falsified on the stated decision rule.

---

## References

1. Piwowar, H. A., and Vision, T. J. (2013). Data reuse and the open data citation advantage. *PeerJ*, 1, e175. https://doi.org/10.7717/peerj.175

2. Colavizza, G., Hrynaszkiewicz, I., Staden, I., Whitaker, K., and McGillivray, B. (2020). The citation advantage of linking publications to research data. *PLOS ONE*, 15(4), e0230416. https://doi.org/10.1371/journal.pone.0230416

3. Eysenbach, G. (2006). Citation advantage of open access articles. *PLoS Biology*, 4(5), e157. https://doi.org/10.1371/journal.pbio.0040157

4. Gargouri, Y., Hajjem, C., Lariviere, V., Gingras, Y., Carr, L., Brody, T., and Harnad, S. (2010). Self-selected or mandated, open access increases citation impact for higher quality research. *PLOS ONE*, 5(10), e13636. https://doi.org/10.1371/journal.pone.0013636

5. Langham-Putrow, A., Bakker, C., and Riegelman, A. (2021). Is the open access citation advantage real? A systematic review of the citation of open access and subscription-based articles. *PLOS ONE*, 16(6), e0253129. https://doi.org/10.1371/journal.pone.0253129

6. McKiernan, E. C., Bourne, P. E., Brown, C. T., Buck, S., Kenall, A., Lin, J., et al. (2016). How open science helps researchers succeed. *eLife*, 5, e16800. https://doi.org/10.7554/eLife.16800

7. Wulder, M. A., Loveland, T. R., Roy, D. P., Crawford, C. J., Masek, J. G., Woodcock, C. E., et al. (2019). Current status of Landsat program, science, and applications (and the benefits of the free and open Landsat data policy). *Remote Sensing of Environment*, 225, 127-147. https://doi.org/10.1016/j.rse.2019.02.016

8. Zhu, Z., Wulder, M. A., Roy, D. P., Woodcock, C. E., Hansen, M. C., Radeloff, V. C., et al. (2019). Benefits of the free and open Landsat data policy. *Remote Sensing of Environment*, 224, 382-385. https://doi.org/10.1016/j.rse.2019.02.015

9. Wilkinson, M. D., Dumontier, M., Aalbersberg, I. J., Appleton, G., Axton, M., Baak, A., et al. (2016). The FAIR Guiding Principles for scientific data management and stewardship. *Scientific Data*, 3, 160018. https://doi.org/10.1038/sdata.2016.18

10. Data Citation Synthesis Group (2014). Joint declaration of data citation principles. https://doi.org/10.5281/zenodo.7356758

11. Fenner, M., Crosas, M., Grethe, J. S., Kennedy, D., Hermjakob, H., Rocca-Serra, P., et al. (2019). A data citation roadmap for scholarly data repositories. *Scientific Data*, 6, 28. https://doi.org/10.1038/s41597-019-0031-8

12. Callaway, B., and Sant'Anna, P. H. C. (2021). Difference-in-differences with multiple time periods. *Journal of Econometrics*, 225(2), 200-230. https://doi.org/10.1016/j.jeconom.2020.12.001

13. Goodman-Bacon, A. (2021). Difference-in-differences with variation in treatment timing. *Journal of Econometrics*, 225(2), 254-277. https://doi.org/10.1016/j.jeconom.2021.03.014

14. Sun, L., and Abraham, S. (2021). Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. *Journal of Econometrics*, 225(2), 175-199. https://doi.org/10.1016/j.jeconom.2020.09.006

15. Borusyak, K., Jaravel, X., and Spiess, J. (2024). Revisiting event-study designs: robust and efficient estimation. *Review of Economic Studies*, 91(6), 3253-3285. https://doi.org/10.1093/restud/rdae007

16. Roth, J. (2022). Pre-test with caution: event-study estimates after testing for parallel trends. *American Economic Review: Insights* (and AER 2019 working version). https://doi.org/10.1257/aer.20180609

17. Rambachan, A., and Roth, J. (2023). A more credible approach to parallel trends. *Review of Economic Studies*, 90(5), 2555-2591. https://doi.org/10.1093/restud/rdad018

18. de Chaisemartin, C., and D'Haultfoeuille, X. (2020). Two-way fixed effects estimators with heterogeneous treatment effects. *American Economic Review*, 110(9), 2964-2996. https://doi.org/10.1257/aer.20181169

19. Rosenbaum, P. R., and Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. *Biometrika*, 70(1), 41-55. https://doi.org/10.1093/biomet/70.1.41

20. Stuart, E. A. (2010). Matching methods for causal inference: a review and a look forward. *Statistical Science*, 25(1), 1-21. https://doi.org/10.1214/09-STS313

21. Austin, P. C. (2009). Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. *Statistics in Medicine*, 28(25), 3083-3107. https://doi.org/10.1002/sim.3697

22. Henderson, J. V., Storeygard, A., and Weil, D. N. (2012). Measuring economic growth from outer space. *American Economic Review*, 102(2), 994-1028. https://doi.org/10.1257/aer.102.2.994

23. McMillan, M. S., and Rodrik, D. (2011). Globalization, structural change and productivity growth. *NBER Working Paper No. 17143*. https://doi.org/10.3386/w17143

24. Landefeld, J. S., Seskin, E. P., and Fraumeni, B. M. (2008). Taking the pulse of the economy: measuring GDP. *Journal of Economic Perspectives*, 22(2), 193-216. https://doi.org/10.1257/jep.22.2.193

25. North, D. C. (1990). *Institutions, Institutional Change and Economic Performance.* Cambridge University Press. https://doi.org/10.1017/CBO9780511808678

26. Kurtz, M. J., Eichhorn, G., Accomazzi, A., Grant, C. S., Murray, S. S., and Watson, J. M. (2000). The NASA Astrophysics Data System: Overview. *Astronomy and Astrophysics Supplement Series*, 143(1), 41-59. https://doi.org/10.1051/aas:2000170