Technical Summary: Competitive Satellite Placement and the Geography of Orbital Risk

1. Problem Statement

The paper addresses the uneven spatial distribution of satellites in the geostationary orbit (GEO). While some orbital longitudes are densely crowded with multiple operators, others remain largely empty. This spatial pattern is consequential for three reasons:

1. Resource Allocation: It impacts slot allocation and spectrum coordination under International Telecommunication Union (ITU) rules.
2. Global Coverage: It determines the geographic distribution of communications services.
3. Orbital Sustainability: Satellite placements dictate the future geography of orbital debris. Debris accumulates where satellites are placed; therefore, understanding the drivers of satellite placement is a prerequisite for predicting long-run orbital risk.

Existing literature on orbital debris typically relies on engineering models of post-launch collision, disposal, and end-of-life behavior. This paper argues that a prior, often overlooked question remains: how much of the future geography of congestion and debris is encoded in the initial competitive location choices of operators?

2. Methodology

The authors utilize the geostationary orbit as a "near-ideal laboratory" because it is physically one-dimensional (a ring of 360 degrees) and governed by explicit, observable allocation rules (first-come, first-served).

The Competitive Entry (CE) Model

The core of the analysis is a structural model of sequential, profit-maximizing entry:

• Demand Driver: Demand at each longitude is driven by the gridded population within the satellite's commercial footprint (approx. ±40° longitude). The model tests two demand specifications: raw population vs. GDP-weighted population (income).
• Entry Mechanism: Operators enter the market sequentially. Each new entrant selects the orbital slot that maximizes its market share given the positions of all previously placed satellites.
• Market Sharing Rule: Consumers (population) are served by the nearest satellite. If distances are equal, demand is split randomly.
• Path Dependence: Early entrants secure high-demand arcs. Subsequent entrants face a trade-off: locate in a low-demand, empty region or co-locate in a high-demand region and share the market. The model posits that the marginal gain of sharing a large market often exceeds the gain of capturing a small, empty market entirely, leading to clustering.

Data and Validation

• Data Sources: The study uses the ITU Compliance Assessment Monitor (ITU-CAM) database for satellite positions (390 active commercial satellites post-2006; 513 inactive payloads). Demand data comes from the Gridded Population of the World (GPWv4) and G-Econ (GDP).
• Exclusions: Chinese population is excluded from the demand measure because regulatory restrictions prevent Chinese consumers from accessing commercial GEO services, making them an unaddressable market for international operators.
• Validation Strategy:
  1. In-sample Fit: Comparing the model's predicted density against observed active satellite density ($R^2$).
  2. Out-of-Sample Prediction: A "walk-forward" test predicting individual slot choices for 431 new commercial satellites (post-2000) based on the orbital state at the time of launch, compared against a fitted conditional logit model.
  3. Long-Run Simulation: A 21-year forward simulation (2000–2021) placing 400 new satellites to test aggregate distribution predictions against a naive persistence baseline.
  4. Debris Prediction: Testing if the model predicts the distribution of inactive payloads (debris).
  5. Placebo Tests: Rotating the population distribution to ensure the fit relies on specific east-west alignment, not just lumpy data.

3. Key Results

A. Population Drives Placement, Not Income

The CE model using population as the demand driver achieves a strong fit with observed active satellite distribution ($R^2 = 0.64$, Spearman $\rho = 0.82$). It successfully reproduces dense clusters over South/Southeast Asia, Europe, and the Americas, and sparse coverage over the Pacific and Central Asia.

In contrast, a model using GDP-weighted demand performs significantly worse ($R^2 = 0.11$). This indicates that operators prioritize total population size over purchasing power. The model correctly predicts the high density of satellites over India (a high-population, moderate-income region), whereas the income-weighted model fails to capture this. Robustness checks using an independently estimated income elasticity of technology adoption ($\hat{\gamma} \approx 0.88$) confirm that any positive income weight deteriorates the model's fit; the best fit occurs at $\gamma \approx 0$ (pure population).

B. Predictive Power Outperforms Fitted Models

In walk-forward out-of-sample tests, the structural CE model (with no parameters fitted to slot-choice data) predicts individual slot choices with a mean rank of 140 out of 360. This outperforms a fitted conditional logit model (mean rank 154) and a random baseline (mean rank 180). This suggests the structural model captures the underlying mechanism rather than overfitting historical noise.

C. Long-Run and Debris Prediction

• Aggregate Deployment: In a 21-year forward simulation, the CE model predicts the aggregate distribution of new entries ($R^2 = 0.35$), outperforming a naive persistence benchmark ($R^2 = 0.09$) by a factor of four.
• Debris Accumulation: The model predicts the distribution of inactive payloads (debris) with $R^2 = 0.44$. This demonstrates that the geography of orbital risk is partly a downstream consequence of competitive entry patterns. Regions with high competitive entry (e.g., the India arc) are predicted to accumulate the most debris.

D. Efficiency and Welfare

The paper finds that the current orbital distribution is relatively efficient. While a Pareto-improving reallocation (increasing total welfare without hurting any operator) was possible when the constellation was small ($N=25$, potential gain of 15.8%), this "efficiency window" closed rapidly. By $N \approx 150$ (and certainly at the current $N \approx 500$), no single-satellite reallocation exists that improves aggregate welfare while preserving all operators' market shares.

4. Contributions and Significance

The paper makes three primary contributions:

1. Mechanism Identification: It establishes that the uneven geography of orbital infrastructure is driven by a simple logic of sequential competitive entry under heterogeneous demand, rather than coordination failure or simultaneous optimization.
2. Predictive Framework: It provides a validated, parameter-free structural model that predicts both active satellite placement and the resulting geography of debris accumulation using only population data and physical constraints.
3. Policy Implications:
   • Coverage Equity: The "underserved" regions (e.g., sub-Saharan Africa, Central Asia) are not unprofitable per subscriber but are structurally bypassed because their populations are small relative to global demand peaks. Competitive dynamics will not spontaneously close these gaps.
   • Debris Risk: The regions most at risk for long-run debris accumulation are those that competitive dynamics will continue to target. This creates a self-reinforcing hazard where commercial attractiveness leads to congestion and increased operational risk.
   • LEO Relevance: The authors suggest this mechanism is now scaling to Low Earth Orbit (LEO) mega-constellations, implying that ground tracks will similarly concentrate over high-density population regions.

The paper concludes that the spatial distribution of orbital risk is not merely an engineering problem but a reflection of economic fundamentals and competitive entry patterns. Understanding these drivers allows for better anticipation of future congestion and debris hotspots.