A Physico Economic Model of Low Earth Orbit Management

Environmental and Resource Economics (2020) 77:695–723
https://doi.org/10.1007/s10640-020-00515-z
A Physico‑Economic Model of Low Earth Orbit Management
Sébastien Rouillon1
Accepted: 3 October 2020 / Published online: 17 October 2020

© Springer Nature B.V. 2020
Abstract
We analyze the externality caused by the accumulation of space debris, focusing on the
long-term equilibrium induced by a constant rate of satellite launches. We give conditions
such that the long-term population of functioning satellites is an inverted-U shape function
of the launch rate. We compare typical ways of managing the orbit. The maximum carry-
ing capacity is the maximum population of satellites that the space sector can sustain in the
long run. The physico-economic equilibrium occurs under open-access to the orbit. The
optimal policy maximizes the present value profit of the space sector. Finally, we discuss
the use of standard economic instruments (command-and-control, tax and market) to regu-
late space activity in order to achieve an optimal outcome. A numerical application based
on a realistic calibration illustrates all results.
Keywords Space economics · Orbital debris · Sustainability
Mathematics Subject Classification L1 · L9 · Q2
1 Introduction
Since the launch of Sputnik in 1957, the number of objects in orbit around the Earth, either
operational satellites or space debris, has increased considerably (ESA 2019). Unfortu-
nately, the population of space debris has now reached levels that are beginning to dam-
age the space sector (Weinzierl 2018) and could even compromise its activity (Kessler and
Cour-Palais 1978; Cordelli et al. 1993; Dolado-Perez et al. 2015). Space debris remains
in orbit for long periods of time (i.e., decades or centuries, depending on altitudes) and
poses a threat for existing operational satellites (i.e., risk of damage or breakup by way of
collision). Moreover, the process propagates itself, with past collisions contributing to the
accumulation of space debris.
Electronic supplementary material The online version of this article (https://doi.org/10.1007/s1064

0-020-00515-z) contains supplementary material, which is available to authorized users.
* Sébastien Rouillon
sebastien.rouillon@u‑bordeaux.fr
1
GREThA ‑ UMR 5113, University of Bordeaux, Avenue Léon Duguit, 33 608 Pessac Cedex,
France
13
Vol.:(0123456789)
696 S. Rouillon
In previous decades, the rise of the “New Space” has ushered in an era in which
commercial companies play a major role in space activity (Weinzierl 2018). Revenues
currently exceed $300 billion yearly, with the bulk of it going to the satellite market
(i.e., around $220 billion yearly). In this context, the threat of space debris was soon
recognized as critical by the international community (UN 1999). In 2007, the United
Nations General Assembly endorsed the Space Debris Mitigation Guidelines (UN
2010), recommending to limit the debris released by launch vehicles and spacecrafts,
to avoid intentional destructions, and to minimize the risk of breakups, explosions and
collisions. These measures are not legally binding under international law. Neverthe-
less, they have been formalized into a set of international standards published by the
International Organization for Standardization (Stokes et al. 2017). The degree of com-
pliance in recent years is documented by the European Space Agency’s Annual Space
Environment Report (ESA 2019). Importantly, recent simulations (Drmola and Hubik
2018; Somma et al. 2019) show that, even if these guidelines are followed, space debris
will continue to accumulate in the future unless the satellite launch rate is drastically
reduced.
In this paper, we construct a stylized physico-economic model of space environment
and activity. Following a tradition which has been productive in fishery economics (Gor-
don 1954; Schaefer 1957), we consider the long-term equilibrium of the orbit, induced by
a constant rate of satellite launches. The expected lifetime of a satellite is affected by tech-
nical failures and/or breakups by collisions. In the long run, the risk of collision depends
on past launches, as space activity generates additional debris, either directly (i.e., rocket
bodies, released objects) or indirectly (i.e., debris fragments generated by explosions or
collisions), and collisions become more frequent as the density of orbital debris increases.
Therefore, our model postulates that the long-term rate of satellite breakup by collision can
be formalized as an increasing function of the launch rate.
Within this framework, the expected population of operational satellites in the long-
run equilibrium of the space environment depends on the launch policy planned today.
We derive conditions such that the curve representing the long-term expected population
of active satellites as a function of the launch rate has an inverted-U shape. This occurs
when the long-term risk of collision is an increasing and convex function of the launch
rate, becoming arbitrary large for sufficiently large launch rates. Importantly, these condi-
tions are quite likely according to existing physical models (Kessler and Cour-Palais 1978;
Somma et al. 2019).
Interestingly, our result recalls a well-known finding in the economics of fisheries (Gor-
don 1954; Schaefer 1957), showing that the catch rate in the long-term equilibrium of the
ecosystem is an inverted-U shaped function of the fishing effort. This leads us to look for
several properties related to this literature. We calculate the launch rate which maximizes
the long-term expected population of operational satellites, subsequently referred to as the
orbital maximum carrying capacity. We also derive the physico-economic equilibrium
launch rate, which would presumably emerge under open-access. Typically, it is defined
as the launch rate such that the satellite operators make no profit and, therefore, have no
incentive to change their strategy. Finally, we define the socially optimal launch rate, that
maximizes the expected surplus of the satellite operators.
We compare these three situations to each other. A physical overexploitation of the
orbit occurs if the launch rate in the physico-economic equilibrium is larger than the
one maximizing the long-term expected population of operational satellites. We prove
that this happens if the rent from a functionning satellite is high enough, if the cost of a
launched satellite is small enough and/or if the discount rate is small enough. An economic
13
A Physico‑Economic Model of Low Earth Orbit Management 697
overexploitation happens when the launch rate in the physico-economic equilibrium is

larger than the socially optimal launch rate. We show that this is always the case.
Finally, we show how economic instruments can be designed to implement a social opti-
mum, considering in turn an aggregate launch quota, a tax per launch and a market of
individual transferable launch quotas. We argue that the first instrument is likely to induce
perverse effects if critical characteristics of satellites cannot be monitored. By contrast, the
latter two instruments can implement an optimal plan. The corresponding (pigouvian) tax
level or equilibrium price of a quota then induce the satellite operators to internalize the
external cost of a marginal launch.
The economic analysis dealing with orbital congestion and satellites destruction by col-
lisions is recent and quite sparse. To the best of our knowledge, it is only considered by
Adilov et al. (2014), Adilov et al. (2018), Klima et al. (2016), Macauley (2015), Muller
et al. (2017), Rao and Rondina (2019) and Rao et al. (2020).1
Adilov et al. (2014) propose a model of horizontal differentiation (Salop 1979), where
firms locate satellites around the orbit and consumers have preferences for satellites close
to their location. The game has two periods, with launches taking place in period one and
damages to satellites occurring in period two. This setting allows to analyze private incen-
tives to launch satellites and to mitigate space debris. The model predicts that profit-max-
imizing firms launch too many satellites and choose technologies which create too many
debris.
Adilov et al. (2018) construct a simplified dynamic model of the space environment,
with functionning satellites hit at a rate proportional to the quantity of orbital debris. Under
open-access, firms launch satellites as long as the expected profit from a marginal one
is positive. Assuming adaptive anticipations regarding the future rent and collision risk,
Adilov et al. (2018) derive several properties of the space sector equilibrium behavior.
In particular, they show that the equilibrium launch rate first increases with the stock of
orbital debris, for low levels of congestion, before it decreases above a certain threshold.
Finally, they predict that space debris render the orbit economically unprofitable before the
occurrence of the Kessler Syndrome.2
Klima et al. (2016) use an extensive simulation model to consider space debris removal
missions by the US, Europe and China. They assume that each country can initially com-
mit to remove zero, one or two pieces of debris per year. They simulate the effect of each
scenario on the evolution of space debris and the resulting risk of collision between 2016
and 2165. Finally, they use these projections to calibrate an empirical game in normal
form. Their main result is that there exists pure strategy Nash equilibria, where only one
player commit to remove one piece of debris per year, and mixed strategy Nash equilibria,
where the players randomize between removing zero or one piece of debris per year.
Macauley (2015) develops a simplified model of debris accumulation and formalizes
the externality generated as a reduction in a spacecraft’s lifetime productivity. The main
finding is the calculus of the external cost from adding one satellite to the initial fleet. A
numerical application shows that it could range from 0.06 up to 2.13% of the cost of one
satellite.
1
Sandler and Schulze (1981) and Weinzierl (2018) mention this issue. However, they do not analyze it
explicitly.
2
The Kessler syndrome is a scenario such that the density of objects in low Earth orbit becomes so high to
render the use of satellites infeasible (Kessler and Cour-Palais 1978).
13
698 S. Rouillon
Muller et al. (2017) analyze the choice of a monopolistic firm regarding the number
of satellites to be launched initially, knowing that they may be destroyed by collision in
the future. The probability of destruction is assumed given and constant over time. In par-
ticular, it is invariant with the number of satellites launched initially and damaged sub-
sequently. The company provides and sells some service, as long as at least one satellite
remains in order. Under these assumptions, Muller et al. (2017) show the private loss in
expected profit caused by the destruction of satellites is slightly larger than the cost of
replacing them.
Rao and Rondina (2019) present a long-run economic model of Earth orbit, assuming a
perfectly competitive market for satellites. They compare a scenario where profit-maximiz-
ing firms have open-access to one where a fleet planner owns all satellites. They consider a
non-standard definition of open-access, such that firms are limited in their launch choice by
the requirement to own at most one satellite at a time. They state several properties of the
open-access equilibrium (i.e., conditions for interior or corner solutions, comparative stat-
ics, steady states and stability). They finally propose an expression for the external cost of a
marginal satellite and show that open-access leads to too many satellite launches relative to
the socially optimal policy if the latter is positive.
Rao et al. (2020) use an applied model, based on Rao and Rondina (2019) and cali-
brated on physical and economic data from 1957 to 2015, to determine the trajectories of
open-access and optimal launches until 2040. They simulate two scenarios, one without
orbital debris removal, the other with 50% orbital debris removal from 2029 onwards. By
comparing the latter, they illustrate the failure of purely technical and managerial solutions
to the orbital debris problem, as the space sector is encouraged to launch more satellites as
soon as the profitability of the orbit is restored. They finally calculate an optimal tax, capa-
ble of implementing an optimal trajectory under open-access.
The present paper extends the previous literature in several directions. First, our model
allows a definition of the orbital maximum carrying capacity. This concept is new in the
literature. Second, we complement the analysis of private incentives to exploit the orbit for
profit making under open-access. This issue is addressed in Adilov et al. (2014, (2018) and
Muller et al. (2017). However, they assume imperfect competition (horizontal differentia-
tion or monopoly) and/or a finite time horizon. Here, we consider perfect competition and
focus on the long-term stationary state of the space environment. With the rapid develop-
ment of a private space economy (Weinzierl 2018), the assumption of perfect competi-
tion among satellite operators is becoming increasingly relevant.3 Besides, our focus on the
long-term equilibrium is valuable because it allows a better representation of orbital sus-
tainability. Rao and Rondina (2019) also analyze open-access assuming perfect competi-
tion and infinite horizon. However, they use a non-standard definition of open-access, with
firms allowed to only operate one satellite, and postulate that satellites have an infinite life
span unless destroyed by collision. In this paper, we assume instead that firms can own any
number of satellites, each with limited operational lifetime. Third, our model allows to for-
malize a sustainable and optimal management plan for the space activity. Such a concept
3
As documented by Dos Santos Paulino and Le Hir (2016), the satellite market has been marked recently
by the arrival of numerous new entrants, who have overcome traditional barriers to entry thanks to the min-
iaturization of satellites. Dos Santos Paulino and Le Hir (2016) regard small satellites as a disruptive inno-
vation (Christensen 1997). Relying on more modular and standardized technologies, they are simpler and
faster to design and construct. As a result, they display lower performance, but can be offered at a lower
price, thus attracting new costumers.
13
can also be found in Adilov et al. (2014) and Muller et al. (2017), but their definition is
limited to finite time horizon and thus does not formalize long-term sustainability. Here,
we define an optimal policy in the long-term stationary state. Rao and Rondina (2019) and
Rao et al. (2020) also describe an optimal policy in an infinite horizon model. However,
they do not explicitly characterize it and use simulations instead. Here, we explicitly calcu-
late the optimal policy and obtain general properties. Finally, we give policy recommenda-
tions to mitigate orbital debris, dealing with command-and control, a tax per launch and a
market of individually transferable launch quotas. This issue is partially covered by Adilov
et al. (2014), where they derive a optimal tax schedule. Their characterization is not valid
in our setting, since it is meant to correct two market failures simultaneously, namely, the
imperfect competition induced by product differentiation and the externality caused by
space debris.
This paper is organized as follows. Section 2 sets out the physical model and states the
notion of maximum carrying capacity. In Sect. 3, we construct our economic model to
derive the physico-economic equilibrium and socially optimal outcome. Section 4 shows
how standard economic instruments can be used to implement an optimal outcome. Sec-
tion 5 is dedicated to a numerical application based on a simplified specification and a cali-
bration using real data. Section 6 presents two refinements of our reference model, dealing
respectively with active debris removal and entry barriers. In Sect. 7, we conclude and
present possible extensions for future researches. Supplementary material and proofs are
supplied in an “Appendix”.
2 The Physical Model
Consider a given spherical shell of low Earth orbit (e.g., altitudes between 800 and
1000 km).4 Assume that the space sector continuously launches q satellites per period.5
The operational lifetime of a satellite orbiting the Earth depends on technical (i.e., fuel
exhaustion, components failures) and environmental (i.e., collisions) events. Below, it is
formalized as a positive random variable T, distributed according to an exponential prob-
ability distribution function 1 − e−(𝜆+𝜇)T , with rate parameter 𝜆 + 𝜇. The rate of defect by
technical failure is denoted by 𝜆, which is considered as exogenous. The rate of breakup by
collision is denoted by 𝜇 and endogenously determined, as explained below.
In our stylized model, following a tradition derived from the economics of fisheries
(Gordon 1954; Schaefer 1957), we focus on the equilibrium of the orbital environment,
resulting in the long run from an assumed constant launch rate q. In this way, suppose that
there exists a relationship x = 𝜙(q), giving the long-term number of space debris, once
the orbit has reached the corresponding stationary state. Assume further that there exists
4
The Earth orbit is divided into three regions, differing by their physical properties and uses by the space
industry. Low Earth orbit lies at altitudes between 200 and 2000 km, medium Earth orbit, between 2000
and 35,786 km, and geosynchronous Earth orbit, at exactly 35,786 km. Our model considers a given spheri-
cal region of low Earth orbit. To simplify, the orbital environment surrounding it is assumed in a stationary
state, ensuring that the quantity of space objects crossing from outside is constant through time.
5
Satellites differ in many respects (mass, surface, power, etc.). In our model, we consider a standardized
prototype, supposedly designed to minimize costs given available technologies. We also assume constant
returns to scale, which implies that this template can be produced at different scales. Then, the launch rate
can be viewed as a continuous variable. This allows to account for the diversity of satellites, while keeping
the model tractable.
13
700 S. Rouillon
a relationship 𝜇 = 𝜑(x), giving the instantaneous risk of destruction of a satellite by col-

lision as a function of the number of space debris. Overall, this formalization drives us to
conclude that the risk of collision in the long run can be written as 𝜇 = f (q), where we let
f (q) ≡ 𝜑(𝜙(q)) for the sake of notational convenience.
Many factors are likely to determine the shape of this relationship. Some are under the
control of the satellite operators (e.g., altitudes, technological innovations, mitigation and
remediation measures). Others are out of control (e.g., atmospheric density, solar activ-
ity). These factors are left implicit in order to focus on the economic intuitions. Still, we
describe in “Appendix A1” a physical model showing how to rationalize our approach and
how the neglected factors would intervene.
Below, we simply postulate the following:
Assumption 1 f (0) = 0, f � (q) > 0 and f �� (q) > 0 for all q > 0, and limq→Q f (q) = ∞ for
some Q > 0.6
Assumption 1 is illustrated by Fig. 1. The intuitions supporting it are the following. Due
to orbital decay, all debris and satellites eventually fall back to the Earth. Therefore, if the
launches stop for a sufficiently long period of time, the orbit will ultimately be cleaned up
(Eichler and Reynolds 1997). In that state, the first satellite launched afterwards bears no
risk of collision. Then, as the launch rate increases, the long-term collision rate is increas-
ing at an increasing rate. This is supported by existing prospective simulations (Somma
et al. 2019). Finally, the existence of a threshold launch level after which the long-term rate
of collision becomes infinite, echoes the Kessler syndrome (Kessler and Cour-Palais 1978).
In our analytical framework, the expected population of operational satellites at any time
t0 converges in the long run to:
n(q) ≡
t0
�−∞
qe−(𝜆+𝜇)(t0 −t) dt, with 𝜇 = f (q).
Indeed, by assumption, the launch rate is equal to q forever. The probability that a satellite
launched at time t < t0 remains operational till time t0 is e−(𝜆+𝜇)(t0 −t). The expected popula-
tion of operational satellites at time t0 is the sum of all generations t of satellites launched
previously (i.e., q per period) weighted by their probability of being operational at time t0
(i.e., e−(𝜆+𝜇)(t0 −t)). Integrating the above expression, we obtain that:
q
n(q) =
𝜆 + f (q)
. (1)
This expression can also be seen as the product of the rate of satellite launches, q, times the
long-term expected lifetime of a satellite, d(q) ≡ 1∕(𝜆 + f (q)).
We can show the following:
6
We have assumed in this paper that Q is finite. Note however that all results remain unchanged if Q is
infinite, provided that the last condition in Assumption 1 is replaced by limq→∞ f (q)∕q = ∞. In a supple-
mentary material posted online, we fully solve the model for f (q) = 𝜃q2 and f (q) = 𝜃q, with 𝜃 > 0. The
former illustrates the case where Q = ∞ and limq→∞ f (q)∕q = ∞. The latter illustrates the case where
Q = ∞ and limq→∞ f (q)∕q < ∞.
13
Property 1 Under Assumption 1, n(0) = limq→Q n(q) = 0 and there exists q such that
0 < q < Q and n� (q) ⋛ 0 iff q ⋚ q.
Proof of ( Property 1 The ) first part is immediate. The derivative of n(q) is

n� (q) = 𝜆 + f (q) − qf � (q) ∕(𝜆 + f (q))2. The numerator is decreasing (as its derivative is
−qf �� (q) < 0). Note that n� (0) = 1∕𝜆 > 0. Assume that n� (q) ≥ 0 for all q. Then n(0) < n(q)
for all q, which contradicts the fact that n(0) = limq→Q n(q) = 0. Therefore, there exists q
such that 0 < q < Q and n� (q) ⋛ 0 iff q ⋚ q . ◻
Under Assumption 1, the curve representing the long-term expected population of

operational satellites typically has the unimodal shape shown in Fig. 2. It is equal to 0 for
q = 0 and q = Q. It is first increasing and then decreasing in q, thus admitting a maximum
for some q between 0 and Q. The reason why we get this inverted-U shape is because
the expected long-term population of functioning satellites is the product of the satellites
launch rate, q, times their expected lifetime, d(q). In the long run, increasing the launch
rate has an ambiguous effect, because this simultaneously reduces their operational lifetime
(Fig. 2).
Now, let us consider the following definition:
Definition 1 The orbital maximum carrying capacity n is the maximum expected

number of active satellites that the space sector can sustain in the long term. Formally,
n ≡ maxq n(q). The corresponding expected operational lifetime of a satellite is d ≡ d(q).
( )
Using Property 1, the maximum carrying capacity n is equal to n q . It can be main-
tained with a launch rate set equal to q forever.
3 The Economic Model
In this section, we propose an economic model in order to investigate two main issues.
First, it is important to understand the private incentives to launch new satellites, as this
potentially governs the future of the orbital environment. Second, it is useful to propose a
normative criterion for characterizing a socially optimal launch policy.
The basic players in our economic model are satellite operators. They order satellites
from manufacturers and request launchers to place them in orbit. They then lease the ser-
vices produced by their satellite fleet to consumers. It is assumed below that the entry and
exit of satellite operators into and out of the market is free and without cost. The maximum
number of potential operators is assumed to be sufficiently large.7 Each of them is capable
of managing a large fleet of satellites.
7
Using the UCS database (cf. footnote 12), we found through an Internet survey that 67 operators owned
the low Earth orbit satellite fleet in 2018 and that 23 of them were founded after 2010. The data collected
are provided in a supplementary material available online.
13
702 S. Rouillon
Fig. 1 Long-term risk of colli-

sion
The economic parameters at stake are the following. The unit cost of a satellite is
denoted by c.8 A satellite generates a cash-flow per operating period denoted by p, equal to
the rents paid by consumers for the services provided, less operating costs.9 The economic
agents are assumed to discount time at the rate of time preference 𝛿.
In order to investigate the private incentives to launch satellites, let us first examine the
situation where the orbit is under open-access. By definition, this means not only that the
number of satellite operators can vary, through entries or exits, but also that each of them is
free to send as many satellites as they wish. Presumably, they will do so as long as they can
expect to achieve a positive benefit per launch.
A firm launching a satellite at any time t0 bears an immediate cost c and anticipates to
receive a cash-flow p during its whole operational lifetime. Given that it discounts time at
the rate 𝛿, the expected present value v(q) of its satellite is defined by:
∞( )
v(q) ≡
t
�t 0 �t0
pe−𝛿(𝜏−t0 ) d𝜏 (𝜆 + 𝜇)e−(𝜆+𝜇)(t−t0 ) dt, with 𝜇 = f (q).
Indeed, if it remains operational until time t > t0, it will generate a cash-flow equivalent to
∫t pe−𝛿(𝜏−t0 ) d𝜏 in present value.10 It fails at time t with probability density
t
0
(𝜆 + 𝜇)e−(𝜆+𝜇)(t−t0 ). By definition, the satellite expected present value is the sum of the
product of the last two terms over all possible failure times t > t0. The expression above
can be integrated, leaving:
8
Cost c is the counterpart of the remuneration of subcontractors for the manufacture and placing in orbit of
a satellite. It therefore includes the cost of design, construction, testing, launch and insurance. Furthermore,
to implicitly reflect any market power of subcontractors, cost c may also be assumed to include a subcon-
tractor’s mark-up.
9
The rent of an active satellite is assumed a parameter in our model. Several arguments can justify this, at
least as an approximation. First, competition in the market for satellite services goes beyond the space sec-
tor, as most of these are also available by terrestrial means. Second, the region considered in our model is
assumed to be sufficiently small relative to the entire low Earth orbit so that the number of satellites placed
there can only have a negligible effect. Third, since the lifetime of a satellite ranges from approximately
5–15 years, the addition or removal of satellites in a given year has a negligible effect compared to the size
of the fleet of satellites already in orbit.
Upon integration, we calculate that ∫ pe−𝛿(𝜏−t0 ) d𝜏 = p 1 − e−𝛿(t−t0 ) ∕𝛿.
10 t ( )
t0
13
Fig. 2 Long-term expected num-

ber of satellites
p
v(q) =
𝛿 + 𝜆 + f (q)
. (2)
We can derive the following:
Property 2 Under Assumption 1, v(0) = p∕(𝛿 + 𝜆), limq→Q v(q) = 0 and v� (q) < 0 for all
q.
Proof of Property 2 The first part is immediate. The derivative of v(q) is

−f � (q)p∕(𝛿 + 𝜆 + f (q))2 < 0. ◻
Property 2 is illustrated in Fig. 3, showing the typical shape of the expected discounted
value v(q) of a satellite placed in orbit. The intuition behind it is straightforward. In the
long run, the insertion of additional satellites eventually induces a larger population of
debris, thus increasing the collision risk. This reduces the expected operational lifetime
of satellites and, therefore, their expected present value. The latter varies from p∕(𝛿 + 𝜆)
when q = 0 (i.e., in the long run, the orbit is clean and there is no collision risk) to 0 when
q ≥ Q (i.e., in the long term, the orbit reaches the Kessler syndrome and satellites are
immediately destroyed).
Now, recall that under open-access, our postulate is that private agents will launch new
satellites as long as they can expect a positive benefit.11 Formally, this condition writes
v(q) > c. Conversely, they will cease launching if they anticipate a negative benefit. For-
mally, this condition writes v(q) < c. Accordingly, the space sector will be satisfied with its
strategy if and only if the launch rate satisfies v(q) = c.
We summarize in the following definition our postulate regarding the behavior of pri-
vate agents under open-access:
Definition 2 A physico-economic equilibrium is reached when the expected present

value of a satellite equals its overall cost. Formally, q = q∗ such that v(q∗ ) ≡ c. The cor-
responding expected population of operational satellites is n∗ ≡ n(q∗ ). The corresponding
expected operational lifetime of a satellite is d∗ ≡ d(q∗ ).
11
Either incumbent operators will expand their fleet or entrants will build up their own satellite fleet.
13
704 S. Rouillon
Fig. 3 Expected present value

and cost of a satellite
We call this situation a physico-economic equilibrium because it characterizes a situa-

tion where both the orbit is in a stationary state and economic agents are satisfied with their
plan.
We show the following proposition:
Proposition 1 If p∕(𝛿 + 𝜆) > c, there exists a unique physico-economic equi-

librium. It is characterized by a launch rate q = q∗ such that 0 < q∗ < Q and
v(q∗ ) = p∕(𝛿 + 𝜆 + f (q∗ )) = c. (If p∕(𝛿 + 𝜆) ≤ c, no physico-economic equilibrium exists,
in the sense that launching a satellite is never profitable).
Proof of Proposition 1 From Property 2, v(q) is strictly decreasing and takes val-
ues between v(0) = p∕(𝛿 + 𝜆) and limq→Q v(q) = 0. If p∕(𝛿 + 𝜆) > c, then
v(0) = p∕(𝛿 + 𝜆) > c > 0 = limq→Q v(q). By continuity, there exists a unique q∗ such that
0 < q∗ < Q and v(q∗ ) = c. Clearly, if p∕(𝛿 + 𝜆) ≤ c, then v(q) < c for all q > 0. ◻
Proposition 1 is illustrated by Fig. 3, assuming that p∕(𝛿 + 𝜆) > c. The physico-eco-

nomic equilibrium is at the intersection between the curve representing the expected rev-
enue, v(q), and the horizontal line plotting the overall cost of a satellite, c. Knowing that
v(q) is decreasing, only one such point exists. Moreover, the physico-economic equilibrium
launch rate q∗ lies between 0 and Q.
As f � (q) > 0, the physico-economic equilibrium rate of launching q∗ is increasing in
p and decreasing in c, 𝛿 and 𝜆. Also note that q∗ < Q for all configurations of parameters.
Hence, rational agents have incentive to avoid the Kessler syndrome, as emphasized by
Adilov et al. (2018).
The following property follows directly from the definition of a physico-economic
equilibrium:
Property 3 In a physico-economic equilibrium, d∗ = 1∕(p∕c − 𝛿) > 0.
Proof of Property 3 By Proposition 1, q∗ satisfies v(q∗ ) = p∕(𝛿 + 𝜆 + f (q∗ )) = c. Elemen-

tary algebra yields d∗ = d(q∗ ) = 1∕(𝜆 + f (q∗ )) = 1∕(p∕c − 𝛿). ◻
In other words, under open-access, the expected operational lifetime of a satellite only
depends on economic parameters c, p and 𝛿. For instance, since the cost of launch increases
with altitude, satellites placed at higher altitudes should have a longer operational lifetime,
13
all other things being equal. The UCS Satellite Database corroborates this positive cor-
relation.12 Indeed, as reported by the satellite operators, the expected lifetime of satellites
is on average 6.63 years for the low Earth orbit, 9.38 years for the medium Earth orbit and
13.47 years for the geosynchronous Earth orbit. This observation is also made by Hiriart
et al. (2009) and Rao and Rondina (2019).
It is interesting to compare the physico-economic equilibrium with the maximum carry-
ing capacity. A physical over-exploitation of the orbit occurs when the equilibrium launch
rate q∗ is higher than the launch rate q maximizing the long-term expected population of
operational satellites.
Our results are explained in the following property:
( )
Property 4 q∗ ⋛ q if and only if p∕c − 𝛿 ⋛ qf � q .
( )
Proof of Property 4 By definition, q satisfies n� q( =) 0 and( q)∗ satisfies v(q∗ ) = c. Here,
these conditions are usefully rewritten as 𝜆( +) f q = qf � q and 𝜆 + f (q∗ ) = p∕c − 𝛿
respectively. Then, q∗ = q when p∕c − 𝛿 = qf � q . Moreover, as f � (q) > 0, q∗ is increasing
in p∕c − 𝛿 . ◻
In other words, the orbit is over-exploited when the rent of a satellite is high, the overall
cost of a launched satellite is small and/or the economic agents are not too impatient. Then,
according to Property 1, the long-term population of operational satellites will be smaller
than the maximum carrying capacity, because the satellite operators launches too many sat-
ellites, each with a shortened lifetime caused by space debris. It is worth noting that inverse
cases exist where the equilibrium launch rate q∗ is lower than the launch rate q inducing the
maximum carrying capacity.
Finally, we propose a normative criterion to characterize a socially optimal launch
plan. At any time, the space sector launches q satellites. In the long-term stationary state
of the space environment, each satellite placed in orbit is worth v(q) and costs c. Over-
all, each generation of satellites therefore periodically generates a social surplus equal to
S(q) = (v(q) − c) q. Below, our postulate will be that the social objective should be to max-
imize it.13
We summarize our definition of an optimal policy in the following:
Definition 3 An optimal launch policy maximizes the social surplus generated peri-
odically by the satellites operators. Formally, q = qo where qo ≡ arg maxq S(q). The cor-
responding expected number of operational satellites is no ≡ n(qo ). The corresponding
expected operational lifetime of a satellite is do ≡ d(qo ).
Clearly, under open-access, the satellite operators do not adopt an optimal launch rate.
Indeed, a physico-economic equilibrium is characterized by v(q∗ ) = c, which implies that
S(q∗ ) = 0. Hence, the behavior of the satellite operators yields a full dissipation of the
12
The UCS Satellite Database is managed by the Union of Concerned Scientists, available at https://www.
ucsusa.org/nuclear-weapons/space-weapons/satellite-database. It currently lists 2062 active satellites (July
8, 2019).
13
Alternatively, one may propose to maximize the cash-flow pn(q) − cq of the satellite opera-
tors. Note that our approach embeds it as a special case. Indeed, given that v(q) = p∕(𝛿 + 𝜆 + f (q)) and
n(q) = q∕(𝜆 + f (q)), the two objectives coincide when 𝛿 = 0.
13
706 S. Rouillon
social surplus. This is another illustration of the tragedy of the commons (Hardin 1968).
Individual firms pursuing maximum profit neglect future externalities generated by their
satellites, as they expect that induced profits will be captured by others.
We can show the following proposition:
Proposition 2 If p∕(𝛿 + 𝜆) > c, there exists a unique optimal policy, characterized

by a launch rate q = qo such that 0 < qo < Q and S� (qo ) = v(qo ) + v� (qo )qo − c = 0. (If
p∕(𝛿 + 𝜆) ≤ c, the optimal orbital policy is qo = 0).
Proof of Proposition 2 By definition of S(q) and v(q), we have

( )
p
S(q) = − c q. (3)
𝛿 + 𝜆 + f (q)
By differentiation, we obtain
𝛿 + 𝜆 + f (q) − f � (q)q
S� (q) = p −c
(𝛿 + 𝜆 + f (q))2
and
( )
��
f �� (q)q(𝛿 + 𝜆 + f (q)) + 2f � (q) 𝛿 + 𝜆 + f (q) − f � (q)q
S (q) = −p .
(𝛿 + 𝜆 + f (q))3
For an interior solution, an optimal launch policy must satisfy S� (q) = 0. If p∕(𝛿 + 𝜆) > c,
as S� (0) = p∕(𝛿 + 𝜆) − c > 0 and S� (q) < 0 for q sufficiently close to Q,14 this condition has
a solution q = qo such that 0 < qo < Q. Observe that it necessarily belongs to the set where
𝛿 + 𝜆 + f (q) − f � (q)q ≥ 0 (as otherwise, S� (q) < 0). This implies that S�� (q) < 0 in the rel-
evant region. Then, as S� (q) is strictly decreasing, S� (q) = 0 has unique solution determin-
ing a maximum of S(q). Clearly, if p∕(𝛿 + 𝜆) ≤ c, then v(q) < c for all q > 0, implying that
q = qo = 0 is optimal. ◻
It is worth comparing the physico-economic equilibrium to the optimal policy. From an

economic perspective, the orbit is overexploited when the physico-economic equilibrium
launch rate q∗ is higher than the optimal launch rate qo.
Our result is stated in the following property:
Property 5 The physico-economic equilibrium always induces an economic over-exploi-

tation of the orbit. Formally, q∗ > qo for all configurations of parameters.
Proof of Property 5 By definition, v(q∗ ) = c and v(qo ) + v� (qo )qo = c. Given that qo > 0
and v� (qo ) < 0, v(q∗ ) = c < c − v� (qo )qo = v(qo ). Finally, as v(q) is decreasing (see the
proof of property 2), this implies that q∗ > qo. ◻
14
To show this, note that f (q) − f � (q)q is null for q = 0 and is decreasing (as its derivative is −qf �� (q) < 0).
Thus, f (q) − f � (q)q ≤ 0 and S� (q) ≤ p(𝛿 + 𝜆)∕(𝛿 + 𝜆 + f (q))2 − c for all q. Finally, the right-hand side of
the latter inequality has a limit equal to −c for q → Q.
13
Adilov et al. (2014) find a similar result in a different framework, where horizontally
differentiated firms have private incentives to launch too many satellites in a two-period
model. Their result cannot be directly compared with Property 5, as it brings together both
issues of imperfect competition and negative orbital externalities. Moreover, they use a
model with two time period, whereas we focus on the long-term stationary state.
Figure 4 is useful to illustrate the determination of q∗ and qo. It decomposes the social
surplus S(q) in its two components (i.e., v(q)q and cq). The curve v(q)q and the line cq
respectively give the aggregate expected revenue in present value and cost from launching
q satellites.15 The vertical distance between them gives S(q). By definition, the physico-
economic equilibrium is attained when the space sector generates no surplus. Graphically,
this occurs at q = q∗, where the two curves intersect. The optimal policy is attained when
the distance between the two curves is maximum. Graphically, this occurs at q = qo, when
the parallel to cq is tangent to v(q)q.
Figure 5 gives another way to find q∗ and qo graphically. It is identical to Fig. 3, except
for the curve representing the expected revenue of a marginal launch, v(q) + v� (q)q (i.e., the
derivative of v(q)q). The proof of Proposition 2 shows that this curve is decreasing as long
as it is positive. It lies below the curve representing the expected present value of a satel-
lite, v(q), as it includes the damage generated by a marginal satellite (i.e., −v� (q)q > 0).
The physico-equilibrium launch rate q∗ is found when the decreasing curve v(q) intersects
the horizontal line c. Here, the optimal policy qo is attained when the decreasing curve
v(q) + v� (q)q intersects the horizontal line c.
4 Policy Implications
In this section, we describe standard economic instruments (command-and-control, tax

and market) to regulate the satellite operators’ activity. In reality, the use of such measures
would be hampered by the international dimension of the problem. Just like other global
commons (ozone layer depletion, climate change, international fisheries, etc.), an interna-
tional agreement is needed before any policy can enter into force.16
A first instrument would be to issue an aggregate quota, limiting the number of satellites
allowed for launch per period. Based on our results, it should be set equal to qo to achieve
an optimal outcome. However, if some satellite characteristics are not fully monitored, this
policy should be used with caution, as it could have perverse effects. Indeed, in our model,
we have considered satellites as standardized units (mass, surface, power, etc.), implicitly
designed efficiently to minimize costs. In pratice, one way to circumvent the quota may be
for the space sector to design upgraded satellites, supplying improved services to consum-
ers in counterpart of higher rents. In addition, the implementation of an aggregate quota
15
From the proof of Proposition 2, v(q)q is concave as long as it is increasing. It can be either convex or
concave afterwards.
16
An anonymous referee suggests that an interesting extension of this section would be to take into account
imperfect information in the model. We believe that this could indeed open up two avenues for future
research. On the one hand, in the manner of Weitzman (1974), taking into account the asymmetry of infor-
mation on satellite manufacturing would make it possible to better enlighten the choice between fee and
quantity-based pricing instruments. This is also suggested in Rao et al. (2020). On the other hand, taking
into account the scientific uncertainty on the functioning of the orbital environment, and in particular on the
conditions that trigger the Kessler syndrome, would make it possible to bridge the subject of orbital debris
to the literature on the precautionary principle.
13
708 S. Rouillon
Fig. 4 Expected revenue and cost

of a launching campaign
Fig. 5 Physico-economic equilib-

rium versus optimal outcome
would render this strategy profitable, with the life span of satellites increasing as the state
of the orbital environment improves. If such a race for sophistication begins, the policy will
fail, both because new series of satellites will be more expensive than efficient units and
could also contribute more to the accumulation of space debris. To conclude, if an aggre-
gate quota is implemented, not only the number of launches needs be monitored, but also
all technical characteristics relevant from the perspective of costs and debris accumulation.
This might be necessary, even if technology standards generally have detrimental effects in
terms of static and dynamic efficiencies (Adilov et al. 2014).
Another instrument would be to impose a tax rate 𝜏 per satellite launched. Using our
results, we show that it should be set equal to 𝜏 o ≡ −v� (qo )qo to implement an optimal out-
come. This (pigovian) level coincides with the expected loss of revenue caused by a mar-
ginal satellite. We can verify that it actually implements an optimal outcome as follows.
From Proposition 1, the physico-economic equilibrium q∗ induced by any tax rate 𝜏 satisfies
v(q∗ ) = c + 𝜏 . From Proposition 2, the optimal allocation qo satisfies v(qo ) + v� (qo )qo = c.
If 𝜏 = 𝜏 o, it follows that v(qo ) = c + 𝜏 and q∗ = qo, given that the physico-economic equi-
librium is unique by Proposition 1. The underlying intuition can also be discussed using
Fig. 6. Initially, the physico-economic equilibrium is at the intersection between v(q) and
c. If the unit cost of a satellite is increased by Δ, the physico-economic equilibrium moves
at the intersection between v(q) and c + Δ. In Fig. 6, the latter coincides with the opti-
mal outcome, since Δ = −v� (qo )qo by construction. Importantly, it should be noted that a
tax per satellite should not foster perverse incentives to design upgraded satellites, con-
trary to command-and-control regulations. The reason is because in a physico-economic
13
Fig. 6 Optimal policy (where

Δ = −v� (qo )qo)
equilibrium, the satellite operators make no extra profit and building uneconomical satel-
lites could only worsen the situation.
A last instrument would be to implement a system of individual transferable quotas.
Anew, given our previous results, the total number of satellites allowed for launch must be
set equal to qo. The aggregate quota is divided among the satellite operators participating in
the market. Firms can either launch satellites up to the amount of quota they hold, or they
can sell their quota to others. Let us denote by 𝜌 the price of a quota. Each firm faces the
tradeoff of launching a marginal unit, which is worth v(q∗ ) − c, or selling a quota, which is
worth 𝜌. Clearly, in a market equilibrium, it must be the case that v(q∗ ) − c = 𝜌. As a result,
if the aggregate quota is qo, we will have q∗ = qo and 𝜌 = v(qo ) − c. Moreover, given that
v(qo ) + v� (qo )qo = c by Proposition 2, the equilibrium price will be 𝜌 = 𝜌o ≡ −v� (qo )qo,
thus reflecting the external cost of a marginal launch. In Fig. 6, for any total allowable
launch rate q, the equilibrium price is equal to the difference between v(q) and c. In par-
ticular, it is equal to Δ = −v� (qo )qo when qo quotas are allocated. Note that the physico-
economic equilibrium then coincides with the optimal outcome, if one includes in the cost
of a satellite the opportunity cost of the foregone quota. Like the pigovian tax, this policy
should not create perverse incentives to launch upgraded satellites. On the one hand, they
would be more expensive than standard ones. On the other hand, in case where these tech-
nological improvements favors space debris accumulation, they will lead to a depreciation
of the price of individual transferable quotas.
5 Numerical Application
In this section, we propose a specification and a calibration of our model. The parameters
are chosen to fit with actual data, whenever available, in order to give realistic orders of
magnitude. A sensitivity analysis is performed, both in terms of alternative calibrations and
specifications.
Consider the following benchmark specification of f (q), giving the rate of satellite
breakup by collision, at the long-term equilibrium the orbital environment:17
17
The alternative quadratic (i.e., 𝜃q2) and linear (i.e., 𝜃q) specifications are considered in a sensitivity anal-
ysis below. Parameter Q is infinite in these cases.
13
710 S. Rouillon
with 𝜃 > 0 and 0 ≤ q < Q.

q
f (q) = 𝜃 ,
Q−q
All analytic results are provided in a supplementary material available online (i.e., launch
rates, expected populations and expected operational lifetimes, associated with the maxi-
mum carrying capacity, the physico-economic equilibrium and social optimum).
Our calibration of the physical parameters relies on Farinella and Cordelli (1991). They
implement a long-term simulation of the proliferation of space debris, assuming a forever
rate of insertion of 100 satellites per year in low Earth orbit (i.e., all altitudes between 200
and 2000 km). Following their physical model, the annual rate of collision of an intact
satellite is the product of a constant coefficient, equal to 3 ⋅ 10−10/year/fragment, times the
number of orbiting debris fragments larger than 1 cm. At the end of their 500 years simula-
tion time span, they calculate that the latter will reach 5 ⋅ 108 units (read on Fig. 2, p. 372).
As a result, they predict a collision rate about 0.15 per year within 500 years.
In order to use this prediction of Farinella and Cordelli (1991), we must first know how
the satellites placed into low Earth orbit are usually allocated among altitudes. Indeed,
recall that our model focus on a subregion of low Earth orbit. In particular, we are inter-
ested below in altitudes between 800 and 1000 km, which are known to have already
reached a “runaway” level (Kessler and Anz-Meador 2001; Kessler et al. 2010).18 The DIS-
COS database lists a total of 6111 active payloads launched into low Earth orbit since the
origin of space activity.19 This amounts to about 100 units per year, as in Farinella and
Cordelli (1991). The shell between 800 and 1000 km accounts for approximately 10% of
this overall traffic, with a total of 593 active payloads launched and an average of about 10
units per year.
To summarize, under the business-as-usual scenario, corresponding to a launch rate
into low Earth orbit of 100 satellites per year, including 10 satellites per year at altitudes
between 800 and 1000 km, the long-term collision rate is estimated at about 0.15 per year.
In our setting, this is formally written as f (10) = 0.15 and Q > 10. Accordingly, we will
calibrate our model by treating f (10) as a parameter, with 0.15 per year as benchmark esti-
mate, and by choosing Q higher than 10.20
Our calibration of the economic parameters relies on the following data. Somma et al.
(2019) argue that a typical satellite placed into low Earth orbit has an active lifetime of
8 years. According to Dos Santos Paulino and Le Hir (2016), the cost of designing, build-
ing and testing a satellite is between $100 and $400 million. By adding the launch price,
we will set the overall cost of a satellite at $300 million. In 2018, the total revenue from
satellite services was $126,500 million (Satellite industry association 2017). With a total
population of 1,457 operational satellites (based on UCS database, including all altitudes
between 200 and 35,786 km and all users), the average revenue per satellite was $87.53
million. Since this includes not only commercial, but also civilian, government and mili-
tary users, it is probably an underestimation. As a result, we will round up the rent for a
satellite to $100 million per year. Finally, we will use a rate of discount equal to 5% a year.
18
A “runaway” environment is such that no equilibrium is possible, as long as the background population
of intact objects remains constant (Kessler and Anz-Meador 2001; Kessler et al. 2010).
19
The DISCOS database is managed by ESA’s Space Debris Office, available at https://discosweb.esoc.
esa.int/web/guest/home. The data were kindly provided to us by Vitali Braun, space debris engineer at ESA/
ESOC Space Debris Office, on May 22, 2019.
20
Note that f (10) = 10𝜃∕(Q − 10) implies that 𝜃 = f (10)(Q − 10)∕10.
13
Based on this information, we propose to use the benchmark calibration summarized in

Table 1. The corresponding numerical results are displayed in Table 2. They are explained
below while commenting Figs. 7, 8 and 9.
Figure 7 represents the expected population of satellites n(q) as a function of the launch
rate q. The top of this curve at point A gives the maximum carrying capacity. It is equal
to n = 36.47 satellites, sustainable if the satellite operators launch q = 10.68 satellites per
year forever. Point A0, the coordinates of which are the average number of launches since
1957 (10 satellites/year) and the population of active satellites orbiting between 800 and
1000 km in 2018 (57 operational satellites according to UCS database), is displayed for
the sake of comparison. The reason why point A0 is well above the curve is that the current
risk of collision is still negligible today, because the orbit has not yet reached its stationary
state. Indeed, all existing simulation models predict a large increase in space debris in the
future under the status quo scenario (Drmola and Hubik 2018; Eichler and Reynolds 1997;
Farinella and Cordelli 1991; Lafleur 2011; Somma et al. 2019). In other words, the current
orbit state is probably unsustainable, as the risk of collision is expected to increase in the
future, which will graphically move point A0 down to the curve. Points A∗ and Ao respec-
tively plots the outcomes arising in the physico-economic equilibrium and social optimum.
Note that no physical over-exploitation of the orbit occurs in the physico-economic equilib-
rium, since point A∗ lies at the left of point A.
Figure 8 illustrates the physico-economic equilibrium and the social optimum. The
inverted U-shaped curve displays the expected present value v(q)q of each generation of q
satellites. The line represents the total cost cq. Their intersection at point A∗ determines the
physico-economic equilibrium. From Table 1, it is attained with q∗ = 10.32 launches per
year. The dashed line parallel to cq is used to locate the socially optimal policy. By con-
struction, the distance between v(q)q and cq is maximum at point Ao when it is tangent to
v(q)q. From Table 2, it is attained with qo = 5.02 launches per year.
Figure 9 plots the expected lifetime of a satellite d(q) as a function of the launch rate q.
It is decreasing from 8 years when q = 0 to 0 year when q = 25. The outcomes associated
with the maximum carrying capacity, the physico-economic equilibrium and the social
optimum are respectively displayed by points A , A∗ and Ao. The long-term expected opera-
tional lifetimes are d = 3.42 years, d∗ = 3.53 years and do = 5.51 years, corresponding to
launch rates q = 10.68 satellites per year, q∗ = 10.32 satellites peryear and qo = 5.02 satel-
lites per year respectively.
From the policy perspective, we have also calculated the economic instruments evoked
in Sect. 4. Under a command-and-control approach, the aggregate quota of satellites
allowed for launch must be set equal to qo = 5.02 satellites per year. The pigovian launch
tax is 𝜏 o = $131.93 million/satellite. Finally, under a system of individual transferable quo-
tas, the number of allowed launches must be limited to qo = 5.02 satellites per year. In a
market equilibrium, the price of a quota is 𝜌o = $131.93 million/satellite. This amount can
be compared with the overall cost of a satellite, which is between $100 and $400 million
(Dos Santos Paulino and Le Hir 2016).
Finally, in order to provide a sensitivity analysis, we randomize the six parameters of our
calibration (i.e., f (10), Q, 𝜆, c, p and 𝛿), by letting them vary by more or less 20% around
their baseline value.21 We implement 56 = 15, 625 random calibrations and calculate the
21
Precisely, we draw f (10) ∈ [0.12, 0.18], Q ∈ [20, 30], 𝜆 ∈ [0.10, 0.15], p ∈ [80, 120], c ∈ [240, 360] and
𝛿 ∈ [0.04, 0.06], according to uniform distributions.
13
712 S. Rouillon
Table 1 Benchmark calibration

Coefficient Estimate Unit Sources
f (10) 0.15 %/year Farinella and Cordelli (1991)

Q 25 satellites/year Conjectured
𝜆 1/8 %/year Somma et al. (2019)
c 300 $millions/year Dos Santos Paulino and Le Hir (2016)
p 100 $millions/sat./year SIA (2017) and UCS (2019)
𝛿 5 %/year Conjectured
Table 2 Numerical results

Launch (sat./y.) Population (sat.) Lifetime (year) Marg.Dam. ($m./y.)
Max. carrying capacity 10.68 36.47 3.42 –

Phys.-eco. equilibrium 10.32 36.44 3.53 –
Socially optimal policy 5.02 27.65 5.51 131.93
Fig. 7 Maximum carrying capac-

ity (units: sat./year; sat.)
Fig. 8 Physico-economic equi-

librium versus social optimum
(units: sat./year; $millions)
13
Fig. 9 Expected lifetime of a

satellite (units: sat./year; years)
Fig. 10 Sensitivity analysis (over 15,625 simulations)
Fig. 11 Alternative specifications
13
714 S. Rouillon
corresponding outcomes to give a representative picture of all possible configurations.22

The results are represented in the form of box-and-whisker plots in parts a to d of Fig. 10,
respectively.23
Consider first the interior of the boxes, which include 50% of all simulations by defini-
tion.24 Half of the time, the maximum carrying capacity n lies between 34.60 and 39.00
active satellites, attained with a launch rate q between 10.06 and 11.26 satellites per year.
The physico-economic equilibrium launch rate q∗ exhibits a larger variability, as it varies
between 8.72 and 11.88 satellites per year. The resulting long-term expected population n∗
varies between 33.17 and 37.91 satellites. Finally, the optimal launch rate qo generally lies
between 4.25 and 5.78 satellites per year. This allows to sustain an expected population of
active satellites no varying between 24.14 and 31.21 units in the long run. Note that this is
not that far from the other situations, despite a launch rate dramatically lower. This is due
to the longer expected lifetime do that the optimal policy allows in the long run, between
5.18 and 5.97 years. This is to be compared with an expected lifetime d between 3.19 and
3.73 years in the maximum carrying capacity situation, and d∗ between 3.08 and 4.05 years
in the physico-economic equilibrium. Finally, the damage of a marginal satellite launch at
the social optimum varies between 103.63 and 164.74 $million/satellite.
Let us now take a look at extreme values. To be concise, we focus only on very high
launch rates and marginal damages. Filtering the 15,625 simulations to sort out all cases
such that the physico-economic equilibrium launch rate q∗ is larger than 16.01 satellites
per year leaves us with 16 outcomes (i.e., 0.1% of all cases). Interestingly, they coincide
with parameter configurations such that the optimal launch rate qo is larger than 7.42 satel-
lites per year. Reciprocally, filtering the 15,625 simulations to select all cases such that the
socially optimal launch rate qo is larger than 7.66 satellites per year yields 16 outcomes
(i.e., 0.1% of all cases). Likewise, they coincide with configurations such that the physico-
economic equilibrium launch rate q∗ is larger than 15.10 satellites/year. Overall, this means
that launch rates are simultaneously very large in the physico-economic equilibrium and in
the social optimum. Looking for a general pattern in the parameters inducing these cases,
we observe that all parameterizations satisfy f (10) ≤ 0.14, c ≤ 255.31 and p ≥ 113.40
(and no clear-cut regularity for the other parameters). Hence, they occur only under the
conjonction of favourable physical (i.e., f(10) is small enough) and economic (i.e., c is
small enough and p large enough) conditions. Finally, filtering the 15,625 simulations to
retain all cases such that the marginal damage in the social optimum is larger than $261.35
million/satellite draws out 16 outcomes (i.e., 0.1% of all cases). Again, looking for a gen-
eral pattern in the parameters generating these cases, we find that 𝜆 ≤ 0.11, c ≤ 281.19 and
p ≥ 114.39 (and no clear-cut regularity for the other parameters). Accordingly, a large pig-
ovian tax or equilibrium price of a launch quota occurs only if the rate of technical failure
of a satellite is small enough, if the cost of a satellite is small enough and if the annual rent
of a satellite is large enough.
Finally, we check the sensibility of our results with respect to the specification of f (q).
We complement the benchmark specification with a quadratic (i.e., f (q) = 𝜃q2) and linear
22
All simulations, implemented with Excel, are available upon request.
23
The boxes plot the first and third quartiles. The band inside the box is the median. The cross represents
the mean. The diamonds represent the first and ninth deciles. The ends of the whiskers are the lowest datum
still within 1.5 interquartile range of the lower quartile, and the highest datum still within 1.5 interquartile
range of the upper quartile. “Outliers” are plotted as individual points.
24
Also outcomes inside the segments joining the diamonds represent 80% of all data.
13
(i.e., f (q) = 𝜃q) specification. The calibration is made to ensure the total comparability
of all outcomes. Thus, we use the benchmark estimates in Table 1, with the exception of
parameter Q, which is infinite by definition with these specifications. Figure 11 plots the
curves of f (q) obtained under the three specifications calibrated this way.
The numerical results are presented in Table 3. Each cell gives the outcomes in the order
corresponding to the benchmark, quadratic and linear specifications respectively. We focus
on the launch rates and marginal damages at the social optimum.25 With the exception of
the launch rate in the maximum carrying capacity situation with the linear specification, all
outcomes are not very different. Moreover, their variability remains in orders of magnitude
comparable with previous results. The maximum carrying capacity with the linear speci-
fication is atypical because Property 1 does not apply in this case. Indeed, the expected
population of operational satellites is then always increasing, converging asymptotically to
the maximum carrying capacity as the launch rate tends to infinity.
6 Extensions
We propose two refinements to our previous analysis. The first one introduces the possibil-
ity of cleaning up the orbital space through active debris removal. The second one relaxes
the assumption of market fluidity to take account of barriers to entry into the satellite ser-
vices market.
6.1 Active Debris Removal
The different active debris removal methods currently in development (e.g., laser, ion
beam, sail) to clean up polluted orbital regions are surveyed by Mark and Kamath (2019).
Here we amend our benchmark model to take these into account. In particular, we give an
estimate of the value of removing space debris, which is essential to identify promising
technologies.26
Pursuing our methodology, we focus on the long-term equilibrium of the orbital envi-
ronment. We thus consider a constant launch rate q and a constant debris removal rate r.27
Clearly, the number of space debris and the risk of collision are expected to be an increas-
ing function of q and a decreasing function of r. Actually, we show in “Appendix A1” that
the long-term risk of collision is given by 𝜇 = f (q − r∕𝛼) if q ≥ r∕𝛼 and 𝜇 = 0 otherwise.
Parameter 𝛼 > 0 is the number of debris fragments to be removed every year to offset the
long-term pollution of a marginal satellite launched every year.28
Going through the same steps as in previous sections, the social surplus generated
periodically by the satellite operators is given by S(q − r∕𝛼) = (v(q − r∕𝛼) − c)q. As the
expected present value v(q − r∕𝛼) of a satellite is a decreasing function, active debris
25
Complete results, including sensitivity analysis, are proposed in supplementary material posted online.
26
The issue of active debris removal is considered in Rao et al. (2020). However, they cannot assess its
economic value under their model.
27
We measure active debris removal in units of debris fragments (i.e., objects generated by collisions or
explosions). However, the technologies in development can target either small or large objects. Based on
the underlying physical model in “Appendix A1”, large space objects (i.e., intact end-of-life satellites and
rocket bodies) are equivalent to 𝜂 = 10, 000 debris fragments in the long run (Lafleur 2011).
28
Note that our benchmark model corresponds to the special case where r = 0.
13
716 S. Rouillon
Table 3 Numerical results with Launch (sat./y.) Marg.Dam. ($m./y.)

alternative specifications
Max. carrying capacity 10.68, 9.13, ∞ –
Phys.-eco. equilibrium 10.32, 10.27, 10.56 –
Socially optimal policy 5.02, 5.06, 4.43 131.93, 168.62, 114.04
removal increases the long-term social surplus generated by satellite operators. Precisely,
by first-order differentiation with respect to r, we get 𝜕S(q − r∕𝛼)∕𝜕r = − v� (q − r∕𝛼)q∕𝛼,
which gives the social benefit from removing a marginal debris fragment given that r
debris fragments are removed each year.
Under open-access, it should be clear that the space sector has no private incentive to
remove space debris, since no market is likely to emerge to remunerate this service. Indeed,
a satellite operator paying for the removal of one piece of debris would share the benefit
with all current and future satellite operators in the market. It follows that the physico-
economic equilibrium will be unchanged, even if active debris removal activity is socially
worth.
We finally calculate the social value of active debris removal in the physico-economic
equilibrium (i.e., for q = q∗ and r = 0). Formally, it is thus given by − v� (q∗ )q∗ ∕𝛼 . Again,
we rely on the calibration of Sect. 5. Based on Lafleur (2011), we also pose 𝛼 = 1370
debris fragments/satellite. Accordingly, in the physico-economic equilibrium, the social
value of eliminating one piece of debris fragment each year is equal to 19,031 $/fragment,
21,805 $/fragment and 10,902 $/fragment, under the benchmark, quadratic and linear spec-
ifications respectively.
6.2 Entry Barriers
Our benchmark model postulates that under open-access, either incumbent satellite opera-
tors or new entrants will launch satellites as long as they can expect a positive benefit from
them. However, this is relevant only if we postulate that firms can enter or exit the market
freely in the long run.
An alternative assumption is where the number of incumbent operators is given in the
short run and new entrants have to bear a setup cost F to participate in the market. Without
loss of generality, we normalize the latter such that each entrant can operate a fleet of at
most one satellite. Then, the overall capacity of the sector will vary in proportion to the
number of firms entering or exiting the market.
A company that pays F to enter the market at any time t0 expects to incur the cost k(q)
(defined below) of maintaining a fleet in order to receive in return the cash flow p peri-
odically. Presumably, it will do so as long as p∕𝛿 − k(q) > F . The expected cost in present
value k(q) of building up and replinishing a fleet of one satellite is defined by the recur-
rence relation:29
∞ ( )
∫t 0
k(q) = c + k(q)e−𝛿(t−t0 ) (𝜆 + 𝜇)e−(𝜆+𝜇)(t−t0 ) dt, with 𝜇 = f (q).
29
It is implicitly assumed here that each entrant anticipates remaining active forever.
13
Indeed, each entrant expends c to have its first satellite in orbit. The latter fails at time t
with probability density (𝜆 + 𝜇)e−(𝜆+𝜇)(t−t0 ). The satellite operator then finds itself back to
the starting point, which costs k(q) by definition. The expression above can be integrated
and solved, giving:
𝛿 + 𝜆 + f (q)
k(q) = c.
𝛿
Rearranging these results, we finally can show using (2) that new satellite operators will
enter as long as v(q) > c + 𝛿F∕(𝛿 + 𝜆 + f (q)).
Conversely, incumbent companies will eventually die whenever they cease to replenish
their satellite fleet, because the launch of new satellites remains unprofitable for a suffi-
ciently long period of time. Formally, this condition writes v(q) < c.
In this setting, the definition of a physico-economic equilibrium has to be modified. It is
reached for any q = q∗ such that c ≤ v(q∗ ) ≤ c + 𝛿F∕(𝛿 + 𝜆 + f (q)). Indeed, this character-
izes a situation where both the orbit reaches a stationary state and the number of incumbent
firms is constant, all working at full capacity.
This amended definition deserves several comments. To begin with, there exists an
infinite number of physico-economic equilibria. Formally, q∗ can take any value between
v−1 (c + 𝛿F∕(𝛿 + 𝜆 + f (q))) and v−1 (c). Clearly, some of these are more favourable than the
others in terms of the state of the orbital environment. In particular, the existence of entry
barriers can only reduce the risk of overexploitation. Finally, the physico-economic equi-
librium which eventually will emerge may depend on the orbit and market initial states and
on the speed at which firms react to changes in the profit perspectives.
7 Conclusion
In this paper, we have developed a simplified physico-economic model of orbital envi-

ronment and space activity. Focusing on the long-term equilibrium, we have assumed
that the risk of a satellite breakup by collision is an increasing and convex function of the
launch rate, becoming arbitrarily large for large enough launch rates. Then the long-term
expected population of functioning satellites is shown to be a inverted-U shaped function
of the launch rate. This setting allows to define the orbital maximum carrying capacity,
the physico-economic equilibrium launch rate (that would emerge under open-access) and
the socially optimal launch rate (that maximizes the surplus of the satellite operators). All
three outcomes have been compared, in order to highlight conditions of either physical or
economic orbital overexploitation. Finally, we have described economic instruments (com-
mand-and-control, tax, market) to regulate the space activity.
This work paves the way for much future research. An interesting theoretical extension
will be to build a physical model of the orbital environment, with three state variables,
one for the debris population, one for the end-of-life satellite population and one for the
active satellite population.30 The objective will be to use optimal control and differential
game theory to derive an optimal and a non-cooperative orbit exploitation trajectory. The
main contribution will be to focus on transition dynamics, instead of stationary states.
Another contribution will be to consider complementary strategies other than launch rates
30
A version is analyzed in “Appendix A1”.
13
718 S. Rouillon
for managing orbit congestion, such as automatic de-orbiting capabilities and active debris
removal. Ideally, the model should remain simple enough to be resolved analytically. Under
this condition, it will be possible to derive general results about the optimal design of satel-
lites and conditions for cost-effectiveness of space debris removal techniques.
In the longer term, applied research will be needed to design realistic models, better taking
into account the diversity of space debris and the relationships between the different altitudes
of the Earth’s orbit. At this global scale, a physico-economic equilibrium will be defined as a
state such that every region of the Earth’s orbit is exploited until all economic rents are dis-
sipated. However, the externalities generated by space objects will depend on their destination
altitude, due to the Earth’s attraction and atmospheric drag, determining their post-mission
residence time. In addition, altitude will play an asymmetric role, since satellites placed in
higher orbits represent a future threat to those placed lower, while the opposite is not true.
These mechanisms deserve to be studied and evaluated if appropriate economic incentives are
to be developed to ensure optimal exploitation of the Earth’s orbit.
Acknowledgements I would like to thank Vitali Braun from ESA/ESOC Space Debris Office for providing
data on past launches into low Earth orbit classified by altitude. I am indebted to Alan de Brauw, Tanguy
Bernard, Antoine Bouet, Pauli Lappi, Akhil Rao and Marc-Alexandre Sénégas, for their useful comments on
early drafts.
Appendix A
A Stylized Physical Model of the Orbital Environment
We describe here a physical model of low Earth orbit, based on Farinella and Cordelli (1991)
and Lafleur (2011). They introduce a system of two coupled differential equations, one for
the population of big objects and the other for the population of debris fragments. Big objects
include intact satellites, either active or inactive, and rocket bodies. They typically have a
cross-section of a few square meters and a mass of hundreds of kilograms. Debris fragments
are objects generated by collisions or explosions, with a size of a few centimeters and a mass
of a few grams, capable of causing catastrophic breakup when impacting a satellite.
The main difference of our model with Farinella and Cordelli (1991) and Lafleur (2011)
is that we divide the set of big objects into two parts, in order to distinguish active satellites
from other inactive big bodies. This implies introducing a third differential equation for
the population of operational satellites. Unless assuming infinitely lived satellites, as Rao
and Rondina (2019) do, this extension is necessary to formalize the operational lifetime.
Another difference of our model is that we focus on a subregion of the low Earth orbit,
whereas they treat it as a whole homogenous system.
We use the following notations:
q(t) = rate of active satellites launched;
r(t) = rate of debris fragments removed;
r(t) = rate of inactive satellites removed;
x(t) = population of debris fragments;
y(t) = population of inactive satellites;
z(t) = population of active satellites.
13
We formalize the evolution of the orbital environment with the following dynamical
system:31
⎧ x(t)
̇ = 𝛼q(t) − r(t) + s(t) − 𝛽x(t) + 𝜂𝜑(x(t))(y(t) + z(t)), x(0) = x0 ,
⎪
⎨ y(t)
̇ = 𝛼q(t) − r(t) + 𝜆z(t) − 𝜑(x(t))y(t), y(0) = y0 ,
⎪ z(t)
̇ = q(t) − (𝜆 + 𝜑(x(t)))z(t), z(0) = z0 .
⎩
The first differential equation describes the evolution of the population of debris fragments,
x(t). The first two terms, 𝛼q(t) and r(t), result from space activity. On the one hand, 𝛼q(t)
fragments are released as a byproduct of satellite launches (i.e., explosion of rocket bodies
and space objects), with 𝛼 > 0 the number of fragments per launch. On the other hand, r(t)
fragments are removed, as a result of active debris removal activities.32 The third term, s(t),
reflects pieces of debris coming from nearby altitudes, either scattered after collisions or
decayed by the atmospheric drag. Inversely, the fourth term, 𝛽x(t), represents the decay of
the stock of debris fragments due to the atmospheric drag, with 𝛽 > 0 the inverse of their
average orbital lifetime. The fourth term, 𝜂𝜑(x(t))(y(t) + z(t)), gives the addition of debris
fragments generated by collisions of debris fragments with intact satellites (either active or
inactive), with 𝜂 > 0 the numbers of fragments per collision and 𝜑(x(t)) ≥ 0 the rate of col-
lision per unit of intact satellite.
The second differential equation represents the evolution of the population of inactive
satellites, y(t). The first two components, 𝛼q(t) and r(t), refer respectively to the rate of
inactive satellites released by launching activities (i.e., rocket upper stages), with 𝛼 > 0
their number per launch, and disposed of through debris removal activities. The third term,
𝜆z(t), represents operational satellites arriving at the end of their lifetime, thus becom-
ing inactive, with 𝜆 > 0 the inverse of their average operational lifetime. The last term,
𝜑(x(t))y(t), refers to the number of non-operational satellites destroyed as a result of
collisions.
The last differential equation gives the evolution of the population of operational sat-
ellites, z(t). The first component, q(t), is the result of lauching activity by the space sec-
tor. The last term, (𝜆 + 𝜑(x(t)))y(t), refers to the number of satellites that cease to oper-
ate, either for technical reasons (i.e., fuel, breakdowns) or environmental reasons (i.e.,
collisions).
Now, let us consider constant rates of satellite launches and debris removals forever,
i.e., q(t) = q, r(t) = r and r(t) = r for all t. Assume further that the nearby altitudes are in a
stationary state, implying that s(t) = s for all t. We wish to calculate the stationary states of
31
To simplify, we neglect the orbit decay of intact satellites, and collisions between debris fragments (i.e.,
x–x) and between intact satellites (i.e., y–y, y–z and z–z). The same simplifying assumptions are also used
by Farinella and Cordelli (1991) and Lafleur (2011). They are justified as follows. Orbit decay of intact
satellites is found to be negligible by Lafleur (2011). Collisions between debris fragments generate second-
order fragments too small to cause catastrophic breakup. Collisions between intact satellites is negligible
according to simulations by Lafleur (2011). This is even more true for operational satellites, which can per-
form avoidance manoeuvres, knowing that intact satellites are large enough to be traced by radar.
32
Active debris removal activities have attracted attention recently and might become profitable in the
future (Somma et al. 2017). Although not considered in our benchmark model, active debris removal can be
seen as implicit in our framework, as long as the effort of debris removal remains constant over time (either
in absolute value or relative value, i.e. as a ratio of the stock of debris). It is briefly discussed in an exten-
sion in Sect. 6.1.
13
720 S. Rouillon
the dynamical system, i.e., x(t) = x∗, y(t) = y∗ and z(t) = z∗ such that x(t)
̇ = y(t)
̇ = z(t)
̇ =0
for all t.
Note that the corresponding system
̇ = 𝛼q − r + s − 𝛽x∗ + 𝜂𝜑(x∗ )(y∗ + z∗ ) = 0,
x(t)
̇ = 𝛼q − r + 𝜆z∗ − 𝜑(x∗ )y∗ = 0,
y(t)
̇ = q − (𝜆 + 𝜑(x∗ ))z∗ = 0,
z(t)
is equivalently written as
(( ) )
̇ + 𝜂 y(t)
x(t) ̇ = 𝛼q − r + s + 𝜂 𝛼 + 1 q − r − 𝛽x∗ = 0,
̇ + 𝜂 z(t)
( )
̇ = 𝛼 + 1 q − r − 𝜑(x∗ )(y∗ + z∗ ) = 0,
̇ + z(t)
y(t)
̇ = q − (𝜆 + 𝜑(x∗ ))z∗ = 0.
z(t)
It follows that
(( ) )
∗
𝛼q − r + s + 𝜂 𝛼 + 1 q − r
x = ,
𝛽
∗
( )
(𝜆 + 𝜑(x )) 𝛼q − r + 𝜆q
y∗ =
𝜑(x∗ )(𝜆 + 𝜑(x∗ ))
and
q
z∗ = .
𝜆 + 𝜑(x∗ )
Assuming that q > 0, 𝛼q − r + s ≥ 0 and 𝛼q − r ≥ 0, all results are strictly positive. The
last two inequalities mean that the space sector removes less space debris than periodically
generated. This assumption is physically necessary
( in)the long run.
To economize on notations, let 𝛼 ≡ 𝛼 + 𝜂 𝛼 + 1 and r ≡ r + 𝜂r . Parameter 𝛼 is the
number of debris fragments to be removed each year to offset the long-term pollution of a
marginal satellite launched each year. The aggregate r is the rate of active debris removal,
measured in units of debris fragments. Then, we can rewrite the long-term number of
debris fragments as x∗ = (𝛼(q − r∕𝛼) + s)∕𝛽.
The above closed-form expressions rationalize our model in Sect. 2. Assuming implic-
itly that r = 0 and s = 0, we have postulated the existence of a function x = 𝜙(q), giving
the long-term number of debris as a function of the launch rate q. Our calculus here sug-
gest that a possible specification is 𝜙(q) = x∗ = 𝛼q∕𝛽 . Also, we have defined a function
𝜇 = f (q), giving the long-term collision risk as a function of the launch rate q. Our calculus
here show that a possible specification is f (q) = 𝜑(x∗ ) = 𝜑(𝛼q∕𝛽).33 Finally, we have cal-
culated the long-term expected population of satellites as n(q) = q∕(𝜆 + f (q)). The fact that
this coincides with the expression of z∗ obtained here confirms again that the dynamical
system considered above is a way to rationalize our approach in the main text.
Also, let us note that the above stationary solution highlights some assumptions left
implicit in the main text. The function f(q) depends not only on natural parameters:
33
In Sect. 6.1, the generalization for the case where r > 0 follows directly.
13
• the physical laws of the orbit, here formalized by the rate of decay of debris fragments,
𝛽 , and the function describing the rate of collision, 𝜑;
• the state of nearby orbits, represented here by the debris fragments coming from other
altitudes, s;
but also on artificial parameters under the control of the space sector:
• the number and types of objects released per launch, 𝛼 and 𝛼 ;

• the number of debris fragments generated by a destroyed satellite, 𝜂;
• the number of space debris removed periodically, r and r.
We can supply conditions such that the stationary solution is locally stable. Linearization
of the dynamical system around it gives
̇ ⎤
⎡ x(t) ⎡ x(t) − x∗ ⎤ ⎡ 𝛼q − r + s ⎤
̇ ⎥ = A⎢ y(t) − y∗ ⎥ + ⎢ 𝛼q − r ⎥,
⎢ y(t)
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
̇ ⎦
⎣ z(t) ⎣ z(t) − z∗ ⎦ ⎣ q ⎦
where
⎡ −𝛽 + 𝜂𝜑� (x∗ )(y∗ + z∗ ) 𝜂𝜑(x∗ ) 𝜂𝜑(x∗ ) ⎤

A=⎢ −𝜑� (x∗ )y∗ − 𝜑(x∗ ) 𝜆 ⎥.
⎢ ⎥
⎣ −𝜑� (x∗ )z∗ 0 − 𝜆 − 𝜑(x∗ ) ⎦
A necessary and sufficient condition for this linearized dynamical system to be stable is for
matrix A to have all its eigenvalues with a negative real part. The corresponding character-
istic polynomial is34
( ( ) )
det (A − 𝜈I) = −(𝜈 + 𝜆 + 𝜑(x∗ )) 𝜈 2 + 𝛽 + 𝜑(x∗ ) − 𝜂𝜑� (x∗ )(y∗ + z∗ ) 𝜈 + 𝛽𝜑(x∗ )
The eigenvalues of matrix A are35

𝜈1 = −𝜆 − 𝜑(x∗ ),
( ( ) )
1 √ − 𝛽 + 𝜑(x∗ ) − 𝜂𝜑� (x∗ )(y∗ + z∗ )
𝜈2 =
2 + (𝛽 + 𝜑(x∗ ) − 𝜂𝜑� (x∗ )(y∗ + z∗ ))2 − 4𝛽𝜑(x∗ )
and
( ( ∗ � ∗ ∗ ∗
) )
𝜈3 =
1 √ − 𝛽 + 𝜑(x ) − 𝜂𝜑 (x )(y + z ) .
2 − (𝛽 + 𝜑(x∗ ) − 𝜂𝜑� (x∗ )(y∗ + z∗ ))2 − 4𝛽𝜑(x∗ )
All eigenvalues have negative real part if and only if
34
To avoid any confusion, recall that the notation 𝜆 refers in our model to the rate of technical failure of an
operational satellite. This is the reason why we denote the eigenvalues by 𝜈 , departing from the usual nota-
tion.
35
The last two conjugate eigenvalues can be either real or complex, depending on the sign of the term
under the square root.
13
722 S. Rouillon
𝛽 + 𝜑(x∗ ) − 𝜂𝜑� (x∗ )(y∗ + z∗ ) > 0.
Using
( )
𝛼+1 q−r
y∗ + z ∗ = > 0,
𝜑(x∗ )
this condition is equivalently written as
𝜑� (x∗ ) 𝛽 + 𝜑(x∗ )
∗
< (( ) ).
𝜑(x ) 𝜂 𝛼+1 q−r
Multiplying by x∗, we obtain

x∗
𝜎(x∗ ) < (𝛽 + 𝜑(x∗ )) (( ) ),
𝜂 𝛼+1 q−r
where we define 𝜎(x) ≡ 𝜑� (x)x∕𝜑(x), that is, the elasticity of the rate of collsion 𝜑(x). Since
(( ) )
∗
𝛼q − r + s + 𝜂 𝛼 + 1 q − r
x = ,
𝛽
we have
) ≥ ,
x∗ 1
(( )
𝜂 𝛼+1 q−r 𝛽
which implies finally that

𝜑(x∗ )
𝜎(x∗ ) < 1 + .
𝛽
This shows that a sufficient condition for stability of any stationary solution is that 𝜎(x) ≤ 1
for all x. However, given a stationary solution x∗, a weaker condition will be necessary,
depending on the ratio of the rate of collision, 𝜑(x∗ ), with the rate of decay of debris frag-
ments, 𝛽.
References
Adilov N, Alexander PJ, Cunningham BM (2014) An economic analysis of earth orbit pollution. Environ
Resource Econ 60(1):81–98
Adilov N, Alexander PJ, Cunningham BM (2018) An economic “Kessler Syndrome”: a dynamic model of
earth orbit debris. Econ Lett 166:79–82
Christensen C (1997) The innovator’s dilemma: when new technologies cause great firms to fail. Harper
Business, New York
Cordelli A et al (1993) Future collisional evolution of earth-orbiting debris. Adv. Space Res. 13(8):215–219
Dolado-Perez JC, Pardini C, Anselmo L (2015) Review of uncertainty sources affecting the long-term pre-
dictions of space debris evolutionary models. Acta Astronaut 113:51–65
Dos Santos Paulino V, Le Hir G (2016) Industry structure and disruptive innovation: the satellite industry. J.
Innov. Econ. Manag. 20(2):37–60
Drmola J, Hubik T (2018) Kessler syndrome: system dynamics model. Space Policy 44–45:29–39
13
Eichler P, Reynolds RC (1997) Synergistic use of debris environment models for flexible, very long-term
projections. Acta Astronaut 40(12):843–858
ESA (2019) ESA’s annual space environment report
Farinella P, Cordelli A (1991) The proliferation of orbiting fragments: a simple mathematical model. Sci
Glob Sec 2:365–378
Gordon HS (1954) The economic theory of a common-property resource: the fishery. J Polit Econ
62(2):124–142
Hardin G (1968) The tragedy of the commons. Science 162(3859):1243–1248
Hiriart T et al (2009) Comparative reliability of GEO, LEO, and MEO satellites. In: Proceedings of the
international astronautical congress, IAC-09.D1.6.1
Kessler DJ, Cour-Palais BG (1978) Collision frequency of artificial satellites: the creation of a debris belt. J
Geophys Res 83:2637–2646
Kessler DJ, Anz-Meador PD (2001) Critical number of spacecraft in low earth orbit: using satellite frag-
mentation data to evaluate the stability of the orbital environment. ESA 473:265–272
Kessler DJ et al (2010) The Kessler syndrome: implications to future space operations. Adv Astronaut Sci
137:201
Klima R et al (2016) Space debris removal: a game theoretic analysis. Games 7:20. https://doi.org/10.3390/
g7030020
Lafleur JL (2011) Extension of a simple mathematical model for orbitaldebris proliferation and mitigation.
In: Advances in the astronautical sciences, vol 140. pp 11–173
Macauley MK (2015) The economics of space debris: estimating the costs and benefits of debris mitigation.
Acta Astronaut 115:160–164
Mark CP, Kamath S (2019) Review of active space debris removal methods. Space Policy 47:194–206
Muller C, Rozanova, O, Urdanoz, M (2017) Economic valuation of debris removal. In: 68th international
astronautical congress. National research council, 2011. Limiting future collision risk to spacecraft: an
assessment of NASA’s meteoroid and orbital debris programs. The National Academies Press, Wash-
ington, DC. https://doi.org/10.17226/13244
Rao A, Rondina G (2019) Debris and collision risk in the orbital commons. Manuscript. https://akhilrao.
github.io/assets/working_papers/Cost_in_Space.pdf
Rao A, Burgess MG, Kaffine D (2020) Orbital-use fees could more than quadruple the value of the space
industry. Proc Natl Acad Sci 117(23):12756–12762
Salop SC (1979) Monopolistic competition with outside goods. Bell J Econ 10:141–156
Sandler T, Schulze M (1981) The economics of outer space. Nat Resour J 21:371–393
Satellite Industry Association (2017) State of the satellite industry report. pp 1–33
Schaefer MB (1957) Some considerations of population dynamics and economics in relation to the manage-
ment of marine fishes. J Fish Res Board Can 14:669–81. https://doi.org/10.1139/f57-025
Somma GL, Colombo C, Lewis HG (2017) A statistical LEO model to investigate adaptable debris control
strategies. In: 7th European conference on space debris, Darmstadt, Germany. https://www.researchga
te.net/publication/319123620_A_Statistical_LEO_Model_to_Investigate_Adaptable_Debris_Contr
ol_Strategies
Somma GL, Lewis HG, Camilla C (2019) Sensitivity analysis of launch activities in low earth orbit. Acta
Astronaut 158:129–139
Stokes H et al (2017) Status of the ISO space mitigation standards. In: Proceedings 7th European conference
on space debris. https://conference.sdo.esoc.esa.int/proceedings/sdc7/paper/979/SDC7-paper979.pdf
UN (1999) Technical report on space debris. New York
UN (2010) Space debris mitigation guidelines of the commitees on the peaceful uses of outer space. Vienna
Union of Concerned Scientists (2019) Satellite database. https://www.12FL02ucsusa.org/nuclearweapons/
space-weapons/satellite-database. Accessed 8 July 2019
Weitzman M (1974) Price versus quantity. Rev Econ Stud 41(4):477–491
Weinzierl M (2018) Space, the final economic frontier. J Econ Perspect 32(2):173–192
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
13

A Physico Economic Model of Low Earth Orbit Management

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Physico Economic Model of Low Earth Orbit Management

Uploaded by

Copyright:

Available Formats

Environmental and Resource Economics (2020) 77:695–723

A Physico‑Economic Model of Low Earth Orbit Management

Accepted: 3 October 2020 / Published online: 17 October 2020

Keywords Space economics · Orbital debris · Sustainability

Mathematics Subject Classification L1 · L9 · Q2

Electronic supplementary material The online version of this article (https​://doi.org/10.1007/s1064​

overexploitation happens when the launch rate in the physico-economic equilibrium is

2 The Physical Model

a relationship 𝜇 = 𝜑(x), giving the instantaneous risk of destruction of a satellite by col-

Proof of ( Property 1 The ) first part is immediate. The derivative of n(q) is

Under Assumption 1, the curve representing the long-term expected population of

Definition 1 The orbital maximum carrying capacity n is the maximum expected

3 The Economic Model

Fig. 1 Long-term risk of colli-

Fig. 2 Long-term expected num-

We can derive the following:

Proof of Property 2 The first part is immediate. The derivative of v(q) is

Definition 2 A physico-economic equilibrium is reached when the expected present

Fig. 3 Expected present value

We call this situation a physico-economic equilibrium because it characterizes a situa-

Proposition 1 If p∕(𝛿 + 𝜆) > c, there exists a unique physico-economic equi-

Proposition 1 is illustrated by Fig. 3, assuming that p∕(𝛿 + 𝜆) > c. The physico-eco-

Property 3 In a physico-economic equilibrium, d∗ = 1∕(p∕c − 𝛿) > 0.

Proof of Property 3 By Proposition 1, q∗ satisfies v(q∗ ) = p∕(𝛿 + 𝜆 + f (q∗ )) = c. Elemen-

Proposition 2 If p∕(𝛿 + 𝜆) > c, there exists a unique optimal policy, characterized

Proof of Proposition 2 By definition of S(q) and v(q), we have

It is worth comparing the physico-economic equilibrium to the optimal policy. From an

Property 5 The physico-economic equilibrium always induces an economic over-exploi-

In this section, we describe standard economic instruments (command-and-control, tax

Fig. 4 Expected revenue and cost

Fig. 5 Physico-economic equilib-

Fig. 6 Optimal policy (where

with 𝜃 > 0 and 0 ≤ q < Q.

Based on this information, we propose to use the benchmark calibration summarized in

Table 1 Benchmark calibration

f (10) 0.15 %/year Farinella and Cordelli (1991)

Table 2 Numerical results

Max. carrying capacity 10.68 36.47 3.42 –

Fig. 7 Maximum carrying capac-

Fig. 8 Physico-economic equi-

Fig. 9 Expected lifetime of a

Fig. 10 Sensitivity analysis (over 15,625 simulations)

Fig. 11 Alternative specifications

corresponding outcomes to give a representative picture of all possible configurations.22

6.1 Active Debris Removal

Table 3 Numerical results with Launch (sat./y.) Marg.Dam. ($m./y.)

In this paper, we have developed a simplified physico-economic model of orbital envi-

A Stylized Physical Model of the Orbital Environment

• the number and types of objects released per launch, 𝛼 and 𝛼 ;

⎡ −𝛽 + 𝜂𝜑� (x∗ )(y∗ + z∗ ) 𝜂𝜑(x∗ ) 𝜂𝜑(x∗ ) ⎤

The eigenvalues of matrix A are35

All eigenvalues have negative real part if and only if

𝛽 + 𝜑(x∗ ) − 𝜂𝜑� (x∗ )(y∗ + z∗ ) > 0.

Multiplying by x∗, we obtain

which implies finally that

You might also like

Electronic supplementary material The online version of this article (https://doi.org/10.1007/s1064

2 The Physical Model

3 The Economic Model

6.1 Active Debris Removal