Enterprise in orbit: The supply of communication satellites

Abstract

This study develops a general understanding of the evolution of the Western World commercial communication sate-llite supply industry. Initial information classifies this industry as an oligopoly with the vendors as price setters. Over three generations, the technical attributes of communication satellites are advancing. Taking the hedonic regression approach, the customer’s willingness to pay for characteristics is calculated. Exploring the relationship between price and complexity, it is demonstrated that in the short run, the oligopoly structure of this industry is accompanied by a simple form of cost plus price setting, whereas in the long run, the engineering satellite pricing ‘rule of thumb’ applies.

Keywords:

• Hedonic prices
• Technical change
• Mark-up pricing
• Oligopoly

Acknowledgements

This work is the result of studies for a D.Phil. in economics at the University of Oxford under the supervision of Prof. Paul A. David who deserves my sincere gratitude. I would like to thank William Harvey of Spar Aerospace Ltd., Canada, for letting me use their copy of the database compiled by Roger Rusch (1992 Edition of Communications Satellite Data Bases), the Canadian and European Space Agencies for graciously letting me use their highly specialized libraries and the Social Sciences and Humanities Research Council of Canada for funding this research. I would also like to acknowledge helpful comments from my D.Phil. examiners, Bronwyn Hall and Patrick Cohendet, from the participants of the 1999 EARIE conference, from the participants of the 1999th edition of ESSID, particularly Paul Geroski and José Mata, and from an anonymous referee. All remaining mistakes are my own.

Notes

¹The data utilized for the present study cover 29 years and include contracts for commercial communication satellites from 1964 to the end of 1992. This represents the data available at the time of the study and were not updated for the following reasons. First, it would have taken a considerable amount of time to update it accurately. Secondly, the decision to exclude from this study the contracts signed from 1993 onwards resides in the dramatic change, after this date, not only in the structure of the industry but also in the technology, introduced by the new smaller and cheaper ‘constellations’ of LEO satellites such as the Motorola Iridium and Qualcomm Globalstar programs. On the manufacturing side of the satellite industry, the advent of these constellations of satellites implies the introduction of production lines for the fabrication of more than 60 identical satellites compared with an average of two satellites per contract prior to 1993. Considerable economies of scale can therefore be obtained and the price per spacecraft should decrease dramatically. Moreover, these satellites require less power for successful transmissions, as they are closer to the surface of the planet, making savings on the technology as well. On the demand side, i.e. the supply of communication satellite services, the launch of these LEO world-covering constellations of satellites should give a competitive advantage to their owner and drive some less competitive players out of the market, especially in the field of mobile telecommunications. In addition, from a technology point of view, these satellites are in a class of their own. They orbit the earth on a much lower orbit and, to reach their orbital positions, smaller launch vehicles costing substantially less can be used. Being much smaller and lighter, these satellites can also be launched in bunches, thus further reducing the launch costs. In any case, as the price considered here corresponds to satellite contracts, they would appear as outliers in this study because although in terms of number of satellites, they contribute a substantial amount, they still represent a very small number of contracts in the database.

²Secondary sources of data, used to complete and correct the main database, comprise the European Space Directories (1992, 1994), which provided general technical attributes data about satellites planned or in operation, in the beginning of the 1990s: number of satellites purchased, geostationary transfer orbit mass, geostationary orbit mass, number and frequency band of transponders and effective isotropic radiated power. The third edition of the Space Satellite Handbook (Curtis, 1994) was used to compare the planned and actual launch dates. The third edition (1995/96) of the World Satellite Yearly (Baylin, 1994) as well as Matra Marconi 1993/94 and 1995/96 editions of the World Communication Satellite Directory (Matra Marconi Space 1993 and 1995) provided invaluable information to complement and update the Rusch database. Euroconsult’s 1993 edition of the World Space Markets Survey, Ten Year Outlook (Euroconsult, 1994), was a third source of technical data, providing numbers and frequency band of transponders as well as a great understanding of the satellite market.

³For the purpose of the graph, the separation into six sub-periods originates from the waves of satellite orders which are characterized by stagnation in 1971, 1975 and 1984 and a slow down in orders in 1979 and 1988.

⁴Three types of services are generally considered in this industry: fixed satellite services (FSS) used for standard telecommunications and broadcasting feeds, broadcasting satellite services (BSS) reserved for direct broadcasting services to the end user’s satellite receiver and mobile satellite services (MSS) commonly utilized for air (aircraft) and sea (boats) communications.

⁵Please consult Table A1 in the appendix for a description of the frequency bands to which these letters correspond.

⁶For the purpose of the article, the three overlapping time periods chosen to represent the three generations are 1964–79, 1976–84 and 1980–93.

⁷Deflated by the US consumer price index. The majority of satellites included in this study were manufactured in the USA (See Tab. A2 in the appendix). Berndt (1991), based on the work of Montgomery and Day (1985), suggested that in calculating learning effects, an overall deflator is preferable to an industry-specific output deflator, such as the PPI, because if the latter is used, the learning curve effect may be washed out owing to the price deflator already capturing a substantial portion of the learning-induced productivity gains in that particular industry. Ultimately, this hedonic analysis was used to calculate learning effects (although not presented in this article). It is believed that using the PPI for communication equipment (PCU366) would cloud the hedonic results. In addition, the CPI is used as a deflator on the grounds that groups of years are used as dummy variables as opposed to single years (and sometimes as many as 4 years grouped together). The price of these groups of years needs to be in constant prices. These are the main reasons for using a deflator on the price as opposed to none as is commonly used in hedonic calculations.

⁸This last variable, the number of satellites ordered in a given contract, has decreased steadily over the years (see Tab. A3 in the appendix).

⁹The method employed by engineers accounts for every single electronic component—number of transistors, square inch of the solar cell, number of batteries and so on—included in the design. It also incorporates the costs of engineering models, flight models as well as the actual spacecraft. Engineering and flight models are preliminary prototypes of the spacecraft developed to test its specifications and are therefore not part of the final product delivered to the customer. The engineering method of evaluation for the cost of subsystems uses a mass^complexity functional form for each satellite subsystem summed over all subsystems.

¹⁰Dummy variables for various manufacturers, customers and producing and buying countries were introduced in the regressions and none were significant. Hedonic regressions for different manufacturers and customers were also performed and the tests for structural change mildly rejected the hypothesis of a common regression between manufacturers from Europe and the USA. Owing to the small number of satellites produced in Europe at any period of time, the power of the Chow tests is very low and the regional distinction is questionable. Consult Table A2 in the appendix for a breakdown of satellites built and owned by different producers, owners. The distinction between manufacturers and customers will therefore be ignored in the rest of the article.

¹¹The number of satellites ordered in each contract is treated as exogenous here and therefore should not create the simultaneity bias expected in such functional forms. If the dependent variable was the total value of the contract, i.e. the price of all the satellites ordered in the contract, then its dependence on the number of satellites ordered would be obvious. By using the price of each satellite as the left-hand side variable, potential bias is avoided. Studying the bidding for satellite contracts, Beaudry (1994) discovered that customers, in general, specify the number of satellites they require in the request-for-proposal to the potential bidders. Having this variable decided by the sellers prior to the bid and the contract signature, designates this variable as exogenous to the bidding.

¹²In this article, the functional form used to estimate satellite costs in industries to the total launch mass of the satellite raised to the power of the complexity of the entire satellite is generalized. It is a somewhat different method unless the mass of each sub-system is identical, and indeed it is not. The Eng functional form is only loosely inspired from the method used in industries.

¹³A value significantly different from zero would violate our exogeneity assumption, but this is not the case.

¹⁴This result is obviously obtained from the Econ model, but we will also see in the next section that this is also the case for the Eng model which yields a positive value of complexity to the power of which the Mass is elevated.

¹⁵This rule of thumb consists of a price per kilogram of the satellite weight.

¹⁶There is an increasing tendency in the industry to use standard platforms such as the HS-601, a three-axis stabilized platform built by Hughes Aircraft, or the GE-3000, another three-axis stabilized platform built by Lockheed Martin. This, however, does not preclude firms from entering the communication satellite market. In addition, no evidence of economies of scale relative to the use of standard platforms was found in the analysis of the data. This usage of standard platforms probably contributed to the reduction in development time observed over time.

¹⁷The coefficients of the dummy variable AmpType are all negative, implying that customers are willing to pay less when satellites only use ‘old’ travelling wave tube amplifiers, as opposed to a combination of this old technology with the ‘new’ solid state power amplifiers.

¹⁸This can be seen from the negative value of all transponder frequency proportion coefficients, implying that hybrid mutli-band satellites have a positive effect on the price.

¹⁹The inclusion of this variable follows from the idea that the longer it takes to build a spacecraft, the more complex it is likely to be.

²⁰The price bears a constant proportionality relationship to the costs if and only if: (a) the regression line goes through the origin, or (b) the regression line is shifting at the same rate as the complexity is changing through time. In the present case, (b) is the most likely explanation.

Additional information

Notes on contributors

Catherine Beaudry

Tel.:[phone] (ext. 3357); Fax:[phone]; Email: [email protected]