Survey procedure Survey sample

P.C. Mayer r Energy Economics 22 2000 319]330 322 generator capacity minus the capacity of the largest generator under planned operation always greater than peak load. Generators under planned operation refer to those not under planned maintenance. Another reflection of economies of scale is the installed reserve margin required for given reliability. Installed reserve margin is the difference between system capacity and annual peak load, here measured as percent of annual peak load. Larger scale allows a lower percent margin } thus, lower unit cost } for given reliability but the relationship is also influenced by the size of existing generators. The larger the largest generator } or perhaps the largest two, three, four or... generators } the greater the reserve margin required for given reliability. 1 As can be seen from comparing the first and fourth columns of Table 1, capacity of the largest generator and system peak load are highly correlated. Such correla- tion is expected since a power system is expected to take greater advantage of the economies of scale of individual generators with greater system load. This correlation limits the statistical estimation from a cross-section of power systems to combining impact of generator size and peak demand. The estimation is through regressing the percent intended reserve margin as a function of a ratio between the capacity of the largest generator and peak demand. Other variables, Ž . besides the ratio between capacity of largest generator s and peak, might affect intended reserve margin. The system size, measured by peak load, might matter. To operate a smaller system a lower intended reserve margin might be chosen because a smaller system may require more expensive generator capacity to provide redundancy. Also tested were regressions including having hydropower and using computer simulation programs to estimate system reliability; these variables may influence intended reserve margin. Nature and technology provide different operational uncertainties for hydro generation than for fossil-fuel generation. Furthermore, seasonal variation in hydro capacity is likely since tropical islands often have wet]dry seasonality. Use of computer programs to estimate reliability may reflect an attitude toward reliability or directly influence the choice of intended reserve margin. No hypotheses are made for the direction of the impact of these two variables.

3. Survey procedure

The generic cover letter and mail survey questionnaire for utilities believed to Ž operate more than one independent power system lie in Appendix A. As seen . from the letter, the survey’s purpose was benchmarking. For utilities known to operate only one system, the cover letter and survey questionnaire differed slightly: 1 This is a case of a firm’s production options depending on the firm’s historical investment decisions. Ž . Smith 1996 provides a thorough discussion of the impact of previous investment on current options. P.C. Mayer r Energy Economics 22 2000 319]330 323 the letter omits reference to multiple systems and the questionnaire omits the request for the system name since the utility, alone, is adequate identification. Some written responses required telephone calls or fax correspondence to clarify their meaning. Questionnaires with cover letters were sent to the Hawaiian utilities and the utility membership of the Pacific Power Association and Carllec, the association of Caribbean power utilities. Before mailing, the questionnaire and cover letter were tested with three of these utilities. The three utilities used for testing are in the sample. The survey covered slightly fewer than 46 utilities and an uncertain number of independent power systems. Some of the 46 addresses were duplications through a utility being a subsidiary of another. The number of power systems operated by many of the utilities is uncertain.

4. Survey sample

Responses covered 39 independent power systems and 24 utilities; responses useful for statistical analysis covered 37 systems and 22 utilities. The parts of the responses ordered by system peak load used for statistical analysis are listed in Table 1. The survey requested ‘Please list highest-rated unit capacities; please list at least the three largest’. Five in the sample of 37 used in the statistical analysis listed only the two largest } four probably had only two generating units } and one listed only one. Note that statistical analysis using the largest unit will give similar results to analysis using the sum of the two or three largest units; the correlation of capacity between any two of the three largest units is greater than 99.

5. Survey analysis