A final possibility for predicting mature cable penetration is to utilize the projections provided by the system operators themselves. Fortunately, such projections have been provided on a confidential basis 11 to the NCTA by large multiple system owners. We shall consider this alternative in the next section after testing the existing statistical demand models against new data.

3. Operating Costs

Operating costs are defined as those annual, recurring, noncapital expenditures required to provide cable service to subscribers. Unfortunately, precise data on these costs are not normally available from the nation's operating systems. Most systems do not report financial statistics to the public since their securities are not publicly traded. Moreover, systems operating under the new FCC regulations have not had sufficient experience in complying with these rules to provide good estimates of their contribution to operating costs. And, summaries of their financial reports to the FCC have not as yet been made public.

Because of these difficulities, most analysts of cable system profitability are forced to rely upon Comanor-Mitchell data--collected during an NCTA study--for estimates of operating costs. We shall be forced to do the same, despite the criticism which has been leveled at their estimates for being too high12 We are especially interested in examining the influence of system size upon operating costs per subscriber, and we shall stress the importance of calculating rates of return for only those systems within the range of efficient scale of operation. Very small, inefficient systems must be excluded from any analysis of profitability of "typical" cable systems.

11Don Andersson, op. cit.

12For example, see John J. McGowan, Roger Noll, and Merton J. Peck's technical memorandum prepared for the Sloan Commission, 1971.

As in the case of revenues, the timing as well as the magnitude of capital costs are critically important to the calculation of rates of return. The longer a system spreads out its initial investment in a distribution system, the higher is its rate of return for a given pattern of subscriber growth. We shall see that Mitchell assumes that all investment expenditures except house drops are completed at the beginning of the first year. Alternative assumptions which we employ are more realistic.

The capital cost of a cable system is largely dictated by the geographical characteristics of the area served. The density of living units, the necessity of burying utility cables, the type of topography, and the distance from transmitters of signals to be imported are among the most important determinants of the cost of capital facilities per home passed by the cable. But, the average capital cost per home subscriber is obviously dependent also upon the percentage of homes actually subscribing; hence, the predicted rate of penetration is especially important in determining capital outlays per subscriber.

The cost of constructing a mile of cable distribution plant varies considerably from location to location. It is virtually impossible to posit a single estimate for aerial construction or for underground construction, and it is for this reason that we expect subscribers in different communities to face quite different charges. For the purposes of this analysis, however, we shall be forced to adopt Mitchell's technique of positing a typical system and to utilize the same cost data for all calculations.

To summarize this section and to anticipate our criticisms of Mitchell in the next section, the major determinants of rate of return are the prediction of ultimate penetration, the path which the approach to maturity takes, the future pattern of revenues per subscriber, and the magnitude and timing of capital expenditures. On most of these points, Mitchell's assumptions dictate results which, in our view, are biased in the direction of low profitability. We shall correct these sources of bias in preparation for calculating the rates of return which typical cable systems may be expected to enjoy in future years.

C. SELECTION OF APPROPRIATE VALUES FOR THE VARIABLES

1. Calculation of Internal Rates of Return

Mitchell does not inform us of his precise methodology for calculating the yield on cable investment; he merely states that returns are calculated under the assumption of fixed equipment life but infinite franchise life. Thus, he simply replicates the cable system periodically in order to calculate rates of return, but he fails to tell us about the precise timing of revenues and expenditures. Our experiments with Mitchell's data suggest that he must have assumed that all capital expenditures are undertaken immediately preceeding the first day of the year and that revenues and operating costs are recorded at the beginning of each year. As a more satisfactory approach, we have chosen to center all revenues and operating costs at the middle of each year-a compromise which is compelled by the sheer difficulty of allowing for continuous flows and the discounting of these flows--but we retain the assumption that capital outlays occur at the beginning of the year in which they are incurred.

In addition, we calculate returns under the assumption of four-generation life. All expenditures are replicated in the same year they fall due in the first generation, and subscription levels are allowed to grow at the same rate in the second, third and fourth generations-namely, 2 percent per annum. Thus, our computation algorithm

solves for r in the following equation:

Our assumption of only four-generation life--of fifteen years each--will have little effect upon the rate of return; one dollar of net revenue sixty years hence, discounted at a rate of 15 percent, has a current discounted value of less than one-tenth of one cent.

a.

Subscription Revenues. Among the most conservative and downward-biased parameters used by Mitchell is his estimate of subscription fees. He posits a monthly fee of $5, plus a secondary fee for additional sets equal to $1, which is chosen by only 20 percent of homes. Thus, he estimates subscribers to contribute an average of only $62.40 each year.

In the random sample of 81 cable systems collected for the purpose of testing the predictive power of various demand equations (to be described below), the average subscriber fee is $5.25 per month or $63 per year. The number of second connections is not known, nor is the secondary subscription fee in most instances. However, published data show that 41 percent of all homes have at least two receivers, and

the proportion is growing. We shall assume that 41 percent of subscribers 13

attach two receivers to the cable at an additional cost of $1 per month.

Surprisingly, Mitchell assumes that subscription fees will not grow at all in the infinite life of the franchise. Our sample systems demonstrate an average of 1.2 percent annual growth in monthly fees in the past five years, and this rate of growth is likely to increase in lagged response to the recent inflation. Therefore, in one variant of our calculations we shall utilize a 2 percent growth rate for subscriber fees for the entire forty-five year period. The allowance for revenue growth, among other things, allows us to treat inflation symmetrically--since we use Mitchell's operating-cost data which contain significant escalator provisions in several of the labor categories.

131972-1973, Television Factbook.

20-344 O-73-22