|
|
|
|
 |
|
|
|
|
|
|
|
 |
 |
 |
|
||||
|
Calls to mobiles: Economic depreciation |
|
||||||
 |
 |
 |
|
||||
 |
|
|
|
 |
|
||
|
|
|
|||||
|
Entrants and barriers to entry Competitor constraints from both entrants and incumbents Presentation by Analysys of economic depreciation 1 The purpose of Oftel’s cost modelling is to derive the most appropriate cost figures to form the basis of regulated termination charges. Regulation must ensure that operators are given the opportunity to recover their efficiently incurred costs, including a reasonable return on investment. Since investment by operators is ongoing and the charge control period is significantly shorter than the period over which investment costs would be recovered, the timing of the recovery of costs is an important issue. In Oftel’s model to derive the cost of mobile termination, economic depreciation has been used as the means to annualise costs and so influence the timing of cost recovery. This annex explains the theory of economic depreciation and how it has been applied in this case. Economic depreciation seeks to set the optimal profile of cost recovery over time by mimicking the operation of a competitive market. In Oftel’s view the costs derived from this analysis provide the best available figures to use as the basis for regulated charges. 2 In terms of capital costs, bottom-up models work by identifying the investment costs of building a network, which then need to be annualised appropriately to yield annual capital costs. One approach to annualisation is to use an accounting approach, such as straight-line depreciation plus return on capital employed. Another approach, which is commonly used in bottom-up modelling, is an annuity, in which the annual capital charge, the sum of depreciation and return on capital employed, is the same in each year of the defined asset life. This assumes a ‘back-loaded’ depreciation profile (ie more depreciation later in the asset life), which is generally inappropriate for telecoms assets. This is because real prices of telecommunications assets tend to be declining, meaning that future entrants will be able to purchase cheaper assets and so incumbents will typically wish to ‘front-load’ unit cost recovery. This fundamental deficiency of annuities can be ameliorated, by adding a ‘tilt’ to the annuity to bring forward some of the depreciation to earlier in the asset life. 3 In principle, the correct approach to annualisation of investment costs is to use economic depreciation. Economic depreciation is the change in economic value during the year. Economic value is the asset’s earning power, ie the discounted present value of expected future revenues from the output produced by the asset, less the present value of associated future operating costs. The function of economic regulation is generally to mimic the outcome of a competitive market. So, the appropriate starting point for the economic depreciation profile for regulatory purposes is the pattern of the recovery of the investment costs, if the relevant market was competitive/contestable at every point in time*. *This annex focuses on the discussion of capital costs, but there is a similar treatment of operating costs under economic depreciation. 4 Another way of thinking about economic depreciation is that it involves a cash-flow analysis to answer the question: what time-series of prices, consistent with trends in the underlying costs of production, yield an expected net present value of zero (ie normal profit)? Cash-flow analysis is an established business technique and Oftel would expect that it would often be used by telecoms operators in investment appraisal. 5 To develop economic depreciation profiles requires that assumptions are made about a range of factors, including forecasts of variables (such as the future price of the Modern Equivalent Asset), which are not straightforward to derive. However, economic depreciation allows the sensitivity of the results to a range of different assumptions to be tested within a framework that is conceptually sound. 6 The economic depreciation profiles in the LRIC model reflect assumptions about the following key variables:
7 The lower the discount rate, the lower the financing cost of investment that needs to be recovered in any year. The greater the expected future reductions in the price of the MEA, the more that depreciation needs to be front-loaded. The incumbent will only be able to compete against future entrants and earn a reasonable return, if it brings forward cost recovery. The more that operating costs of an asset increase with the age of the asset (eg increasing maintenance costs due to wear and tear), the more that depreciation should be brought forward. The effect of utilisation on the profile of cost recovery is discussed further below. 8 A set of slides explaining the basic economic depreciation calculations is at Attachment 1. This is an updated version of a presentation given by Analysys at a LRIC Working Group meeting on 20 March 2001. 9 The overall question to be answered in the economic depreciation analysis is:
10 The purpose of the analysis is to mimic the effects of a competitive/contestable market, because this provides an appropriate benchmark for regulation. So, the profile of cost recovery should be identified by constructing a ‘competitor constraint’, ie addressing the question:
11 Such a market involves the absence of super-normal profits. This is a critical feature of the analysis for two reasons.. First, all super-normal profits are competed away in a competitive/contestable market. Second, that otherwise the figures derived would include super-normal profit and so would not be true measures of cost. 12 The next question is to establish how the competitor constraint should be defined. Two broad types of competitor constraint can be identified:
13 Here, the term ‘competitive’ is used to refer to actual competition among incumbents that is sufficiently vigorous to remove super-normal profits, even in the absence of entry or the threat of entry. ‘Contestable’ is used to refer to potential competition from new entrants that is sufficiently strong to remove super-normal profits, even in the absence of competition among incumbents (eg even if there were a sole incumbent). 14 The question, which is addressed in detail below, is:
15 It should be noted that to answer this question requires the construction of a hypothetical scenario, because Oftel has concluded that the markets for mobile termination are not competitive. The hypothetical scenario is important, however, because it provides an appropriate benchmark for regulated charges. The discussion below is about the considerations to be taken into account and the choices to be made when constructing the hypothetical scenario. Entrants and Barriers to Entry 16 One important factor influencing the nature of competition between incumbent and entrant is the utilisation that can be achieved by each, or the total output over the asset lifetime available to new entrant and incumbent. This could be assessed by comparing:
17 Type (ii) difference might arise from changes in market demand. For example, if the size of the market is growing over time, there is a benefit to entering later because output (average utilisation) over the asset life will be larger. This affects the appropriate cost recovery profile for the incumbent(s). If the incumbent knows that it will face competition from an entrant next year, which is able to produce more output over the asset lifetime than the incumbent’s current asset, then it knows that it must recover sufficient of the cost this year that it can compete next year with the entrant next year and still be profitable overall. The result is to front-load the cost recovery (per unit of output). 18 However, type (i) difference has a deeper implication, in the sense that it might seem to imply that one type of market player becomes uneconomic. If entrants could produce more output over the asset lifetime, then incumbents might be unable profitably to compete against entrants. Or, more plausibly, if entrants would expect to produce less output in a given set of calendar years, then entrants would be unable profitably to compete against incumbents who are pricing to earn only normal profits. In theory, entrants would be able to survive in the market only if incumbents are earning super-normal profits in this period of time. Another way of describing this situation is that there are material barriers to entry and so the market is not perfectly contestable. 19 Such a barrier to entry might arise because of the time that it takes for a new entrant to build out a network from scratch and acquire customers. The incumbent would have gone through this process in the past, but now can replace an asset and immediately obtain high utilisation, because it already has an established network and customers. 20 One way to specify the competitor constraint would be the contestable market approach. It could be assumed for the purposes of the analysis (even if this represents a departure from reality) that entrants never experience a type (i) difference compared to incumbents. In a contestable market entrants face no barriers to entry and so would always be able to achieve the same utilisation as the incumbent(s) in any calendar year. So, for illustration, assume that the incumbent invested three years ago and achieved 50% utilisation in its first year of operation and 75% in its second year before reaching 100% in the current year. The contestable market approach would mean that the entrant in the current year would be assumed to achieve 100% in the current year, its first year of operation (and so has greater type (ii) efficiency than the incumbent). 21 Competition from potential entrants to a contestable market would be sufficient to ensure the removal of super-normal profit (whatever the number of incumbents or the nature of competition among them). The incumbent would be unable to defer depreciation when utilisation is low. If input costs (MEA price and operating expenses) were constant, then the economic depreciation profile under contestability would be a constant annual cost recovery (in £) each year. The unit cost (or price) would be inversely proportional to utilisation. 22 Although contestability provides a feasible answer to the specification of the competitor constraint, the price/unit cost profile that it implies seems unattractive. When utilisation is very low, the price/unit cost is very high and vice versa. It also involves an assumption about new entrants that seems very unrealistic. 23 A second way to specify the competitor constraint would be to assume a competitive market, ie sufficient competition among incumbents to remove super-normal profit (whatever the barriers to entry faced by entrants). 24 However, it appears that this approach would not define a unique price path as the equilibrium solution. Instead, an infinite number of price paths over the period can be derived, each of which will solve the model in the sense that they enable all incumbents full and exact cost recovery. Some examples are shown in Figure 1 for the simple, illustrative case of a two-year asset with utilisation of 50% (or 50 units of output) in the first year and 100% (100 units) in the second. Purely for the purpose of illustration, it is also assumed that there is a constant MEA price of 150, and that operating costs and the cost of capital are zero. One solution involves a constant price, once the utilisation pattern has become established (in Figure 1 this is at a price of 1, the asset price of 150 divided by 150 units of output over the two years). The other solutions involve prices that oscillate around this constant-price path. 25 There is a logic for regarding the constant set of prices as the equilibrium solution, if the oscillating patterns would be undermined by entry. A firm that tried to charge a price above 1 would be unable to sustain it, because it would be profitable for another firm to undercut it. But it would not be profitable to set a price below 1, because the firm would only recover its investment cost if it was able to charge a price above 1 in the following year (the other year of the asset’s life). For example, incumbents could recover their cost by charging 0.5 this year and 1.25 next year. But, the distinction between competition among incumbents and against new entrants is important here. The incumbents know that they would have to compete against new entrants next year and a price above 1 next year would not be sustainable, because it would be undercut by new entrants investing in that year. Therefore, in the current year the firm could not sustain a price above 1 and it would not be willing to set a price below 1, because of the threat of future entry.
26 So, it is not competition among incumbents that would make the constant-price path the equilibrium solution of the model, but the ability of new entrants to undermine any other price path. However, if entrants faced a barrier to entry, ie the market was not perfectly contestable, there would not be a unique solution*. * There may be factors which would reduce the number of paths in the feasible set. For example, if there is an alternative use for the asset, ie the investment is not sunk, then the opportunity cost sets a price floor. Interdependence between demand in different time periods might also make some paths strictly preferable for the incumbents than others. A third factor, the threat of entry, is discussed below. Competitor Constraints from both Entrants and Incumbents 27 A third approach to specify the competitor constraint would be to take account of the effect of both potential competition from new entrants and actual competition among incumbents. This is the approach that Oftel has adopted and it is assumed that:
This has some similarities to the second approach, discussed above, but with a reduction in the number of price paths in the feasible set of solutions, because of the constraint imposed by entrants. 28 The entrants would be assumed to suffer a barrier to entry of type (i), because of the time that it takes to build out a network from scratch and acquire customers. The incumbent would have gone through this process in the past, but now can replace an asset and immediately obtain high utilisation, because it already has an established network and customers. One possible assumption is that the entrant is assumed to experience the same utilisation profile, given by years since the date of entry, as the incumbent experienced in the past. Another possible assumption would be that the entrant’s utilisation profile in terms of years since entry is more favourable than the utilisation profile experienced in the past by the incumbent, because of the growth in the size of the market over time. 29 Even entrants facing a barrier to entry would impose some constraint on the pricing of incumbents, but they would not constrain prices to a level where all of the incumbents’ super-normal profit would be removed. So the constraint imposed by entrants would define a set of feasible price paths for the incumbents that would not be undermined by entry. Competition among incumbents would impose another constraint and further restrict the feasible set to those paths that would yield no super-normal profit. 30 But over what period of time should the constraint of no super-normal profit be assessed? The proposed approach is that the incumbents would earn no super-normal profit over the long run, which is longer than one asset lifetime. For illustration, assume that the period modelled is equal to two asset lifetimes. Prices for the later investment are assumed to be as high as permitted by the constraint imposed by entrants (who face a barrier to entry). Such prices would allow the incumbents to earn super-normal profits on the later investment. But, because of competition among incumbents, they would be prepared to make a loss on the earlier investment. The size of the loss on the earlier investment would exactly offset the super-normal profit earned on the later investment. 31 Could this be the outcome of a competitive market? It might be thought that competition among incumbents would remove the possibility of an incumbent earning a loss on one investment and super-normal profit on another. One of the incumbents could choose not to make the earlier investment which is loss-making and so would seem to be able to undercut the prices of the other incumbents on the later investment which yields super-normal profit. However, such a strategy would not be feasible. The reason is that, if the firm did not make the earlier investment, at the time of the later investment it would be an entrant, not an incumbent. It would therefore face the same barrier to entry as any other entrant, ie it would still have to go through the process of building out its network and acquiring customers. 32 Why would the incumbents not set prices for the earlier investment that avoided a loss, eg prices that yielded normal profit? The answer is because of competition among incumbents. By choosing to make the earlier investment, the firm would gain an advantage, ie it would become an incumbent and so in later years would not need to go through the process of building out its network and acquiring customers. This means that, when it comes to make its replacement investment, it would be in a favourable position compared to entrants and so be able to earn super-normal profits. 33 Now consider two incumbents: one that only takes account of its current investment and sets prices to earn normal profit on that investment; and another that also looks ahead to the profits available in the future period of the replacement investment. The second incumbent would be willing to undercut the prices of the first incumbent, because it knows that overall over the extended period it would still be able to earn at least normal profit. Given a competitive market, this process of bidding down the price would continue until the size of the earlier loss exactly offset the extent of the later super-normal profit. Overview 34 To recap, the question being addressed is: what would the path of costs (prices) be if the market were competitive? This is a hypothetical, given that the market is not competitive (and indeed, if the market were competitive, regulatory intervention would be inappropriate). Rather, the competitive outcome is being used to derive the appropriate benchmark for regulation. 35 The nature of the hypothetical competitive-market benchmark that is assumed is:
36 Entrants are assumed to face a barrier to entry, in the sense that their utilisation in a given calendar year soon after the date of entry is assumed to be lower than the incumbents’ (or previous generations of entrants’) utilisation in that same year. The algorithm allows the height of the barrier to entry faced by each generation of new entrants to be varied by changing a parameter value, which is referred to as the ‘competitor constraint’ parameter*. For example, a declining barrier to entry over time could be assumed, ie in an initial period of years, each successive generation of potential entrants would be able to obtain greater utilisation quicker than the previous generation, due for example to growth in the size of the market. After a given date, this improvement could be assumed to cease and so all subsequent generations of entrants could be assumed to have the same utilisation profile (face the same barrier to entry). *The competitor constraint parameter is in the form of a percentage change figure each year. This is consistent with an entrant utilisation profile (which yields normal profit over the long run), although it does not uniquely define only one such profile. 37 The algorithm derives a solution by requiring that the incumbents (and all generations of entrants) earn normal profit over the long run (assuming perfect foresight). There is discounting so that each year in the future has a smaller ‘weight’. The model is set up to calculate economic depreciation using one of two time horizons. It can be run with the period of complete asset lifetimes over 50 years (eg a 40-year horizon for an asset with a lifetime of 20 years; a 45-year horizon for an asset with a lifetime of 15 years). Or it can be run with an infinite time horizon (using a perpetuity calculation). The two versions of the model yield very similar results (the unit costs with 50 year horizon are only about 0.5% higher than with an infinite time horizon). Detailed description 38 The algorithm calculates the ‘revenue required’ for incumbents and entrants to earn normal profits separately for each of capital investment and operating expenses. This separation simplifies the modelling of the two types of underlying input cost trends: capital costs (MEA prices) and operating expenses. Operating expenses are allowed to vary both with time and with the age of the asset. 39 There are three parts to the calculations in the algorithm for each of capital investment and operating expenses. The calculations for capital investment are discussed initially. 40 The first part of the calculation is to identify the price in the last year* that would allow normal profit to be earned by an entrant in that year, achieving the long run utilisation and paying the long run MEA price. This is referred to as the ‘base price’. *The last year is the final year of the 50 year time horizon (strictly, the last year of the last complete asset lifetime within then 50 year horizon). In the infinite horizon version of the model, perpetuities are calculated assuming that conditions remain constant after this 'final' year. 41 If the incumbents (and all generations of entrants) were to charge this base price in every year, they might not obtain full cost recovery for two reasons:
42 The algorithm derives an amount that needs to be added on to the base price to allow for each of these factors. So, the price/unit cost path that is derived comprises:
43 The second part of the calculation, the utilisation component of the price/unit cost, is calculated as follows. To earn normal profit, the incumbents would require higher prices in the earlier years than the base price, because they achieve utilisation rates below the long run level. The amount of the cost recovery required to achieve normal profit is the present value of the loss that would be incurred by the incumbent if it were to charge the base price in every year. The profile over time of the utilisation component of the price is assumed to follow the competitor constraint parameters. For example, if the competitor constraint parameter were set to -5% in every year, the utilisation component in any year would be 5% lower than in the previous year, ie a tilt would be introduced to the utilisation component, raising it in the earlier years and reducing it in the later years (compared to competitor constraint parameter values of 0%). 44 The utilisation component calculation, therefore, depends on:
45 The third part of the calculation, the input price component of the price/unit cost, is calculated as follows. To earn normal profits, the incumbents would require higher prices than the sum of the base price and the utilisation component, because they would (typically) purchase their assets at higher prices than the MEA price in the last year. The amount of the cost recovery required to achieve normal profit is the difference between the present value of the capital expenditure of the incumbents over the long run and the present value of the revenue that they would obtain if the price charged in every year were the base price plus the utilisation component. 46 The profile over time of the input price component reflects a variable referred to as the ‘output value shortfall’, which is defined in each year as the utilisation in that year multiplied by the excess of the MEA price in that year over the MEA price in the final year. The input price component is derived by distributing the required amount of cost recovery across years in proportion to the output value shortfall. For example, the input price component would be zero for years in which the MEA had reached and stabilised at its final year value (because the output value shortfall would be zero in that year). This is appropriate, because that generation of entrants would pay the same MEA price for all future asset purchases and so would impose a constraint on the incumbents, preventing them from recovering additional costs in that year (and subsequent years). 47 Also, for two years that had the same MEA price (which was higher than the MEA price in the last year) the input price component would add the same amount to the price/unit cost even if the utilisation were different between those years (because the input price component is proportional to utilisation). It seems reasonable to smooth the price/unit cost in this way and avoid fluctuations related to utilisation*. *Such fluctuations would be experienced in a perfectly contestable market. But, the nature of the competitive market assumed here is not perfectly contestable, in part because of its apparently unattractive feature of price fluctuations with utilisation. 48 The input price component calculation, therefore, depends on:
49 The calculation of economic depreciation in relation to operating costs is similar as for capital costs with a base price, a utilisation component and an input price component. One difference is that in the model some types of operating costs are assumed to fall/increase at a constant rate (rather than becoming constant between years as for MEA prices). In the operating cost part of the economic depreciation calculation, this rate of change is reflected in the profiles of the base price, the utilisation component and the input price component. 50 The price or unit cost in each year is given by the sum for both capital and operating costs of the base price, the utilisation component and the input price component. This algorithm represents one reasonable approach to modelling the type of competitor constraints assumed from new entrants and among incumbents. 51 It is a key feature of the economic depreciation algorithm that unit cost (or price in a competitive market) does not fluctuate between years because of different utilisations experienced by operators. If all input costs (MEA prices and operating costs) were constant over time, then a constant path of unit costs would be derived from the economic depreciation algorithm. 52 The level of this constant unit cost would be affected by a number of factors. For example, it would be higher if average utilisation over the long run were lower, and it would be higher if a larger cost of capital were used as the discount rate. 53 One set of factors affecting the shape of the profile of unit cost over time is the profile of input costs - for example, the rate at which MEA prices are changing are important assumptions. For most assets MEA prices are assumed to decline in real terms, but the MEA price of some assets has an upward trend (eg site acquisition). The profile of operating costs also has an effect on the pattern of unit costs over time. 54 Another potential influence on the pattern of unit costs over time is the set of assumptions about the ‘competitor constraint’ parameters in each year, discussed above. Oftel has taken a conservative approach by assuming that the competitor constraint parameter is 0% in every year, ie it is assumed that each generation of entrants would be unable to gain higher utilisation levels faster than the profile of utilisation experienced by the incumbent MNOs (or by earlier generations of entrants). This is conservative because there has been substantial growth in the size of the market since the entry of the MNOs. There is an argument for choosing negative competitor constraint parameters in at least some of the years, to reflect the ability of later entrants to benefit from the larger market size by being able to obtain higher average utilisation over the long run. However, the assumption of 0% in every year has been adopted in the absence of better information about the likely utilisations that would be experienced by entrants. 55 For the purposes of cost analysis to obtain regulated charges, in principle the correct approach to annualisation of investment costs is to use economic depreciation, because it reflects the profile of costs in a competitive market. The approach adopted by Oftel to the modelling of the competitive constraint takes account of the effect of both potential competition from new entrants and actual competition among incumbents. It is assumed that:
56 The cost figures used by Oftel to set the termination charge caps have been derived from this approach to economic depreciation using reasonable assumptions, such as for operating costs, asset prices and competitor constraints.
Attachment 1 Presentation by Analysys of Economic Depreciation The question to be addressed is: What time-series of prices, consistent with trends in the underlying costs of production, yield an expected NPV of zero over the period of interest? The inputs
First calculate the total expenditure (we will initially assume a lifetime of 10 years)
...then calculate the total relative output value (assuming the same lifetime of 10 years)
Divide one by the other to yield the unit price for a relative output value of 100%
Multiply this by the relative output value in each year to yield annual revenues
Economic depreciation is then the difference between revenues and operating expenses
Check that everything is consistent!
Notes re implementation in the Analysys Model of Mobile Network Costs:
Costs of higher input prices = Extent to which earlier entrants have to pay input prices higher than those implied by the long-run
|
|
||||||
|
|
||||||
 |
 |
|
|||||
 |
|
||||||
