Transportation Industry Marketing Research
Background and Executive Summary
The public transport policy focus in Adelaide, South Australia is on the beginning of aggressive tendering to develop productive efficiency and competence, recover financial performance and decrease fiscal pressure. Further than this initiative, the question of public transport subsidy is also an issue of critical policy importance requiring attention ( 1995). In particular, what is the suitable level of subsidy in Adelaide which can be justified on economic efficiency bases? This paper contributes to the declaration of this question by estimating most favorable bus subsidy in Adelaide, based on the user economies of scale (UES) argument for subsidy.
The reason of the UES argument is as follows. Bus transport is typically distinguished by unvarying returns to scale in producer costs, with marginal and average producer costs consequently being identical. If pricing was based purely on producer costs, well-organized pricing at marginal cost would hence bring a financial breakeven outcome, and as a result no need for subsidy. As (1972) first pointed out, conversely, the determination of an efficient outcome necessitates deliberation of all costs, both producer costs and user costs. The existence of economies of scale in some user costs then generates a case for subsidy.
The pertinent user costs are time costs which are related to the frequency of service. The relevant time costs are waiting time used up at bus stops, and any other user delays caused by services departing at times which do not go with users. An augment in frequency, and thus the scale of operation, will cause these user costs to fall for all users. The marginal user cost of supplementary frequency is then below average user cost (i.e. there are scale economies in user costs) (or ‘user economies of scale’, UES). Consequently, as in all cases of economies of scale, efficient pricing at marginal social cost will result in a financial deficit and a need for subsidy.
By means of focusing on UES, the subsidy results presented here should be seen as part only of a broader investigation of subsidy which gives appropriate weight to the range of alternative subsidy arguments including as an equity instrument, for energy conservation, due to the option value of public transport, and as an instrument to manage road congestion on un-priced roads. In spite of this, a number of reasons exist for focusing on the UES subsidy argument. First, many of the other arguments proposed in support of subsidy do not have a sound basis ( 1984). Second, the UES argument has been a central element in the major economic efficiency based studies of subsidy overseas (e.g. 1972; 1979; 1982; 1987; 1987), thus far, surprisingly, it has received little attention in Australia. Third, the subsidy argument which has received most focus in Australia over the years, that subsidy is a second-best policy for managing road congestion when roads are un-priced, is being increasingly recognized as being of limited importance, particularly in relatively low road congestion cities like Adelaide. The UES argument, hence, appears to be a significant subsidy argument which has received limited attention in Australia, suggesting that focus on this topic is therefore warranted.
Marketing Research Question/Objective, Components
The evaluation in this paper makes a number of contributions beyond previous Adelaide work ( 1985, 1986; 1990; 1990):
· It presents a more complete disaggregated analysis: ‘s work was undertaken at the highly aggregated entire network/all day average level, whilst only considered a single route in the peak period. Although considered both peak and off-peak, his work was limited to only a couple of bus routes. On the contrary, this paper hypothesizes subsidy for the majority of the network for both peak and off-peak, and does so by working upwards from a considerable level of disaggregation.
· It utilizes an enhanced user behavioral choice model: The work, the main piece of previous UES subsidy estimation in Adelaide, makes the plain hypothesis that users arrive at bus stops randomly, i.e. with no reference to a timetable. Nevertheless, this is a rather incomplete and improbable model of user behavior ( 1991). use a more realistic model, but lacks strong theoretical underpinnings. The sensitivity of subsidy results to the nature of user behavioral assumptions and models suggests re-estimation for Adelaide using a superior user cost model is justified. A logit choice model, in which users choose between random and planned behavior, is used here.
· Optimization with explicit acknowledgment of two market segments, full fare paying versus concession users. This contrasts with preceding work which has modeled a single average bus user.
The paper also tests the robustness of ‘s (1990; 1992) conclusion that optimal UES subsidy is small in size. ‘s analysis focused on peak periods only. The analysis tests the robustness of the conclusion in off-peak periods.
TABLE 1
Bus Corridors
Corridor Average
Route Length (kms)
1. Outer North 19
2. Inner North 12
3. Outer NorthEast 23
4. Inner NorthEast 12
5. Eastern 10
6. Stirling Hills 22
7. SouthEast 8
8. Mitcham/Happy Valley 21
9. Outer South 20
10. Inner South 13
11. SouthWest 16
12. Western 15
13. NorthWest 11
User economies of scale (UES) are driven by frequency-related user costs (e.g. waiting time). It is useful, therefore, to distinguish between these and other user costs (e.g. in- vehicle time, walk time, etc). Accordingly, the subscripts F and O will refer to frequency-related user costs, and other user costs, respectively. Thus:
Research Design
Framework for the Analysis
Level of Disaggregation
The bus network was modeled as a series of corridors, where a corridor is a collection of routes which serve a similar geographical area. A total of 13 corridors (listed in Table 1) are analyzed. Two time periods are modeled, Peak (PK: 6am-9am, plus 3pm-6pm, on working weekdays) and Off-Peak (OP: the interpeak period 9am-3pm on working weekdays, evenings (6pm-12am), weekends, and public holidays).
Policy settings were allowed to vary between corridors and time periods to reflect variations in user cost (the determinant of user economies of scale (UES)) and variations in demand. To enable meaningful comparison of results between corridors, the unit level of analysis within each corridor is the representative route. Aggregate results for each corridor are subsequently derived by factoring up accordingly.
Producer Cost
[C.sub.p] [C.sub.p] is a function of service vehicle-kms, VK. Urban buses display constant returns to [C.sub.p] with respect to VK (, 1988; , 1990; , 1993), thus [C.sub.p] / VK = [Delta][C.sub.p] / [Delta]VK and is constant. Noting that VK = 2[L.sub.1]F, where [L.sub.r] is route length and F is service frequency, then:
User Cost, [C.sub.u]
[C.sub.u] is the total time user cost incurred across all users. User time costs are experienced by all users. For example, if catching a bus results in users having to wait for the arrival of a bus, this wait applies to all users who want to use the service. The convention is, therefore, to define the user cost experienced by each user as the average cost, [AC.sub.u]. Thus:
[C.sub.u] = [AC.sub.u] q (2) where q is route patronage (boardings per hour).
Demand, q
Route patronage, q (boardings/hour), is given by q = q(g) (9) where g, the generalized cost of travel (money plus non-monetary costs [in money units]).
Concessions
Current concession fares (which deliver discounts of 50 per cent) for certain groups of travelers (students, the aged, the unemployed, etc) are maintained. (8) Two users groups are modeled: regular full fare paying users, and concession users, denoted by the subscripts 1 and 2. [P.sub.1] and [P.sub.2] are the prices paid. The Social Justice Price Discount, SJPD, received by concession users is: SJPD = [P.sub.1] – [P.sub.2] and the Social Justice Discount Factor, SJDF (%) is: SJDF = SJPD/[P.sub.1] = ([P.sub.1] – [P.sub.2])/[P.sub.1].
Rearranging then yields: [P.sub.2] = (1 – SJDF)[P.sub.1]. (14)
The Optimization Problem
The Optimization problem is cast as a peak load problem ( 1957; 1966). The standard approach of allocating capital costs (bus capital costs) to the peak ( 1989) is used. Service levels in the peak and off-peak can be made to vary independently by having different peak and off-peak bus frequencies. For simplicity, it is also assumed that demand cross elasticities between time periods are zero. Overall therefore, peak and off-peak periods can be analyzed as separate optimizations, identifying optimal policy settings for each period.
Current bus size, N = 78 is used.(10) The policy variables to be optimized are price (P) and frequency (F). A range of previous studies have considered this bus optimization problem ( 1976, 1985, 1985, 1987, 1992, 1993).
Optimization with Non-Distortionary Public Finance
Consider the representative route within any given corridor for any given time period. (ES = [CS.sub.1] + [CS.sub.2] – S, S = [C.sub.p] – [P.sub.1][q.sub.1] – [P.sub.2][q.sub.2] and s = S/q where ES is economic surplus, CS is consumer surplus, s is unit subsidy and S is total subsidy. With [P.sub.2] related to [P.sub.1]).
A key feature of the model is that it models the mode of user behaviour. Users can choose between:
· planned behavior, through use of timetables (obtained at an information cost I) to coordinate their arrival time at the bus stop with scheduled bus departure times; and
· Random behavior, where users choose not to consult a timetable, resulting in users arriving at bus stops at times which are randomly related to scheduled bus departure times.
Data Analysis
Comparison of Optimal and Current Settings
In what way do optimal outcomes compare with the outcomes in each corridor/time period, and thus what magnitude of the change is required to attain optimal positions? Except for a small number of exceptions (the Eastern, South East and Stirling Hills corridors in the off-peak), a consistent pattern emerges. The changes required to move from the current situation to optimal outcomes have the following characteristics:
· Reductions are required in frequency (F). Thus at the current frequencies, marginal time costs are generally not big enough to justify the higher marginal producer costs of increasing frequency. ES can therefore be increased by allowing frequency to decline which raises time costs by less than the resulting rise in producer cost.
· With frequency falling, the likelihood of random behaviour falls, and therefore so does the proportion of random users, R.
· Like frequency, moving to optimal outcomes also requires patronage, q, to fall.
· Optimal fares are higher than existing fares. Although frequency falling causes [AC.sub.u] to rise, fares must also rise in order to achieve the required patronage reductions.
· Average unit subsidy, [s.sub.ave], falls. To understand why, consider Figure 2 which represents a typical situation in either period. With [q.sub.0] [greater than] [q.sup.*], this suggests we are currently at a point like a in Figure 2. If we assume momentarily that [q.sub.0] is being produced in the least costly manner, so that ATC at [q.sub.0] is LAC([q.sub.0]), then moving to point b would result in both MB and AC rising, but with the former by more so, and thus s (= ATC – MB) falling. To the extent that production at [q.sub.0] is less than fully cost efficient (i.e. frequency is not being optimized at [q.sub.0]), so that AC ([q.sub.0]) [greater than] LAC ([q.sub.0]), this would merely reinforce the decline in s brought about from moving to the optimum at point b.
· Total subsidy (S) falls. The reduction in both [s.sub.ave] and q explains this result.
· The changes are greater in magnitude in the peak than the off-peak, and the changes required to move to an optimum in a distortionary public finance setting are of significantly greater magnitude than those required to reach an optimum in a nondistortionary public finance setting.
The exceptions to the above general pattern of outcomes are the Eastern, South-East and Stirling Hills corridors in the off-peak. In the Eastern corridor, fares also fall along with q, F and [s.sub.ave]. The reason for this is that the required fall in frequency (F) (and thus associated rise in [AC.sub.u]) is rather substantial, thus, with only a modest fall in patronage (q) required (and thus a modest rise in generalized cost (g) to achieve it), scope exists for fare (P) to fall. In the South-East corridor, optimizing F requires it to be increased marginally. In addition, additional patronage can be accommodated. To achieve this, fares need to fall. Referring again to Figure 2, this type of situation would exist at a value below [q.sup.[*.sub.NDPF]]. The rightward move to [q.sup.[*.sub.NDPF]] results in MB and AC falling. The former must fall by more to attain the fall in s reported in Table 1. The Stirling Hills corridor reports rises in q, F and P, and a fall in [s.sub.ave]. The feature of the base case for this corridor is the very low frequency and high headway. Thus [AC.sub.u] is disproportionally higher than that which occurs in other corridor/time periods. The substantial rise in frequency in moving to a non-distortionary public finance optimum therefore causes a very significant fall in [AC.sub.u]. The size of this fall in [AC.sub.u] is so significant that substantial fare rises are required to limit the size of the growth in patronage.
It is interesting to compare the changes suggested here for frequency and fares with those suggested by (1985, 1986), who concludes in his study of urban public transport subsidy in Australia’s major cities that:
”This study indicated that there are benefits to be derived from a reduction in the level of public transport services in many of the Australian cities, and a switch of the subsidies saved to finance lower fare levels. . .this conclusion was reached for both bus services and rail services” ( 1985, p.77)
The study finds that, on the whole, frequency must fall and fares must rise. Thus there is consistency with respect to how frequency should move, but not fares. Even though the two studies propose fares should move in opposite directions, there is no conflict in this result if one contrasts the objectives of the two studies. considered the changes required in fares and frequency to maximize the return from subsidy given existing subsidy levels were to be maintained. Therefore, his findings were not unconstrained outcomes, but rather sub-optimal outcomes from an optimization subject to a subsidy constraint. With total subsidy given, a reduction in frequency, and thus a rise in [AC.sub.u], allows fares to fall. Alternatively, in the current study the changes in fares and frequency move us to optimal outcomes, with subsidy reducing significantly rather than remaining fixed. With subsidy falling, fares need to rise (even though [AC.sub.u] falls with the fall in frequency).
Findings and Recommendations
Results under Random Behaviour
It is also practical to consider the non-distortionary public finance results derived from using the random user behaviour model, rather than the logit model. The random model has been widely used in past analyses, and in the most significant previous Adelaide study (1990). Table 2 reports results using the random model for the Eastern corridor (the NDPF(R) lines).
The results show that using the random model has a significant impact in both peak and off-peak on optimal unit subsidy and total subsidy, yielding values well above those generated with the logit model. An important implication of this is that one of ‘s (1990) main conclusions should be treated with some skepticism. finds that popular arguments, which claim that current subsidy levels in Adelaide are too high, are not supported by his optimal results. This conclusion is based, however, on the random user behaviour assumption, which tends to over-estimate subsidy in cases where planned behaviour, or a mix of planned and random behaviour, might be expected. As indicated earlier, the anecdotal experience in Adelaide is that planned behaviour is the dominant form of user behaviour. Thus, ‘s results may be significantly biased by the strong random user behaviour assumption built into his analysis. The same can be said of most past subsidy analyses ( 1992).
This paper has presented a disaggregated analysis of optimal outcomes in the Adelaide bus system. The advantage of a disaggregated approach has been the ability to establish policy settings which differ across the system to reflect variations in demand and cost differences, together with differences in user economies of scale, the key focus of this study.
The final consideration is the off-peak robustness of ‘s peak conclusion that UES subsidy will be small (see introduction). It is essential to note that draws his conclusion based on analysis in a world of distortionary public finance, with the marginal distortionary cost of public finance the same as that considered here, .40 for each raised. The analysis here found that in such a world of distortionary public finance, total peak plus off-peak subsidy is also rather small. Therefore, when public finance is distortionary, ‘s conclusion is robust in both the off-peak, and also on an overall basis. It is important to note however that the overall size of optimal subsidy, and ‘s conclusion, is strongly dependent on the cost of public finance. Improvements in the way public finance is raised to make it less distortionary will increase the level of optimal bus subsidy.
Limitations
The focal point of the study is limited to buses, the dominant mode of public transport in Adelaide (and most other Australian cities).
Bibliography and Referencing
Credit:ivythesis.typepad.com
0 comments:
Post a Comment