Universal service through roll‐out targets and licence conditions: lessons from telecommunications in South Africa

Abstract

A priority for the post‐apartheid government was the extension of basic infrastructure services to the vast majority of citizens that were not serviced under apartheid. The Reconstruction and Development Programme set objectives for each of these utilities that would be achieved in the first decade of democracy, while departmental policy aimed to find means of achieving these targets. The strategy of choice in most sectors was one of ambitious roll‐out targets being set for utility operators. Targets were set for individual residential service (‘universal service’) and for community service outside of individual homes (‘universal access’). While most utilities remained under public ownership, in telecommunications there was partial privatisation of the incumbent Telkom and the entry of privately owned mobile cellular operators. This article examines how roll‐out targets and licence conditions for universal service have performed in this sector where private operators exist. It examines the failure of the Telkom licence and draws out some lessons for policy.

1 LICENCE CONDITIONS AND ROLL‐OUT TARGETS IN TELECOMMUNICATIONS

A universal service or access licence condition involves imposing a target for the roll‐out of either residential or community access over a predetermined time period as part of an operating licence. The target for this service will be uneconomic customers (e.g. low‐income households), uneconomic areas (e.g. high‐cost rural areas) or uneconomic services (e.g. payphones). In the case of telecommunications, this would take the form of a number of residential telephone lines, a number of payphones or, even more recently in South Africa, a number of internet laboratories in schools.

When universal service or access conditions are placed on a private operator that will affect its profitability, the regulator or government department has to grant that company some concession if it is to attract private operators into the market. This might include a reduction in its licence fees (or any other fees and taxes it pays), a period of limited or no competition that allows the firm to make abnormal profits and cross‐subsidise, or a reduction in the productivity factor in a price cap, which also allows abnormal profits. The approach is to determine what the specific licence obligations the regulator or government department want will cost, and to look at means to increase the profitability of the firm by the same amount through one of the strategies cited above.

The alternative to licence conditions is to have operators contribute to a Universal Service Fund (USF). Such a fund is used to ‘purchase’ universal service from operators or directly fund consumers. Universal service or access can be purchased through either much the same process as a licence condition – costing the service delivery and then contracting an operator – or it can be awarded through an auction. In an auction, operators bid for a subsidy to implement universal service goals and the bidder that requires the least subsidy wins.

1.1 Roll‐out targets and licence conditions in Phase 1 of telecommunications reform in South Africa

In telecommunications, the establishment of new political priorities came at a time of increasing pressure through the World Trade Organisation to liberalise the sector from both domestic business, multinationals and foreign governments. While labour and the African National Congress may have preferred continued state ownership of the public switched telecommunications service (PSTS) operator Telkom, its high level of indebtedness severely reduced its ability to realise the level of infrastructure investment required to achieve its social objectives. The fiscus was also not in a position to fund such investments because of its own high debt levels and large annual budget deficit. The policy outcome was a pragmatic compromise, with Telkom being partially privatised (30 per cent) and granted a five‐year exclusivity period within which it had to roll out infrastructure to meet the social objectives.

To ensure that the public switched telephone network's (PSTN) exclusivity period fulfilled the goals of infrastructure roll‐out, strict licence conditions were placed on the network provider. These included rolling out 2,81 million new lines over the period, of which two‐thirds would be in underserviced areas and for priority customers (Table 1). Financial penalties would be imposed for failure to reach these targets. Telkom would pay R450 per line for the first 100 000 lines missed and R900 per line for each extra line missed. If it missed priority customer targets the penalty per unit would be R4 500, for schools R900, public payphones R2 250 and villages R1 125. The targets were set on the basis of the teledensity level South Africa should have, given its per capita income level. The expectation was that the country could aim for a teledensity of 20 phones per 100 people – double the teledensity at the time. This target was translated into a specific target in terms of the number of lines. The mobile operators were not given specific roll‐out targets because they were licensed prior to the consultative policy process, and were considered a luxury service that did not have mass appeal.

Licence obligations for operators in Phase 1

	Roll‐out obligations	Community service obligations
Telkom	2,69 million lines brought into service of which: 1. 1,676 million are in underserviced areas 2. 20 246 are for priority customers 3. 3 204 are for villages	1. 120 000 payphones
MTN	1. 60 per cent population coverage in two years 2. 70 per cent population coverage in four years	1. 7 500 community service telephones in underserviced areas over five years 2. Low community service tariff
Vodacom	1. 60 per cent population coverage in two years 2. 70 per cent population coverage in four years	1. 22 000 community service telephones in underserviced areas over five years 2. Low community service tariff
Sources: Telkom, MTN and Vodacom licence agreements (see, for example, Department of Communications, 1997; Icasa, 2002a, 2002b).

The universal service policy embodied in the roll‐out targets for Telkom was complemented by universal access policies. These included payphone roll‐out targets for both Telkom and the two mobile operators, MTN and Vodacom. All three operators also contributed to a USF. However, the contribution to this fund was capped at R20 million per year – R10 million for Telkom and R5 million for each mobile operator. The Universal Service Agency (USA) was established to manage the USF, which was to be used exclusively for subsidies to assist needy people towards the cost of access to telecommunications and/or for Telkom and other operators to assist them in roll‐out (South Africa, 1996).

1.2 Licence conditions in Phase 2 of telecommunications reform

The second phase of the reform process began in 2001 with steps to determine the reforms that should follow after the PSTN's exclusivity period ended in May 2002 (South Africa, 2001a). The policy direction that emerged was one of gradual, ‘managed’ liberalisation and not an immediate opening up to competition in all quarters. The primary components of the policy were as follows:

1.	Second national operator (SNO). The introduction of a single facilities‐based competitor with a full PSTS licence that would include Easitel (the internal telecommunications arm of the state electricity operator, Eskom), Transtel (the internal telecommunications arm of the state transport operator, Transnet) and a black economic empowerment partner.
2.	Single carrier of carriers licence. Sentech, the state broadcasting signal‐distribution company, would be licensed to provide international gateway services to other operators only, and not directly to the end‐user.
3.	Single multimedia licence. Sentech would be licensed to build a network to transport media content (e.g. internet, video, data).
4.	Numerous underserviced area licences (USALs). A number of small and medium‐sized enterprise operators would be licensed to provide local loop PSTSs to areas with a teledensity of less than 5 per cent.
5.	The appointment of a Board by the Minister to oversee the USA.
6.	The inclusion of school internet access as a new goal for universal access and the provision of a new e‐rate that provides a 50 per cent discount on calls for dial‐up internet access in schools.
7.	The addition of fixed‐mobile services to the PSTS licence for both Telkom and the SNO. Fixed‐mobile services allow the subscriber to link to the PSTN from either a fixed or mobile device, but the latter does not allow call handover to other cells, thus limiting mobility to the local exchange area.
8.	The establishment of the next phase in the reform process (2005) where additional entrants and resale would be examined.

In parallel with this policy process, a third mobile cellular operator was licensed. The initial reform direction anticipated that an additional cellular provider might be viable by 2001, and so built in a review in 2000 which resulted in the initiation of a licensing process for the third operator. After considerable delays and litigation (see Ayogu & Hodge, 2002, for details), the new entrant, Cell C, finally began services in November 2001.

Licence obligations for operators entering during the second period of reform are listed in Table 2.

Licence obligations for operators in Phase 2

	Roll‐out obligations	Community service obligations
Cell C	1. 8 per cent geographic coverage in five years; 40 per cent with roaming agreements 2. 60 per cent population coverage in five years; 80 percent through roaming agreements in one year	1. 52 000 community service telephones in underserviced Areas over seven years 2. Low community service tariff
Sentech (carrier of carriers)	None	None
Sentech (multimedia)	None	1. 500 Internet laboratories in rural schools over five years
SNO (proposed)	Coverage of all metropoles in five years	30 000 community service telephones in rural areas over ten years
	1. 80 per cent of territory in ten years	1. 2 500 Internet laboratories in rural schools over ten years
Source: Operator licences (see, for example, Icasa, 2001, 2003).

Only the SNO and the new mobile entrant were given roll‐out obligations. The perceived success of mobile licensing in Phase 1 meant there was no great need to tamper with the formula. The result was roll‐out obligations similar to those of the first two operators, but less demanding. With the existing high national coverage, the emphasis was on network competition rather than additional geographic coverage.

For the SNO a very different approach is followed, compared with the Telkom roll‐out targets. Here the emphasis is on geographic coverage and points of presence rather than specific numbers of lines. The regulator has accepted that the only viable entry strategy is to target business users in the metropoles for the first few years until the operator is profitable and has a revenue stream for reinvestment in broader roll‐out (Madyibi, 2003). Again, because the focus is on building network competition rather than bringing new areas into service, different types of obligations are appropriate.

The community service obligations have also shifted in direction. First, far greater demands are being made on the delivery of community service to compensate for lower roll‐out demands. This demonstrates a shift to focusing on universal access and not service (as in the Telkom licence). Secondly, the obligations are no longer limited to payphones but also include internet laboratories in schools.

More recently it was announced that cellular operators would have extensive universal service obligations imposed on them for 15‐year access to the 1 800 MHz and 2,4 GHz spectrum. Initially a one‐off licence fee of R700 million was mooted, but a deal was struck that reduced the licence fee to R100 000 per annum per frequency pair used, a R5 million annual fee, 5 per cent of net operating income and a list of universal service obligations. These included the supply of 250 000 free cellphones over five years, provision of 4 million free Security Identity Module (SIM) cards over five years, more public payphones, internet laboratories in schools and multipurpose community centres (Business Day, 2003).

2 A LESSON IN ROLL‐OUT OBLIGATIONS: THE FAILURE OF TELKOM RESIDENTIAL TARGETS

The failure of Telkom's network expansion policy during the first phase of the reform is now relatively well known. As documented in Table 3, the failure of the universal service policy was not due to an inability to finance and physically roll out the required number of lines, but rather because of Telkom's inability to sustain subscription to these lines.

Network expansion of residential fixed line (000s), 1997–2002

	Total lines	Annual roll‐out of lines under programme	Annual net change	Cumulative lines rolled out	Cumulative addition to operating lines
1997	4 258
1998	4 645	354	387	354	387
1999	5 075	495	430	849	817
2000	5 492	621	417	1 470	1 234
2001	4 961	630	−531	2 100	703
2002	4 925	570	−36	2 670	667
Source: Telkom (n.d.).

Rapid initial expansion occurred until 2000, in line with the roll‐out targets set in the licence. However, this growth only lasted until the 2000/1 financial year when Telkom made a business decision to clamp down on bad debt and enforce more strictly the timely payment of accounts (Telkom, 2001). This precipitated mass disconnections from the PSTN that went beyond normal churn. In 2001, a total of 1 160 000 lines were disconnected, resulting in an actual decrease of 530 000 in the total number of active lines despite 630 000 new lines being rolled out. In 2002, a further 606 000 lines were disconnected, resulting in a more modest decline of 36 000 lines in total despite an additional 570 000 new lines being rolled out.

Given the massive disconnections from the fixed‐line network, the policy of roll‐out targets for exclusivity has to be considered a failure. The cost of policy failure can be summarised as the cost of wasteful investment in the lines that were disconnected, and the deadweight loss to society from the higher prices sustained during the exclusivity period to finance the investment. The cost of the wasted investment has been estimated at R17 billion (Hodge, 2003), based on Telkom's figures on the cost of line roll‐out to underserviced areas.

With regard to the policy of roll‐out targets, the source of the problem is essentially twofold. In the first place, a limited information set is used to determine roll‐out targets in an uncertain environment, making their suitability subject to enormous potential error. Secondly, roll‐out targets as they are implemented at present – whether for residential use or payphone access – are inflexible instruments, and thus cannot be adapted to the potential market changes used in their estimation in the first place.

2.1 Targets are subject to uncertainty

The fact that Telkom could physically meet the target number of lines but was unable to sustain the demand for them suggests that the targets themselves were inappropriate. Roll‐out targets are determined through a simulation of expected demand and cost for the operator over the roll‐out period. The simulation is an information‐intensive exercise that of necessity makes use of estimates or assumptions about the value of certain key parameters. The estimation of what are suitable targets is sensitive to the accuracy of the assumptions made and also to unanticipated shocks to any of the variables. The expected variation will increase the longer the period over which the targets are set, and the extent to which the sector or consumers undergo change in that period. If assumptions are inaccurate or the demand/cost dynamics of the sector change unexpectedly in the roll‐out period, the targets will no longer be suitable. This may result in targets not being met or sustained, or in targets being easily achieved. The former is a far greater problem for social delivery because investment is wasted in the process. The latter is still problematic because it implies that there has been a distribution of surplus from consumers to producers as producer concessions exceeded actual cost to the producers.

In a dynamic sector such as telecommunications one should expect a high level of uncertainty, making any simulation exercise subject to considerable speculation. This uncertainty is compounded by the fact that it would be difficult to arrive at good estimates of the potential demand amongst unserviced low‐income consumers. The result is that the targets set in the licence had the potential to be inappropriately high or low. With hindsight, one can argue that they were set inappropriately high given the many changes that took place in the market since 1996 when the targets were set. What follows is a discussion of some of these changes and their impact on Telkom's ability to achieve the targets.

2.1.1 Target assessment

Telkom was set the target of having expanded residential lines by 2,69 million by 2002. As there were about 2,95 million residential lines in existence in 1997 when the roll‐out period started, this target translates into 5,24 million residential lines by 2002. The exclusivity period only offered a temporary period of cross‐subsidisation because of the expectation of competition at the close of the period, and due to the understanding that contributions to the USF would not be used extensively to subsidise individual consumers, if at all. This means that network expansion could not target uneconomic or unprofitable consumers because they required ongoing subsidy. Rather, it had to target profitable consumers in profitable areas who were denied access by apartheid, or profitable customers in uneconomic areas where the additional cost of expanding the network would be covered by the temporary subsidy. The roll‐out targets would be achievable and sustainable if the size of these two groups of profitable customers exceeded the targets.

Table 4 assesses the potential market for residential fixed‐line services in 2003 based on the affordability of the service.

Potential market analysis for fixed line residential services, 2003

	1995 IES expenditure patterns		1999 IES expenditure patterns
Decile	Expenditure on telecommunications as a % of income	Average spending (rand)	Expenditure on telecommunications as a % of income	Average spending (rand)^1
1	0,51%	3,71	1,18%	8,61
2	0,65%	6,79	1,59%	16,61
3	1,06%	16,12	2,10%	31,93
4	1,30%	25,61	2,71%	53,61
5	1,58%	41,78	3,42%	90,41
6	1,77%	63,67	4,10%	147,29
7	1,96%	101,15	3,94%	202,74
8	2,46%	187,99	4,07%	310,89
9	2,24%	273,74	3,98%	487,86
10	2,01%	552,83	4,22%	1162,21
Potential market analysis
Deciles that can afford		6–10		5–10^3
Potential market^2		5 950 000		7 140 000
Target/potential		88%		73%
Actual penetration of potential		50%		41%
Notes: 1. Nominal expenditure growth from 1995 based on Reserve Bank consumption expenditure growth figures. 2. This is based on the number of deciles that can afford the service times the number of households (11,9 million). 3. Decile 4 has been excluded because although expenditure can cover the monthly fee, it leaves only sufficient spending to make six minutes of calls, which means it is unlikely that households in this category will subscribe.

Affordability is determined by whether the household's expenditure on telecommunications (as a percentage of income) is above the fixed fee, with a reasonable amount of expenditure left for usage. The fixed fee used is that of prepaid Telkom service because this is the lower of the two and one assumes that low‐income consumers would take up this service. Using 2003 Telkom tariffs, any household spending less than R480 per month on a fixed‐line telephone service would prefer the prepaid to the postpaid option. Two different estimates of expenditure as a percentage of income are used for the analysis – the 1995 Income and Expenditure Survey (IES) (StatsSA, 1996), and a 1999 update of this estimate by WEFA Southern Africa (2000). The 1995 figures are probably on the low side because the lack of availability of telephone services to low‐income groups may have limited their expenditure on communications. The 2000 IES (StatsSA, 2002) has similar estimates to the 1995 IES, but problems with that database suggest they might be inaccurate.

The results of this analysis suggest that the targets set were reasonably challenging – requiring a penetration of the potential market of between 73 and 88 per cent. To put this in perspective, the penetration rate of the potential market for 1995 was 62 per cent and for 1997 was 65 per cent. As noted in the table, actual penetration of this potential market was considerably lower at between 41 and 50 per cent, depending on assumptions over expenditure figures. As penetration rates were previously around 65 per cent, this suggests considerable unanticipated changes in consumer demand for fixed‐line residential services during the roll‐out period. Two such changes are analysed below – demand substitution to cellular service and the impact of rate rebalancing.

2.1.2 Demand substitution

A primary source of unanticipated changes in consumer demand came from the introduction of cellular services, in particular the prepaid cellular package. Hodge (2003) takes a detailed look at why consumers with low expenditures on communications would prefer a prepaid cellular service to a fixed‐line residential service. In essence, although the former has considerably higher usage rates than the latter, it also has no monthly fees. Therefore the actual usage rate of cellular service, including contribution to fixed fees, is lower than that for fixed lines for low usage levels. Hodge (2003) estimates the point where spending is equalised for the same number of minutes to be about R85 in 2002, using certain assumptions about usage patterns. Table 5 updates these estimates using 2003 tariff rates and the actual usage patterns experienced by the operators as per their regulatory tariff filings. The average of usage patterns is almost the same as the previous estimate of R82 for prepaid fixed lines to a prepaid cellular service. As the prices of each product impact consumers' calling pattern, the estimate is higher for those on cellular networks (they call other cellular subscribers more often) than for those on the fixed‐line networks.

Fixed line‐cellular spend equalisation levels for different usage patterns, 2003

	Prepaid Telkom‐Cellular spend equalisation level	Postpaid Telkom‐Cellular spend equalisation level
Telkom usage patterns	R65 (32 min.)	R102 (51 min.)
Cellular usage patterns	R98 (49 min.)	R157 (79 min.)
Average of usage patterns	R82 (41 min.)	R130 (65 min.)

Using these estimates of consumer preferences for cellular over fixed‐line services, one can reassess the targets set for Telkom by assessing the impact that cellular service has on the potential market for fixed‐line connections. One can expect that consumers who are able to afford the Telkom line but spend at a level that makes them prefer cellular will always adopt cellular rather than fixed‐line services, thus eliminating them from the potential market for a fixed line.

Table 6 presents the results of this adjustment of the potential market for fixed lines based on the introduction of the cellular prepaid service. Two estimates are presented for each expenditure pattern used in the initial analysis. One estimate uses the average spend equalisation level for prepaid fixed line and cellular, and the other a higher estimate of this spend equalisation level, substituting one decile higher away from Telkom. The use of a higher estimate is justified because other factors enter the consumption decision on technology choice, and all of these factors favour a choice of cellular over fixed‐line service. The advantages of a cellular service include a significantly lower waiting time for a phone; the additional utility from mobility; no financial penalties for reconnection; a second‐hand market for phones, which reduces connection charges; per‐second billing for the first minute; and the additional utility from free added service and handset features (e.g. voicemail, caller identity and directory). (See Hodge, 2003, for a more detailed discussion.)

Revised potential market analysis for fixed‐line service based on the introduction of cellular prepaid service, 2003

	Low estimate*		High estimate**
	1995 IES expenditure patterns	1999 IES expenditure patterns	1995 IES expenditure patterns	1999 IES expenditure patterns
Deciles switching to cellular	5	4	5 and 6	4 and 5
Potential market for fixed‐line service with cellular substitution	4 760 000	5 950 000	3 570 000	4 760 000
Target/potential	110%	88%	147%	125%
Actual penetration of potential by fixed‐line service	62%	50%	83%	62%
Notes: *Makes use of the average prepaid fixed line‐cellular spend equalisation level. **Assumes that one more decile will prefer the cellular service due to other factors.

The results suggest that this demand substitution has had a significant impact on Telkom's ability to realise the targets set in its licence. In only one case – for the low estimate of demand substitution and the 1999 IES expenditure patterns – does the revised potential market remain above the set roll‐out targets. In all other cases the potential market for fixed‐line service shrinks below the targeted number of lines for Telkom. The scenario of the high 1999 IES estimate for expenditure and the high estimate for cellular demand substitution seems relatively realistic, as the actual penetration rate of this potential market of 62 per cent is in line with the 1995 and 1997 potential market penetration rates achieved by Telkom.

The result of this demand‐side substitution is reflected in access figures for different household income levels using household survey data in Table 7.

Percentage household ownership of a telephone by income group and location, 1998 and 2001

Monthly income	Fixed line	1998 Mobile	Total*	Fixed line	2001 Mobile	Total*
Up to R499	5,6%	0,1%	5,6%	5,6%	4,9%	9,7%
R500–R899	7,0%	0,3%	7,1%	9,4%	6,2%	14,7%
R900–R1 399	15,3%	0,5%	15,4%	15,0%	9,3%	22,4%
R1 400–R2 499	31,3%	2,0%	32,3%	23,6%	15,1%	35,5%
R2 500–R3 999	51,9%	6,4%	54,8%	41,3%	24,0%	56,3%
R4 000–R5 999	66,3%	12,6%	71,2%	52,8%	36,4%	73,3%
R6 000–R9 999	77,4%	22,9%	83,1%	62,7%	48,6%	85,7%
R10 000+	87,9%	42,2%	92,7%	75,9%	59,7%	93,9%
Location
Rural	4,4%	1,3%	4,8%	6,4%	8,8%	13,2%
Urban	49,1%	10,7%	51,4%	43,5%	28,6%	57,6%
All households	29,2%	5,8%	30,6%	28,7%	20,7%	39,9%
Note: *Some households will have both fixed‐line and mobile connection, making the total less than the sum of the fixed‐line and cellular ownership columns. Source: SAARF (1998, 2001).

The table shows that while household ownership of a telephone has improved dramatically from 30,6 per cent of households in 1998 to 39,9 per cent in 2001, it has been entirely due to mobile networks. In fact, the proportion of households with fixed lines has actually declined over this period (from 29,2 to 28,7 per cent). In contrast, the proportion of households owning a cellular phone increased from 2,14 per cent in 1995 to 20,73 per cent in 2001. For 11,2 per cent of households, the cellular phone was their only connection to the telephony networks, up from only 1,5 per cent in 1998. It is therefore safe to argue that cellular service – and not fixed line – has been the success story of improved access in South Africa despite the huge investments made in fixed‐line service during the exclusivity period.

2.1.3 Rate rebalancing

A number of commentators have identified rate rebalancing as another reason why Telkom experienced such high disconnection rates during the roll‐out period. Rate rebalancing in preparation for competition results in the price of local calls increasing and the price of long‐distance calls decreasing. This has a detrimental effect on residential access because such users typically make far more local than long‐distance calls, resulting in an increase in the average bill. However, the demand for a phone is dependent on both access charges (installation cost and monthly exchange access) and the price of calls or usage (including local, national long‐distance and international long‐distance calls). Consumers therefore weigh up the usage and non‐usage utility received from having access; if it exceeds the cost of access, they will choose to subscribe. Thus the impact of rate rebalancing on the consumer is a combination of changes to access fees and usage tariffs.

The nominal and real changes in Telkom tariffs from 1997 to 2003 are documented in Table 8. A listing of the changes to each usage tariff (local, long distance, mobile and international) does not provide a useful picture of how a consumer's monthly expenditure is likely to change as a result of the tariff increases. Instead, Table 9 lists an average tariff per minute, determined by multiplying the tariff for each service by the proportion of total call minutes that the service covers for the average consumer. These proportions are taken from Telkom's annual tariff filings. International calls are excluded because the Telkom filings do not list the average tariff. However, such calls make up a small proportion of average residential call minutes and are likely to be even lower for low‐income households – which is the focus of this assessment.

Real changes in tariffs for fixed line, 1997–2003

	Monthly fees	Average tariff per minute	Average tariff per minute (cell use at 1997 level)
Postpaid 1997	R44,69	R0,18	R0,18
Prepaid 2003
• Nominal	R47,23	R0,51	R0,38
• 1997 rand	R32,13	R0,35	R0,26
• Real change	−28%	94%	49%
Postpaid 2003
• Nominal	R76,20	R0,54	R0,42
• 1997 rand	R51,83	R0,37	R0,29
• Real increase	16%	106%	66%

Estimations of the change in usage and monthly bills of residential consumers under tariff rebalancing

		Monthly expenditure on communications in 1997
	R40	R50	R60	R70	R80	R90	R100
1997 tariffs
Usage (minutes)	No usage	30	87	145	202	259	316
2003 prepaid tariffs (1997 rand)
Usage (minutes	21	26	75	125	174	223	272
New spending (R)*		41,80	60,00	78,20	96,40	114,61	132,81
% change in spending		−16%	0%	12%	21%	27%	33%
Welfare change (R)		5,32	−8,31	−21,94	−35,57	−49,20	−62,83
2003 prepaid tariffs (cell use at 1997 levels) (1997 rand)
Usage (minutes)	27	27,43	79,08	130,73	182,38	234,04	285,69
New spending (R)*		40,08	55,06	70,04	85,02	100,00	114,98
% change in spend		−20%	−8%	0%	6%	11%	15%
Welfare change (R)		7,82	−1,12	−10,05	−18,99	−27,92	−36,85
Notes: A constant price elasticity of usage of −0,2 was used. *Includes both usage expenditure and monthly fees.

The tariff changes show that although the average usage tariff increased significantly from 1997 to 2003 (up 94 per cent for prepaid users and 106 per cent for postpaid users), there was a real decrease of 28 per cent in the monthly access fees for those consumers taking up the prepaid Telkom service. As monthly fees are commonly used as a measure of affordability, this would suggest that fixed‐line residential service actually became more affordable during the exclusivity period. However, a more complete picture needs to determine how the rise in usage charges affected the monthly bills of consumers. This can be achieved by fitting a telecommunications demand function to the household at each initial expenditure level and determining the change in behaviour from the increase in the average tariff. This change in usage behaviour can then be translated into a change in spending by applying the new average tariffs and monthly access fees. Table 9 shows the results of such an exercise, making use of a constant elasticity demand function with an elasticity of –0,2 and real price changes from 1997 to 2003. The choice of elasticity was based on estimates for Peru (Torero et al., 2001) and is in line with expectations that usage is relatively price inelastic. Sensitivity tests show that the results are not that sensitive to higher and lower elasticity levels.

The impact of tariff rebalancing has been twofold. First, high real increases in usage charges would increase consumers' expenditure on usage and further decrease their welfare through a reduction in actual usage. Secondly, real decreases in the monthly fee will reduce their expenditure on access and increase their welfare. The total impact on expenditure and welfare will depend on how significant each item (usage spend and access fees) is in the consumption bundle. The results in Table 9 show that consumers who have a low monthly spend are actually better off from the rate rebalancing – the drop in access fees outweighs the increase in usage spending. Households that spent R60 or less in 1997 (R90 in 2003) saw no increase in spending, and those who spent less than R53,90 actually saw a welfare increase. (The lower spend figure for no losses in welfare are due to the deadweight loss from lower usage levels.) There is also a welfare increase for consumers who could not subscribe to a residential service at 1997 levels but can do so under the new tariff structure (i.e. consumers who spent between R32,13 and R44,69 at 1997 prices or R47,23 to R65,70 at 2003 prices). These results are contrary to the common wisdom on the impact of tariff rebalancing. Of course, this is a sector where one would expect real decreases in tariffs due to technological and efficiency changes. Consumers are, therefore, still worse off than they could have been.

Tables 8 and 9 also present an analysis of real increases in tariffs and their expenditure/welfare effects if cellular calls as a proportion of total calls from fixed phones remained at their 1997 level. This is another unanticipated change in the market resulting from the success of cellular usage that has significantly impacted on the average tariffs of fixed‐line users. Cellular call rates from fixed lines are considerably higher than both local and long‐distance call rates, and their proportion of non‐international call minutes has increased from 2,3 to 15,5 per cent. This increase in cellular calls has contributed to a 45 per cent increase in real tariffs for postpaid and a 40 per cent increase in real tariffs for prepaid fixed‐line service (out of a total of 94 and 106 per cent, respectively). If this change did not take place, the 1997 spending level at which consumers would not increase their expenditure from tariff rebalancing would rise to R70 (or R100 in 2003 prices), up from the current R60 (or R90 in 2003 prices).

These results do not fundamentally change our assessment of the suitability of the roll‐out targets. That analysis was based on the affordability of the fixed fee, which has decreased under rate rebalancing, and it was a rather blunt analysis using income deciles rather than a greater level of subdivision of the market. However, a more complete analysis would determine at what usage levels consumers do get sufficient surplus on their usage to cover the fixed fee. This approach would set a minimum spend that makes a fixed line both affordable and desirable. This approach was loosely used in the case where a decile whose mean expenditure on communications lay only marginally above the fixed fee level was excluded. However, a more scientific approach needs to be used. In this case, the unanticipated increase from greater cellular calls would impact the affordability benchmark and the suitability of the target.

2.1.4 Other demand uncertainties

Other sources of possibly unanticipated demand changes that may also have impacted on the suitability of the roll‐out targets include increasing preference for payphone access, reductions in expenditure on communications, and lower than expected income growth. One can also add to this list a range of cost uncertainties.

A further source of demand substitution that the operator most probably faces is an increasing preference for public payphones. In its decision on whether or not to subscribe to a fixed line, a household will take into account the relative cost of cellular service, but also the relative cost of and utility from payphone use. Payphone use does not come with a fixed monthly fee, which means that a household's total expenditure on such a service will be lower than for a residential phone, despite the fact that average usage charges for payphones are higher than for a residential service. However, the fact that a household is not able to receive calls except by prior arrangement and that payphones are often some distance from the home result in lower utility. The decision on whether or not to acquire a residential phone in cases where households are able to afford it will depend on how the consumer weighs up the utility gains and losses of payphone use. While a determination of what this point may be for certain income deciles is beyond the scope of this article, it is apparent that the rapid roll‐out and greater access to payphones that form part of the universal access policy must have lowered the disutility from payphone use for households (the average distance to a payphone for any household would have been reduced through this roll‐out). As such, it must have increased the expenditure point below which households prefer using a payphone to a residential service.

Another demand uncertainty is the percentage of expenditure households spend on communications. This percentage is not independent of changes in preferences and above‐inflation price changes for more essential items. This is particularly relevant to poor households that are the target of universal service policies such as roll‐out targets. These households have limited degrees of freedom around expenditure on more essential items such as food, shelter and transport. Above‐inflation increases in the prices of these items will cause the household to reduce expenditure on all other items, including communications. The extent to which expenditure on an item is reduced depends on the priorities and preferences of the household. While expenditure on communications saw an increase from 1995 to 1999, the trend thereafter might have been towards a reduction in such expenditure because of high above‐inflation increases for food and fuel, and therefore transport for South African households. This would change the size of the potential market for residential phone lines and also result in some households that previously could afford a phone to disconnect (as was observed in 2001/2).

Finally, income growth may not have been as high as estimated in the initial drafting of the roll‐out targets. In the assessment of the targets above, the author has assumed that all income deciles experienced the same level on expenditure growth. However, there is sufficient evidence to suggest growing inequality in South Africa and that the lower‐income deciles are not maintaining the same level of income growth as the top deciles.

2.2 Roll‐out targets are inflexible

2.2.1 Inflexible targets

Although it has been argued that roll‐out targeting as a policy is full of enormous uncertainty, it is still possible to improve the policy even if the estimation process for the targets cannot be improved. After all, one cannot entirely fault the bureaucrats for failing to anticipate the enormous changes to the market from the introduction of a cellular service, particularly a prepaid one. However, one can fault them for drawing up an inflexible contract with the operator that does not allow renegotiation of the targets in the light of changes to the market.

Once partial privatisation has occurred and the private partner has management control, the licence with all its conditions becomes a binding contract. An optimal contract in the face of considerable uncertainty must accommodate such uncertainty through increasing the flexibility of and scope for renegotiation in the contract. It must also put in place effective monitoring of those aspects of the contract that are subject to such uncertainty. The licence for Telkom did neither. The roll‐out targets were laid out with the appropriate fines but no scope for renegotiation. This meant that once it became apparent that large numbers of subscribers were disconnecting, the programme continued to be implemented. If it had been stopped, some of the wasted investment could have been saved to the benefit of consumers. Even more wasteful investment could have been saved if effective monitoring had been in place. While the regulator was tasked with such monitoring, it failed to do so and relied on annual accounting for targets by Telkom (Mathyssen, 2003). The regulator negotiated a deal with Telkom that allowed indebted customers to not be disconnected but moved to the prepaid package (and repay debts) only in June 2002 (Business Day, 2002). By this time, around 1,7 million lines had already been disconnected.

2.2.2 Inflexible technology

A further source of inflexibility in meeting the roll‐out targets was the limited use of different technologies. The policy makers at the time granted a degree of flexibility by permitting Telkom to make use of fixed‐wireless technologies, which had been shown to entail a lower cost for low‐density areas because they eliminated the need for lengthy cabling for fixed‐wire technologies. However, in such a technologically dynamic sector as telecommunications, this level of flexibility was inadequate. As it turned out, low‐income consumers were best served by mobile cellular technologies where the low fixed costs associated with connecting a customer meant that operators were able to offer low usage schemes with low or no monthly rental but higher call rates. The cost of connecting an individual consumer is important in determining tariff structures in network industries, as it determines the minimum monthly access fee that an operator is willing to charge to recoup the fixed costs. Operators will connect customers who do not contribute to the common costs of the network as long as their revenue covers the incremental costs of connection. The lower these incremental connection costs, the lower the revenue requirement from the customer and so the lower the fixed fee.

In contrast, the high cost of connecting an individual consumer for fixed‐line service limits the extent to which the monthly fee can be reduced. For fixed‐wireless technology the cost of connecting a customer is reduced by the cost of laying the line to the home, but it still requires trained technicians to install the receiver in the home, thus raising the individual connection costs above that of cellular service. Cellular operators have also benefited from the positive externalities associated with offering a mobile service. Extension of coverage to provide mobility for the core contract customers meant that towns and villages which by themselves had insufficient demand to warrant a transmission tower were now profitable based on the sum of their demand and the ‘passing trade’ of mobile customers from urban areas.

Another technology that has been used effectively to service low‐usage customers in other countries – the so‐called fixed mobile – emerged in the 1990s. Fixed‐mobile service permits mobility within the range of a single transmission tower but no mobility between transmission towers. Again, a feature of this technology is the low cost of connecting an additional subscriber, enabling a tariff package more suitable to low‐income consumers. Telkom only gained the right to this technology in the legislative changes of August 2001 – by this stage the roll‐out period was almost over and it could not be operationalised.

The limited technological flexibility that Telkom enjoyed for the roll‐out period was also constrained by tariff inflexibility. The lower connection charges offered by fixed‐wireless service provided scope for a low‐usage scheme below that of the prepaid option that Telkom introduced. However, if Telkom were to implement this, limitations on price discrimination would have forced it to offer the same price to all potential subscribers. Given that such a tariff scheme designed for fixed‐wireless service may not be profitable for fixed‐wire service, the inability to price‐discriminate would have prevented Telkom from introducing such a low‐usage scheme in the first place. It might have been taken up unprofitably by either new or existing consumers in areas already served by fixed‐wire technology.

3 COMMUNITY SERVICE OBLIGATIONS AND COVERAGE REQUIREMENTS

The problems of the Telkom roll‐out targets have been very visible. However, they were also unique in that they prescribed residential rather than community service roll‐out. The fulfilment of the other obligations appears to have been more successful. The cellular operators have far exceeded their coverage requirements and the community service and payphone roll‐out has been completed. This does not imply that they have been entirely successful because they may not have been achieved at least cost to the government and may not have been placed in areas of most need. If a government commits itself to universal access for political reasons, its objective should be to provide such access as efficiently as possible, implying least cost and effective targeting of those in need.

As noted previously, easy achievement and exceeding of targets imply that the costs to the operators have been below those of the concessions granted to them. The result is that universal service is more costly than it should be, with a distribution of surplus to producers from consumers, who pay higher prices or are denied additional tax revenues. This may have been the case with the cellular licences where coverage exceeded the licence conditions (coverage reached 90 per cent of the population by 2000) and the number of community phones. There is more likely to be a conservative (and therefore higher) estimation by government bureaucrats of the potential cost of the obligations to the operator. This is because of information asymmetries over cost and expected revenue information that place the operators in a stronger bargaining position. Also, the successful delivery on these obligations is an easier and more visible means by which voters can assess the performance of bureaucrats and politicians, while a higher cost than necessary is less visible. This gives bureaucrats an incentive to err on the side of caution when assessing costs to ensure that the operator delivers on the obligations.

Furthermore, the delivery by the operators themselves may not have been the method with the least subsidy available to the industry as a whole. The least subsidy depends both on the cost of providing the service that will differ with different technologies, and on whether the operator can exploit operating externalities in an area to reduce the cost. Licences are currently technologically constrained, especially those of mobile operators. Even the relatively broad technological freedom of the fixed‐line and underserviced area licences does not include the use of mobile technology. As noted previously, the coverage externalities of operating a network selling mobility means it might have coverage in certain remote and unprofitable markets that can be exploited through community payphones. Similarly, for certain areas, the USALs – which could capture the economies of scope from operating in the area and benefit from special treatment that lowers their costs – would be strong contenders to provide some of these community service phones at a lower price. It is also questionable whether an operator such as Telkom would exploit the least‐cost technology at its disposal in all areas. While the operator has a choice of technology, it is likely to choose an appropriate technology for each area and then be locked into that choice. The fact that a large part of the country is on fixed‐line technology already would also limit the operator's choice in those areas.

The other potential problem with these seemingly successful roll‐outs is that the distribution of the phones by the operators may not have been in line with areas with the greatest need. There has not been effective control and monitoring over where the phones are placed beyond rather broad definitions of underserviced areas – which gives operators an incentive to place phones where there is least cost to themselves within these broad constraints. This might partially solve the problem raised above, although it still leaves the problem of concessions being higher than costs. The downside, however, is that areas of most need may not be serviced and a lack of coordination amongst operators may lead to a clustering of phones in urban and semi‐urban areas with high revenue potential. For instance, operators in Pretoria townships are said to be placing phones in areas where there is proven demand through the profitable use by small shop‐owners. Such placement of phones reduces the demand for the independent provision of a phone service by the shop‐owners, often forcing them to stop their service. Clearly this does nothing to improve access, as the areas are already serviced.

4 WHY HAS THE USE OF LICENCE OBLIGATIONS CONTINUED?

4.1 Licence conditions in Phase 2

The lack of a residential roll‐out target for the SNO and USALs in the second phase of reform shows that there has been some learning from the Telkom roll‐out failure. However, licence conditions for universal access still remain an important feature of the delivery of a universal service. Furthermore, the recently announced obligations for the cellular operators for spectrum access have elements of residential provision (free handsets and SIM cards) that appear more vulnerable to failure. Given the demonstration that poor users are adopting cellular above fixed‐line service, it is tempting to welcome the switch in strategy to put more focus on cellphone access. However, the strategy of continuing to use licence obligations to achieve this still raises questions. As already noted, licence obligations are less flexible and are subject to great uncertainty because of the lengthy period over which they are implemented. There is also concern over whether the government is getting a good deal, given the concessions on spectrum fees it has made – both in terms of least cost and sustainability.

While cellular phones may currently seem to be the best means of reaching the poor at the least cost, are we not making the same mistake as Telkom did in assuming that no better alternative will arise? In fact, it is entirely plausible that the fixed‐mobile technologies which the USALs are likely to use might offer a better package for consumers. These technologies have the benefit of low consumer connection fees, enabling a lower monthly access fee. They should offer lower call rates, given the lower technical requirements for cell handover that increases the cost of mobile networks. These technologies have enjoyed enormous success in India amongst poor households and at competitive call rates. Should we not be collecting the spectrum fees from the cellular operators and buying fixed‐mobile phones for the poor? There is at least sufficient uncertainty to make it seem imprudent to commit to one technology for the next five years.

4.2 Why stick to licence conditions?

The initial use of licence conditions can be understood in terms of the policy and institutional environment at the time. First, the policy of exclusivity seemed to make sense as a means to meet a number of competing government objectives. Not only did it force the operator to focus on universal service issues, but it also enabled the state operator to improve its value, provided time to establish a regulator and ensured that consumers were not faced with an immediate rate rebalance. As there would be no competition for subsidies beyond the mobile service (which was assumed to be unable to compete for low‐income households), it seemed to make little sense to use fund contributions to a universal service managing agency which would then disburse them back to the operator. Administrative costs would be minimised by internalising the process with the operator, and there would be no informational or bargaining gains from putting it with the USA. Furthermore, as the Agency still needed to be established, using it to disburse funds would delay the roll‐out of universal service further. The initial intent was to make greater use of the USA and USF contributions in the second phase of reform where competition for subsidies was feasible, needy individuals would have been identified and capacity was established in the USA to manage the process.

4.2.1 Agency failure

An important reason why the Department of Communications may have decided to continue to use licence conditions in the second period was probably that the USA suffered institutional failure on a similar scale to the Telkom roll‐out disaster. The Agency failed to deliver on its mandate to develop universal access and service targets for the country and use the USF monies to enhance access. The precise reasons for the Agency's organisational failure are not apparent. However, while it initiated a process of defining the goals for universal access, this task was never completed during the first phase of reform (USA, 1998, 1999). The USA also decided to put all its eggs in one basket and back the establishment of telecentres with all the USF monies. Telecentres are facilities that offer access not only to voice telephony, but also to other information and communication technologies, such as fax machines, scanners, computing facilities and the internet. According to Benjamin (2001), the USA initially set targets for hundreds of telecentres but costs were higher and progress slower than expected. The first centre was only established in 1998 and only 34 were operational by 2000. Many of these telecentres have subsequently failed.

Benjamin (2001) argues that the South African‐styled telecentres were inappropriate for delivering broad access to telephony because they provided far more than was necessary and so raised their costs dramatically. A typical South African telecentre costs about US$40 000 and includes five computers with internet access amongst other communication devices. The higher cost has limited the number of telecentres, thereby limiting the extent of improvement in telephony access brought about by the USF. Building telephony access would require a focus on lower cost payphones that could be more broadly deployed.

In essence, the managing Agency misunderstood demand and miscalculated costs of delivery, resulting in an unsustainable programme that did not improve access significantly. Tasking this Agency with even greater levels of responsibility for universal service would be political suicide. The second phase of reforms tried to address the institutional failure of the USA through the appointment of a Board to oversee the operations of the Agency. It also tried to address some of the problematic aspects of universal access roll‐out by tasking the USA with establishing a geographical information system (GIS) to keep record of operators' roll‐out of payphones and telecentres, and to determine the best use of funds for achieving the targets. The GIS would allow them to guide operators as to where to install their payphones in future according to greatest need.

However, the damage was already done in that a second round of licence conditions was imposed and the contributions to the USF limited. During the latter part of the initial reform period, Icasa was also required to determine operators' future contribution to the USF. The amendment to the Act provided an upper limit of 0,5 per cent of annual turnover (South Africa, 2001b). Icasa's initial recommendation was to have operators who did not have community service obligations stipulated in their licences contribute the maximum of 0,5 per cent, and those who did have obligations 0,4 per cent (Icasa, 2002a). However, this position was contested on the basis that there was no clear guideline from the USA as to what it would use the funds for. Because the Agency lacked a clear plan for improving access and a history of wasteful projects, Icasa was forced to reconsider its recommendation and subsequently dropped the initial recommendation for a flat 0,2 per cent contribution by all operators (Msimang, 2003). Hence agency failure in the first phase of reform diminished the universal service provisions in the second phase.

4.2.2 Bureaucratic budget maximisation

In part, licence obligations also remain because of the desire of the regulator and the Department of Communications to have more funds and greater control over the development of the sector. Currently, licence fees and spectrum fees raised in the sector are taken into the government's general fund, which the national Treasury oversees. A member of Icasa has admitted that by imposing licence obligations, the regulator could keep more money in the sector for development than would be the case if the sector were at the mercy of a budget allocation from Treasury. This is a clear case of bureaucratic incentives to try and keep as much funds under Icasa's control rather than permit these funds to enter the general ‘pot’ and let the political process determine their best use.

5 CONCLUDING REMARKS

The goal of any government that has universal service priorities should be to deliver them, and to do so at the least cost. Licence conditions as a means to achieve universal service have been shown to be problematic in the telecommunications sector. The failures of this approach seem to be greater, or at least more visible, with residential targeting than with community access. The particular failures in South Africa's case have hinged on a lack of flexibility being built into the roll‐out contracts, a lack of flexibility in technology use and non‐achievement of the lowest cost for the fulfilment of these obligations. Targets set for a number of years are inherently risky, given the many assumptions that must go into their calculation and the degree of market change that can occur over the period. This is especially true of a dynamic sector such as telecommunications.

This situation suggests that the use of licence conditions can be improved by agreeing to more flexible contracts so that both parties can respond to market changes or poor prior assumptions made in drawing up the original targets. Improved monitoring can also ensure that problems are identified early and that operators fulfil their obligations in the manner in which they were intended. However, it also suggests that the use of an alternative system of universal fund contribution and the purchase of universal service from operators might offer a better solution. Yet, it too can suffer from agency failure. Furthermore, while policies such as the auctioning of subsidies by a USA may seem appropriate on paper, they could also be problematic in practice, through collusion amongst bidders. At the very least they should offer greater flexibility, as contracts are not drawn up for long periods and least‐cost technologies can be sought for delivery.

Some of the problems that have plagued the roll‐out of universal service obligations in telecommunications are also relevant to other utilities with different ownership structures and technological options. For instance, the ten‐year National Electrification Programme (NEP) that also prioritises household access has experienced similar problems. Initial estimates of the cost of and expected revenue from the NEP were well off the mark. Eskom calculated the initial costing based on delivery to mainly urban and semi‐urban customers. However, as municipalities had jurisdiction over these areas, Eskom was forced to connect greater numbers of rural households to reach its targets – thus raising the cost of provision.

The utility also miscalculated demand and therefore the revenue from the service. Eskom provided a prepaid tariff with no monthly fees so as not to exclude any households from affording the service. This tariff would cover the fixed costs of connection if consumers used 350 kwh of electricity each month. However, these households' average consumption stands at a third of this value, contributing to a ballooning subsidy for electrification. Reasons for this are both the higher tariffs for prepaid use and the fact that consumers did not switch from alternative energy uses for cooking and heating. According to the 2001 Labour Force Survey (StatsSA, 2001), 83 per cent of households use electricity for lighting but only 51 per cent use it for cooking. This demand substitution problem is partly due to the costs of purchasing new electrical appliances. There have also been many ‘quiet disconnections’, where a connected household stops using the electricity source. In 2000, the electricity regulator estimated that roughly 6,8 million households had been connected, while the IES (StatsSA, 2002) puts the number of households spending on electricity at roughly 6 million. This suggests that 800 000 households have quietly disconnected from the programme. As soon as competition is introduced into the electricity market and household subsidies are reduced, Eskom will be forced to charge monthly access fees, which should see a large increase in such disconnections.