Andrew Sears <email@example.com>
Massachusetts Institute of Technology
One of the most important regulatory issues facing the Internet is the issue of access fees to the telephone network. At one extreme are some phone companies in the United States that seem to argue that the phone system will collapse soon if access fees are not changed. At the other extreme are some Internet access providers that seem to argue that the whole Internet will collapse if access fees are changed. This paper argues that neither of these extremes is correct. This paper examines the economics of Internet access and explains the strategic significance of access fees to the telephone network. The paper then draws some conclusions of the implications of this for business strategy and regulation. The paper concludes that while access fees may need to be adjusted slightly to reflect costs, they should be adjusted on a flat-rate basis rather than using per-minute fees. The paper examines some proposals on how this could be done while maintaining economic efficiency and avoiding any harm to the advancement of universal service.
Keywords: economics, access fees, regulation, strategy, pricing, billing, IAP, ISP, Internet telephony, conferencing, LEC, FCC, ACTA, cost.
The issue of access fees has been closely coupled with the development of Internet telephony--the ability to transmit real-time voice over the Internet. In March of 1995, America's Carriers Telecommunications Association (ACTA) filed a petition to ban Internet telephony before the Federal Communication Commission (FCC) . It now appears that the FCC will not act on that petition directly; however, the key issue of the petition still exists, which is access fees. The FCC recently proposed new access fees for long distance carriers, but chose not to change access fees for the Internet for right now. The FCC did issue a Notice of Inquiry to request information about issues related to Access Fees for IAP's. This means that while the IAP's do not have to pay access fees in the short term, the long term issue has yet to be decided.
The Local Exchange Carriers (LEC's) and long distance resellers argue that the existing access fees for Internet Access Providers (IAP's) are unfair and should be changed. Currently, long distance carriers and resellers are required to pay a per-minute access fee to the LEC's for every call. This access fee is charged at both ends of the call and is about 5 cents per minute. Based on a recent ruling from the FCC, these access fees will be decreased significantly, although the LECs have succeed in having this change put on hold by order of the Supreme Court. There are two main purposes for these access fees. The first purpose is for the long distance carriers to compensate the LEC's for using their networks. The second purpose is to advance universal service by allowing long distance calls to subsidize local access, so everyone can afford access to telecommunications. These access fees are a significant source of revenue for the LEC's, and make up approximately 40% of the costs for the long distance carriers.
The main controversy is over the fact that IAP's are not required to pay access fees. When AT&T was broken up, a new class of service providers, called Enhanced Service Providers, was created to make it possible for others to provide value added services. Service providers under this classification were not required to pay access fees, and since IAP's are considered Enhanced Service Providers, they are exempt from paying the fees. They simply lease lines in bulk from the LEC and pay monthly fees rather than paying a large per minute access fee.
There are several reasons why LEC's and resellers are arguing to increase access fees . The first reason is that they argue that existing access fees for IAP's are priced below cost, which puts LEC's at an unfair disadvantage. The second reason is that the advent of Internet telephony has called into question whether the IAP should really be classified as an Enhanced Service Provider. The main argument of the LEC's is that if IAP's continue to grow and Internet telephony becomes popular, then it could undermine access fee structure which subsidizes local service. LEC's argue that this could limit the advancement of universal service, so access charge reform is needed. Some arguments go so far as to say that IAP's should pay the same per-minute access fees as the long distance carriers.
One of the key points of the LEC's argument is that the holding time for a call to an IAP is much longer than the average holding time. This holding time is measured hourly using what is called "hundred call seconds" (CCS). The maximum holding time is 36 CCS (since there are only 3600 seconds in an hour), so as the average CCS approaches this number then call blocking becomes likely. While call blocking from a long distance carrier can be tolerated occasionally, the LEC's argue that call blocking on a local level is intolerable because it could block 911 calls. They argue that there are large costs involved with upgrading their network to avoid call blocking, and they should be compensated for those costs. They also argue that the flat rate pricing model of the Internet is economically inefficient because it encourages users to stay online for long periods of time and only exists because the access fees are lower than they should be. The following chart comes from  and is the result of a LEC study on the access rates of IAP's.
|Peak Hour CCS||Peak Hour|
|IAP's||26 CCS||10 PM|
|Central Office Avg||12 CCS||5 PM|
|Business Customers||3 CCS||4 PM|
Table 1. Peak line usage in hundred call seconds (CCS).
The counter argument to the argument presented by LEC's and long distance resellers comes from IAP's, software and hardware companies. One of the best overviews of these arguments can be found in , which is used extensively as a source for this section. The first aspect of the IAP's argument is that they are an Enhanced Service Provider, and should receive the exemption. They argue that they provide value added services, and although some people may chose to use the Internet for voice calls, there is no cost effective way they can distinguish data from voice. The IAP's argue that the success of the Internet and its rapid growth show that the Enhanced Service Provider exemption is an incredibly successful policy, and should not be changed because it is working. They argue that to add access fees at this point would stifle growth.
One key aspect of their argument is that the FCC should not try to apply outdated regulation to new technology. They argue that the current access fee model is outdated and inefficient. They argue that the LEC's claims about the cost of IAP's using their network are incredibly exaggerated, and are only a ploy for the LEC's to gain control of the rapidly growing IAP market. They show statistics that the average time users are online is not the high numbers that the LEC's imply even with flat-rate billing. They argue that the number of modems available in the IAP provide a self-regulating factor to avoid overloading the network. They argue that even America Online, with 8 million customers, can only have a maximum 400,000 users on at a time, which is not a significant load on the network. They show that IAP's only allow for about 1 in 20 users to be online, while the LEC's build for 1 in 7 users. This means that IAP's would overload their modems before the LEC's lines would be overloaded, so they argue that the IAP's could not overload the phone network. A summary of AOL's access statistics from  is as follows:
|AOL||Avg. Time per Day||Avg. Time Per Month|
|Before Flat-Rate||14 min/day||7 hr/month|
|After Flat-Rate||32 min/day||16 hr/month|
Table 2: Usage rates for AOL before and after instituting flat rate pricing.
Another aspect of the IAP argument involves ISDN (Integrated Services Digital Network). The IAP's argue that LEC's are using their monopoly power to overprice ISDN, which limits the growth of higher speed connections to the Internet. They argue that ISDN does not cost much more than a regular phone line, and they provide examples of phone companies that are offering ISDN at low rates. The IAP's suggest that the FCC create incentives for LEC's to provide ISDN at lower rates. Not all IAP's seem entirely opposed to changing access fees, but they all feel strongly that any access fees should be flat-rate, and not based on per minute usage.
To understand the current debate it is first useful to separate out the points on which one can find an emerging consensus. It can be taken as a given that the Universal Service Fund will exist and it must not be undermined. This fund currently is used to subsidize telecommunications service where costs would be prohibitively high. Another thing that can be considered as a given is that Internet access does increase the load on the local telephone network, which may require increased revenue to offset increased cost. The third point of consensus is that the growth of the Internet should not be slowed by regulation, although some companies may like this to happen for strategic reasons. The fourth point is that the LEC's could use access fees to give themselves an enormous competitive advantage in the IAP market.
As much as the LEC's would like to frame this debate as being "universal service vs. IAP's," that is not what the debate is about. The debate is about securing strategic advantage through regulation. In a capitalist society, it can only be expected that both sides would try to secure strategic advantage through regulation. The debate is not about leveling the playing field as both sides might argue, it is about tilting the playing field to the advantage of one side. Universal service can be upheld without threatening IAP's or slowing the growth of the Internet, as will be explained later in the paper, so it should not be a key point of debate.
The importance of this regulatory debate is that access fees have enormous strategic significance. If the access fees are set to the advantage of the LEC's, then others will have a difficult time competing with the LEC's in providing Internet access. If the access fees are set to the advantage of the IAP's or LD carriers, then the LEC's will have difficulty competing in providing Internet access.
There are several possible parts of an access fee to consider: a one-time flat fee, a monthly flat rate and a per minute usage fee. A one-time fee could be used to cover initial costs for an IAP to interconnect, such as running additional lines or purchasing equipment. The fee could also be distributed over time through a monthly flat-rate fee. A usage fee is charged on a per minute basis. Usage fees are different because they require that the user pay "per unit" of service (which is a minute), but the other two fees are flat rate.
LEC's need prices to be able to recover their costs, whether they use per-minute, usage-based fees or flat-rate fees. A LEC currently recovers most of its cost using monthly flat-rate fees and from the per minute fee for long distance. It is possible to instead recover its cost either through a one-time fee or per-minute usage fees. Economics teaches us that the economically efficient thing to do is to price the per-minute, usage-based fees at what is called the long run, average incremental cost (AIC). The long run, average incremental cost is the cost to add another minute of service capacity to the network averaged over a long period of time. The AIC is extremely important to regulators because it provides information on how prices should be set for maximum economic efficiency.
The key strategic aspect of the access fee is how the per-minute usage-based price compares to the AIC. Because LEC's provide their own access circuits, they do not have to pay a "price" for the circuits, but instead pay the cost of the AIC to build the circuits. If the price is set above the AIC then the LEC will have a competitive advantage in providing Internet access because the IAP's would have to pay much more for the same access lines. If the price is set below the AIC then the IAP could have the strategic advantage depending on how large this difference is, as well as the flat-rate fee and the amount of usage (in minutes). In either case, the side with the strategic advantage on the per-minute price would want to be sure that Internet users would stay connected as much as possible because the more they are connected the greater the strategic advantage. The flat rate fees also play a strategic role, but not nearly as significant. In general, if the cost for access of one group is much lower than for the other, then they will have the strategic advantage. The equation for cost of access is as follows:
IAP: CIAP=Pflat+Pusage*Q where Q = Number of minutes
Currently the per-minute usage fee (Pusage) for IAP's is zero, and the AIC is not zero, although it may be fairly close to zero. This means that the IAP's could gain an advantage as compared to the LEC's from increasing their usage. However, IAP's do incur a cost for having more of their lines tied up, and it is unclear whether increasing usage would provide much of an advantage if any at all. If LEC's were to charge a per minute fee, then they could have a significant advantage by encouraging more time online. With current access fees of 2.5 cents per minute, this would add $1.50 per hour to the cost of IAP's, which would essentially give the LEC a monopoly on Internet access.
A rough estimate of the competitive advantage can be determined for two possibilities for access fees: no fees and fees set to 2.5 cents/minute. Assuming 20 users per line and a monthly rate of $40 per line, the monthly rate Pflat would be $3 per user. Using AOL's average of 16 hours/month and assuming an AIC of .5 cents/min, the cost for each option is as follows:
The LEC's might like to argue that they have to increase the per-minute fee to recover revenue, but it is possible for them to recover their costs using a flat rate fee. If a significant per minute fee is used, then the LEC's will have a strong competitive advantage. What is presented here is a "back of the envelope" calculation, and the actual numbers may vary by a few dollars. This means that it is possible that the LEC's should receive some compensation to level the playing field; however, this compensation does not have to be very large if it is even needed at all and it can be done on a flat rate basis. Figure 1 graphically illustrates how the different options for access fees would affect the "playing field" of monthly cost per user:
Figure 1. A graphical illustration showing proportionately how different proposals would affect competitive advantage.
The Internet Telephony Consortium at MIT has been modeling the cost of Internet access, particularly to investigate the cost of Internet telephony and the effect of access fees. The first model focuses on costs seen by the IAP and was developed by the author and students of the TPP91 Class at MIT. This section is not meant to reproduce that model, but to examine the results from it, although a detailed explanation of the first iteration of the model is available in .
The model breaks up the cost into four components: local, IPOP equipment, upstream and other operations costs. The local cost component is the price charged by the LEC for both incoming and outgoing leased lines. The IPOP equipment consists of the capital costs only of the equipment used to provide Internet access. The upstream cost consists of the price paid (or internal cost) to an upstream provider for a high bandwidth connection to the Internet. Other operations cost covers all other costs incurred by the IAP.
The results of the model showed that the cost to the IAP for providing Internet access is approximately 1.8 cents per minute, which is broken down as shown in Figure 2. It is important to note that cost does not scale linearly with the number of minutes, but in fact scales at a much lower rate. In this sense, using a per-minute cost is misleading, which will be explained in the paper where the AIC of connection time is examined.
Figure 2. Cost distribution of Internet Access from an IAP's perspective.
Figure 3 shows how the cost percentages for IAP's change if an access fee of 2.5 cents/min is added. The main point of this diagram is that if an access fee were charged then it would dominate all other costs. Because the LEC's would only pay this fee internally, this would given them an extreme competitive advantage.
Figure 3. Cost distribution seen by IAP paying an access fee of 2.5 cents/min
To better understand the real costs of Internet access it is necessary to examine the AIC for Internet access as seen by the IAP. The purpose of this section is to provide a better understanding of the basic economics of Internet access, and explain some of the incentives that might result from these economics that might affect market or regulatory strategy. For the IAP, there are two long run average incremental costs to consider: cost of adding connection time (AICconnect) and the cost of adding bandwidth used (AICBW). In considering AICconnect, it will be assumed that the IAP is purchasing incoming lines at T1 rates and that equipment is being depreciated over three years, and the only costs that will change are the cost of the local leased lines and the cost of the equipment.
AICconnect monthly = Plocal+Cequip = 40 + 1000/36 = $67.80 per month per line or $3.39 per user per month
Using AOL's current usage figures, the results per user per minute for connection cost are as follows:
AICconnect minute = 67.80/20/(16*60) = $.0035 per user per min
Using T3 rates given by BBN in 1996, a rough estimate for AICBW is as follows.
AICBW = $1.50 per kbps increase of average bandwidth for a month
Another way of looking at this number is try to estimate the average cost per bit based on BBN rates for Internet capacity. This number is only an average, and although it might be useful to estimate the order of magnitude of the average cost per bit on the Internet, it could be misleading if used as a standard cost for all bits (see explaination below). Since this is the average over the month, the cost per bit can be found by dividing by the time in a month giving:
AICbit = $3.5*10-8 per bit or 3.5 cents/Mb
Assuming AOL's access rates and assuming the average usage is 2 kbps, then average cost per user for bit transmission is:
AICBW user= $4.03
The following table summarizes these results. The third row shows a first order approximation of the costs seen by the IAP if an access fee of 2.5 cents per minute were charged.
|per month per line||per month per user||per minute per user||per Mb|
Table 3. Long run average incremental cost seen by IAP.
There are several points to consider fully to understand what these numbers mean. The first is that AICconnect+fee is a first order approximation, and would be lower because usage would decrease as the price increased. To calculate this value more accurately would require knowing the elasticity of demand for Internet access, which is not easily available. Another point is that the value for AICBW given here applies only to an ISP that purchases "upstream" capacity. Because of the rapid changes in Internet pricing, these values have probably decreased significantly in recent months. It can also be assumed that IAP's providing their own upstream capacity would have a much lower AICBW, and this value will be much lower still for carriers with their own facilities. It is for this reason that many predict that IAP's will need their own facilities to be competitive in the long run.
AICbit is about as close as you can come to the mythical "cost per bit" on the Internet. This number represents an average and only applies to IAP's purchasing upstream capacity. One important point is that this number does not apply everywhere on the network. In particular, transcontinental links are likely to have a much higher AICbit, and LAN's will have an AICbit of zero. This difference has been termed by Mike O'Dell of UUNet at a meeting of the Internet Telephony Consortium in the following way: transcontinental bits are "gold" bits, bits over long distance lines are "silver" bits, while bits over local lease lines are "bronze" bits and LAN bits are free. Another point is that the cost of the leased lines that the Internet runs on is included in the AICbit and AICbw used here; however in calculating this number, it was assumed that the IAP would be connecting to a local T3, so the cost of the lease line for interconnection would be insignificant. A final point to note is that the real cost of bandwidth is for peak capacity, and during times without congestion, the economic cost of bandwidth is zero .
Other clarifications also need to be made for AICconnect. For an organization that has a LAN, AICconnect is essentially zero, while the AICBW should be about the same as for an IAP. For connection time, the main cost seen by the IAP is to provide the number of lines to support their peak hour. This means that in economic terms, the cost of connection time is zero during non-peak times when lines are free.
It is also important to consider how these costs might change in the future. One part of the cost (laying down wires or fiber or wire) is fairly fixed and is not likely to change much over time. Wireless technologies may help alleviate this problem, but it is not clear that they will be significantly more cost effective for high bandwidth connections. Costs which involve technology (particularly processing speed) are likely to decrease significantly according to Moore's law, which says that the cost/computing ratio decreases by a factor of two every two years. Because there is currently an excess of fiber deployed, and the capacity of fiber is largely determined by the equipment at the edges, it is likely that the AICbit will continue to decrease rather significantly. Because much of the cost in the local loop is fixed, AICconnect is likely to remain relatively flat with a slight decrease as equipment becomes lower in cost.
Which of these two costs is dominant (Cbw or Cconnect) has large strategic implications. If AICconnect is increased using an inflated per minute access fee, then the LEC will be given the strategic advantage. If it the price is set below cost then the IAP will have a strategic advantage. There is a second key point of significance that may seem less obvious: the higher bandwidth available to the user, the greater Cbw becomes. If AICbw becomes a dominant factor, then those who can provide raw bandwidth at the lowest cost (LD carriers) will have a strategic advantage. This last point should not be considered as "tilting the playing field," but as letting those who are the most competitive get the profits they earn. The following table illustrates how increasing the bandwidth on the local loop will affect cost of bandwidth:
|Avg. BW||28.8||ISDN 128k||T1: 1.54M||T3: 45M|
Table 4. Effect of increasing bandwidth on increasing the "bandwidth" component
The exact values in this chart are only an estimation, but the order of magnitude is what is important. The cost of installing a local loop is relatively bandwidth independent. The LEC will have an incentive to keep its percentage of total cost of communications as being the dominant portion giving it greater control over the market. If the LEC provides a high bandwidth local connection, then other players (LD carriers and equipment providers) will be able to capture most of the value. Until the LEC's are able to build their own LD networks, they do not have an incentive to increase the bandwidth in the local loop such as promoting ISDN.
One of the issues that both sides are arguing about is that of economic inefficiency. The questions arise of what the economic inefficiency in the current system is and what would the economic inefficiency be if IAP's paid normal access fees. There are two levels of efficiency: efficiency in pricing the access lines and efficiency in the IAP's pricing.
For pricing of IAP's, what constitutes economic efficiency is a complex issue. Often IAP's are criticized for inefficiency in their flat rate pricing because the optimal pricing would have a per minute fee for connection time equal to AICconnect. It is possible that flat rate fees help the IAP's achieve economies of scale that may not be possible otherwise so they can upgrade from T1 to T3 and OC3 rates which are more cost effective. Another point to consider is that the cost of billing on a per minute basis can become a significant cost component to IAP's, and is likely to be even more significant to users (which is why users prefer flat rate billing). Taking billing costs into consideration, it is likely that flat rate billing for Internet connectivivity is the economically efficient thing to do based on the current techology of the Internet.
Economic theory also supports this view, and Gong and Srinagesh predicted in  that because Internet access is largely a commodity service, Bertrand price competition is likely to emerge. Bertrand price competition is when companies compete on price, and the economic equilibrium is that everyone's price should equal marginal cost, which should stabilize the same value. This theory could explain why the pricing for Internet access in the United States has leveled off at $19.95 per month for unlimited access. If Gong's theory is correct, then this should be the marginal cost of providing a month of Internet access to one more user. Whether this theory is correct is not the point; the point is that it is widely agreed that the market for IAP's is fairly competitive. If flat rate billing is the economic equilibrium that results from market forces in a competitive market, then economic theory tells us that this is what is economically efficient. In summary, all this talk about flat rate billing currently being inefficient is not only uncertain, but economic theory can explain why a flat rate is the most efficient pricing method.
This is not meant to dispute the many people who argue for usage-based pricing [15, 16] because actually there are two seperate debates. The first debate is whether flat rate billing is currently economically efficient. This is the debate that I refer to when I promote flat rate billing. This is also the debate that has implications for local access fees. The second debate is whether flat rate billing will still be efficient as higher bandwidth connections and bandwidth reservations become available. On this debate, it is clear that some form of usage-based pricing will be necessary, although much of its complexity may be hidden from the user . The author of this paper is a strong advocate of usage-based pricing in the second case involving new technologies, while feeling that there are strong economic reasons opposing usage-based pricing in the first case involving current IAP pricing. LEC's may try to use the results of this second debate involving new technology to argue that the current flat-rate pricing of IAP's is inefficient. This second debate has nothing to do with local access fees, but involves changes in techology not directly related to the access mechanism.
The pricing of access lines to the telephone network can be considered separate from the pricing of Internet access. For access lines, the economically-efficient, usage-based price should be set to average incremental cost of access to the phone network (AICaccess). This number is one that has been well studied  as well as other studies, although there is much dispute over what its real value is. Without knowing this number, it is impossible to calculate the economic inefficiency of various schemes for access fees.
To summarize, it seems likely that both sides in this argument have valid points: the LEC's may need increase revenue to cover increased costs of the IAP's, but the access fee model for LD carriers is not the right mechanism to do this and would be disastrous to competition if implemented. If a per-minute fee were instituted, then it would need to be very low, and it is likely that billing cost for the per-minute fee would make it lose most of its increased efficiency.
One possible solution is to consider is what happened between Bell Canada and Canadian IAP's . Bell Canada found a regulatory loophole which allowed them to triple the price they charged IAP's for their access fees. When Bell Canada tripled their price they experienced what could be termed as a "revolt" from IAP's, and were pressured into creating a new tariff for IAP's. The new tariff is called their "ISP Link" service, and offers a stripped-down access line for $22/month. This new tariff appears to be a compromise on both sides and has alleviated the current problem; however Canadian IAP's still face Telecom Public Notice 96-19, which could raise access fees again. This example shows two important points. The first is that a compromise can be reached which meets the needs of both parties. The second point is that raising access fees dramatically is likely to create a public uproar that neither a phone company or a regulatory body would like to face.
A second solution has been proposed by Jerry Hausman, a prominent economist from MIT . His solution is to replace the inflated access fee with a flat-rate subscriber line charge (SLC). He estimates that the SLC would need to be increased by only $3.50 per month. His proposal to maintain universal service would be to provide targeted subsidies toward those with low income who may not be able to pay the increased fee. His estimate was that this could save over $1 billion through increased economic efficiency. With the current size of the Internet market, this savings could be even greater than when this calculation was made.
A third solution is to use Hausman's proposal but apply it only to IAP's rather than all users. This would mean that IAP's would pay an increased flat rate monthly fee to cover the increased cost they incur on the phone network. Exactly what this flat fee would be set to is a question that is beyond the scope of this paper. Another suggestion is to have the IAP's pay a flat rate to cover cost, but have that flat rate adjustable from time to time, if costs change either through different usage patterns or technological changes. This allows the IAP to face their real AIC, and allows them to set their prices accordingly.
But it is important to note that IAP's already pay non-residential rates, which on average are profitable for the LEC's. It may be that the IAP's overuse these lines so that they are well above the average cost making them unprofitable. The point of this argument is that it is unclear as to what the real costs are for the IAP, and it may be possible that they are already paying more than the costs. In considering various solutions, the objective should be to consider how to compensate the LEC's for any increased costs without tilting the playing field. The LEC's will try to argue that the only solution is to impose a per-minute access fee, but that can give the LEC's enormous strategic advantage and stifle competition.
In summary, the debate over access fees should be considered from the perspective that different groups are trying to secure competitive advantage through regulation. It is possible that with the current access fees IAP's and LD carriers currently have a slight competitive advantage over the LEC's. The size of this competitive advantage is likely to be small and could be eliminated by charging slightly higher flat-rate monthly access fees to IAP's. This could be done in a way that makes IAP's responsible for their own increased use to avoid economic inefficiency. Imposing usage-based, per-minute fees has the potential to give the LEC's a very large competitive advantage while causing the US economy to lose billions of dollars due to inefficiency.
Special thanks go to the members of the TPP91 class at MIT, Brett Leida, Donya Krairit, and James Wahl, for their work with me in constructing the cost model for Internet access and Internet telephony. Thanks also go to Joe Bailey, Lee McKnight, and David Clark for their talks which helped me to establish a framework for the economics of the Internet.