EPUBs are an experimental feature, and may not work in all readers.

Paper presented at AAUP/ARL Symposium on Electronic Publishing, November, 1994

In thinking about the pricing of electronic publications we are facing very difficult questions in a context of remarkable unclarity. Even the product to be sold is not easily defined.

In this paper, I will consider the role that prices should play, then focus on the central reason why pricing electronic publications is difficult, look at some of the general strategies that are available in this kind of situation, and then review three examples of electronic publication pricing that demonstrate desirable features. I will also pay some attention to the role that libraries play in the solution of this pricing problem.

The first question then is what do we want the price of an electronic publication to do. Frequently the issue of pricing is discussed in the context of cost recovery. This is, in some ways, not a bad starting point as it certainly defines a minimum requirement — without cost recovery from some source or another there are not going to be any electronic publications.

But should we be content with prices that recover costs? Indeed is that even a meaningful standard?

It does assume that costs are well-defined and easily agreed upon. It implies that some auditor could examine a publishing operation and easily determine exactly what each publication costs. The awkward fact is that there is no precise answer to that question. Every publisher, every information supplier, produces many products. At the same time many of the expenses of those organizations are not directly related to any one product. Of an average publisher's costs something like 60% fall into this category. Those costs can only be ascribed to individual products according to some conventional but basically arbitrary rule.

Even if we somehow surmount that obstacle and all agree on a suitable allocation of costs to individual electronic products, we may still have disagreements about which costs are to be included in the total to be recovered.

In particular we need to think about the cost of capital. Running a business whether profit or non-profit requires capital. This is because products are manufactured at one time, delivered at another and paid for at yet another. Suppliers and staff have to be paid before customers pay their bills. That requires that some one lend money to bridge that financing gap. Publishers also buy equipment, much more now than they used to, and probably even more in the future, that needs to be paid for on purchase, but it lasts many years giving up its value and repaying the initial outlay over its life time. The publisher needs to borrow money to finance that investment.

Is the interest that publishers pay to the lenders a cost? Something that the moral auditors will allow to be recovered from our customers? Does a cost recovery price include that cost?

I am not trying to make a big point about the cost of capital, but I am trying to point out by that example that the principle that prices should be based on cost recovery is not simple, unambiguous or necessarily uncontroversial.

Even with resolution of the definitional problems, is cost recovery an appropriate basis for setting prices? Although I work in a non-profit and very much believe in the non-profit approach, I worry about the lack of incentives that cost reconvery implies, incentives to innovate, to grow and to improve products. Our society primarily allocates resources according to profitability. By excluding profits from the publishing of academic work, are we not risking slower innovation and slower growth and a persistent of shortage of resources? After all, would we have fast processors, and all the other technology which makes these discussions worth having, if Intel, for example, were operating on a cost recovery only basis. I am afraid the answer is very obvious. The organizations who make profits attract extra resources, money to invest and grow.

Just as the price plays a significant role in generating investment and thus supply, so more obviously it plays a role in determining demand. Of course a high price deters some buyers and a low price attracts more buyers. Is then a low price always better? There is a limit.

With each cut in the price, someone extra decides to buy the good. Who is the new buyer? It is someone who did not buy at the previous higher price because they did not think the good worth that price to them. Now their valuation of the good is higher than the new price and therefore the good is worth buying.

So as we reduce the price we bring in buyers who place a lower and lower value on the good. How far should we go in making the good attractive to them? It makes no sense to make it available to people who place a value on it which is less than the value of the labor and materials used in making the good. We could use the labor and materials in better ways making other products. In other words the price should not be reduced to below the cost of producing the good.

It sounds as though we are back to pricing for cost recovery. Not quite. Because when we think more precisely about the cost that sets a floor on the price, we see that it is not the average cost of making the good, but the cost of making the last example of the good. It rarely costs the same to make every example of a good, the cost sometimes increases as we make more and frequently decreases as we make more. When we are wondering whether to cut the price one more step and attract some more buyers, the question is whether those extra buyers will put a value on the good higher than the costs of making the examples of the good for them. I will refer to this cost as the incremental cost.

So we should lower price and expand output until the price is just equal to the incremental cost of making the last example.

This is a principle that we should strive towards as we think about the pricing of information goods. It is also a principle that immediately leads us into another conundrum. The nature of the costs of providing information goods means that such a price based on incremental costs will not provide the manufacturer with enough income to cover his total costs. By now this is a familiar result to most people so I will just sketch the argument.

Simply, there are substantial costs in making the first copy of a publication and very low costs in making subsequent copies. A price that is close to the incremental cost of making another copy will not make any contribution to the first copy costs that have been expended. We are therefore faced with a very interesting problem: how to combine two conflicting objectives. The first is to make the publication available at a price close to incremental cost and thus provide an incentive for everyone to buy it who places a value on it that is greater than the value of the resources used in making it. And the second is to ensure enough income to the manufacturer to give him or her the incentive to incur the first copy costs and make the publication possible.

While this has always been a challenge in the paper world, the challenge is dramatically enhanced in the digital world. First copy costs can be much higher as we embed audio and visual materials and build elaborate links within the document. While the cost of additional copies is lower or even zero given the ease of electronic copying.

The general form of this problem is a common one. A classic case much studied by economists is electricity supply where the power stations and distribution network involve an enormous expense but once the customer has received the first unit of electricity subsequent ones can be supplied at very low cost. A conventional approach to this problem — known as a two-part tariff — is to charge the customer a high price for the first few units and a decreasing charge per unit as he or she buys more. Decisions about discretionary uses are thus made on the basis of a price close to the low incremental production cost. For this solution to work, each consumer must buy multiple units of the product and each consumer must be identifiable to the supplier and maintain a relationship with the supplier.

Thinking now about those two characteristics in the context of publishing, it would seem that the electronic book is not going to lend itself to two-part tariff pricing as customers usually buy only one copy of each book and there is no continuing relationship between publisher and customer. Likewise for the paper journal, the subscriber buys one subscription. But the journal in electronic form does perhaps permit such an approach. The subscriber would pay an access fee at the beginning of the year to cover the publisher's first copy costs and then a much lower charge for every article selected and delivered.

The approach I have just suggested is an example of what is also called country club pricing. You pay a substantial fee for membership, but then pay the incremental cost for each use of any of the facilities.

The other standard strategy to solve the general economic problem is known as price discrimination. You will recall in my explanation of the basic problem that I presumed some customers valued the good more highly than others. If the high valuation customers could be segmented from the low valuation customers, a higher price could be charged to the former than to the latter. The lower price would be close to incremental cost while the high price would permit first copy costs to be recovered.

An example of this in publishing is the delayed paperback. First, the high priced hardcover is published for those with a strong and urgent need for the book. Then when the high valuation customers have bought and the first copy costs have been recovered, the paperback is brought out with a price that is close to the incremental production cost.

While I am aware of no electronic publications overtly using timing to separate high valuation from lower valuation customers, I do notice that some of the magazines that claim to be also available on the Internet in fact seem to delay making the latter version available. Thus those who pay a non-zero price get the news and information earlier than those who seek it at zero price.

The essential element of such pricing schemes is a method for segmenting the customers into valuation groups — there could be more than two. Selling the same product at more than one price is not going to work unless there is a way of determining which customers get to pay which price. This calls for creative thinking. By the way under the Robinson-Patman Act there does have to be some cost justification for such price differentials and any especially ingenious scheme would need to be considered in the light of that Act.

Another way to segment customers is to bundle the good with another good or service. If the added item is generally wanted by the high valuation customers but not by the low valuation customers, then the combined package can be priced so as to recover not only the cost of the bundled item but also some of the first copy costs of the basic item. A possibility here for a software manufacturer would be to sell the complete package with sophisticated help menus for one high price and the bare software for a lower price. The business user for whom time is money will want the help, the individual user may be more willing to seek a solution by trial and error. Or one could envision selling a data resource of some kind and bundling with it access to an online service providing frequent updates. A document with hypertextual add-ons might be sold for one price with access to the add-ons and another lower price without. Bundling is definitely a route to explore in endeavoring to solve the problem of recovering first copy costs.

So far I have carefully eschewed discussion of the library in all this. We could think of the library as just another customer for a publication. Because they serve many users, they will in general be customers who place a higher valuation on a publication. So libraries are a distinguishable category of high valuation customer. Thus we see the logic for, and also the practicability of, the common practice of charging libraries a subscription price for a journal that contributes to first copy costs while the individual subscription rate is closer to incremental cost. We can also consider libraries as customers who put a premium on the durability of the objects they buy. Thus we see that bundling a durable binding with the book is another way of charging that category of higher valuation user a price that contributes to first copy cost recovery.

But libraries can be thought of in quite another way, not as straighforward customers but as intermediaries in the transaction between originator and final reader. To recover first copy costs, the publisher has in many circumstances to price far above incremental costs. The various devices I have been discussing will not always be applicable. So there will be cases where many potential users will find the price of the information good above their valuation of it. They will be deprived of its benefit even though their valuation is higher than the incremental cost of making the good for them. The library steps in. It pays the publisher's full cost recovery price and then makes the book or journal available at zero price. Thus solving, indeed over-solving, the pricing problem.

Although there is more to say about libraries on this topic. I want now to look at some examples of actual pricing schemes for electronic publications and then return to libraries and related topics at the end.

The first scheme I want to present is the Johns Hopkins University Press approach to pricing journals included in Project Muse. The admirable feature of their work is the clear analysis of costs. They have carefully distinguished between those costs that are common to both print and electronic forms, those specific to the paper form and those specific to the electronic form. This is an essential step on the way to any well-thought out pricing scheme for a product being published in more than one form. In the Project Muse case, taking the paper subscription as $100, it is estimated that $60 of that is for items that are medium independent and $40 are for costs specific to the paper edition. By contrast for the electronic edition, the specific costs are down by about one quarter to $30, so the electronic edition price is $90. Where a customer wishes to have both paper and electronic editions, the combined price includes the base price once, and the specific costs of each edition. The overall figure is $130. While this distinguishes between the common costs and the medium-specific costs in an attractively clear way, it does not fully distinguish between first copy and incremental costs and some first copy costs are in the medium-specific figures.

Although not presented as an explicit two-part tariff structure, the pricing for MIT Press's Chicago Journal of Theoretical Computer Science actually does fit this scheme. The journal is distributed electronically and so most of the costs are first copy costs. The basic price is an institutional one of $125 explicitly designed to cover the first copy costs. Within a subscribing institution, access and printing out individual articles is permitted without further charge. So for the individual user the price is equal to the incremental cost of zero for another electronic copy. There is also an individual subscription rate of $30. As any customer involves some overhead and transaction costs, a non-zero price is appropriate.

The third example is the new pricing structure for Mathematical Reviews. This is to my mind the best example as it makes very clear to the customer the difference between the first copy costs and the incremental costs. The subscribing institution pays a $3595 data access fee which pays for 'building and maintaining the database from which all Mathematical Reviews products are derived'. I quote from their publicity. Then any subscriber pays a product delivery fee which is the incremental cost for the particular product e.g. an annual $315 for the paper form or $520 for the CD version. This is by the way a huge journal. What I think distinguishes the Mathematical Reviews scheme is the very clear explanation of the underlying logic — the distinction between first copy and incremental costs, and the distinction between incremental costs for different delivery modes. I incline to believing that as we launch electronic products such clarity in presenting the logic of what we are doing is very important.

A merit common to all three cases that I have described is simplicity. As we struggle with the pricing problem it will be tempting, at least to people like me, to construct complex structures. That temptation has to be resisted both to ensure the prices are just workable for seller and buyer, but also so that the customer can understand the logic and see that the schemes have integrity.

That principle of simplicity also bears on a much more fundamental problem. What is the product? I have carefully and deliberately skirted that awkward issue but it cannot be entirely avoided. Are we selling a file that is transferred to the customer, or are we selling access to that file? Is access charged by time or by the access? The more I try to resolve this type of question the more difficult it all seems to become. Despite it meeting none of the standards I have established in this paper, I find myself thinking that Ted Nelson's idea of a price label on every digital packet and a charge for every movement across the network may be the only practical solution.

But I want to get back to the question of the library and its role as the agent who makes the publication available at zero price. This is possible because the library is subsidized rather than being expected to recover its costs from its users. In general, one possible route to a solution of the first copy cost problem is through subsidies. This is being argued vigorously for academic works by Stevan Harnad in his 'subversive proposal'. For most of us who deal with university budgets regularly, the assumption that we should plan on extra subsidies seems rather hopeful. Of course if the actual publications' prices decline to or close to zero then perhaps the requisite subsidies could be liberated from library budgets.

Apart from all the obvious naivety of such a remark, there is actually a fundamental problem. Universities and communities support their libraries because the benefits are concentrated locally. But subsidies to a publisher bring benefits which are difused widely at least nationally and often globally. So why, we can hear a Provost asking, should her university spend substantial amounts of scarce resources to make publications available at zero or low price throughout the world? This is not, I think a minor debating point, but a very serious problem for a rational reappraisal of the system and the funding of a new system for disseminating scholarly work.

The answer to the provost's doubts might well be based on the local benefits of providing for the publication needs of the local faculty. Then the funds that drive scholarly publishing are allocated according to the needs of the author. Is this desirable? I think not and that brings me back full circle to the incentive effect of the pricing system.

As we consider pricing in the electronic environment, I believe that we need to maintain some dependence by the producer on demand from the final user. At present in much of scholarly publishing, and in university presses in particular, we have a a very healthy balance between the incentives of the market place — choosing the publications that will sell, and the incentives of the academy — choosing publications of intrinsic importance. A complete dependence on subsidies eliminates the influence of demand. And, if those subsidies are directed to support the publication of local authors, we will also see a rapid slackening of the standards of scholarly importance.

It all comes down to incentives. As we think about pricing particular products or as we think about the general shape of of pricing structures for electronic producs, the crucial issue is one of incentives. We need incentives that on the supply side encourage innovation and investment, and that encourage a balance of emphasis between intrinsic academic quality and the needs and interests of the readers. And on the demand side we need to create pricing systems that make the material being disseminated available to all who would find it valuable. While I think the examples I have cited are good and set patterns that could well be followed, I see considerable opportunities for the ingenious to devise new systems that work more effectively to meet both the demand and supplyside objectives.


Author Information

Colin Day is the Director of the University of Michigan Press.


Notes

*I owe this insight to Scott Bennett.