Dearing's figures for the expansion of networked desktop computers seem affordable but Lawrie Schonfelder finds neglected costs that will stretch budgets
The Dearing report stresses the importance of communications and information technology in the development of higher education, but is remarkably deficient in concrete estimates of the cost of its recommendations. Nevertheless, it does give some fairly precise figures in one very important area, where it recommends additional expenditure to provide enhanced computing facilities for students.
Unfortunately, the cost estimates are seriously flawed because they fail to include some very real and high cost elements that are essential if continuing effective use is to be made of the recommended expanded provision.
The report estimates that higher education institutions overall currently provide networked desktop computers (NDC) for student use at a rate of approximately one NDC for every 15 students. This is rightly recognised as far too small a provision if students are to be able to make use of IT as an essential tool in the learning process. The report recommends that this access ratio be improved to 1:5 over a four-year period by the installation of an additional 130,000 machines.
The report recommends an NDC should be a well-configured multimedia PC with good network connection to suitable server facilities. It estimates a unit cost for fully installed systems of Pounds 2,800 and suggests that additional funds of Pounds 100 million are needed for each of the first two years, Pounds 90 million for the next two years and Pounds 60 million a year thereafter.
A superficial look at these estimates would suggest that they are not unreasonable. A new learning resource centre containing 280 networked PC systems has recently been installed at the University of Liverpool so some actual costs can be used to compare with Dearing's estimates.
The following table shows the actual costs:
Multimedia PC client 1,800
Furniture (desk & chair) 130
Local networking 200
Additional servers, printers, etc. 170
Additional software 100
Some of these are marginal costs in that these 280 new machines were being added to an existing backbone network already supporting some 600 NDCs. No contribution to the cost of this backbone is included. Also there is a smaller unit cost from the servers and software items because of the already well-established infrastructure. The Dearing unit cost estimates are therefore reasonably adequate. However, they fail to take into account the cost of the accommodation.
Each NDC requires approximately three square metres of environmentally controlled office quality space. To provide this in new buildings of, say, up to two floors and in rooms of 40 seats each costs approximately Pounds 1,000 per square metre of usable space. Given that there is very little unused space in higher education institutions, new space will be required to house these additional facilities.
Dearing therefore has underestimated the actual additional cost of providing adequate student access by a factor of two.
Another serious omission from the report is any realistic estimate of the recurrent cost implications of an expansion on this scale; expanding the number of NDCs from a ratio of 1:15 to 1:5 is an increase in the number of installed client machines by a factor of three. An expansion on this scale must have significant recurrent cost implications.
The first and most obvious recurrent cost is that of physical maintenance. Computers and networks do break down now and then. When they do they have to be repaired. The cost of this for computing equipment is typically about 8 per cent of purchase price. Leaving out the furniture and installation costs from the above unit costs provides a unit maintenance cost of Pounds 182 annually, which implies a cost of to the sector of Pounds 24 million (for 130,000 PCs).
In addition, there is a cost for heating, lighting and servicing the new building space. This is estimated to be about Pounds 30 per square metre, implying an additional cost to the sector just short of Pounds 12 million.
The next cost is one that is frequently omitted because it is not usually recognised as a recurrent cost except by accountants. This is the cost of update and replacement.
Computing equipment of this type has a very finite lifetime as an effective usable device. Five years is almost certainly the maximum that can be expected for the type of equipment and the function it is required to perform. If we assume that the buildings, furniture and local networking (cables, and so on) have much longer lifetimes this still leaves a unit cost of replacement of Pounds 2,070. If a linear amortisation model is used this gives an annual unit cost of Pounds 414, which translates into an annual cost to the sector of Pounds 54 million. The final missing cost element is probably the most important but is also the most difficult to quantify. This is the additional support service cost.
A trebling of the number of supported client machines will not mean three times the number of support staff in higher education institutions computing service departments. However, expansion on this scale cannot be achieved without a significant increase in the size of such departments. Software has to be installed and updated, students have to be trained in how to use the facilities, they have to be helped employ the facilities effectively and rescued when they make mistakes, and someone has to manage the logistics of running client/server installations of the very considerable sizes envisaged. As a tentative guesstimate of the extra support costs, it could be argued that one additional member of support staff could be needed for every 200 NDCs. At a ratio of 1:5 this would work out at one support person for every 1,000 student users.
The average gross cost of a member of support staff is also difficult to estimate as such staff generally perform other duties besides supporting student computers but a figure of, say, Pounds 25,000 covering salary plus overheads is probably not too far wrong. This gives an additional overall additional support cost of Pounds 16 million.
Adding all these recurrent costs together gives an additional annual cost of Pounds 106 million. Assuming that the Dearing estimate of the ongoing Pounds 60 million was supposed to cover the recurrent cost element, this too has been seriously underestimated.
In both cases, capital and recurrent, additional costs have been underestimated by close on a factor of two.
Lawrie Schonfelder is director of the computing services department, University of Liverpool.