IEEE TRANSACTIONS ON SOFTWARE ENGINEERING, VOL. SE-13, NO. 9, SEPTEMBER 1987 Software and Hardware in Data Processing Budgets VIJAY GURBAXANI, MEMBER, IEEE, AND HAIM MENDELSON Abstract-This paper develops a microeconomic framework for the determination of data-processing budgets over time, and, in particu- lar, the allocation of these budgets between software and hardware. The model dynamically balances the value and cost of information ser- vices, given the prevailing cost trends. It regards software and hard- ware as inputs to the process of producing information services, and identifies the complementarity and substitution between them as major determinants of the efficient budget allocation. The theory provides a basis for understanding the budgeting process and for predicting future trends, and is applied to actual budget data. Index Terms-Cobb-Douglas production function, cost trends, data processing budgets, programmer productivity, software engineering economics, substitution and complementarity. I. INTRODUCTION THE prevailing trends in data processing budgets are the subject of an ongoing debate among researchers, analysts, and managers [4]-[7], [12], [16]-[18]. There is a widespread belief that software represents a growing proportion of data processing expenditures. This assertion is often depicted by an S-shaped curve demonstrating a significant increase in the software and software-mainte- nance portions of the budget over time and a correspond- ing decline in the hardward portion (see [18] for the his- tory of this curve). This view was challenged by Frank [16]-[188], who contends that the actual data show a sur- prising constancy of the budget shares allocated to soft- ware and hardware over time, based on the results of an- nual user budget surveys [22]. The empirical evidence in these surveys shows relatively constant proportions of software and hardware expenditures in user budgets, and seems difficult to explain in light of the prevailing cost trends. The behavior of the budget-proportion curve is closely related to the trends governing the costs of software-de- velopment and hardware over time, which are notably dif- ferent. The fast technological improvements in semicon- ductor technology brought about a sharp decline in hardware costs [27]. In contrast, software-development is still labor-intensive, and while there has been some prog- ress in programming productivity, it has been relatively slow. This disparity in the software and hardware cost trends intuitively suggests the classical S-curve relation- ship: as hardware gets cheaper without a corresponding Manuscript received January 2, 1985; revised November 3, 1986. This work was supported in part by the IBM program of support for education in the Management of Information Systems. V. Gurbaxani is with the Graduate School of Management, University of California, Irvine, CA 92717. H. Mendelson is with the William E. Simon School of Business Admin- istration, University of Rochester, Rochester, NY 14627. IEEE Log Number 8716547. decline in software-development costs, software expenses are expected to consume an increasing percentage of the data processing budget. This scenario seems plausible, and hence the evidence of constant budget shares is puz- zling. Frank himself seems surprised at the "remarkable steady state of staff and hardware costs as a percentage of total costs, even as the total expenditure increases from year to year" [18]. An important ingredient in the evolution of this debate is the lack of a theory explaining the determination and allocation of data processing budgets. The main objec- tives of this paper are to offer a theoretical framework which underlies the budget allocation process and to study its implications. We develop a model that explains the dynamics of data processing budgets, examining both the determination of the overall budget and its allocation be- tween software and hardware. We consider a net-value maximizing firm which balances the value of information services against their costs (cf. [29]). 1 To capture the dy- namics implied by the cost trends, we study the behavior of a firm operating over a multiperiod horizon, treating software and hardware as inputs in the production of in- formation services. The model identifies the importance of software-hard- ware substitution in determining the scale and allocation of data processing budgets. While hardware and software are complementary inputs (i.e., an investment in one re- quires an appropriate investment in the other to achieve the desired benefits), they are also substitutable in that varying ratios of the two can be used to produce any given output. Our results suggest that the shape of the budget proportion curve is determined by the relative importance of software-hardware complementarity versus the substi- tution between them. In particular, we show that if the production of information services is modeled by the widely used Cobb-Douglas production function, the bud- get shares allocated to hardware and software will remain constant, while the overall size of the budget will grow over time. This is consistent with Frank's proposition that the software-hardware budget shares remain constant de- spite the growth of overall data processing expenditures. Thus, our results provide a theoretical basis for the seem- ingly enigmatic empirical evidence. Our model also suggests that when in fact, software and hardware are not substitutable, software becomes a bot- 'The reader who is unfamiliar with the application of microeconomics to computer management is urged to follow [29], [25] which also discuss the reasons for using net-value maximization as a proper objective. 0098-5589/87/0900-1010$01.00 © 1987 IEEE 1010