The "software crisis" is characterized by an inability to develop software on time, on budget, and within requirements. A seminal paper in quantitatively assessing software costs predicted that software costs would come to consume an increasingly larger share of system costs (Boehm 73). Figure 1 resembles a graph presented by Dr. Barry Boehm, illustrating this prediction. Figure 1, however, is derived from a mathematical model yielding this curve, not Boehm's paper. This model shows that as long as productivity increases faster in hardware than in software, the percentage of system costs spent on software is bound to increase. This trend is not a manifestation of a software crisis, but the inevitable result of trends in hardware which are unlikely to be duplicated in software or almost any other industry.

The model was developed based on techniques developed by the economist Luigi Pasinetti (1993). Suppose hardware and software can each be measured in a standard unit, and that a computer system consists of a unit of hardware and a unit of software. Suppose the cost of producing both hardware and software is proportional to the amount of labor used in producing the respective outputs. Let l _h denote person-years required to produce a unit of hardware, and let l _s be the number of person-years required to produce a unit of software. Assume productivity grows at the rate P _h in the hardware industry and at the rate P _s in the software industry:

Define P as the proportion of system costs spent in software in the production of a computer system. The rate of growth of this proportion is given by Equation 2:

Given a rate of growth of productivity in hardware exceeding the rate of growth of productivity in the software industry, software will consume an ever increasing proportion of system costs, as in Figure 1.

Assuming the above argument is correct, is increasing emphasis on software process improvement therefore misdirected? No, economic reasoning shows it is appropriate. Optimally, management should continue to add additional funds to a software process improvement program as long as the marginal decrease in the expected cost of software exceeds the marginal cost of a software process improvement program. If the ratio of the marginal costs of hardware and software process improvement programs remain unchanged, then management should direct proportionality more to software process improvement as software costs become a greater proportion of system costs. Progress in software process improvement technology may also increase the desirability of funding such programs.