NIST GCR 02-834
Benefits and Costs of ATP Investments in Component-Based Software

In many respects, the firms engaged in software development are unlike companies typically envisioned in basic economics theory.  Although they sell their products in competitive markets, their ability to differentiate their products gives them a good deal of market power in new, highly differentiated product areas and thus the ability to raise prices above marginal and average cost.  Because of the nature of the R&D process, however, software development companies must commit significant funds far in advance of achieving either technical or commercial feasibility.  These pre-commercial fixed costs, along with low marginal production costs and highly differentiated products, produce many of the features of natural monopoly markets.  The rapid pace of technological development in software limits the scope and duration of the market power the innovating firms exert, however, as expanded product markets and competitor companies inevitably emerge around the new idea and product. 

Many of the companies involved in component-based software are small start-ups that face the types of financing constraints discussed in Section 3.2.2.  Because the bulk of their expenditures occur prior to their earning any revenues, indeed before technical feasibility has even been established, these firms have difficulty obtaining capital from loans or equity participation.  Funding by ATP or another source of patient capital is thus the only way that these technology development projects can be undertaken.

In this section, we present an economic model based on natural monopoly markets, building on the qualitative results from the first phase of this study.  We then discuss how this general model was operationalized as a means to estimate the economic benefits from the ATP focused program.  Benefits to customer/users of the products sold by the ATP-funded firms are estimated in addition to benefits to the ATP-funded firms.  This methodology was used to analyze the economic contribution of the eight case studies of successful component-based software firms.  Details of each of the case studies, including economic performance calculations, appear in Appendix B; an analysis of the entire ATP "portfolio" and with cross-cutting qualitative results comprises Section 5 of this report.

When companies develop technologies to be embodied in new software products, R&D costs are incurred well before any commercialization activities commence.  Once products are released, firms hope to recoup these up-front R&D costs by charging a premium above the marginal cost of production.  When the firms make the initial decision to invest, the expected value of their future profits must be sufficient to amortize the development costs.  However, firms may also be concerned about the possibility of competitors entering the market.  As a result, we expect that originating firms will price their products at a point that is high enough to cover the fixed and variable costs of production, but low enough to slow the entry of other firms into the market. 

This type of market can be conveniently analyzed as a natural monopoly (i.e., a market characterized by high fixed costs of entry and low marginal costs of production).  Figure 4-1 presents a diagram that illustrates this analysis.  It shows the demand curve for component-based software products (D) and the associated marginal revenue curve (MR).  A small, constant marginal cost of producing each unit of software is shown (MC), as is the average total cost of production (AC).  Fixed costs per unit, which are downward sloping and approach zero at high levels of output, are omitted here for clarity.  The AC curve shows the key characteristic of natural monopoly markets, that average costs fall with increasing production.  This effect gives a large advantage to high-volume producers and favors those firms that are first into the market.

Figure 4-1.  Market Equilibrium for Natural Monopoly Software Product

If a single firm could be assured of maintaining a monopoly, either through control of proprietary technology or by legal protection, it would set price at Pm and expect to sell Qm units of its product.  In most software markets, however, this high price and profit level would attract profitable entry from other firms.  The would-be monopolist will actually be worse off than if it had set a slightly lower price.  Firms that had not incurred the R&D costs would have substantial incentive to enter and capture the market with a similar product at a lower price.  To deter entry, the innovator firm will charge a price of Pnm and produce Qnm units of output.  This price allows the innovator no profit, because price equals average cost but also permits no profitable entry.

Neither of these equilibrium points is socially optimal, however.  Society would be better off if production were pushed out to Q* and the price charged were equal to marginal cost, denoted here as P*.  Producers will never willingly operate at this level, however, because price would be well below average cost-the firms would not be able to recoup their fixed R&D costs.  In the absence of total fixed cost subsidization by the government, this first-best outcome cannot be obtained in a market setting.

Several potential effects of ATP funding can be illustrated in this model.  One effect of ATP's paying for a portion of technology development in the component-based software industry is a reduction in the fixed costs borne by the firms.  As a result of ATP's support, average variable cost is reduced, while the marginal costs are unchanged. 

This concept is presented in Figure 4-2, with the reduction in fixed costs driving average costs down from AC to AC´.  When this occurs, an entry-deterring firm will increase production to Qnm´ and allow the price to fall to Pnm´.  Social surplus increase from the initial state, although quantity is lower and price higher than the social optimum because firms still need to cover their (lower) fixed costs of production.

Figure 4-2.  Effect of R&D Subsidy on Software Product Equilibrium

ATP funding may produce an additional effect that augments the increase in social surplus.  By promoting the development of a component-based market, demand for components may increase.  This effect is presented in Figure 4-3, where demand shifts from D to D´ and average costs fall from AC to AC´.  In this diagram, the shifting out of the demand curve is accompanied by an increase in the quantity produced, and price and output move even closer to the efficient level. 

Figure 4-3.  Effect of Demand Increase on Software Product Equilibrium

Previous work has suggested that ATP involvement may increase demand for component-based products and tools through network externality effects, by conferring scientific credibility on the new technology, or by demonstrating a commitment to continued financial support.  For example, as the number of available components grows, each newly developed component makes existing components more valuable, a positive externality which is magnified by ATP's efforts.  Information disseminated by ATP and project participants may convince applications developers of the inherent advantages of a component-based software design approach; the continued funding of new projects may, in turn, convince these same developers that the array of components and tools will continue to increase in the future.  In the detailed case studies, RTI assessed the extent to which the funded firms have seen demand shift.

Return to Table of Contents.

When attempting to measure the economic benefits created by the ATP projects, we see that the natural monopoly in the product market ensures that the social benefits will be shared between the innovating firm and its customers.  In each period that a product is sold, the component-based software firm earns a producer surplus, equal to the difference between price and marginal cost, multiplied by the number of units sold.  Figure 4-4 shows producer surplus in a single year for a hypothetical component-software firm selling products at zero marginal cost.

Figure 4-4.  Consumer and Producer Surplus Calculation for Component Software Product

By discounting the producer surplus for every period back to the project's start, along with the firm's R&D investments, measures of private return (profit) can also be calculated.  To calculate benefits to users of the products sold by the ATP-funded firms, we need to capture two additional sets of financial flows:  benefits that spill forward to the component firms' customers and public expenditures by ATP.  In this section, we discuss how we measured each of these flows.  For each project, the combined producer and consumer surplus comprise the total social benefit.

4.2.1  Measuring Benefits Captured by Customers

To directly estimate the benefits captured by the downstream firms, we needed to measure the impact of the software components on reducing customers' production costs.  This is an empirically difficult proposition and could be approached in at least two ways.  One method would be to interview customers of each of the products, as Mansfield did in his classic industry studies (Mansfield et al., 1977).  A second approach, the one taken in this project, is to question the innovating firms directly about how much their customers would be willing to pay for each product using the ATP-funded technology.

The consumer surplus generated by sales in a market can be thought of as the benefit received by buyers of a product over and above the amount they must pay to purchase it.  On a demand graph such as Figure 4-4, it can be measured as the area below the demand curve and above the price line.  In standard public economics analysis, consumer surplus is not an exact measure of the social benefit from the product in question, because the buyers would enjoy some surplus from an alternate use of their money.  The net social surplus would more properly be measured as the difference between the benefit received from consuming the product and that of the next best alternative.

In the present case, however, the "consumers" in question are actually firms that use software components in their own products.  When we measure consumer surplus in such a market, it represents the cost savings incurred by purchasers of the components.  Those firms that can reduce their software production costs the most will have the greatest willingness to pay (WTP) for the component product.  Some will be willing to pay less.  In combination, the different customer preferences define the product demand curve and enable us to compute consumer surplus.

Because all of the surplus is actually cost reduction, an estimate of consumer surplus is an accurate measure of the public benefit generated by the component.  As a result, it is possible to extract the customer surplus information we require directly from the component producers, under the assumption that they are well informed about the value of their products to their potential and actual customers.  A properly estimated demand curve for the component products contains all of the information about the social benefits that are passed along from innovator to customer firm to final consumer.  This equivalence has been much discussed in the microeconomics literature, especially in the field of welfare economics.  An accessible and cogent demonstration of the duality of surpluses appears in Just et al. (1982).

A second, even more compelling reason for obtaining the needed information directly from the component software firms is the wide range of customers for each of the component-based products and services.  The logistics of contacting tens or hundreds of customers for each product would make this a prohibitively time-consuming and expensive proposition, beyond the scope of this project.  The very nature of software components ensures that purchasers are likely to be diverse, enjoying quite different potential benefits from using the components.  As a result, the most cost-effective source of data on customers' willingness to pay is the software producers themselves, who need to be aware of the valuation of potential customers to succeed financially.

4.2.2  Empirical Measurement of Derived Demand

Having decided to extract data on derived demand from the innovating firms, we faced the issue of the best way to obtain the valuations.  This issue can be viewed as attempting to empirically construct the demand curve like that described in Section 4.1.  Product prices and quantities sold in a given year provided the equilibrium point on the curve.  By assuming a linear form for the demand curve, one additional piece of information allowed calculation of the producer and consumer surpluses.  One method is to ask firms to estimate their customers' maximum WTP, thus establishing a "choke point," or the price at which the first unit will be demanded.  An alternative approach is to estimate the quantity that would be sold if price were lowered by a specified percentage.  We included questions in the case study interviews aimed at informing these two methods.  The case study analyses in Appendix B include estimates derived from one or the other approach, depending on which appeared more reliable for that case.

4.2.3  Measuring Producer Surplus and Profit

As we note above, producer surplus consists of the difference between marginal costs and price for each unit sold during a specified time period.  With fixed costs of development incurred years before revenues are accrued, relatively large producer surpluses are to be expected in software product markets.  By questioning firms about their prices, marginal costs, and production quantities, we expected to obtain the information needed to estimate surpluses in every year of the product's life.  Marginal costs, consisting mostly of duplication and distribution expenditures, were assumed to be constant with increasing volume.

With each product being offered over a multiperiod time horizon, the producer surplus in each year, along with fixed cost R&D investments, can be discounted back to the project's inception to yield an expected profit for the producing firm.  Each funded component software company was asked to report its R&D and related spending for product launch.  Over the entire product life cycle, this profit must be expected to be positive or else the firm will not engage in costly R&D.  In a natural monopoly market with entry-deterring pricing, the expected value of lifetime profits is zero.  When measured empirically, however, we expect that estimated profits may be positive or negative, for a variety of reasons:

• Positive profits will result if the firm sets its price above the entry-deterring level.  For instance, if the firm does not anticipate competition during the life of the product, it may choose a price that will yield positive profits.  If the innovator is incorrect about potential competition, we would expect entry to occur over time.  In several of the case studies discussed in the next section, we calculated positive profits for the component software firms.
• Negative profits might be found if the firm does not correctly account for its fixed development costs or if it is forced by competition to set a price below its average total cost.  One of the firms in the case studies was found to earn a negative profit, although the total surplus generated by the component-based product was significant.
• In addition to these two non-optimal situations is the possibility that RTI might have employed a different discount rate than the component software company did, and its "true" expected profits are positive or zero.  In the small case study sample, we found firms had small positive profits, indicating close to optimal entry-deterring behavior. 

4.2.4  Capturing ATP Expenditures

We used monthly payments data from ATP, which we then aggregated to annual flows in calculating gross and net social returns.  For the smallest firms, ATP funds made up half or more of the capital spent on technology development, while for the larger firms and joint ventures, corporate resources accounted for a majority share.

 4.2.5  Summary of Calculation Methodologies

The evaluation of the economic impact of the component-based software focused program involved calculating several performance measures, both at the firm level and for the entire portfolio of funded projects.  Three of these measures, benefit-to-cost ratio (B/C), net present value (NPV), and internal rate of return (IRR), provide estimates of the net social surplus created by the combined public and private investment.  A more in-depth description of each of the measures follows.

Benefit to Cost Ratio (B/C)

The annual time series of benefits and costs was assembled for each of the eight case study projects.  Letting B_t be the net benefits accrued in year t and C_t the total funding for the project in year t by ATP and industry, then the benefit-cost ratio for the program is given by

(4.1)

where t is the first year in which benefits or costs occur, n is the number of years the benefits and/or costs occur, and r is the social rate of discount.  In this study, r was set at 7 percent, the Office of Management and Budget (OMB) specified level.  Because benefits and program costs may occur at different time periods, both are expressed in present-value terms before the ratio is calculated.

A similar method was used to evaluate the B/C ratio for the entire CBSD portfolio of 24 funded projects.  For the overall program calculation, the results of which appear in Section 5, the sum of the benefits produced each year for the eight projects was included in the summation term in the numerator.  The denominator included outlays of ATP and industry funding, both during and after ATP funding, provided to all 24 projects, a conservative assumption that is discussed in greater detail in Section 5.

Net Present Value (NPV)

The NPV of ATP's contributions to the CBSD projects was calculated as

(4.2)

where the terms have the same meanings as identified for the B/C determination. As before, the overall portfolio NPV was found by summing the discounted net benefits from the eight in-depth studies, less the total focused program expenditures. Any project that yields a positive NPV is considered to have been economically successful. It should be noted that the 7-percent real discount rate required by OMB is a rather high hurdle for project analysis, ensuring that the software projects that showed a positive NPV were quite socially advantageous.

Internal Rate of Return (IRR)

The IRR is the value of r that sets NPV equal to 0 in Eq. (4.2). Its value can be compared to conventional real rates of return for comparable or alternate investments. Risk-free capital investments such as government bonds can be expected to yield rates of return under 5 percent in real terms, while equities seldom return more than 10 percent over an extended period of time. In academic studies of the diffusion of new technologies, however, real rates of return of 100 percent or over have been found for significant advances. It should be noted that for projects in which costs exceed benefits, an IRR cannot be calculated.

Return to Table of Contents or go to next Chapter.

Date created: December 3, 2002
Last updated: August 2, 2005

ATP website comments: webmaster-atp@nist.gov  / Technical ATP inquiries: InfoCoord.ATP@nist.gov.

NIST is an agency of the U.S. Commerce Department
Privacy policy / Security Notice / Accessibility Statement / Disclaimer / Freedom of Information Act (FOIA) /
No Fear Act Policy / NIST Information Quallity Standards / ExpectMore.gov (performance of federal programs)