Consumer preferences and product and process R&D.

Firms often conduct both product and process R&D. Consumer preferences typically play a vital role in determining these R&D choices. The economics literature, however, has mostly ignored the interrelationship between the R&D choices of firms and consumer preferences. I study product and process R&D from the perspective of their relationship with consumer preferences. In doing so, I find a novel distinction between the two kinds of R&D. This distinction can explain several empirical observations regarding firms' choices of the two, and can also potentially enable us to better empirically identify the two.

1. Introduction

* Firms continually try to improve the quality of existing products and create new products--product R&D--and to lower the cost of making existing products--process R&D. For example, automatic data synchronization, better handwriting recognition, and a larger and more readable screen are some recent product innovations in hand-held computers, whereas a better production technique leading to lower reject rates would be an example of a process innovation.

Studies of R&D activities by firms in the economics literature have focused mainly on the interrelationship of R&D and variables such as firm size, market concentration, mode of competition, and market structure--see, for example, the surveys of Cohen and Levin (1989) and Symeonidis (1996)--and have more or less ignored the interrelationship between R&D and consumer preferences. (1) Consumer preferences, however, often play an important and, in many cases, a (if not the) key role in determining these R&D choices. In this article, I analyze the interrelationship between the R&D choices of firms and consumer preferences in a dynamic set-up where firms continually conduct both product and process R&D simultaneously.

I begin by asking whether any difference between product and process R&D arises due to consumer preferences. The difference between product and process R&D is a subtle conceptual question that has not been answered satisfactorily to date--see, for example, Cohen and Levin (1989). I consider a monopolistic industry in an infinite horizon set-up with discrete time. The monopolist faces a discrete-choice model of consumer demand with vertical product differentiation. I solve the monopolist's dynamic optimization problem. I find that, ceteris paribus, the value of a process innovation depends only on the quantity sold, while that of a product innovation depends both on the quantity sold and the marginal buyer's willingness to pay--hereafter WTP--for the product innovation, and hence, also on who buys the product. Consequently, though both serve to increase the per-unit price-cost margin for the product, ceteris paribus, they turn out to be nonequivalent, provided that consumers differ in their WTP for product innovations. Thus, I find a novel economic distinction between the two kinds of R&D based on their relationships with consumer preferences.

Next, I consider whether this distinction can explain empirical observations of the product and process R&D choices of firms. One pattern that has been identified is that firms usually conduct relatively more process R&D over time. In fact, according to Klepper (1996), this is a defining feature of evolution in industries with opportunities for both kinds of R&D.

In my framework, potential buyers of the product differ in their WTP for improvements in product quality. Further, WTP for the product and for quality improvements are positively correlated--i.e., people with a higher WTP for the product (consumers at the higher end of the market) also value quality improvements more than those with a lower WTP for the product (consumers at the lower end of the market). While the distribution of preferences across potential buyers is time invariant in my framework, the WTP of the marginal buyer for quality improvements is lower over time. Hence, the monopolist increasingly devotes more of its R&D effort to making the product cheaper. Put simply, in my model, there is relatively more process R&D over time because consumers want it to be so.

The decline over time in the WTP of the marginal buyer of a product for quality improvements as discussed above is evident in many industries, even from casual observation. Typically, early buyers of a product care mainly about its features or quality and would pay a lot for improvements in them. However, over time, a product is usually sold increasingly to consumers at the lower end of the market who often find its features sufficient for their needs and care more about price reductions. This has happened for PCs in particular and computers in general--see Filson (2002) and Flamm (1988). Thus, my analysis also reveals a link between observed shifts over time in the R&D composition for a product and the composition of its actual buyers in their WTP for quality improvements.

I also develop the cross-sectional implications of my model. It has been found empirically that a larger firm typically conducts more R&D but has a lower R&D productivity and also conducts relatively more process R&D than a smaller firm--these are discussed in detail later. My framework can explain these empirically observed patterns about firm size and R&D choices. Further, I find that the approach here can also potentially help us do a better job of empirically distinguishing product and process R&D than is currently the case.

There are two major conventional approaches to understanding the issues analyzed here about the product and process R&D choices of firms. The first relies on exogenously given changes in opportunities for the two kinds of R&D over time. Within this, there is the dominant design hypothesis--hereafter DDH. This postulates the emergence of standards and a dominant design (a product with standardized features and capabilities) and a resultant decline in the relative scope for product R&D over time--see Utterback and Abernathy (1975). More recent work by Filson (2001) shows that changes in opportunities for product and process R&D over time can explain some features of firms' choices of the two. The DDH and Filson hypotheses differ in that Filson does not postulate an a priori pattern of change over time in opportunities for the two kinds of R&D. The second approach is based on differences between the two kinds of R&D in the extent to which an innovating firm can earn revenues by licensing its outcomes to other firms. From a survey of R&D executives, Levin et al. (1987) find that product innovations are generally easier to protect through patents and to license than process innovations. Cohen and Klepper (1996a) and Klepper (1996) build models that incorporate this feature and explore its effects on the product and process R&D choices of firms. Both assume that returns from process R&D for a firm come through its own output only, whereas that is not so for product R&D. Klepper assumes that a firm conducting product R&D either succeeds or fails. The return from success in product R&D is an exogenously given constant. Cohen and Klepper (1996a) assume that part of the returns from product R&D for a firm is from its application to an exogenously given, constant amount of output outside of that sold by the firm. Hence, there are scale effects in returns to process R&D in both, but no scale effects at all in Klepper's scenario, and scale effects but to a lesser extent in that of Cohen and Klepper (1996a), in returns to product R&D. This difference in scale effects that arises from the difference in the way that returns to the two kinds of R&D are modelled is the key explanatory factor in these models.

My framework explicitly models how returns to both product and process R&D arise and finds scale effects in returns to both. However, the relative magnitude of these scale effects (product to process) varies inversely with the degree of market penetration due to the difference in how product and process R&D are related to consumer preferences. This is the key in my framework. Further, while my framework can easily accommodate exogenously given changes in opportunities for product and process R&D, my explanation is not built on that feature. Hence, any such changes in opportunities for the two kinds of R&D that can explain some feature of the product and process R&D choices of firms would only strengthen my explanation in that regard. Thus, my approach and the previous approaches complement each other.

The rest of the article is as follows. Section 2 lays down the basic framework. In Section 3 I consider the difference between product and process R&D. Section 4 looks at the implications of this difference for R&D choices by firms. In Section 5 I discuss how the difference between product and process R&D found in this article may help in categorizing innovations as product and process innovations. Section 6 considers several alternative scenarios. Section 7 concludes. All proofs are in the Appendix.

2. The model

* The framework of analysis is described as follows.

* Preferences, production, and R&D. Basic set-up. I consider a nondurable product. The product is characterized by a single, one-dimensional product attribute--quality--denoted by q, q [member of] [R.sub.+]. The time horizon is infinite with discrete time periods denoted by t; t = 0, 1, 2,.... Period 0 is the period in which the product is commercially introduced. In each period, there is an exogenously given, time-invariant mass of potential buyers that I normalize to be 1. The price and utility of the product, R&D expenditure, costs, and revenues are all measured in terms of a numeraire good.