Estimating Dynamic R&D Demand: An Analysis of Costs and Long-Run Benefits

Estimating the private return to R&D investment has been a major goal for decades and most of the empirical literature has been built around the knowledge production function. In this framework, investment in R&D creates a stock of knowledge within the firm that enters into the firm’s production function as an additional input factor. Estimates of the effect of this knowledge stock on output provides a measure of the ex post return to the firm’s investment in R&D.

This paper provides a different approach to measuring the private payoff from R&D investment. We develop and estimate a dynamic, structural model of the firm’s decision to invest in R&D and quantify the cost and long-run benefit of this investment. The dynamic model incorporates and quantifies linkages between the firm’s R&D investment, product and process innovations, and future productivity and profits. And it provides a natural measure of the long-run payoff to R&D as the difference between the expected discounted sum of future profits if the firm undertakes R&D versus if it does not. The firm will choose to invest in R&D if this payoff is greater than the fixed or sunk cost they pay to invest in R&D.

We use firm-level data from the Mannheim Innovation Panel (MIP) for German manufacturing industries to estimate the dynamic structural model and to calculate the long-run payoffs to R&D. Comparing across industries for the firm with the median productivity level, we find that the expected long-run benefit of investing in R&D varies from a high of 43 million euros in the vehicle industry and 20 million euros in the chemical industry to a low of about 350 thousand euros in the plastic, non-metallic mineral products, and manufacturing nec industries. By combining estimates of the expected long-run benefit of R&D with the cost of R&D, we also estimate the distribution of net benefits across firms in each industry. We find that the expected net benefit varies substantially across industries and across firms that have already invested in R&D and those that are just starting R&D investments because of the substantial differences in the fixed versus sunk costs. Expressed as a proportion of firm value, our results show, for instance, that the net benefit for the median firm with prior R&D experience varies from 2.4to 3.2 percent across five high-tech industries but varies from -4.6 to 0.6 percent for firms with no previous R&D experience. The negative value implies that the median inexperienced firm would not find it profitable to invest in R&D. Given unexperienced firms find R&D investment profitable and start performing R&D, we estimate a net benefit of 2.0 to 2.4 percent in the high tech industries. These net benefits are substantially smaller, around 0.2 percent in low-tech industries.

The estimated dynamic structural model of R&D demand can be used to simulate how a change in the cost structure of R&D arising from, for example, a tax break or direct subsidy for R&D investment, affects the firm’s investment choice and future productivity growth. We find that, in high-tech industries, a 20 percent reduction in the fixed cost of R&D leads after five years to an average increase of 7 percentage points in the probability a firm invests in R&D and a 4 percent increase in mean productivity. A similar reduction in the cost faced by firms just beginning to invest in R&D, however, has very little impact on the probability of investing or the level of productivity. That is, the two cost changes have very different impacts on firm incentives. Fixed cost reductions encourage all firms to invest. In contrast, the reduction in startup costs encourages new firms to begin investing but also reduces the option value of investing leading some firms to stop their R&D.

Peters, Bettina, Mark J. Roberts, Van Anh Vuong and Helmut Fryges (2013), Estimating Dynamic R&D Demand: An Analysis of Costs and Long-Run Benefits, ZEW Discussion Paper No. 13-089, Mannheim, published in: RAND Journal of Economics.