R&D policy has not adequately modeled the relevant R&D underinvestment phenomena and thus is unable to characterize, explain, and measure such underinvestment in a manner that facilitates effective R&D policy. Four factors can cause systematic underinvestment in R&D-intensive industries, thereby creating patterns of excessive risk: complexity, timing, existence of economies of scale and scope, and spillovers. The impacts of these factors vary in intensity over the typical technology life cycle, so government policy responses must be managed dynamically. In addition to understanding the causes of underinvestment in R&D, the magnitude of the deficiency relative to some optimum must be estimated to enable a ranking of technology areas with respect to expected net economic benefits from a government subsidy. Project selection criteria must therefore be based on quantitative and qualitative indicators that represent the nature and the magnitude of identified market failures. This paper provides an improved conceptual framework for management of R&D policy by synthesizing the broad and segmented literature on R&D investment and then presenting a methodology that facilitates assessing the long-term expected benefits and risks from current and proposed R&D portfolios. A three-stage process is developed: (1) identify and explain the causes of the underinvestment, (2) characterize and assess the investment trends and their impacts, and (3) estimate the magnitude of the underinvestment relative to a perceived optimum in terms of its cost to the economy. Only after all three stages of analysis have been completed can the underinvestment pattern be matched with the appropriate policy response.