This dissertation takes two different research perspectives to address the central theme of agricultural production and productivity. The first two essays focus on household production, which, as the primary form of agriculture to date, not only affects the welfare of individual rural families but also food supplies on a global scale. Agricultural productivity hinges largely upon farmers' choice of technology, inputs, and management strategies. Specifically, the first two essays investigate land fragmentation, a common farming practice worldwide, and evaluate its impacts on agricultural production. Chapter 2 argues that land fragmentation enables farmers to reduce risk by diversifying production among discrete plots of land which may be subject to heterogeneous growing conditions. Using Tanzanian household survey data, this essay finds robust evidence to support a risk-reduction hypothesis and indicates that land fragmentation is positively associated with production efficiency. Chapter 3 develops a production model that incorporates risk, production efficiency, and risk preferences and shows that land fragmentation may encourage risk-averse farmers to increase labor intensity, thereby leading to higher efficiency. It is also shown that exclusion of risk preferences from efficiency analysis may lead to biased or even misleading estimates. The second focus of this dissertation is an assessment of the published evidence on the payoffs to investments in agricultural research and development (R&D). The related two essays focus on methodological as well as policy issues underlying the agricultural R&D evaluation literature. Specifically, Chapter 4 scrutinizes the prevailing internal rate of return (IRR) measure and argues that it is based on implausible assumptions that often lead to inflated estimates of the returns to research. This essay develops a novel method for recalibrating the reported rates of return using a more plausible modified internal rate of return (MIRR) measure and derives more modest estimates. Using the detailed information collected for each R&D evaluation, Chapter 5 examines how the wide variation in the reported IRR estimates can be explained by factors such as research type, research focus, commodity type, institutional aspects of the research, target region, and methodological specifications. The findings have important implications for future agricultural R&D policy as well as R&D evaluation methodologies.