|The Korean government invests approximately 1.17% of the total amount of GDP in national R&D projects annually, which ranks in fourth place among OECD countries. In addition, the government makes steady and persistent efforts to enhance the overall performance of the national R&D system. However, the output from the government R&D activities falls short of expectations. This leads to the inference that there still exists room for improvement. In particular, considering the current status of the national R&D system, a significant breakthrough is becoming increasingly necessary. As a part of the efforts to redesign the national R&D system, we particularly focus on the R&D budgeting system.
In this study, we aim to develop a framework which helps devise the optimal R&D budget allocation. For this purpose, we conduct this study by using a demand-driven approach - we first find issues to be improved in the current R&D budgeting system, and we then try to find the appropriate quantitative approaches with which to address each issue we found. Specifically, to find the challenging issues that need to be addressed, we extensively review prior studies regarding the national R&D budgeting system and conduct interviews with individuals who have an expertise in designing or analyzing the national R&D budgeting system. Finally, we find three issues that can be improved by using an appropriate quantitative method as follows: 1) R&D investment strategy prioritization process, 2) R&D budget ceiling process, and 3) R&D budget allocation and coordination process.
Next, we comprehensively investigate the quantitative methods that can be applicable to the R&D budgeting system. Considering the features of the national R&D budgeting system, we particularly focus on methods that are mainly inspired from the field of operations research. Specifically, methods were finally selected as follows: linear and non-linear programming, goal programming, data envelopment analysis (DEA), analytic hierarchy process (AHP), portfolio matrix, computable general equilibrium (CGE), decision tree (decision analysis and machine learning), stochastic programming, robust optimization, modeling and simulation, genetic algorithm, and artificial neural network.
We then propose a framework which aligns each issue with approaches that can be used to handle the issue. For instance, we design a two-phase approach for an optimal R&D budget allocation and coordination process. It is worth noting that the two-phase approach starts with preparing the key elements (e.g., estimation of the R&D investment, estimation the success probability of the R&D investment, etc.) to be used to allocate R&D budgets. In the second phase, methods specialized in allocating resources are employed depending on the detailed features of the problem what we address.
The two-phase approach is then applied to the one of the current national R&D projects (BrainKorea 21 plus project). We use the data collected from national science and technology information services (NTIS) and develop models by using automatic machine learning algorithm, optimization methods. The results confirm the effectiveness and the usefulness of our approach. Specifically, 13.6% higher performance (# of patents and academic papers) could be achieved with the current budget.
In sum, this study addresses the national R&D budgeting system. Although it can be fairly difficult to enhance the performance of the R&D budgeting system owing to the inherent characteristics of the R&D budgeting system, we contribute to the small amount of existing literature featuring the design of an evidence-based national R&D budgeting system. In particular, our findings shed light on the value of finding meaningful information, which can help improve not only the national R&D budgeting system, but also the overall national R&D system. Despite some limitations, we believe that our study can be a cornerstone of designing a data-driven and evidence-based national R&D budgeting system in Korea.