연구개발투자 관련 연구의 메타분석

Abstract

There have been many empirical studies regarding the economic impact of R&D, but there also has been caused controversy due to the gap between empirical results and observation in real economy. This study reviews and derives limitations in which previous R&D study had and tried to measure the economic impact of R&D using meta regression analysis method. Reviewing previous studies, we find four disputes regarding the estimation of economic impact of R&D. First of all, there are discussions about R&D stock and depreciation rate. It is usually the case that studies use R&D stock in the estimation, rather than a flow variable of R&D investment. Some research employs R&D to GDP ratio or R&D to total sale ratio, but these have a limitation to reflect the characteristics of R&D investment. In using depreciation rate, there is no formal way to incorporate it into the model and it depends on subjective decision. Second one is about the time lag in the model. There exists a gap between R&D investment and the practical use, and it takes times to have an external effect on other industries. Deciding how long it takes time to commercialize R&D investment also depends on researchers’ subjective judgement, varying from 1 to 3 years in most studies. Thirdly, there needs to distinguish the stage of R&D investment. R&D investment can be divided into basic, applied and experimental development research and each has different goals. Few research has taken this into account. Lastly, in analyzing the economic impact of R&D investment, focal point should be on the construction of panel data and data analysis. Panel analysis has an advantage in capturing and controlling cross-sectional/time series properties, therefore, panel data analysis would be suitable in estimating the effect of R&D investment. From the meta regression analysis, the results which we found are summarized as following; first, empirical model should be based on panel data analysis. By using variable methods and model, it is possible to get more reliable results. Secondly, it is better to use a system of equations which is based on theoretical model. In many studies, an estimation model lacks of theoretical background and simply adopts an estimation equations from previous literature. It raises questions regarding the estimation model’s credibility and makes it vulnerable. Lastly, we find that capital and labor variables should be included in an estimation; according to our results from meta regression analysis, when these variables are omitted, there may cause over-estimation of the effect of R&D investment. In analyzing the economic impact of R&D investment, researchers should cast careful attention on panel analysis, a system of equation and inclusion of capital and labor related variable. In addtion, as was pointed out in other studies, intensive research regarding R&D depreciation rate and its time lag should be needed to have more reliable and improved results.