연구개발투자의 경제적 시차효과에 대한 시계열분석

The literature on the economic effects of research and development (R&D) expenditures is based primarily on cross-sectional regressions or panel data regressions with time and firm fixed effects such that the parameters relating R&D to accounting earnings and/or market value are cross-sectionally constant. This paper utilizes firm-specific time-series data to examine whether reported earnings reflect benefits from past R&D expenditures and whether market values impound the economic value derived from the benefits of past and present R&D expenditures. This study estimates a recursive system of two equations: one for earnings and one for valuation. The earnings equation associates earnings with recorded tangible assets, advertising, and R&D expenditures. It intends to extract realized R&D benefits from reported earnings numbers. The valuation equation relates market values of equity to book values, abnormal earnings, and R&D expenditures. It determines how the accounting and R&D numbers are valued in the market. This system of equations assumes that past R&D generates earnings that create market value. Model parameters are estimated for each firm separately. Unlike many previous studies based on cross-sectional and panel data analyses, this study finds very weak empirical support for the economic benefit of R&D expenditures. The results from the earnings equation indicate that reported earnings reflect realized benefits from past R&D for only 10-15% of the sample firms. The results from the valuation equation indicate that R&D affects firm valuation for only less than 20% of the sample firms at best. In a series of additional robustness tests, regardless of earnings measures and the types of variables, very weak empirical support remained to prevail. More specifically, I estimate polynomial distributed lag (PDL) models with varying lag terms both for earnings and valuation equations. Valid data for earnings, prices, book values, R&D expenditures, advertising expenditures, tangible assets for 18 consecutive years were available for 168 firms from 1990 to 2007. In the earnings equation PDL model, the cross-sectional mean and median of the sum of lagged coefficients of total R&D expenditures are of negative values. Moreover, the firms with significantly negative coefficient estimates outnumber their positive counterparts. I find that the aggregate R&D parameter estimates are significantly positive at 0.1 level for only 22 firms out of 168 sample firms. Similarly, in the valuation equation PDL model, the cross-sectional mean and median of the comparable coefficients are negative. The R&D valuation parameter estimates are significantly positive for only 18 firms at 0.1 level. In an additional analysis, I conduct cross-sectional analyses on pooled and yearly cross-sectional samples both for earnings and valuation equation models. Interestingly, the estimated coefficients for both models are significantly positive, largely consistent with many previous researches based on cross-sectional approach. This lack of consistency between the results of time-series and of cross-sectional approaches suggests that either or both approaches could be fraught with technical problems. From the standpoint of time-series context, cross-sectional analysis has a number of disadvantages. The cross-sectional approach makes use of comparisons across firm expenditures on R&D. Comparisons across firms can only isolate the effects of R&D if they control for other influences on earnings or market values that may vary from one firm to another. However, it is not always possible to observe and control for all firm differences that could affect earnings or market values. This possibility of correlated omitted variables bias is one of the primary disadvantages of cross-sectional models of R&D. Futhermore, the coefficient estimates in cross- sectional studies may be influenced by firm outliers, which is likely to be the case with the results of this study. The result of this study does not necessarily imply that time series approach is superior to cross-sectional or panel data approaches and that many of the prior studies are problematic. Nevertheless, this paper raises significant doubts about the seemingly robust finding that R&D and corporate earnings and market values are significantly and positively related. The economic benefit of R&D expenditures is simply not as robust as many of the existing literature appears to indicate on a firm level basis.