이전가격분석과 비교가능성 요인의 유효성- 국내 도매업체를 중심으로 -

The objective of this study is to contribute to the improvement of reliability of a transfer pricing analysis through confirming the basis of the arm’s length principle and verifying the effectiveness of comparability factors. In order to identify the current comparability factors, a case study has been conducted using 20 different transfer pricing analyses from pharmaceutical, luxury brands, semi-conductor and chemical industries. For the empirical study, a research model using independent variables of intercompany sales-to-total sales rato (INTs), intercompany purchase-to-COGS ratio (INTp), sales volume (lnSale), selling expenses-to-sales ratio (ADP) Selling expenses include advertising expenses, promotional expenses, sales commission and entertainment expenses. , personnel costs-to-sales ratio (PSNL), education and training expenses-to-sales ratio (EDT), inventory risks (INVr) Inventory risks have been measured by adding the losses (net of gains) from inventory valuation to the actual inventory losses. and net working capital-to-sales ratio (NWC), has been designed based on the results of such case study. The multiple regression analysis was used to determine whether the above independent variables (i.e., comparability factors) are statistically linked to the profit level indicators (i.e., dependant variables) such as gross margin (GM), operating margin (OM) and Berry ratio (BR), which are generally considered in a transfer pricing analysis for distributors. The 546 samples (distributors) were narrowed down from a total sample size of 2,522 companies. In order to eliminate certain outlier, the Tukey adjustment has been applied. The period for this analysis was 10 years (1999-2008). The results of the study can be summarized as below:First, there were certain cases where INTp was statistically significant to the profit level indicators, which implies that certain intercompany transactions may not be conducted on an arm’s length basis. Thus, we may say that the arm’s length principle at least has certain grounds. Second, PSNL and NWC were confirmed as effective comparability factors. Identifying and confirming PSNL, which has not been generally considered as an effective comparability factor, may have significant implications for the current practice of transfer pricing analysis. In addition, NWC was also statistically linked to the profit level indicators, which implies that the working capital adjustment applied in a transfer pricing analysis is also meaningful. Third, lnSale, EDT and ADP showed inconsistent results. Thus, these factors should be carefully applied in a transfer pricing analysis considering the characteristics of the relevant industry under question. In addition, INVr was not confirmed as an effective comparability factor. Fourth, it was confirmed that the Berry ratio would not be an appropriate profit level indicator for a transfer pricing analysis for ordinary distributors other than those with simple functions or service providers. Fifth, the results of the empirical study for a large-pool of distributors covering all items were different from those for small-pools of each item. In general, the bigger number of samples, the more statistically meaningful results. Thus, even if OECD Transfer Pricing Guidelines states “quantity does not make up for poor quality of data” (Paragraph 1.51), the results of this study show that increasing the number of samples (i.e., comparables) may still be helpful in improving the reliability of analysis under the situation where it is practically difficult to obtain sufficient samples considered to be of good quality. Notwithstanding the above, this study has the following limitations:First, this study tried to develop grounds for research model based on the results of a case study using actual transfer pricing analyses. Thus, the research model may not have a strong theoretical basis. In addition, the model may have an issue of ‘correlated omitted variable,’ which may be an inherent problem of transfer pricing analysis, but such omitted variables need to be further reviewed. Second, this study is conducted based on a statistical analysis using a large-pool of samples. Thus, considering that the actual transfer pricing analysis is generally conducted using a relatively small number of comparables, there can be certain limitations in generalizing the results of this empirical study.