|Abstract: ||近年來，許多研究企圖從「所有權結構」、「財務關係人權利與義務」、「財務透明度」以 及「董事會特性」等構面，去分析影響企業資金成本的因素。雖然這些研究已經有顯著 的成果，然而，基於傳統研究均以最小平方法來進行分析，而無法進一步瞭解「不同資 金成本水準」的公司，影響他們的資金成本的因素，可能有顯著的不同。對於資金成本 很高的公司而言，他們關切的是如何降低資金成本；而資金成本很低的公司而言，他們 則關切如何預防資金成本的變高。對於這些公司而言，影響他們資金成本的因素，可能 與「一般公司」有明顯的不同。 就資金成本的選擇而言，本研究將分析:債務資金成本與權益資金成本。其中，由於我國 的「證券發行人財務報告編製準則」第17 條三十款與四十二款規定必須揭露短期借款 明細與長期借款明細。其中與本研究有關的內容包含: (1) 借款銀行、(2)借款金額、(3) 借款起迄期間、(4)利率別(固定或浮動)、(5)利率。本研究經利用臺彎特有的資料，進行 以份量迴歸法探討影響債務資金成本的因素。至於權益資金成本則將以美國資料為分析 對象。其中權益資金成本的計算方法為: (1) Brav, Lehavy, and Michaely (2004)、(2)Claus and Thomas (2001)、(3)Gebhart, Lee, and Swaminathan (2001)與(4)Gode and Mohanram (2003)的方法計算權益資金成本。自變數則為Francis, LaFond, Olsson, and Schipper (2004) 所關切的盈餘屬性以及Ashbaugh-Skaife et al. (2006)所研究的公司治理變數。|
Purposes (1) To examine whether the relations between the cost of equity capital and CG-variables holds at different points of the conditional distribution of the cost of capital using the quantile regression method, i.e., whether these relations differ for firms with high cost of equity capital from firms with low cost of equity capital. (2) To introduce the method of quantile regression, which has been used widely in the past decade in applied economics, to accounting research? Motivations (1) For example, Francis et al. (2004) have been examined the relation between the cost of equity capital and seven earnings attributes: accruals quality, persistence, predictability, smoothness, relevance, timeliness and conservatism using the traditional ordinary least squares (OLS) estimation method (and Fama and MacBeth (1973) method). As such, the relations they document describe how the mean of the response variable (cost of capital) relates to seven earnings attributes. The relations between other points on the distribution of the response variable and seven earnings attributes may or may not be the same as the mean relation. This inadequacy of the OLS estimates is recognized in an influential textbook by Mosteller and Turkey (1977).1 (2) Quantile regression, developed by Koenker and Bassett (1978), extends the traditional OLS method by estimating the conditional quantile functions – the relation between the response variable and explanatory variables for each quantile of the response variables.2 Quantile regression, thus, allows us to examine the relation between the response variable and explanatory variables not only for firms that behave according to the mean of the response variable (as captured by OLS) but also for firms that behave above or below the mean. In other words, quantile regression presents a much more detailed picture of the relation between the response variable and explanatory variables than does the OLS. To take an analogy, the OLS is like reporting the mean of a sample distribution whereas quantile regression is like reporting the full percentiles of the sample distribution (e.g., min, p10, Q1, median, Q3, p90, max). (3) Why do I use the quantile regression method to examine the relation between the cost of capital and seven earnings attributes? First, Francis et al. (2004) is an influential paper documenting that information risk, as proxied by seven earnings attributes, is a priced risk factor. Second, they state that their findings have 「practical value for investors, researchers, standard setters and managers」 (p. 1007). In particular, their findings support the notion that the cost of capital is linked to earnings attributes, which implies that managers can influence the cost of capital through influencing earnings attributes, especially accounting-based attributes such as accrual quality. Since firms』 cost of capital could be substantially above or below the mean cost of capital of a sample, the mean relation between the cost of capital and earnings attributes documented in Francis et al. (2004) does not necessarily hold for these firms. Quantile regression, on the other hand, can estimate the relation between the cost of capital and earnings attributes for these firms.