查稅評分卡之建立與驗證--以臺灣營利事業所得稅為例

論文提要本論文以財政部財稅資料中心八十八年度之營利事業所得稅資料檔為研究樣本, 經過篩選程序以及刪去不合邏輯值之樣本後, 將受選案查核之資料從母體區分而出, 再自其中區分出電腦選案以及非電腦選案的資料, 並利用此電腦選案抽查資料作為發展樣本製作「查稅評分表」。「查稅評分表」是利用歷史申報資料找出逃稅者的特徵, 並依這些不同的逃稅特徵給予不同的分數, 也就是將各逃稅特徵予以數量化, 然後將各特徵下所得到的分數全部加總起來作為總評分, 讓稅務稽徵機關可以依總評分的高低對納稅義務人的逃稅風險作一排序, 亦即稅務稽徵機關可以依據分數高低來預測納稅義務人的逃稅可能性, 並依此總評分作為選案的依據。在「查稅評分表」的模型建構過程中, 不但能找出逃稅傾向與特徵, 還能利用「查稅評分表」顯示之結果作為選查逃稅案件的依據。當評分結果之總分愈高, 表示其「逃稅風險」較低,較不傾向逃稅; 當評分結果之總分愈低, 表示該營利事業組織具「高度逃稅風險」, 國稅局的稽查人員可以依「查稅評分表」決定是否查緝該營利事業組織之逃稅行為。根據實證結果顯示, 隸屬於台北市國稅局的營利事業之得分較高, 申報別中以會計師稅務簽證申報的逃稅率較低、交際費用率與捐贈費用率則是顯現出在限額內列支比率愈高的逃稅率較低、存貨週轉率亦呈現出愈低者的逃稅率愈高的趨勢等等。且當我們將截切點設為0.5080時, 只要營利事業經過查稅評分表之計算後的總評分小於407 分者, 則該營利事業就被應該列為選案之一, 另外, 由AUC 值高達0.7297 可知「查稅評分表」具有優越的預測能力。所以, 藉由「查稅評分表」來選擇查緝案件不但能改善傳統之主觀查緝, 還能節省人力資源。

關鍵字

逃稅；查稅評分表；訊息值；證據權重

並列摘要

Abstract This research uses business income tax returns filed for the tax year 1999 from The Data Processing Center of the Ministry of Finance in Taiwan. We take several steps to refine this data. Firstly, we delete firms with illogical values. Secondly, all non-audited firms are removed. Finaly, we seperate the data by firms with computerizing audited and firms with noncomputerizing audited. This computerizing audited data called development sample. We use it to produce the Find Evasion Scoring Card. A Find Evasion Scoring Card is a formula for assigning points to enterprise characteristics in order to derive a numeric value that reflects how likely an enterprise is, relative to other enterprises, to experience and tell us if a particular enterprise is more or less likely to evade taxthan another enterprise. We produce the Find Evasion Scoring Card under the assumption of maximizing tax revenue.During the process of producing the Find Evasion Scoring Card, we can not only find characters of evasion, but also use the outcome of the Find Evasion Scoring Card to make decisions. It shows that the firms are with better credit as scores are increasing. That is, they are a group with lower risk of tax evasion. Vice versa, it shows that the firms are with worse credit as scores are decreasing and they are a group with higher risk of tax evsion. Auditors of IRS can make a decision of audit it or not by the scores of the Find Evasion Scoring Card in order to improve the traditional and subjective audit and save human resources. According to the result of our Find Evasion Scoring Card, we found that the error rate reaches 0.6527 as the cutoff-point is 0.5080, that is to say, businesses should be audited by IRS if they have scores below 407. We also found that business income tax returns attested by a certified public accountant (CPA) or firms under the jurisdiction of the National Tax Administration of Taipei have higher scores than others. Under the examined disbursement amount, lower social expense rate or lower donate expense rate have lower scores. Profit-seeking enterprises with lower inventory turnover rate have the inclination of tax evasion. And the AUC value reaches 0.7297 implies that our Find Evasion Scoring Card has dominant predictiveness.

並列關鍵字

AUC ； Information Value ； ROC ； Tax Evasion ； Weight of Evidence

參考文獻

[18] 劉珮芬(2002)《運用決策樹於營利事業所得稅結算申報書選案

[13] 陳明進(2003), 〈稽徵機關稅務查核對營利事業短漏報所得之

[8] 洪吉山(2002)《稅捐申報行為之研究》, 國立成功大學法律學

[31] Hunter W. J. ang M. A. Nelson (1996), An IRS

[33] Meizuh Huang (2004), Integrated Income Tax System,

國際替代計量

查稅評分卡之建立與驗證--以臺灣營利事業所得稅為例

全文下載

主題瀏覽