政大機構典藏-National Chengchi University Institutional Repository(NCCUR):Item 140.119/69638
English  |  正體中文  |  简体中文  |  全文笔数/总笔数 : 88295/117812 (75%)
造访人次 : 23401779      在线人数 : 44
RC Version 6.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
搜寻范围 查询小技巧:
  • 您可在西文检索词汇前后加上"双引号",以获取较精准的检索结果
  • 若欲以作者姓名搜寻,建议至进阶搜寻限定作者字段,可获得较完整数据
  • 进阶搜寻


    请使用永久网址来引用或连结此文件: http://nccur.lib.nccu.edu.tw/handle/140.119/69638


    题名: 加值型營業稅中營業人涉嫌虛設行號之實證分析
    其它题名: An Empirical Study of Business Entity Tax Fraud in the Value-Added Tax System
    作者: 林左裕
    Lin, Calvin Tsoyu
    贡献者: 地政系
    关键词: 區別分析;羅吉斯分析;虛設行號;營業稅
    tax fraud;value added tax;logit analysis;dicrriminant analysis
    日期: 2006.06
    上传时间: 2014-09-03 11:44:49 (UTC+8)
    摘要: 本研究以最近幾年所查獲並經地檢署起訴之虛設行號集團為對象,探討其特徵因子,建立區別模型。其目的是求出區別函數,判別營業人是否屬虛設行號,以決定是否須加強查核。首先找出進項金額、銷項金額、進項金額減銷項金額、實繳稅額、加值率、實繳稅額占銷項金額比率、進貨廠商家數、銷貨廠商家數等八個變數,並檢定是否符合多元常態分配、共變異數及兩母體平均數是否相等。其次以逐步區別分析,篩選出進貨廠商家數、銷貨廠商家數兩個變數具有區別能力,並建立區別模型。另外再加上營業人擅自歇業、負責人異常、租用房屋、循環開立發票等四個名義變數,以無母數區別分析、羅吉斯分析,分別建立區別模型。結果其分類正確率均達91%以上。本研究發現進貨廠商家數在三種方法分析中均顯著; 而加值率並不顯著;其計算公式,分母銷貨為零時,電腦程式產出結果亦為零,似有誤,財稅主管機關宜檢討修正。
    Some business entity in Taiwan usually utilizes falsified transaction uniform invoice to aid others in committing the tax evasion, or to add sales amount to make a loan from financial institutes for the purpose of tax fraud in Taiwan. The object of this study is thus to construct a discriminant model for tax authorities to decide whether there is any necessity to enhance investigation on a business entity that prosecuted by public prosecutor of tax fraud.First, eight quantity variables, including the sales amount, the purchases amount, the differences of between sales and purchases, the payment of business tax, the ratio of value added, the ratio of tax payment over sales, and the number of purchasers and sellers, are selected and tested respectively for normality, equality of group means, equality of covariance matrices, by means of stepwise discriminant. As a result, the number of purchasers and sellers are significant in determining the tax fraud. Second, several qualitative dummy variables are added, including de-registration, sponsor, renting business house, purchase-sale uniform invoices cycle. We employ nonparametric and logistic regression in the prediction of binary dependant variables. The overall classification accuracy is approximately 91-95 perc