學術產出-NSC Projects

Article View/Open

Publication Export

Google ScholarTM

政大圖書館

Citation Infomation

  • No doi shows Citation Infomation
題名 微調BLP: 修改兩項執行BLP需求估計程序時常見的問題
其他題名 Incomplete Price Data and Misspecifications of Covariance Matrix of Random Coefficients in the BLP Model: Simulation Studies and Empirical Applications
作者 胡偉民
貢獻者 財政系
日期 2013
上傳時間 20-Apr-2016 15:41:27 (UTC+8)
摘要 本計畫旨在處理兩個在進行 BLP 估計(現在常用的一種用於估計異質性產品市場需 求的隨機係數離散選擇模型,由 Berry, Levinsohn 與 Pakes 於 1995 年提出,故在此我們 稱之為 BLP 模型)時常面對的問題:『變數值缺失』,與估計出的隨機係數標準差為負值。 為處理『變數值缺失』的問題,我們使用 complete case analysis 的特性,部分的修改了 BLP 的估計過程,而可以使在變數值(特別是價格)缺失的情況下仍能獲得參數估計的 一致性。處理『負標準差』的問題時,我們認定這個問題來自於隨機係數間相互獨立的 設定,而提出一套完整估計隨機係數的協方差矩陣的方法,用以避免估計出負的標準差 的可能性。此外,數值模擬的結果顯示,將 BLP 模型的隨機係數誤設成相互獨立的關係, 可能使參數的估計偏誤,特別是得出負的標準偏差;而即使真實的隨機係數間的相關性 非常小,誤設也可能造成此結果。最後,我們經由了蒙特卡羅實驗來驗證我們的理論結 果和猜想,並使用中國新車市場的汽車銷售資料進行實證分析。
This project intends to resolve two problems commonly encountered while implementing the BLP procedure: the missing value problem and the negative standard deviation estimated through the optimization procedure. To deal with the missing value problem, we utilize an "unbalanced BLP" procedure, which is a modified BLP, can be used to get a consistent estimate of demand parameters when data on some characteristics are not complete. To handle the problem of “negative standard deviation”, we specify and estimate a model with correlated random coefficients which can circumvent the problem that arises from the specification of independent random coefficients. Furthermore, we numerically show that a misspecified BLP model with independent random coefficients can result in biased and nonsensical estimates of the parameters, such as negative estimates of the standard deviations, even if the correlation among the random coefficients are very small. Finally, we conduct Monte Carlo experiments to validate our theoretical results as well as conjectures and make an empirical application using data on the Chinese new car market.
關聯 計畫編號 NSC 102-2410-H004-010
資料類型 report
dc.contributor 財政系
dc.creator (作者) 胡偉民zh_TW
dc.date (日期) 2013
dc.date.accessioned 20-Apr-2016 15:41:27 (UTC+8)-
dc.date.available 20-Apr-2016 15:41:27 (UTC+8)-
dc.date.issued (上傳時間) 20-Apr-2016 15:41:27 (UTC+8)-
dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/85700-
dc.description.abstract (摘要) 本計畫旨在處理兩個在進行 BLP 估計(現在常用的一種用於估計異質性產品市場需 求的隨機係數離散選擇模型,由 Berry, Levinsohn 與 Pakes 於 1995 年提出,故在此我們 稱之為 BLP 模型)時常面對的問題:『變數值缺失』,與估計出的隨機係數標準差為負值。 為處理『變數值缺失』的問題,我們使用 complete case analysis 的特性,部分的修改了 BLP 的估計過程,而可以使在變數值(特別是價格)缺失的情況下仍能獲得參數估計的 一致性。處理『負標準差』的問題時,我們認定這個問題來自於隨機係數間相互獨立的 設定,而提出一套完整估計隨機係數的協方差矩陣的方法,用以避免估計出負的標準差 的可能性。此外,數值模擬的結果顯示,將 BLP 模型的隨機係數誤設成相互獨立的關係, 可能使參數的估計偏誤,特別是得出負的標準偏差;而即使真實的隨機係數間的相關性 非常小,誤設也可能造成此結果。最後,我們經由了蒙特卡羅實驗來驗證我們的理論結 果和猜想,並使用中國新車市場的汽車銷售資料進行實證分析。
dc.description.abstract (摘要) This project intends to resolve two problems commonly encountered while implementing the BLP procedure: the missing value problem and the negative standard deviation estimated through the optimization procedure. To deal with the missing value problem, we utilize an "unbalanced BLP" procedure, which is a modified BLP, can be used to get a consistent estimate of demand parameters when data on some characteristics are not complete. To handle the problem of “negative standard deviation”, we specify and estimate a model with correlated random coefficients which can circumvent the problem that arises from the specification of independent random coefficients. Furthermore, we numerically show that a misspecified BLP model with independent random coefficients can result in biased and nonsensical estimates of the parameters, such as negative estimates of the standard deviations, even if the correlation among the random coefficients are very small. Finally, we conduct Monte Carlo experiments to validate our theoretical results as well as conjectures and make an empirical application using data on the Chinese new car market.
dc.format.extent 1386961 bytes-
dc.format.mimetype application/pdf-
dc.relation (關聯) 計畫編號 NSC 102-2410-H004-010
dc.title (題名) 微調BLP: 修改兩項執行BLP需求估計程序時常見的問題zh_TW
dc.title.alternative (其他題名) Incomplete Price Data and Misspecifications of Covariance Matrix of Random Coefficients in the BLP Model: Simulation Studies and Empirical Applications
dc.type (資料類型) report