本文根據惠譽的銀行個別評等（Fitch's Bank Individual Ratings，FIRs），利用2002至2005年全球91國3881個銀行觀察值為樣本（其中3433個觀察值為付費以取得之信用評等，448個觀察值獲得的是不請自來的信用評等），分析兩類型銀行之信評的相對績效高低來判斷不請自來的銀行信評是否相對低於付費（主動）以取得之信用評等。我們應用Ruin（1973）與Rosenbaum and Rubin （1983，1985a，b）的配對方法，透過將兩類型銀行樣本之財務特性修正至近似，以改善既有文獻中未處理的樣本選擇偏誤問題。我們發現在樣本配對前，獲得不請自來信評之銀行的許多財務績效指標相對低於付費取得信評之銀行，這解釋了為什麼前者之信評相對較低的原因；在樣本配對後即控制了兩類型銀行的財務特性之後，大部分的證據顯示不請自來的銀行信評仍然相對較低。因此本文發現，即使考慮並處理了樣本的選擇偏誤，不請自來的銀行信評仍然存在向下偏誤的現象。
This paper examines whether there is downward bias in Fitch's unsolicited bank credit ratings relative to solicited ratings using a matching approach. The sample for this study consists of those banks that had Fitch's Bank Individual Ratings (FBRs) during 2002-2005. There are 3,881 observations from 91countries, including 3433 solicited ratings and 448 unsolicited ratings. By employing marching methods developed by Rubin (1973) and Rosenbaum and Rubin (1983, 1985a,b), four matching algorithms, Nearest, Caliper, Mahala and Mahala Caliper, are used to match the financial characteristics of two groups of banks in order to correct for sample selection bias. Our evidence shows that before sample matching, banks received unsolicited ratings have weaker financial profiles and this could partially explain why they also get lower ratings. Based on after-matching samples, most of the evidence suggests that unsolicited ratings are still biased downward after controlling for differences in key financial characteristics.
