Page 205 - 36-1

NTU Management Review Vol. 36 No. 1 Apr. 2026

subsequent audits to avoid future litigation (Schmidt, 2009). Such conservatism would
lead defendant auditors to issue more GCOs, causing lower Type II and higher Type I error
rates (Geiger and Rama, 2006).
The three theoretical hypotheses are not necessarily mutually exclusive. In fact,

multiple factors interact, although empirically, only the effects of dominant factors can be
detected (Liu and Elayan, 2015). The empirical evidence provided in this study can help
determine which of the three hypotheses is predominantly influential.

2. Design/Methodology/Approach

In this study, I obtain audit partner litigation data from the Securities and Futures
Investors Protection Center (SFIPC) in Taiwan and bankruptcy data from multiple sources

(e.g., Google searches for reasons for firm delisting, material information filings from
the Market Observation Post System, or information about business registration from the
Ministry of Economic Affairs, etc.). Other data are retrieved from the Taiwan Economic
Journal database. The final sample consists of 27,888 firm-year observations of financially

distressed firms that survived between 1999 and 2023 (the nonbankrupt sample) and 233
firm-year observations of distressed firms that went bankrupt in the following year (the
bankrupt sample).
The empirical model for this study is as follows:

＝ ... , (1)
where GCO is coded as one if the firm receives a GCO in the current period, and zero
otherwise. BEFORE and AFTER are the variables of interest. BEFORE equals one if

the firm is audited by a defendant partner before the partner is sued, and zero otherwise.
AFTER equals one if the firm is audited by a defendant partner after the partner has been
sued, and zero otherwise. Controls stands for a vector of control variables.
I estimate Model 1 using data for the nonbankrupt sample to obtain evidence of GCO
Type I errors. Prior studies have shown that a single audit failure can systematically reflect

the lower overall audit quality of the involved auditor (Chang, Chen, Chou, and Ko, 2016;
Francis and Michas, 2013; Li et al., 2017). Accordingly, the coefficient for BEFORE is
expected to be positive for the nonbankrupt sample, indicating that defendant partners are
more likely to commit GCO Type I errors before litigation than nondefendant partners. In

197