Detecting Falsified Cases in SCF 2004 Using Benford's Law

• Activity Number: 239
• Type: Contributed
• Date/Time: Tuesday, August 9, 2005 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Yongyi Wang*+ and Steven Pedlow
• Companies: NORC at the University of Chicago and NORC at the University of Chicago
• Keywords: falsification ; survey data ; curbstoning interviews ; validation

Abstract

Despite the best efforts and high standards NORC uses to hire interviewers, there is still the risk that interviewers will falsify cases. This study analyzes Survey of Consumer Finances (SCF) 2004 data to ascertain whether Benford's Law can be used to detect cases that had already been determined to be falsified (through "failed" validation). Benford's Law asserts that for any set of continuous variables, the leading digits will follow a certain distribution. We used financial variables from the SCF 2004 questionnaire and compared how closely they followed the theoretical Benford's Law distribution and several modifications of Benford's Law. Each method's effectiveness is judged by how well it segregates the bad interviews and interviewers from the good interviews and interviewers. Even though the large amount of rounding done by the respondents made using the Benford's Law distribution ineffective at spotting suspicious interviews or interviewers, using the distribution of first digits from all cases did work better at segregating the "bad" interviewers, especially under some modifications we tested.