Academic journal article Electronic Journal of Business Research Methods

Militating against Data Fabrication and Falsification: A Protocol of Trias Politica for Business Research

Academic journal article Electronic Journal of Business Research Methods

Militating against Data Fabrication and Falsification: A Protocol of Trias Politica for Business Research

Article excerpt

1 Background

In the context of research, data fabrication is the creation of bogus data that either supplements or substitutes for genuine research data, while data falsification is the deliberate altering of research data (Herndon, 2016) and is symptomatically no different from data fabrication. Data fabrication or falsification - sometimes euphemistically referred to as data "curbstoning" (e.g. Birnbaum, Borriello, Flaxman, DeRenzi, & Karlin, 2013; Clark & Moul, 2004; Kennickell, 2015) - can occur at various stages in the research process, and can be perpetrated by interviewers, enumerators, surveyors, researchers, respondents, or any others participants in the research. There can be no debate that data fabrication and falsification fall within the definition of research misconduct or questionable research practices (Banks, Rogelberg, Woznyj, Landis, & Rupp, 2016) and there can be no justification or rationale for tolerating them in business research.

It is worth reflecting on the potential impact of data fabrication or falsification on other researchers, and research users. The most direct consequence is that the users of the research (and potentially the researchers themselves) are likely to be misled to assume validity and reliability that the research does not merit. Depending on the extent and nature of the fabrication or falsification, results may be obfuscated, exaggerated, or entirely erroneous. Strategic and operational business decisions, as well as future academic research activities will be misinformed while stakeholders and decision makers remain obliviously vulnerable.

It is acknowledged that measures to detect and militate against data fabrication and falsification are most effective when implemented ex ante or during the data collection process, for example through call-backs and face-to-face re-interviewing (Bredl, Storfinger, & Menold, 2011). However, such measures are not always practicable due to timing, logistical and outsourcing arrangements. The focus of this research is on ex post analysis to detect bogus data, and measures to militate against fabrication and falsification in the dataset under analysis. This research will not consider data fabrication in clinical research as the protocols and consequences are typically different from that of business research.

In the remainder of this article, prior research pertaining to data fabrication are reviewed, case studies illustrating data anomalies indicative of data fabrication or falsification are presented, specific statistical tests that could be used to identify data anomalies are highlighted, and a research protocol to militate against data fabrication and falsification in business research is recommended and discussed.

2 Literature review

It may be useful to begin by discussing the relationship between outliers, extreme values, and fabricated data. An outlier has been defined as data "that appears to deviate markedly from other members of the sample in which it occurs" (Grubbs, 1969, p. 1). Grubbs (1969) is specific that an outlying value may either be a legitimate, extreme value that has occurred as a result of the inherent variability of the variable in which case the value must be retained, or may be an anomaly or error in which case the value may be omitted from the dataset. It is noted that fabricated or falsified data may or may not appear as outliers or extreme values, and therefore procedures for detecting outliers (e.g. Grubbs, 1969) are not appropriate for detecting fabricated or falsified data.

2.1 Motivation for fabricating or falsifying data

Bredl et al. (2011) refer to a number of reasons why interviewers might be inclined to falsify survey data. For example, typically interviewers are not involved in the data processing, and do not have a vested interest in the quality of the data and the research output. Interviewers may not be knowledgeable in research design, sampling protocols and research ethics (American Association for Public Opinion Research, 2003) and may not appreciate their impact on the research. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.