Abstract
In their meta-analysis on how privacy concerns and perceived privacy risk are related to online disclosure intentionand behavior, Yu et al. (2020) conclude that “the ‘privacy paradox’ phenomenon (...) exists in our research model” (p. 8). In this comment, we contest this conclusion and present evidence and arguments against it. We find five areas of problems: (1) Flawed logic of hypothesis testing; (2) erroneous and implausible results; (3) questionable decision to use only the direct effect of privacy concerns on disclosure behavior as evidence in testing the privacy paradox; (4) overinterpreting results from MASEM; (5) insufficient reporting and lack of transparency. To guide future research, we offer three recommendations: Going beyond mere null hypothesis significance testing, probing alternative theoretical models, and implementing open science practices. While we value this meta-analytic effort, we caution its readers that, contrary to the authors’ claim, it does not offer evidence in support of the privacy paradox.
Highlights
In their meta-analysis on how privacy concerns and perceived privacy risk are related to online disclosure intention and behavior, Yu et al (2020) conclude that “the ‘privacy paradox’ phenomenon (...) exists in our research model” (p. 8)
In a recent meta-analysis on how privacy concerns and perceived privacy risk are related to online disclosure intention and behavior, the authors conclude that “privacy concern cannot significantly affect disclosure behavior, which confirms that the ‘privacy paradox’ phenomenon [. . . ] exists in our research model” (Yu et al, 2020, p. 7f.)
We introduce and illustrate these two approaches using the data from Yu et al (2020) and from Baruh et al (2017), which is another meta-analysis on the privacy paradox
Summary
The results reported by the authors are erroneous and implausible. First, because leaving out a relationship in a path model effectively constrains it to zero (Kline, 2016), which is unlikely in most cases (e.g., Orben & Lakens, 2020), model fit should increase if we add another path (Kline, 2016). Granted, it is a very small effect (see below) – but if we just use the pvalue for hypothesis testing (e.g., p < .05), which is the approach used in the paper, the conclusion would be to reject the privacy paradox. By focusing on the residual direct effect of privacy concerns on disclosure behavior, the authors in effect claim that risk perceptions, treated as a confounding variable, have no theoretical or empirical role in the privacy paradox framework. This case, presents evidence that leads to the acceptance of the privacy paradox hypothesis (i.e., privacy concerns are unrelated to disclosure behavior)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
