Consumers’ Privacy Concern and Privacy Protection on Social Network Sites in the Era of Big Data: Empirical Evidence from College Students

Abstract

Information privacy and disclosure have been prominent issues revolving around social media. We adopted communication privacy management theory, the persuasion knowledge model, and the technology acceptance model and conducted a survey with 526 subjects and examined their privacy management on Facebook and the conditions upon which their decision to reveal or withhold private information was contingent. The results showed that consumers set up different privacy boundaries for different types of personal information. Social identity information and daily life and entertainment information tended to be shared more freely, while personal contact information was mostly withheld. Knowledge of and concern regarding technology ubiquity and companies’ business strategies involving big data were the strongest predictors of privacy protection behavior and privacy settings on Facebook. Online trust and Facebook intensity also interacted and jointly predicted privacy concerns. This study brought to researchers’ attention how big data are being used by marketers to target consumers.

Keywords:

• Facebook
• big data
• ubiquity
• privacy
• communication privacy management

Notes

Notes

1 A four-point Likert scale was adopted (instead of a five- or seven-point Likert scale) because previous marketing research has shown that, when measuring respondents’ attitudes, eliminating the midpoint may minimize social desirability bias, which arises from respondents’ desire to please researchers or to avoid giving what they perceive to be a socially unacceptable answer (Garland 1991). Nunnally, Bernstein, and Berge (1967) favor eliminating the neutral category in psychometric questions because more answering options may confuse or irritate respondents. Researchers have also observed that when large numbers of respondents choose the midpoint, the results are less likely to reach statistical significance (Clason and Dormody 1994). Therefore, some researchers use four-point Likert scales, deleting the midpoint (Clason and Dormody 1994; Linacre 2002).

2 Varimax rotation method was chosen because it is the most common rotation method of principal component analysis. Varimax rotation maximizes the sum of the variances of the squared loadings of a factor on all variables and keeps the factor loading matrix a simple structure (Dunteman 1989).

3 The Bonferroni correction method was adopted to control Type I error due to multiple comparisons. The acceptance familywise error (.05) was divided by the number of tests (n = 3) for each test (p = .017).

Additional information

Notes on contributors

Wenjing Xie

Wenjing Xie (PhD, University of Maryland College Park) is an associate professor of communication, Communication Department, School of Communication and the Arts, Marist College.

Kavita Karan

Kavita Karan (PhD, University of London) is a professor, School of Journalism, Southern Illinois University Carbondale.