Understanding the Effects of Stigma Messages: Danger Appraisal and Message Judgments

ABSTRACT

Media coverage of health issues has been criticized for creating health stigmas. The model of stigma communication (MSC, Smith, 2007) provides insights into why this is so, but it has two problems: Some of its mediators have not been supported, and it does not do a good job of predicting the transmission of stigma messages (i.e., social transmission). We present a revised model of stigma message effects in which exposure to stigma messages leads to stigma beliefs and stigmatization as a result of a person-oriented danger appraisal. In addition, message judgments—shock value and common ground—are introduced as mediators of the relationship between danger appraisal and social transmission. Participants (N = 200) were randomly assigned to read a health story written either with or without the intrinsic features of stigma messages. The revised model of stigma-message effects was supported: Reading a health news story written with (vs. without) the intrinsic features of stigma messages resulted in greater danger appraisal, which directly predicted stigma-related outcomes and indirectly predicted social transmission through message judgments. Social transmission varied by message judgment: Shocking messages were shared in ways that facilitate diffusion, but common ground messages were shared with influential others, suggesting different means by which stigma as a collective norm may emerge from interactions among community members.

Funding

The research reported in this publication was supported by the National Human Genome Research Institute of the National Institutes of Health under Award Number R21HG007111 and the National Institute on Drug Abuse under Grant P50-DA010075. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Drug Abuse or the National Institutes of Health.

Notes

¹ Etiology should not be confused with susceptibility (e.g., Witte, 1992). Etiology describes why one has the stigmatized condition. Susceptibility describes the likelihood of contracting a problematic condition. Perceived susceptibility is the belief in one’s risk of contracting the problematic condition (Witte, 1992). An experimental test of stigma message content showed that the level of perceived susceptibility was the same regardless of etiology (Smith, 2014).

² Although marks, labels, etiology, and peril may be qualities of health conditions, communicators can choose whether to include or emphasize this content in their messages.

³ For participants who self-identified with one of these four nodes (d, g, h, or i), the expected likelihood of selecting the other node with the same centrality was 1/18, whereas it was 2/18 for those who did not identify with one of these nodes. To address the conditional nature of this measure, we allowed those who self-identified with one of the highest eigenvector centrality nodes (d or g) to select the other (unchosen) node as well as the node with the next highest eigenvector centrality. We used the same procedures for coding the closeness centrality selections.

⁴ Some scholars (Anderson & Gerbing, 1988; Boomsma, 1982) recommend 150-200 as the minimum sample sizes for structural equation models. By including the measurement model, we have more free parameters to estimate than if we had used composite scores. We ran a path model with composite scores, which allowed us to estimate fewer free parameters and to meet the 10:1 recommendation for the ratio of sample size to free parameters (e.g., Kline, 2005); the model generated regression estimates virtually identical to what is presented herein.

In addition, to further test the causal model, we empirically compared reversing the causal order between the message induction, the message judgments, and danger appraisal (instead of our original order of message induction, danger appraisal, and then message judgments). We used the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) to compare the models. Differences in the AIC of more than 3 (Burnham & Anderson, 2002) and in the BIC of more than 6 (Kass & Raftery, 1995) suggest that the models are not equivalent, and the model with the smaller AIC or BIC is considered superior. The estimates for our original model are AIC = 1,956.21 and BIC = 2,246.47; the reversed model estimates are AIC = 2,021.46 or BIC = 2,315.01. The AIC difference is 65.25; the BIC difference is 68.54. Our original model produced the smaller AIC and BIC estimates. These empirical results suggest that our theorized causal order better fit our data.

Additional information

Funding

The research reported in this publication was supported by the National Human Genome Research Institute of the National Institutes of Health under Award Number R21HG007111 and the National Institute on Drug Abuse under Grant P50-DA010075. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Drug Abuse or the National Institutes of Health.

Notes on contributors

Rachel A. Smith

Rachel A. Smith (Ph.D., Michigan State University) is a Professor in the Department of Communication Arts and Sciences at the Pennsylvania State University. Xun Zhu (M.A., Michigan State University) is a doctoral student in the Department of Communication Arts and Sciences at The Pennsylvania State University. Edward L. Fink (Ph.D., University of Wisconsin-Madison) is the Laura H. Carnell Professor of Media and Communication at Temple University.