Temporal shifts in Americans’ risk perceptions of the Zika outbreak

Abstract

Cross-sectional surveys, despite their value, are unable to probe dynamics of risk perceptions over time. An earlier longitudinal panel study of Americans’ views on Ebola risk inspired this partial replication on Americans’ views of Zika risks, using multilevel modeling to assess temporal changes in these views and inter-individual factors affecting them, and to determine if similar factors were influential for both non-epidemics in the USA. Baseline Zika risk scores – as in the Ebola study – were influenced by dread of the Zika virus, perceptions of a near-miss outbreak, and perceived likelihood of an outbreak. Judgments of both personal risk and national risk from Zika declined significantly, and individual rates of news following predicted slower decline of perceived national risk in both cases. However, few other factors affected changes in Zika risk judgments, which did not replicate in a validation half-sample, whereas several factors slowed or increased the rate of decline in Ebola judgments of the U.S. risk. These differences might reflect differences in the diseases caused by these two viruses – e.g., Ebola’s much greater lethality – but more longitudinal studies across multiple diseases will be needed to test that speculation. Benefits of such studies to health risk analysis outweigh the difficulties they pose.

Keywords:

• risk perception dynamics
• temporal change
• Ebola
• Zika

Acknowledgments

We appreciate comments by two anonymous reviewers. This material is based upon work supported by the U.S. National Science Foundation under Grant No. 1644853.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 It is unclear whether this failure reflects the data used, dominated by official statistics and frequency of mass media coverage. Although the authors noted that the three measures were highly correlated, only three of 18 measures across nine hazards were public opinion data and thus “risk perceptions” (∼1955–1985 unemployment; ∼1956–1986 inflation; ∼1965–1987 fear of crime).

2 Note that comparison of early respondents to late respondents in a single cross-sectional survey, or cross-sectional surveys of different samples from the same population at different times, do not study risk perception dynamics in the sense used here, as results may confound changes over time in population responses with other differences between these different-timing respondent groups (e.g., Ibuka et al. 2010). This does not make these other designs illegitimate, but we must grasp these various designs’ differing implications for exploring dynamics of risk perceptions.

3 Specifically, the first survey in the Ebola panel study (Mayorga and Johnson 2019) was launched 3 weeks after the second death in the U.S. from Ebola (November 18, 2014). The five surveys included 815 (December 8–21, 2014), 748 (January 2–12, 2015), 704 (February 2–9, 2015), 666 (March 16–24, 2015), and 625 (May 8–20, 2015) respondents; all May respondents completed all five surveys, or 76.7% of those recruited in December 2014. Median gaps between these surveys for individual respondents were 25, 31, 27, and 54 days, respectively.

4 Two other approaches raised by Loewenstein and Mather (1990) are moot here. Partial adjustment, in which changes in public opinion lag objective changes in risk, cannot be assessed without a clear way to define “lag” when surveys are infrequent and exhibit varying intervals, and objective measures are currently available only in annual figures. Panic does not apply in the sense of a large departure from objective trends, nor is it clear that there is a departure from overall perception trends.

5 Mayorga and Johnson (2019) suggested this lack of a trend in personal Ebola risk perception might be due to a floor effect as well as greater variance in later waves than seen in the U.S. and global risk ratings: on a 1–6 scale, means over four waves were 2.45–2.60.

Additional information

Funding

This material is based upon work supported by the U.S. National Science Foundation under Grant No. 1644853.