In a previous discussion I described how social harassment costs (Borland and Pulsinelli, 1989) might explain why some consumers could hold seemingly contradictory views about science (i.e. accepting certain scientific views related to global warming but rejecting other scientific views related to genetically modified foods).
In my graduate school research I hypothesized that consumers adopt a worldview v (regarding climate change, food preferences, religious beliefs, public policy, etc.) that gives them the greatest level of utility seemingly invariant to evidence supporting some alternative worldview v'.
U(v) > U(v') (1)
One way to to explain this would be to model utility as a function of social harassment 'c'.
U(v, c) > U(v', c) (2)
for c > k
U(v, c) < U(v', c) (3)
for c < k
In this formulation social harassment provides disutility, and would enter the utility function as a negative term. If social harassment is great enough to exceed some threshold 'k', consumers with preferences like those above may choose to ignore scientific evidence that lowers utility by conflicting with their vision or the vision of their peers. The level of 'k' may vary depending on the consumers sensitivity to social pressure.
Some of the implications of this model were that consumers might increase utility and reduce social harassment by avoiding information that conflicts with their world views, they might also seek information that supports utility maximizing views regardless of weight of evidence.
This also seemed to align with a number of ideas supported by findings from behavioral and public choice economics (Caplan, 2007; Kahneman, 2011). For example the idea that beliefs that are irrational from the standpoint of truth-seeking are rational from the standpoint of utility maximization (Caplan, 2007).
In graduate school I attempted to investigate this empirically by developing a survey instrument to measure preferences toward genetically modified foods as well as attitudes toward abortion, climate change, embryonic stem cell research, animal welfare as well as political ideology, education levels, and science knowledge. I found that respondents with a positive view of embryonic stem cell research and those that were more concerned about the impacts of climate change were less likely to accept the safety of genetically modified foods. This is in spite of evidence of the safety of biotechnology or its potential for mitigating the impacts of climate change. However, the sample size was very small and as noted elsewhere a better instrument and structural equation modeling approach might offer a much richer and more rigorous understanding of the latent factors shaping consumer perceptions.
Additionally, the behavioral theoretical utility model above is very general. While this model's predictions could be loosely supported by the empirical work, many untested assumptions remain. For instance, the level of social harassment 'c' and the threshold 'k'. These are abstract latent factors hard to estimate and validate empirically.
However, if we think of social harassment being a function of our exposure to media, social media, and peers, we can begin to frame up an analytical strategy for better understanding these phenomena in the context of social network analysis (SNA). For example, assume two actors, 'A' and 'B' who have preferences similar to (2) and (3) above. And assume a simple network of connections with peers as depicted below:
Each node (depicted as black, white or grey dots above) represents a peer's sentiment toward genetically modified foods. For subject A, strongly influenced by peers with negative sentiments, we might hypothesize that the social harassment costs associated with believing in the safety of biotech crops could be high even in the face of strong scientific evidence (which they may not be aware of, discount highly, or avoid in order to maximize utility). For subject B, social harassment costs in relation to these beliefs might be much lower and likely be imposed rarely by a few peripheral connections. This is just a toy example, but this framework helps motivate a number of questions:
- How exactly should these networks be defined and constructed to properly frame the question/hypothesis I have? Who/what entities should each node represent (people, media outlets, websites, celebrities, scientists, etc.)?
- Connections between nodes are referred to as edges and represent pathways through which information and social harassment costs might flow - should different edges be given different weights as a function of the entity represented by each node? Are there interactions between the type of node and the type of information flowing from it?
- Is there any correlation between network metrics (i.e. degree centrality, eigenvector centrality) and influence on preferences/perceptions?
- What can we learn from previous research in SNA in the area of viral marketing? Are there key nodes that can be influenced?
- What role does network architecture play in information diffusion, influence, and ultimately the level of social harassment costs of a given node (ultimately this is what I would want to quantify to empirically support the theoretical model above)?
- Are there causal inferential approaches with the necessary identification properties allowing us to interpret these effects causally? (see perhaps Tchetgen et al., 2020)
Borland,Melvin V. and Robert W. Pulsinelli. Household Commodity Production and Social Harassment Costs.Southern Economic Journal. Vol. 56, No. 2 (Oct., 1989), pp. 291-301
The Myth of the Rational Voter: Why Democracies Choose Bad Policies. Bryan Caplan. Princeton University Press. 2007
Johnson, N.F., Velásquez, N., Restrepo, N.J. et al. The online competition between pro- and anti-vaccination views. Nature 582, 230–233 (2020). https://doi.org/10.1038/s41586-020-2281-1
Kahneman, D. (2011). Thinking, fast and slow. New York: Farrar, Straus and Giroux.
Eric J. Tchetgen Tchetgen, Isabel R. Fulcher & Ilya Shpitser (2020) Auto-G-Computation of Causal Effects on a Network, Journal of the American Statistical Association, DOI: 10.1080/01621459.2020.1811098
SNA and Related Posts at EconometricSense:
Perceptions of GMO Foods: A Hypothetical Application of SEM
An Introduction to Social Network Analysis with R and NetDraw
GMM, Endogeneity, SNA, Viral Marketing, and Causal Inference
Using SNA in Predictive Modeling
All SNA Posts at EconometricSense
Related Posts and Background at EconomicSense
Consumer Perceptions of Biotechnology: The Role of Information and Social Harassment Costs
Fat Tails, the Precautionary Principle, and GMOs
Defining Consensus Regarding the Safety of Genetically Modified Foods
Comments of Rule for Rules on Gene Editing Technology
