Research Synthesis and Design: Neuroeconomics and Nonmarket Valuation
Carl Mackensen
Kathy Newcomer
Research Methods
Research Synthesis and Design
Summer 2022
I: Introduction
Climate change is a pernicious and ever evolving issue. In order to address the issue, we need to encourage action across the entire strata of society, from individuals to elected officials. The primary goal for this piece is the examination of how we can best motivated citizens to take up the cause of climate change, and hand a mandate to elected officials who would follow through on the subject. The research questions to be addressed include those on what types of cues people respond to, how do demographic groups differ, which brain regions are associated with the presentation and analysis of climate related information, and whether social desirability plays a role. The design of the study is straight forward. A random sample will be taken from the general population, and be presented with a rank ordering task for types of investment. The information will be said to come from different sources, in order to see which source elicits the greatest response. This will be done in an MRI scanner and result in fMRI scans of brain activity. The survey while in the MRI will be a straight forward ranking exercise, with investment in climate change mitigation and prevention compared to other expenditures. In addition, participants will provide demographic information, and participate in lengthy qualitative interviews to better understand where they are coming from. The analysis will consist of a simple regression used to compare the coefficients on dummy variables for each demographic, and treatment group, where each treatment was told the information was provided by a different source. For potential limitations, I hone them down to honesty of responses, causality (whether the read prompt actually causes the response, or if there is an intervening omitted variable), and generalizability, as this study, while random, will only represent the USA, and for full generalizability should be expanded to the entire world, particularly every country that relies on fossil fuels. The next section is a literature review of relevant articles, followed by the conclusion.
II: What Problem is being Addressed?
Climate change is the increase of severe weather outcomes, such as droughts, floods, fires, sea level rise, and more, caused by anthropogenic causes, specifically the burning of fossil fuels that emit carbon dioxide and methane into the atmosphere and warm the planet through the greenhouse effect. This has been occurring since the start of the Industrial Revolution, and the advent of the burning of coal for power and heat. Climate change has reached such a severe point that, should action not be taken to limit the rise in temperatures to 1.5 degrees Celsius, the effects could very well endanger not just all of humanity, but all life on the planet. Action must be taken within a very small window in order to do this, ideally before 2030, with drastic cuts to emissions. This issue poses a classic collective action case, in which there is little motivation for individual action because others have the potential to free ride. If America drastically cuts emissions, but Vietnam does not, Vietnam would be better off at least in the short term, because it would benefit from cheap, dirty fuel sources while emissions would still go down overall. Action is further complicated by the fact that greenhouse gases are a global pollutant, and require action by all emitters to reach the stated emissions targets, unlike other types of pollution such as sulfur dioxide, which is regional and can be addressed by a single country.
For us to make true change, everyone has to change their behavior. We need to not only change the composition of our energy sources to renewable means like solar, wind, and hydro, but also change how we heat and cool our homes, how we transport people, what we eat, and more. In order for public officials to make change on the issue, they need a clear mandate that change is desired. Right now, change is stymied by a number of Republican areas in the USA that still have key constituencies employed by fossil fuel industries. As such, despite the provision of the most up to date science available, elected officials are still hesitant to make change. Some go so far as to outrightly deny the existence of climate change, calling it a ‘Chinese Hoax’ meant to derail the economy. In order for meaningful laws to be passed, the public needs to not just be made aware of the issue, but they have to actively call for action. This is difficult, as people and legislators both have competing priorities and desires. A dollar spent on mitigating climate change is a dollar not spent on anything else.
How, then, can we influence both the public and legislators to take up the issue? A number of groups are attempting to do just this, but in order to influence the public, and thereafter legislators, we need to examine the best means available to effect change.
III: Research Questions
The research questions that I will attempt to address in this study are directly influenced by the goals outlined in the previous section, namely, how best to influence people to care about climate change. To wit, they include the following.
1) What types of informational cues influence people to choose to act to address environmental problems caused by climate change?
2) How does race, age and gender affect how likely people are to choose to act to address environmental problems caused by climate change?
3) What brain regions are associated with the act of weighing environmental investment options?
4) How prevalent is the threat of social desirability on responses to questions about addressing environmental problems caused by climate change?
These questions seek to best understand how we can make a difference. If different cues prompt people to address climate change, they should be tailored to the individual type of person as to be the most effective. Similarly, if there are demographic differences in how people respond to messaging such as broken down by race, age, gender, and so on, then this too should be considered when reaching out to people. The reason why brain regions are particularly interesting in this analysis is complex. Firstly, we would be able to tease out whether someone is lying in their responses, as regions for lying are known and established. Then, we can omit these responses, or follow up with the individual participant and question them further. Secondly, brain regions are interesting as they are very specific in terms of function. We may find that all respondents have a particular region activate for say the valuation process, and as such we would in essence be able to peer into their decision making process. Alternatively, brain regions for traditional values may light up, which would correspond to different outcomes. Brain region analysis is truly coming of age, and hasn’t been used in this context yet. Lastly, for social desirability, we would be able to again via the brain regions see if a comparison is being made, or if the respondent participant is acting of their own values.
IV: Type of Design
As for the purposes of this paper I have unlimited means, access, and abilities to perform my dream analysis, the type of design seeks to examine with the best available measures exactly how we can best motivate people to make change. As such, a randomized group will be selected from the general public for a randomized control trial. This could be done through a lottery system. The question of how to motivate people to actually participate falls outside of the realm of this paper, but conceivably there could be some benefits associated with participating. This could be the prestige of working on such a high profile study, or monetary benefits for participation. However I wouldn’t want the rewards to be so completely outsized that those who don’t participate would feel negatively towards participants. Again, however, this falls beyond the purview of this paper, and we can assume that this randomly selected group would be motivated to participate.
Participants will be split into several groups, each of which will watch a different prerecorded message. The message will be said to originate from either scientists, government officials, activists, or celebrities. After this, participants will be placed in a Magnetic Resonance Imager (MRI) scanner. Functional MRI (fMRI) scans will take place while participants complete a brief series of questions. This allows us to see which area of the brain is working while answering each question.
I take ethics very seriously in considering the type of design to be used in my proposed study. The most pressing ethical method that is available to me, drawn from medical literature and practice, is that of informed consent. Firstly, participants would be made aware of the benefits and costs of participation. Costs would be perhaps boredom from being stationary in an MRI, and some discomfort from being in the machine for probably roughly 30 minutes. The benefits would include adding to critical knowledge, as well as other possible fringe motivators I described above, such as monetary or prestige. Again, though, we can assume that participants will readily take part, and do so earnestly. Aside from the costs and benefits and informed consent, this study should be double blind, with no one who administers it knowing which of the treatments (here, the origin of the message) the participant has been assigned to. All individuating data should be stripped from the final data. In my dream world scenario this would be a project that would fall under the ‘big data’ heading, and as such include the participation of thousands of people not just in the USA, but around the world. However, for a pilot study like the one I am proposing, we can assume that the initial analysis would be restricted to a random sample of Americans and USA residents.
V: The Survey
The survey will be administered while the participants are in the MRI, and fMRI scans are being conducted. Ideally, it would be as short and to the point as possible, so as to not induce fatigue that would influence answers (perhaps people would not take things seriously and just seek to finish the survey as fast as possible to get out of the scanner). In order to accurately gather all of the relevant information, as well as do this in a timely manner, long thought will have to be given to the exact nature of the survey. Inside the scanner, participants have access to a screen that they can see information displayed on, as well as audio through ear phones, though this can be hard in practice as MRIs are relatively loud machines and often participants in scanners wear ear plugs. As such, we would rely on visual information displayed to them. Participants also have access to a small hand held controller that they can operate while in the scanner, in order to get their responses to visually presented information.
First, we will ask whether the participant leans politically left, right, or center. Then, we will ask a series of trade off questions about expenditures on environmental issues. For example, this could take the form of ‘your state has a budget surplus of $1,000 per capita. Should this be spent on climate change mitigation, or public schools?’ This would allow us in the analysis to rank order, as well as put an average price on, how much environmental expenditures are valued.
VI: The Analysis
Following the survey and fMRI scans, I will analyze whether there are any demographic differences among participants in terms of their responses. This would take the form of t tests of averages of or regressions with dummy variables coded on demographic information such as race, sex, age bucket, political affiliation, and so on. This aspect of the analysis should prove straight forward enough in terms of the actual statistical analysis. Ideally, a regression with expenditure on environmental investment would be the dependent variable, and all other information would be coded as independent variables. For the demographic analysis, therefore, it simply becomes a matter of looking at the coefficients on the regression independent variables of interest, as well as their significance levels.
Then I will examine which brain regions are active during the survey. This is the unique nature of fMRI scans, in that they show, through measuring the oxygenation of regions of the brain during the survey in the scanner, which ones are active. I will compare the different groups in terms of their responses. We will be able to rate things like source credibility, including confidence, reliability, truthfulness, and how responses are affected. It is increasingly well established which regions of the brain are responsible for different aspects of experience in a human being. Using this as a second analytical tool would be highly valuable, and be able to give us some insight into the thought pattern of each individual participant, perhaps even beyond what they know of themselves, as well as in the aggregate which regions were turned on or off.
These analyses would answer the first three research questions. The final one, on social desirability, is a little bit more tricky to detangle. The best way to answer this research question would be to compare the differences in what the participant said would be the desired environmental investment with brain regions that light up for the cognition types of self worth or self pleasure. This would prove somewhat challenging, as it is a merging of the two analytical methods that I have kept deliberately separate to date in order to facilitate ease of analysis, but it is not impossible. Firstly, we would find those for whom brain scans indicated they were acting out of social desirability pressures in their responses. Then we would run the same regression outlined above, and see how much they say that they value the environment. This would begin to get us a toe hold into how much these professed environmentalists claim they would like to see change, versus their more sincere counterparts, who would display different brain scans and have different responses.
I would supplement these analyses with as much qualitative work as possible. If the sample is smaller, and time and budget is not an issue, I would send trained interviewers into the participants homes for one on one interviews either before or after the scan, which too would be randomized, as it could affect responses. They would ask questions that would get a full portrait of the individual in terms of their fiscal priorities, experiences with environmentalism, and investment experiences, as well as attitudes on government programs or private sector initiatives, what educational and media sources they regularly consult, how much they value future generations compared to present consumption, and whether they have had any experience with climate related hardship, such as natural disasters. Again, this would be deindividuated so as to protect participants, with such qualitative work only being tied to the number assigned to each individual for the scanning. Interviewers would not know which treatment the individual would be assigned to, and the interviews would randomly take place either before or after the scan and survey. This pre and post treatment interviewing, which is randomized, would add another aspect to the study that we could analyze, namely, seeing whether the interview itself contributed to any changes in attitudes of the participants, all else held constant.
VII: Potential Limitations
I have attempted to make this study as comprehensive as possible in terms of how it is conceived, run, and analyzed. However, as with any human endeavor, particularly new ones, there is the potential for limitations inherent in the activity. Firstly, there is the limit of what we can technically know. This applies to all levels of this project, from the qualitative interviews, to the surveys and self reported data, to the fMRI scans. All are limited in terms of what we can actually infer. Brain scans only go so deep, so to speak, and while some regions are known to be associated with certain thoughts or behaviors, others remain mysterious and only generalities can be spoken of. Also, for a self reported survery, there is always the question of how honest participants will be. I have attempted to address the social desirability aspect, but participants may attempt to deceive or misrepresent themselves in other ways, without even consciously knowing that they are doing so. An additional aspect is that of bias. Different people in the different treatments of the authors of the environmental statements they read could be coming to the study with markedly different outlooks and assumptions about the groups they are assigned to (celebrities, activists, government officials, or scientists). How this information is reported to them could significantly impact the nature of their responses, and thereafter both the statistical analysis as well as the brain scans.
Next there is the question of causality. This is all important in terms of designing something that can be applied to the real world. If the goal is to learn how we can have the biggest impact on climate change through different people presenting the same information, whether the source actually causes the difference between treatment groups may be up for debate. There could be some intervening omitted variable that biases the responses, and analyses. The set up of the study may be such that we internalize this bias, and as such have faulty conclusions, and cannot actually do anything to address the stated issue of clmate change and people’s beliefs about it. I attempt to get at this by having both the quantitative approach of a randomized trial with different treatments, a double blind study, the fMRI scans, and the qualitative interviews, but I’m sure there would be additional things that I have missed.
Last, there is the issue of generalizability. Climate change is a global problem, and requires the active participation of everyone on the planet. Free riding, as stated above, is pernicious, and could infect even the most well meaning of people. If this study, despite being randomized with dfferent treatments and different types of analyses, is restricted to the USA only, as I have proposed for this trial analysis, it would fail to consider the multicultural validity argument, and hence generalizability would be sacrificed. Ideally, this same study should be run in every country in the Global North that is responsible for emissions. As such, different cultures may value different sources of information, or have different brain regions light up on average, or have different back stories in interviews for why they do or do not care about and value the environment. Doing this study globally would allow us to tailor responses and recommendations to each specific area, as there are bound to be differences. Similarly, if we want to initially focus on just the USA, we could do this study specifically for different demographic groups individually, and run the randomized control trial within each group. The responses of and results from Native Americans may be vastly different than those of New England Conservatives, and so on. Again, both the macro and micro administration of this study depend on time and money being available, as well as participants who are willing to take part, but for the purposes of this paper, we can assume that these issues are not present. How something appeals to a person may be highly personal, and vary from person to person. This study may give some general ideas, but is not a panacea.
Before moving on to the reviewing of relevant articles, it also makes sense to examine whether this proposed study meets the standards of different types of validity and reliability. Those to be examined include measurement validity and measurement reliability, internal and external validity, statistical conclusion validity, and multicultural validity. In terms of measurement validity and reliability, I believe that the way the study is set up will return measurements that are both valid and reliable. However, there could be threats to this. If the causal mechanism is not straight forward and there is an omitted variable, or if different people have different neural pathways for how they value things or prioritize things, then these measures could suffer. I believe that I have accounted for this by including the fMRI method, as it accounts for these issues and would allow us to see further what the situation is. While internal validity would be strong for these reasons as well, external validity would suffer because we are limiting ourselves to the USA. This holds true for multicultural validity as well. This can be ameliorated by, as I said, conducting the same study in other countries that currently do and historically have used fossil fuels. Lastly, for statistical conclusion reliability, I believe the analysis is statistically straight forward and conclusion would follow from the analysis detailed. However, again, causality could be an issue. In order to remedy this, we would simply need to control for as many variables as possible in the analysis, which would inform the pre test demographic study, and the post test qualitative interview.
VIII: Literature Review
I conducted a review of six articles related to this study. Predominantly, they were fMRI studies, with a few additional modalities and types of analysis. The first was Neural signatures of betrayal aversion: an fMRI study of trust. Here, people make either 'risky' decisions when facing uncertainty, or 'trusting' ones where they rely on someone else. Differences between these decisions are due to 'betrayal aversion', or not wanting to be betrayed by someone you trust. The research question was how significant is this, and what area of the brain is associated? The research design was while in an fMRI, participants made 82 binary trust decisions of different stakes, choosing to trust or not trust, and then payouts were calculated. The data collection techniques were fMRI scans and game results. 30 investors were sampled. Tabulation of investment winnings by demographic group was done, and the key findings were that significantly more trust was observed when betrayal aversion does not influence decision making. Activity in the right anterior insular cortex was found, as well as the medial frontal cortex and right dorsolateral prefrontal cortex. Measurement validity and reliability were not issues, nor was internal validity, but external validity was an issue as only 30 investors were sampled, which is not representative of the public. The statistical conclusions were valid, but there was little multicultural validity as the sample was not random or representative.
The second article was The neuroscience of investing: fMRI of the reward system. The primary question is what are the correlates of reward system brain regions with experimental behavior on investment, and can a model be made? Subjects played a monetary incentive delay task where repeated trials take place where they make or lose money depending on their ability to pay attention and react quickly. This takes place in an MRI. The results were fMRI scans and game results, and the population was Stanford students. It was found that the NACC and MPFC are associated with rewards, while impulsivity and motivated excitement may be rooted in the NACC. Again, measurement validity, reliability, internal validity and statistical conclusion validity were all fine based on the structure of the study, but external validity and multicultural validity both suffered as the study only looked at Stanford students.
The third piece was Mindfulness training increases cooperative decision making in economic exchanges: Evidence from fMRI. The key question is how does mindfulness training impact the emotional component of economic exchanges? This was a randomized longitudinal design involving mindfulness training or an alternative. FMRI scans of participants playing a game were conducted. The participants were 51 volunteers of equally numbered men and women, with white and black participants of people 'who want to learn to deal with stress issues in everyday life.' It was sought to find brain regions associated with cooperation. It was found that mindfulness increases cooperation, and that the septal region, linked to social attachment, was activated. Measurement validity and reliability are questionable here, as there is no real way to quantify how ‘mindful’ a person is. Internal validity is alright, as is statistical conclusion validity, however external validity and multicultural validity suffer as this was only 51 self selected participants.
Fourth, I looked at Risk patterns and correlated brain activities. Multidimensional statistical analysis of fMRI data in economic decision making study. The question was what neural substrates underlie decisions during risk in investment? Participants were exposed to an investment decision task while in an MRI. A Time series of 3D images of the brain, and the application of a panel version of the dynamic semiparametric factor model (DSFM) were used. FMRI scans and investment results were tabulated. 17 subjects participated who were exposed to an investment decision task. FMRI brain region scans and DSFM, as well as nonparametric statistical modeling, were conducted. Key findings included that decision making is comprised of valuation, comparison, and the final choice. Risk is associated with different brain regions than previously established findings on regions for investment. Measurement validity was alright, but measurement reliability was questionable, as risk is different for each individual. Internal validity and statistical conclusion validity were also fine, however again external validity and multicultural validity suffered as there were only 17 subjects.
Fifth, I read The mere green effect: An fMRI study of pro-environmental advertisements. The question was that purchasing behaviors do not reflect stated sentiment of preferring green products. Do green ads work? In an MRI, participants were exposed to green and standard ads and then rated their preferences and levels of liking the product. FMRI scans during the rating were used. 24 right handed women recruited through a community database was done, and it was found that ratings were more favorable for green ads, but fMRI showed the opposite - participants had more activation in regions associated with personal value and reward (ventromedial prefrontal cortex and ventral striatum) to the control ads. Measurement validity was questionable, as it relied on self reported scores as opposed to some objective measure. Measurement reliability was fine, as was internal and statistical conclusion validity, but again both external validity and multicultural validity suffered due to the limited participant pool.
Lastly, I looked at Identity on social networks as a cue: Identity, retweets, and credibility. The question was how does social media influence people’s source credibility with risk information. A posttest after treatment, where participants looked at social media posts, was administered, in addition to a self report questionnaire. Participants recruited from undergraduate communication courses at a southern research university. 434 people total ranging from 18 to 55 years old, roughly half men and half women. It was found that different online heuristic prompts influence the judgements of competence, good-will, and trustworthiness, while cues for authority strongly influenced credibility. Measurement validity and reliability are questionable, as this required self reporting. Internal validity and statistical conclusion validity were alright, as was external validity given the large sample, but multicultural validity suffered given a lack of diversity.
In sum, there were similarities, as well as differences, between the various studies I examined. The five fMRI studies all went about doing similar things, and taking fMRI scans. For the most part, this centered on investment under different conditions, or rating preferences. The conclusions were varied, with different study set ups finding different regions of the brain lighting up for activity. Another similarity regards the external validity of these papers. Their sample sizes were all quite small, and rarely representative of the public. Further, in some cases, they were self selected. This does not bode well for any generalizability to the broader public, and certainly not to the rest of the world. One strength the studies shared, however, would be that they all had relatively straightforward and reliable designs, with good measurement validity, measurement reliability, internal validity, and statistical conclusion validity. It is hard to knock a decent fMRI study, as they are so in depth and analytical by their nature. The only critique that flows from this would be under the statistical conclusion validity front, in which we could argue that correlation does not equate to causation. Just because a certain brain area is lighting up, it doesn’t necessarily mean that this is being caused by whatever they are doing. There may be intervening omitted processes that take place which are a result of the prompts, which in turn trigger the brain regions. This is a problem for all such studies.
IX: Conclusion
Through analyzing the previous articles, we can come to a synthesis that sets the stage for this novel study. fMRI techniques have not yet been used to examine green investing, though different constituent parts are present in the current literature. This experiment has the potential to address the stated research questions, as well as how honest their answers are. While there are questions regarding the honesty of responses, causality mechanism, and generalizability of conclusions, the outcome of this study would prove highly interesting, and ideally pave the way for further analysis on what motivates people to take up the cause of green investing. Climate change is a serious, and all pervasive issue that will only grow in importance as time goes on. In order to best effect change, this study attempts to get at the root causes of how to motivate people to care about the issue, and in turn, cause legislators and elected or appointed officials to make change. This is done through the structure of the survey and how it is administered, as well as how it is analyzed, and compared to existing literature. With results in hand, I imagine we’d see that certain cue sources for the information provided will be more influential than others, and that efforts should be made to encourage using these sources to reach people. How best to actually do this once we have found out the results lies beyond the purview of this paper, and I leave it to others in the futures.
X: References
Peterson, Richard L., The neuroscience of investing: fMRI of the reward system. Brain Research Bulletin 67 (2005), 391-397
Kirk et al, Mindfulness training increases cooperative decision making in economic exchanges: Evidence from fMRI. NeuroImage 138 (2016) 274-283
Bommel et al, Risk patterns and correlated brain activities. Multidimensional statistical analysis of fMRI data in economic decision making study. Psychometrika Vol. 79, No. 3, 489-514
Lin et al, Identity on Social Networks as a Cue: Identity, Retweets, and Credibility. Communication Studies Vol. 69, No. 5, 2018, 461-482
Aimone JA, Houser D, Weber B. 2014 Neural signatures of betrayal aversion: an fMRI study of trust. Proc. R. Soc. B 281: 20132127
I. Stephanie Vezich, Benjamin C. Gunter & Matthew D. Lieberman (2017) The mere green effect: An fMRI study of pro-environmental advertisements, Social Neuroscience, 12:4, 400-408