Nicola Gennaioli @ Bocconi University
Abstract:
How do people form beliefs about novel risks, with which they have little or no direct experience? We address this question using a 2020 US survey of beliefs about the lethality of Covid. The survey reveals several surprising findings, including most dramatically that the elderly underestimate their own risks, while the young hugely overestimate them. To shed light on the evidence, we present a model in which people selectively and automatically recall past experiences, including those from other domains, and use them to imagine (simulate) the novel risk. In the model, an experience increases perceived risk by making that risk easier to imagine, but decreases perceived risk by interfering with recall of other experiences that may feed imagination. The model accounts for our initial findings, but also yields new predictions based on how non-Covid experiences should shape beliefs about Covid. The model connects average overestimation of unlikely risks with strong disagreement: people with many interfering experiences underestimate risk and are less sensitive to direct experiences with Covid (such as local disease dynamics). We find empirical support for these and other predictions using our survey data on respondents’ Covid and non-Covid past experiences.
Bio:
Nicola Gennaioli is the Romeo and Federica Invernizzi Professor of Finance at Bocconi University in Milan. He obtained his PhD in economicsfrom Harvard University in 2004. He has held research positions in Bercelona, Stockholm and Harvard, and he has been editor of leading economics journals. He works on topics at the intersection of psychology and economics, in particular he studies how salience and memory affect beliefs, decisions, and market activity.
Summary:
Goal: model the process of belief formation
Normal approach: model people as Bayesian agents who update their probability estimates about beliefs
But where do the base events come from?
E.g. COVID and people’s beliefs about it
Argument:
Recall of past experiences from different domains are critical
Personal memory most relevant
Studying psychology of selective memory
Similarity, frequency and interference are critical for choosing what it relevant
Memories seem to simulate possible futures via analogies to past events
Their analysis shows that
People’s forecast of what COVID does are highly dependent on their prior experiences of diseases and people estimate COVID as being similar to those experiences rather than from raw data about COVID itself (e.g. people who experienced health problems will think COVID is more dangerous, while people who didn’t experience health problems but did experience poverty will think COVID is less dangerous)
People tend to be able to remember one major type of experience, which tends to overshadow others
Many different types of experiences: personal, from social network, from news
Survey:
May, July, Nov/Dec 2020
Ask
About Covid’s lethality for themselves or a “normal person”
Health conditions/history
Socioeconomic characteristics
Ask: “what fraction of people have red hair” (probes how much the person over-estimates rare events)
Results:
Older people are less pessimistic about COVID impact on themselves and other
The young actually think they’re more at risk themselves than older people think about themselves
People who over-estimate red-hair are more pessimistic
People who experience more COVID deaths in their state are more pessimistic
People with worse health history are more pessimistic
People who have a history of family hospitalization estimate COVID as being more dangerous than people who live in places with more COVID
Model of thinking about COVID
When you think about COVID you create a model based on your own experiences with health
But you can’t remember everything
The probability that an event will be recalled is proportional to its similarity to event “COVID death”
Prob of remembering event = similarity of event / (sum of similarity of all potentially relevant events)
People will more experiences will have more potentially relevant events, and thus a larger denominator
People with many health adversities in their past (larger numerator) are more pessimistic about COVID
People will many non-health adversities (larger denominator) are less pessimistic
There is interference among multiple relevant experiences. So a person who had COVID will be extra pessimistic if living in a place with few COVID cases but less pessimistic if there are more COVID cases in the place. The idea is that the two experiences (personal COVID and regional COVID) interfere and regional COVID makes your own COVID less salient. Similarly, people who have had COVID are less sensitive to regional changes in COVID rates.
The elderly have more experiences (larger denominator), so they’re less sensitive to any one type of experience (they have so many).
People who over-estimate the number of red-hair people are more sensitive to experience