Cause and effect are normal considerations in daily life. For example, "If I don't wear a jacket, I'll catch a cold"; "If I don't wash my hands, I'll get sick"; or "If a child breathes in cigarette smoke, they'll cough."
However, the same causal connection does not necessarily apply to all people. Physical strength and the risk of increases in severity of medical symptom vary from person to person. Not all people correctly estimate the influence of their choices on themselves and on others. Some may be careless regarding themselves, while others may worry too much. Distortions in risk perception can have negative effects on the individual and on society.
Professor Olof Johansson-Stenman, an environmental economist of the University of Gothenburg, Sweden, in a 2008 paper advocates a framework for showing distortions in individuals' risk perceptions. Our research team developed a research framework for quantitatively estimating distortions in individuals' subjective risk beliefs and named the estimation the "subjective risk belief function." Using this function, it is possible to indicate the level of distortion in risk perceptions of individuals and society as a whole.
Quantifying subjective beliefs
There are two obstacles to public implementation of this research framework. The first obstacle is measuring the subjective risk beliefs. Individuals who are more susceptible to health risks or environmental problems are often low-income groups and people with low education levels, or children, elderly people and women. It is generally difficult to quantify the subjective beliefs of these people and a high degree of prudence is required in collecting information from them.
Professor Adeline Delavande, a development economist of the University of Technology Sydney, Australia, et al. presented a method of using stones and candy to express the level of subjective risk beliefs, allowing for greater visual and intuitional understanding.
The second obstacle is measuring individual-specific true risks. In the case of a disease, one option is the method of having a physician diagnose each person. This method is costly but can be substituted with self-analysis of symptoms by each research subject and using a statistical causal inference method for a diagnosis.
After using these solutions to overcome the two obstacles, we applied the aforementioned framework in the real world and verified what can be found and whether policy implications can be obtained.
Our research targeted household air pollution emanating from primitive household cooking fires, which represents the leading environmental cause of death in many developing countries, and especially in India. Although this is not well-known in Japan, an incredible amount of smoke is generated during cooking at home in developing countries as they use firewood and furnaces for cooking. Health effects and household air pollution caused by such smoke are serious societal problems.
Effects are particularly serious for women who do the cooking and the children around them, and it is estimated that globally, a total of 3.9 million people die annually as a result. Before the COVID-19 pandemic, cooking smoke was considered to be the most serious health risk in the developing world.
To address this issue, the government of India (and gas companies) have endeavored to disseminate the use of propane gas-burning stoves, conducting a policy that provides stove burners free of charge beginning in 2017. Nevertheless, people continue to use firewood.
One of the reasons for the continued behavior was considered to be the misunderstanding of the risks, in other words, the misunderstanding of the risks posed by firewood, or an underestimation of the advantage of propane gas. We conducted a survey to understand the actual status of the misunderstanding. We selected the suburbs of Kolkata, which is not a wealthy city in India, as the subject area. The average monthly income of the 588 households that cooperatively participated in the survey was approximately 12,000 Japanese yen, and 87% of them use firewood or dried cattle feces to cook with. Both generate a large amount of smoke.
We used 10 pieces of candy for the survey of subjective risk beliefs. We asked the survey participants to use the number of candy pieces to express their estimate of the probability of developing any medical symptoms in their eyes and/or respiratory system when using propane gas for cooking for one month. In the same manner, we asked them to express their estimation when using firewood for cooking for one month.
Regarding objective risks, we asked survey participants in detail about their subjective symptoms for the last one month and their cooking methods in the previous month. By using the data obtained, we created statistical estimations of risks posed by cooking with gas and risks posed by cooking with firewood. Of course, there is no guarantee that these show the true risks, but the estimations can be considered to be objective to a certain degree.
Next, we compared subjective risk predictions and objective risks in relation to both cooking with gas and cooking with firewood for each of the 588 survey participants. Through this process, gaps in predictions for the reduced risk of switching from firewood to gas can be made visible for each of them. When tabulating the results, it is also possible to see the trend of perception biases for all of the 588 survey participants.
Variation in prediction gaps
We discovered the following. First, all of the 588 survey participants correctly understood that cooking with gas is better for their health, but they underestimated the effects thereof. The analysis using a causal inference method showed that they predicted a 90-percentage-point reduction of objective risks by switching to gas, but the average subjective prediction was only a 56-percentage-point reduction.
Additionally, the gaps in the risk perceptions of individuals varied. So, who underestimated the risks? We conducted an analysis again regarding this point, but it was almost impossible to explain gaps in perceptions with easily surveyable household attributes (income, etc.). In short, there must be something that can only be ascertained through a survey of subjective risk belief itself.
Political implications can also be obtained through this research. Should the government of India raise people's awareness of the risks posed by cooking with firewood more thoroughly? There would likely be some effect of an information campaign, but as all the survey participants already understood the relevant risks, the expected effects are considered to be limited. Larger effects can be expected from other policy initiatives, such as the government's intervention in decisions on gas cylinder pricing and distribution modes.
We constantly make choices based on causal relationships. The government also selects policies based on their effects. There are cases where causal relationships are underestimated or overestimated based on individuals' assumptions, like the case introduced here, but such assumptions can be confirmed using a causal inference method. It is possible to correct the distortion in subjective beliefs by collecting data and statistically analyzing them.
(This text is based on the following paper: https://doi.org/10.1016/j.jdeveco.2022.103000)
* Translated by RIETI.