One widespread and important innovation is the creation of economic think tanks interested in creating policies based on the insights of behavioral economics. These think tanks have been called “nudge units,” following the Behavioral Insights Team in the UK government in 2010. Working with the ideas popularized by Richard Thaler and Cass Sunstein in their 2008 book Nudge: Improving Decisions about Health, Wealth, and Happiness, these units try to redesign government institutions toward “nudging” people away from their irrational behavior without coercing them. According to the Organization for Economic Cooperation and Development, there are now close to two hundred such units around the world.6
I advocate formalizing some of the intuitive judgment that national leaders already use to acknowledge and harness changing economic narratives. Leaders must lean against false or misleading narratives and establish a moral authority against them. Their first step is to understand the dynamics of the narratives. Their second step is to design policy actions that take account of narrative epidemics. Policymakers should try to create and disseminate counternarratives that establish more rational and more public-spirited economic behavior. Even if the counternarratives are slower to take effect than a more contagious destructive narrative, they can eventually be corrective.
For example, as noted in chapter 10, US President Franklin Delano Roosevelt in his March 4, 1933 inaugural address7 at the bottom of the Great Depression asked people to set aside their fears and spend money. In his first fireside chat, March 12, 1933,8 he appealed to morality, asking people not to withdraw more money than they needed when the banks reopened. He was spinning a narrative of what could happen if unreasoning people with little social consciousness destroyed the economy. We can speculate that President Roosevelt’s request worked because it was based on a moral standard; his chats roughly coincided with upturns in the US economy. However, we do not have a way of quantifying exactly how salient the narratives of the time really were. We would know more, perhaps, if economists had collected better data and conducted more analysis on what people were saying in 1933. If they had, we might now have a better understanding about how to frame such moral-appeal narratives in the future.
A problem in using narratives to forecast economic variables is that human judgment and discourse about narratives tend to be politicized and emotion-ridden. It has been difficult for scholars to research popular narratives, focusing on the core elements that make them contagious, without being accused of taking sides in political, or sometimes religious, controversies. Because many professional economists try to remain nonpartisan, they tend to rely on quantitative, rather than qualitative, observations. However, with modern information technology, economists can now collect data on economic narratives themselves, on their essential elements of meaning, without being overly focused just on words, and they can model the transmission of narratives. If we maintain quantitative rigor, we can make narrative epidemics a part of economic science.
Some may doubt that it is possible to have nonpartisan discussion of economic narratives. However, if we are careful and polite, it should be possible to speak in a nonpartisan way about epidemics of economic narratives. Most people have some instinct about how to speak in a nonpartisan way, and they do so when the occasion demands it. We do not have to go so far in our efforts to be nonpartisan that we exclude study of some ideas and emotions that drive economic changes.
Economic research is already on its way to finding better quantitative methods to understand narratives’ impact on the economy. Textual search is a small but expanding area. A search of the NBER working paper database finds fewer than one hundred papers with the phrase textual analysis. Economists have used textual analysis to document changes in party affiliation (Kuziemko and Washington, 2015), political polarization (Gentzkow et al., 2016), and news and speculative price movements (Roll, 1988; Boudoukh et al., 2013). Much more could be done. For example, economists could carry the historical analysis further into databases of personal diaries, sermons, personal letters, psychiatrists’ patient notes, and social media.
Collecting Better Information about Changing Narratives Should Start Now
Economists must make more serious efforts to collect time-series data on narratives, going beyond the passive collection of others’ words, toward experiments that reveal meaning and purpose. Such great quantities of digitized data are now available that it boggles the imagination. Even so, this vast dataset is minuscule compared to the even vaster universe of human communications that go on every day, most of which are not adequately sampled, described, or understood.
It is important that such data collection be maintained on a consistent basis through decades, so that we can make intertemporal comparisons of major influencing public narratives in the future. There has been relatively little incentive to undertake such a project, because there is little immediate payoff to doing so. Instead, most narrative data collection focuses on immediate interests, such as marketing specific products or predicting upcoming elections.
It is also important to apply creative energies toward such consistent long-term data collection. Understanding people, their behavior, and their thinking may even require the help of psychoanalysts and philosophers.
It will be difficult to combine these two needs, consistency through time and creativity. But we must do so if we are to make real progress in narrative economics.
The first step requires improving existing search engines so that they can better measure the time-varying incidence of narratives. The search engines do not tell us exactly how they determine the estimated total number of hits. Rather, they are designed primarily to help users find articles or information they are looking for. Thus some anomalies pop up when researchers attempt to count the number of references. For example, Google’s search engine instructions say that a search for a phrase should enclose the phrase in quotation marks so that the search is confined to exactly those words in exactly that order. But sometimes including the phrase in quotation marks results in more hits than the phrase without quotation marks. A Google spokesman says that the greater number of hits for the phrase in quotation marks may happen because quotation marks cause Google to “dig deeper” into the database.9 We need to see evidence that such deeper digging is not compromising the accuracy of counts. Google Ngrams is designed to count phrases, and to compare the counts through time, but Ngrams and other search engines could do much more to ensure that users can accurately compare counts through time.
In addition, we should be collecting time-series data about economic narratives at least once a year, ideally more often than that, and on an uninterrupted basis for decades into the future, and in multiple countries and languages. Such data-collection efforts might include the following:
Regular focused interviews of respondents inviting them to talk expansively and tell stories in response to stimulus questions related to their economic decisions. The instructions would ask respondents to tell a story that is interesting or suggestive of causes in the current environment. This is the listening as a research method advocated by Charlene Callahan and Catherine S. Elliott10 and the qualitative research advocated by Michael Piore.11 Some researchers have conducted such research, notably Alan Blinder and his coauthors,12 who interviewed top executives about how they reach decisions about price setting, and Truman Bewley,13 who asked managers about their wage setting. Still more researchers have studied narratives to try to infer motivations of those who decide on fiscal and monetary policy.14 Focused interviews are interviews of individuals that ask them to focus on their understandings and stories related to current behavior. Focused interviews began to be used as research tools in the 1920s and were given a firm foundation by Robert K. Merton and Patricia L. Kendall in 1946.15
Unfortunately, these researchers usually conducted these interviews as one-time-only events, and they did not try to collect long time-series information that would reveal how answers and stories changed through history. If such data had been colle
cted, the entire stories would have been digitized as sections of long time series and preserved for future textual analysis. The data could then have been added to major economic data collections. These include databases such as the Panel Study of Income Dynamics at the University of Michigan Institute for Social Research, the Federal Reserve Board’s Consumer Expenditure Survey, and the Swedish Household Market and Nonmarket Activities database (HUS) at Gothenburg University. Maintaining a consistent research environment through time would allow intertemporal comparisons, though the list of stimuli would have to be augmented as time goes on and as relevant new words and concepts appear. There would likely be some overlap with other surveys, such as those conducted internationally under the International Social Survey Program.16 New efforts could go well beyond the work to date of the University of Chicago General Social Survey17 or the University of Michigan Institute for Social Research,18 which have been useful for many purposes in the past.
Regular focus groups with members of different socioeconomic groups to elicit actual conversations about economic narratives. A focus group is a focused interview done on a group of people. The group interview is especially important for narrative economics since it creates an environment that simulates the very interpersonal contagion that underlies the epidemiology of narratives. The focus group is an important and common research method, typically used by marketers to learn how people in various demographic groups talk among themselves about products or political candidates. In a focus group, the researcher puts together people who likely represent actual groups in human society; participants are typically similar in age, live in the same geographical region, and share other factors that influence social group cohesion. By putting similar people together, the researcher attempts to eliminate barriers of “political correctness” that might inhibit normal conversation in unnatural groups. The focus group leader then facilitates talk about stimulus words related to the subject of the research and records the conversation. Running focus groups requires human judgment on the part of the interviewer. It is an art as well as a science, the art of getting people to think and talk about why they do certain things or hold particular beliefs.
Focus groups are thus experimental situations that could become real observations of the contagion of ideas. Though common, focus groups researchers do not usually seek to provide voluminous data over decades in an attempt to learn about the causes of economic changes. In the case of economic narratives, focus-group participants might be asked to respond to words or phrases such as stock market, bank, unemployed, the real reason to save, or government actions that might impact your future economic welfare or that of your children. Recorded videos of the focus groups might be digitized, and, in the future, possibly even scanned and analyzed by facial recognition and emotionally categorized algorithms.
Focus groups are now recognized as valid tools for research into popular understandings and motivations. Focus groups have their critics,19 for they are often poorly managed, but when done well they are extremely useful. Economists, however, have been extremely loath to use them. Economics and finance are the worst fields for references to focus groups. In the decade 2010–2019, only 0.04% of scholarly economics articles and 0.02% of scholarly finance articles mention the term focus group despite the fact that focus group methods, developed largely by practitioners of marketing science, are much improved in terms of sampling, directing, and experimenting.20
One of the propositions in chapter 8 of this book holds that the economic impact of narratives may change through time, depending on details of the narrative and of the zeitgeist. We saw examples of apparent inconsistencies: The outbreak of World War I caused the US stock market to collapse, while the outbreak of World War II caused the market to soar. The bombing attacks linked with the “big Red scare” in the United States in 1920 were associated with a decline in economic activity, while the 9/11 attacks in 2001 were associated with ample spending and the end of a recession. A timely and appropriately led set of focus groups that homed in on assumptions, emotions, and loyalties might have given us a better understanding of why people behaved as they did.
A historical database of focus groups conducted for other purposes in years past. The Public Opinion Research Archive provided by the Roper Center for Public Policy Research,21 now at Cornell University, has since 1947 amassed a database of opinion survey responses, including the Gallup Data Collection. This archive, however, tabulates answers to individual questions about opinions, questions changing in wording through time and as part of changing questionnaires that provide changing context in terms of other questions asked in the same survey. It does not listen to respondents in their own words and their own thought innovations. The archive is useful, but it is hard to appreciate what elements are contagious or to judge changes in thinking from it. There should be a massive database that asks those conducting focus groups around the world to share the results of past focus group results that may be relevant to understanding changing narratives. It would ask them to share the results of past focus group results that may be relevant to economic narratives. The database administrators would ask permission to publish raw data while remaining suitably respectful of past privacy promises made to participants. The administrators would then find some way (a challenge!) to organize these past focus groups into the closest approximations of computer-searchable time series, which would permit researchers to use the data to plot epidemic curves for specific narratives, as I have done in this book for newspapers and books.
Databases of sermons. Thousands of religious organizations, churches, synagogues, mosques, and the like, must have records of old sermons (derashas, khutbahs, etc.), but databases seem designed for sermon preparation rather than historical research. Sermons are important because they touch on moral values as they seek the deeper meanings in life. Changes in these moral values and value judgments about what is right and wrong are undoubtedly relevant to changing economic decisions.
Historical databases of personal letters and diaries, digitized and searchable. There are the beginnings of such databases already, but we could make a more determined effort to encourage families to donate diaries of deceased family members to such databases. Existing databases do not seem to be based on random samples of the world population with associated personal information. They tend to be assemblages selected for research with a specific purpose, such as research on a single war or social issue in a single country. These are still useful, but better sampling would make for better knowledge on how to generalize results to a broader population.
None of the above-listed data collections is likely to reach the desired scope in the academic research mill any time soon. The payoff to such research is far in the future, and the judgment of such resources is too hard to formalize. Academic research conducted by individuals, who are under pressure to “publish or perish,” is unlikely to start data-collection efforts that will help us understand the relatively rare, but serious, depressions and financial crises that occur from decade to decade, but perhaps no more than twice in a lifetime.
Many survey organizations have been collecting some of the data outlined in the wish list above. They should be funded to do so systematically and consistently through time. I have collected such data on a small scale, with questionnaire surveys of both individual and institutional investors about the stock market, since 1989. There are parallel surveys in Japan and China. Also, Karl Case, and now Anne Kinsella Thompson, and I have been doing surveys of US homebuyers and their perceptions of the market for single-family homes since 1993. The early surveys received support from the US National Science Foundation, with later surveys supported by the Whitebox Foundation and the Yale School of Management. The questionnaires for these surveys include open-ended questions with space that invites respondents to write a sentence or two. The questions are designed to stimulate respondents to think about what is motivating them, so that their responses can be analyzed in perpetuity. Since I started these survey projec
ts, I have seen other survey organizations pursue sometimes similar objectives, and then stop. New survey tools like SurveyMonkey and Qualtrics are encouraging a proliferation of surveys but not a consistent strategy that is pursued over long periods of time.
As of this writing, there does not appear to be much support for the routine collection of historical data in a form that will allow, decades hence, a truly comprehensive study of the dynamics of economic narratives.
Tracking and Quantifying Narratives
Research today needs improvement in terms of tracking and quantifying narratives. Researchers have trouble dealing with a set of often-conflicting narratives with gradations and overlaps. Even the simplest epidemic model shows that no narrative reaches everyone. In addition, the spread of a particular narrative may be largely random. The meanings of words depend on context and change through time. A story’s real meaning, which accounts for its virality, may also change through time and is hard to track in the long run.
There is also the perpetual challenge of distinguishing between causation and correlation. How do we distinguish between narratives that are associated with economic behavior just because they are reporting on the behavior, and narratives that create changes in economic behavior?22
Economic researchers have to grapple with the same issues that have troubled literary theorists who try to list the basic stories in all of literature, who attempt to distill what defines these stories and makes them contagious (see chapter 2). At any time in history there are many contagious stories, and it is hard to sort through them. Literary scholars run the risk of focusing on details of the stories that are common just because the events are familiar in everyday life. They also face the difficulty of accounting for changes through time in the list of stories.
Narrative Economics Page 31