Literary Studies and Narrative
Thinking about economic narratives brings economists to a corner of the university with which they are often unfamiliar: the literature department. Some literary theorists, inspired in part by psychoanalysis, the archetypes of Carl Jung11 and the phantasies of Melanie Klein,12 have found that certain basic story structures are repeated constantly, though the names and circumstances change from story to story, suggesting that the human brain may have built-in receptors for certain stories. John G. Cawelti (1976) classifies what he calls “formula stories” with names like “the hard-boiled detective story” or the “gothic romance.” Vladimir Propp (1984) found thirty-one “functions” present in all folk stories, with abstract names like “violation of interdiction” and “villainy and lack.” According to Ronald B. Tobias (1999), in all of fiction there are only twenty master plots: “quest, adventure, pursuit, rescue, escape, revenge, the riddle, rivalry, underdog, temptation, metamorphosis, transformation, maturation, love, forbidden love, sacrifice, discovery, wretched excess, ascension, and descension.” Christopher Booker (2004) argues that there are only seven basic plots: “overcoming the monster, rags to riches, the quest, voyage and return, comedy, tragedy, and rebirth.”
According to literary theorist Mary Klages (2006), structuralist literary theory considers such efforts to list all basic stories as “overly reductive and dehumanizing.”13 Although she dismisses other scholars’ lists of basic plots, she asserts, “Structuralists believe that the mechanisms which organize units and rules into meaningful systems come from the human mind itself.”14 Peter Brooks (1992) says narratology should be concerned with “how narratives work on us, as readers, to create models of understanding, and why we need and want such shaping orders.”15 Well-structured narratives, Brooks argues, “animate the sense-making process” and fulfill a “passion for meaning,”16 and the study of narratives naturally leads to psychoanalysis.
Russian literature scholar Gary Saul Morson recently collaborated with economist Morton Schapiro in Cents and Sensibility (2017), in which they argue that a better appreciation of great novels—which bring us close to the essence of human experience—would help improve the modeling of economic life.
Neuroscience, Neurolinguistics, and Narrative
Narratives take the form of sequences of words, which makes the principles of linguistics relevant. Words have both simple, direct meanings and connotations, in addition to metaphoric use. Modern neurolinguistics probes into the brain structures and organization that support narratives.17
Contagious narratives often function as metaphors. That is, they suggest some idea, mechanism, or purpose not even mentioned in the story, and the story becomes in effect a name for it. The human brain tends to organize around metaphors. For example, we freely incorporate war metaphors in our speech. We say an argument was “shot down” or is “indefensible.” The human brain notices these words’ connection to war narratives, although the connection is not always a conscious one. The connection enriches the speech by suggesting other possibilities. So when we speak of a stock market “crash,” most of us are reminded of the rich story of the 1929 stock market crash and its aftermath. Linguist George Lakoff and philosopher Mark Johnson (2003) have argued that such metaphors are not only colorful ways of writing and speaking; they also mold our thoughts and affect our conclusions. Neuroscientist Oshin Vartanian (2012) notes that analogy and metaphor “reliably activate” consistent brain regions in fMRI images of the human brain. That is, the human brain seems wired to respond to stories that lead to thinking in analogies.
Consilience Calls for Collaborative Research
The dazzling array of approaches to understanding the spread of narratives, briefly summarized in this chapter, means that collaborative research between economists and experts in other disciplines holds the promise of revolutionizing economics. Particularly important are the ideas and insights of epidemiologists, whose models successfully forecast the future trajectory of disease epidemics and explain how to counteract these epidemics. As we will see in the next chapter, economists can adapt these epidemiological models to improve their own models and forecasts. The marriage of economics and epidemiology is our first example of consilience in this book.
Chapter 3
Contagion, Constellations, and Confluence
Before we embark on a study of how economic narratives go viral, it is helpful to consider how bacteria and viruses spread by contagion. The science of epidemiology offers valuable lessons and may help explain how the story of Bitcoin (and many other economic narratives) went viral.
Let us consider diseases first, caused by real viruses. Consider as an example the major Ebola epidemic that swept through West Africa—Guinea, Liberia, and Sierra Leone—between 2013 and 2015. Ebola is a viral disease for which there is no approved vaccine or treatment, and it kills most people who contract it. Ebola spreads from person to person via body fluids. Its infectiousness can be lowered through hospitalization and quarantine, and through proper handling and burial of the dead.
In Figure 3.1 we see a typical example of an epidemic curve, for Ebola, in a community, this from Liberia. Note that the number of newly reported Ebola cases has a hump-shaped pattern. The epidemic first rises, then falls. The rising period is a time when the contagion rate, the rate of increase of newly infected people, exceeds the recovery rate plus the death rate. During the rising period, the rise in the number of infected people due to contagion outnumbers the fall in the number due to recovery or death. The process is reversed during the falling period. That is, the fall in the number of infected people due to recovery or death outweighs the rise in the number due to contagion, putting the number infected into a steady downward path marking the termination of the epidemic.
FIGURE 3.1. Epidemic Curve Example, Number of Newly Reported Ebola Cases in Lofa County, Liberia, by week, June 8–November 1, 2014
We will see many examples of economic narratives whose prevalence in digitized databases follows a similar hump-shaped pattern. Source: US Centers for Disease Control and Prevention.
After the epidemic started, contagion rates of the Ebola virus eventually fell for various reasons, notably the heroic efforts of Médecins Sans Frontières (Doctors Without Borders), more than a hundred nongovernmental organizations, and individuals who risked their lives to lower the contagion in Africa. According to the World Health Organization, health-care workers were twenty-one to thirty-two times more likely to catch the disease than the general population there, and there were 815 confirmed and probable cases of health-care worker infection as of 2015. Most of these workers died.1
Contagion, Recovery, and Decline
Efforts to lower contagion rates by avoiding contact with sick people are hardly new. The history of quarantines extends back at least to 1377 when the city of Venice imposed during a plague a thirty-day isolation period on arrivals by sea, and then a forty-day isolation period for travelers by land (the word quarantine derives from the Latin word for forty). The world has also seen occasional attempts to increase contagion as an act of war, as with the catapulting of dead bodies of plague victims into a fortified city at the Siege of Caffa, 1346.2
Another mechanism for a declining contagion rate is a decrease in the pool of susceptible people. This pool decreases through time because many people who had the disease are now immune to it (or dead). This mechanism, modeled in the appendix (p. 289), occurs even if no health-care workers take action to contain the disease, as in long-ago epidemics before modern medicine. Eventually, those epidemics ended before everyone was infected.
When the contagion rate is lower than the recovery rate plus the death rate, the disease does not disappear immediately. The contagion rate is not reduced to zero. All that is necessary to conquer the epidemic is to lower the contagion rate below the recovery rate. Unless the contagion rate is zero, there will still be new cases of the disease, but the total number of sick people declines, gradually tailing off to zero,
at which point the epidemic ends.
We are talking here of the average contagion rate and average recovery rate, averaging over many people. However, both the contagion rate and the recovery rate can differ greatly from one individual carrier to another. A relatively small percentage of super-spreaders can infect many people. One such super-spreader was Mary Mallon, “Typhoid Mary,” who a century ago spread typhoid fever to at least 122 people over an interval of years.3 In the context of narratives, most of us may not be contagious enough for long enough to cause an epidemic without the presence of these super-spreaders, and because of a small fraction of super-spreaders the average contagion rate can be much higher than the typical contagion rate. Today’s narrative super-spreaders may be enabled by marketing using accelerated analytics, such as recently provided by NVIDIA Corporation or Advanced Micro Devices, Inc., which is invisible to most of us. So we can’t always accurately judge the contagiousness of a narrative by our own fascination with it.
Both the appearance of the disease epidemic at a given time and place, and the decline in the epidemic after its peak tend to be mysterious. Many factors influence the contagion rate and recovery rate, factors that may be hard to document. For example, the ultimate reason for the recovery could be a change in the weather, which is more readily documented, or it could be a decrease in the number of encounters between people that allow for transmission of the disease, which might be hard to document. Or it might be some combination of the two. The changes need not be big or obvious.
We can apply this same model to epidemics of economic narratives. Contagion occurs from person to person through talk, whether in person or through telephone or social media. There is also contagion from one news outlet and talk show to another, as they watch and read one another’s stories. Once again, the ultimate causes of the epidemic might not be obvious. Fortunately, most economic narratives do not result in deaths, but the basic process is the same. The “recovery plus deaths” variable in the medical model is simply recovery, loss of interest in the narrative, or forgetting in the economic model we are developing. Economic narratives follow the same pattern as the spread of disease: a rising number of infected people who spread the narrative for a while, followed by a period of forgetting and falling interest in talking about the narrative.4
In both medical and narrative epidemics, we see the same basic principle at work: the contagion rate must exceed the recovery rate for an epidemic to get started. For example, when Ebola is found to have infected hundreds of people in one town and virtually nobody in another, the explanation could be some inconspicuous factor that made Ebola contagion rates higher in Town #1 than in Town #2, putting the Town #1 contagion rate above the recovery rate at the beginning of the epidemic. Meanwhile, in Town #2, there is no epidemic because the contagion rate isn’t quite high enough to offset recovery. Similarly, with narrative epidemics there may be two different narratives, one with some minor story details that make it more contagious than the other. The minor story details make the first narrative, and not the second, into an epidemic. Let’s apply this insight to the Bitcoin narrative.
Contagion of the Bitcoin Narrative
Figure 3.2 plots the frequency of appearance in news articles of the words bimetallism and Bitcoin. This figure is not a plot of a price but rather an indicator of public attention. Both bimetallism and Bitcoin represent radical ideas for the transformation of the monetary standard, with alleged miraculous benefits to the economy. Each word is a marker for a constellation of stories that include not only stories of theory but also human-interest stories. The plots for both words look quite similar, and each is similar to a typical infective curve as seen in Figure 3.1. We haven’t seen a definitive end of the Bitcoin narrative yet, as we did with bimetallism; only time will tell.
FIGURE 3.2. Percentage of All Articles by Year Using the Word Bimetallism or Bitcoin in News and Newspapers, 1850–2019
There is a remarkably similar epidemic pattern to the two popular “bi-” monetary innovation narratives a century apart and similarity to the disease epidemic curve in Figure 3.1. Source: Author’s calculations using data from ProQuest News & Newspapers.
We will discuss the remarkable bimetallism epidemic at length in chapter 12, along with other narrative epidemics. For now, it is enough to know that bimetallism and Bitcoin both invoke monetary theory. In both cases, an enormous number of people began to regard a particular innovation as cool, trendy, or cutting-edge. In both cases, the contagion is represented by a hump-shaped curve resembling an epidemic curve. In contrast, in Figure 3.2, the curves look more spiky (that is, compressed left to right) because the figure plots more than a century of data, beyond the virulent periods. In fact, the bimetallism and Bitcoin narratives played out over years, rather than weeks as in the case of Ebola, but the same epidemic theory applies to all three. In the case of bimetallism, we also see a smaller secondary epidemic in the 1930s, during the Great Depression, but it never amounted to much. It was like a secondary epidemic of a disease.
So narrative epidemics really mimic disease epidemics. And it is more than just that. It is interesting also to note that there are co-epidemics of diseases and narratives together. Medical researchers in the Congo during a 2018 outbreak of Ebola linked the high contagion to narratives reaching the population. Over 80% of the interviewees said they had heard misinformation that “Ebola does not exist,” “Ebola is fabricated for financial gains,” and “Ebola is fabricated to destabilize the region.” For each of these statements, over 25% said they believed the narrative. These narratives discouraged prevention measures and amplified the disease.5 The two epidemics fed on each other to grow large.
The appendix to this book looks at theories and models from epidemiology, including the original 1927 Kermack-McKendrick SIR model, to help explain the spread of economic narratives. These models divide the population into compartments: susceptible to the disease (S), infected and spreading the disease (I), and recovered or dead (R). All of the models feature contagion rates and recovery rates. We can think of Figures 3.1 and 3.2 as evidence on the number of infectives (I). These models tend to predict hump-shaped paths for an epidemic, like that in Figure A.1 in the appendix, page 291, even if there is no medical intervention at all. The epidemic will eventually start weakening because the percentage of the population that has still not been exposed to the disease is declining, bringing down the contagion rate below the recovery rate.
In the appendix we will see also that the time to peak and the duration of an epidemic can vary widely, determined by model parameters. The Ebola epidemic ran for a matter of months in a given locale, but we should not assume that all epidemics must follow that same short timetable. In other words, the Ebola epidemic could have stretched on for years if the initial contagion rate had been lower, so long as contagion did not fall below recovery.
For example, epidemiologists have described the acquired immune deficiency syndrome (AIDS) caused by the human immune deficiency virus (HIV) as not very contagious, and they have recommended that health-care professionals should not shrink from treating HIV patients for fear of catching it.6 AIDS tends to be transmitted only in certain circumstances involving unsafe practices. AIDS has been a slow epidemic, developing over decades, even slower than the bimetallism and Bitcoin epidemics, and it is able to grow despite low contagion because it has a smaller recovery rate: an HIV-infected person can continue to infect others for many years.
The Contagion of Economic Models
In 2011, Jean-Baptiste Michel and a team of coauthors published an article in Science providing evidence that mentions of famous people in books tend to follow a hump-shaped pattern through time, rising, then falling, over decades rather than months or years. They amplified their conclusions in a book, Uncharted: Big Data as a Lens on Human Culture, by Erez Aiden and Jean-Baptiste Michel (2013).
The same patterns seem to apply to economic theories. In chapter 5 we consider the contagion of one of these narratives, the Laf
fer curve, a simple model of the relationship between tax rates and the amount of tax revenue collected. But let us first note briefly that these patterns apply even to “highbrow” economic theories that circulate primarily among professional economists. Figure 3.3 shows Google Ngrams results for four economic theories: the IS-LM model (published by Sir John Hicks in 1937), the multiplier-accelerator model (Paul A. Samuelson, 1939),7 the overlapping generations model (Samuelson, 1958), and the real business cycle model (Finn E. Kydland and Edward C. Prescott, 1982). All show hump-shaped patterns similar to those of disease epidemics.8 For our purposes here, it doesn’t matter what is in these theories. None of them has been proven completely right or wrong. They are all potentially interesting. Each of them is a story whose popularity followed the expected path of an epidemic.
For three of the models, the epidemic first became visible more than a decade after the model was introduced, a phenomenon that we also see in the medical-epidemic framework, where epidemics may go unobserved for a while after very small beginnings. The number of cases may be growing steadily percentage-wise, but the disease fails to be widely noticed until the number of cases hits a certain threshold. In practice, the long lag between the publication of an economic theory and its eventual strong epidemic status represents a time interval over which the model evolves from something regarded as peculiar and thought provoking into something that is clearly correct and recognizably great. Over this gestational interval, other scholars in the discipline increasingly appreciate the model, and the epidemic spreads through academic rituals, such as paper presentations at seminars and major conferences.9 Eventually the models make their way into textbooks. Still later, the model is talked about enough that the news media begin to feel that it should be mentioned, and people outside of the economics profession who pride themselves on their general knowledge begin to feel they should know something about it. But in this late stage of the epidemic, the model may begin to lose some of its contagion. Some people begin to consider it stale and unoriginal even if it has merit, while others end up forgetting about it completely.
Narrative Economics Page 4