Introduction for Bubbles and Crashes
INTRODUCTION
“WE’RE LOSING MONEY FAST ON PURPOSE, to build our brand,” Toby Lenk, chief executive officer of eToys.com, proudly proclaimed. Lenk claimed that revenues were increasing an astounding 40% monthly. While most consumer purchases were still made in buildings called “stores,” in Toby Lenk’s world, the new economy had arrived. It was February 2000 and eToys was trading at $86 a share, implying an enterprise valuation of $7.7B, 35% more than bricks-and-mortar industry leader Toys “R” Us. Lenk believed he understood: the internet was changing the business world; traditional retailers would soon be a thing of the past; we would soon be buying groceries, or at least toys, in our underwear. The new economy was inevitable.
This was an astounding proposition given that in 1999 eToys’ revenues were $30 million. In 1999, Toys “R” Us took in $30 million in a single day. Not to mention, Toys “R” Us was profitable, earning $376 million that year, with a respectable, if not particularly remarkable, margin of 6.2%.1
The key to e-commerce was to buy high and sell low, in order to generate volume. With volume, costs would decline and profits would ensue. The revenue growth of eToys’ was extraordinary. These revenues came from “eyeballs,” or website traffic. Investors fit this fact into a narrative that justified losses to attract this traffic: get big fast. Build it, and they will come, costs will drop, and profits will follow!2 Get big fast was a narrative shared by the entire dot-com sector.
Meanwhile, Fortune magazine reporter (and later TechCrunch editor) Erick Schonfeld, was struggling with a different question: How much is a customer worth? In the heady days before costs had dropped to support profits, it was all guesswork. For example, in February 2000, a few weeks before the dot-com crash, a Yahoo! customer was valued at three times the value of an Amazon customer. To make sense of this, investors came up with stories to justify stock market valuations. The margins of Yahoo! would be higher than Amazon’s because online advertising is not as competitive as retail. And while pricing power had proved considerably stronger in advertising than in retail, Yahoo! was a long way from winning the online advertising space (if you don’t believe us, just Yahoo! it).
Was eToys overvalued? If it was, then we might have a bubble. More precisely, if eToys was worth more than the sum total of all the profits that it would make in the future, it would be a bubble. Toby Lenk didn’t think so. And who was to say he was wrong? To support his cause, Lenk proclaimed himself the expert: despite his lack of experience in retail, he was “a grizzled veteran.”3 He had a story too! According to Lenk, the e-commerce market was a land grab, and eToys was grabbing land and worrying about the rest later.4
For eToys, getting big fast required overcoming multiple challenges. The organizational challenges of building a multibillion-dollar business, which are difficult in any low-margin business, would be insurmountable for most new ventures. Timing the build-out of infrastructure to match the unpredictable growth in demand while buying high and selling low further complicated the challenge. The audacity of the bet, trying to sell all toys to all people, instead of focusing on a high-margin niche to start, complicated the mission. By November 2000, the game was almost over. eToys’ stock had fallen from $86 to $6.25 a share, and the “get big fast” narrative was showing cracks.5 Without investors who were willing to continue to make sense of the world through Lenk’s narrative, there would be no way for the company to assemble the funds it needed to survive, let alone grow. With its stock further falling to trade at $.09 a share, eToys shut down in March 2001.6
The eToys story was built on the “get big fast” narrative. And while the magnitude of eToys’ rise and fall is exceptional, the fact that it was built on a story is not. Generally, entrepreneurial capitalism is built on narratives that strive to make sense of imagined futures. These narratives, or stories, do much more than interpret the present; they shape the future. Not all narratives are equal. The logic of capitalism constrains which narratives will be convincing and to whom. For example, all investments require supporting narratives that are plausible to someone, but only a subset of these narratives produce eToys-style bubbles. Hence, understanding why and how narratives, and in particular speculative narratives, form is critical to understanding when there are—and when there are not—bubbles.
eToys was just a subplot in a much larger narrative that included other parallel subplots such as Webvan (groceries), Value America (general retail), CDNow (compact discs) and, of course, Amazon. com.7 These stories had a magnificent effect on the financial markets. The plot accelerated on August 9, 1995, when the browser company Netscape had its initial public offering. That day, the NASDAQ Composite Index closed at 1,005. On March 10, 2000, driven by a host of eToys-like subplots in the larger “get big fast” narrative, the index peaked at 5,132, more than 500% higher. Two and a half years after that, on September 23, 2002, the same index closed at 1,185, marking a loss of nearly 77% from its peak. This decline wiped out $4.4 trillion in market value. Accounting for inflation, it was not until January 2018 that the NASDAQ recovered its value.8
This collapse was much more severe in the tech-heavy NASDAQ than in the broader Dow Jones Industrial Average, which collapsed from 14,164 to 6,547.05 (a mere 54% decline), or the Standard & Poor’s 500 which fell from 1,516 to 800 (only 48%). If we look exclusively at a dot-com index the contrast is even starker. An index of four hundred dot-com stocks increased tenfold from the end of 1997 to March 2000, only to lose 80% of its value in the following nine months.9 The dot-com bubble was concentrated almost exclusively in, well, dot-com and closely related sectors.10
The events of the dot-com era fit into a long line of boom and bust episodes in the prices at which these types of assets change hands. Historical boom and bust episodes, popularly known as “bubbles,” often define their economic eras. For example, relative to the size of the British economy in the mid-nineteenth century, the “Railway Mania” bubble was several times the size of the dot-com bubble. The Roaring Twenties and, subsequently, the Great Depression scarred an entire nation; it was almost two generations before the next major speculative episode hit Wall Street in the form of the “’tronics” boom in the 1960s.11 Bubbles are important, undeniable facts of life for citizens living under entrepreneurial capitalism. However, bubbles are both inefficient (from a strictly economic perspective) and potentially damaging to the individual interests of those who are caught up in them. Our inability to avoid bubbles suggests that our understanding of them is incomplete.
A closer look at the investors in dot-com firms on the NASDAQ reveals additional curiosities. First, inexperienced investors threw around large sums of money. Many retail investors, usually viewed as less experienced than professional investors, were trading in dot-com firms.12 These investors were particularly bullish on dot-com firms and took bigger risks. For example, investors trading on E*Trade—the online, no-frills brokerage catering to retail investors—were over seven times more likely to trade on margin than investors who kept their assets with the full-service brokerage Merrill Lynch.13 One suspects that these margin investors not only were trading online but also were more invested in internet stocks. Second, many Wall Street investors were also inexperienced. While only 12% of professional money managers were younger than the age of 35 in 1997, these younger, less experienced mutual fund managers were more likely to invest in technology stocks than were their more seasoned colleagues.14 Third, many of those providing the initial funding to the dot-com firms that later went public were also inexperienced. From 1990 to 1994, the share of investments made by venture capitalists in the business for less than five years was 10%.15 By the year 2000, recent entrants to the VC space made 40% of all VC investments. Fourth, the entrepreneurs themselves were inexperienced. In earlier work, together with our student Michael Pfarrer, we estimated that between 1998 and 2002, fifty thousand would-be entrepreneur-millionaires founded dot-coms.16 We do not have good statistics on whether dot-com founders themselves were first-time entrepreneurs, but we do know that none of these founders had ever built an internet business—no one had.
What was the lure of dot-coms for investors? Why did they think their investments in dot-com ventures would pay off? For one, it seemed clear that the internet was going to be big. It was flashy, in the news, and most of all already familiar—investors used the new technology. Unlike products and services that targeted industrial buyers, the World Wide Web engaged Main Street, which made its potential value quite tangible to many of those who chose to invest. For example, investors in eToys could purchase toys on eToys.com. As we have documented extensively elsewhere, with economist David Miller, investors thought they knew that the “get big fast” narrative was a good bet.
In retrospect, it proved quite difficult to imagine and implement business models that turned the internet, the next big thing, into profitable businesses. As a young business school professor, David would ask his students questions like “How are entrepreneurs expecting to ‘appropriate’ or capture part of the value that was being created by the internet?” Students often responded that generating a positive bottom line was no longer a relevant business metric. Investors and entrepreneurs were fighting for “eyeballs,” not dollars. These entrepreneurs, analysts, and investors (and, apparently, students) believed that they understood the new economy. It was an urgent land grab, and the land was inherently, inevitably valuable. This confidence is puzzling, given that in the late 1990s few dot-com businesses had generated profits. There was still profound uncertainty about how to value them. It was not merely unknown if and how such metrics would translate into bottom-line profits—it was unknowable.17
The eToys story epitomizes the interaction of unknowability and consequent narratives that are used to divine the unforeseeable future. Understanding this interaction provides clues as to how to identify when a bubble is occurring and, perhaps, how to avoid the most destructive excesses of rampant speculation. For a given opportunity, is it known which business models will be profitable? Can we identify why entrepreneurs, investors, and analysts believe what they believe? Are such beliefs based on real, relevant past experience, or are they simply guesses? Do the players proclaim the future with certainty? Are investors and entrepreneurs making similar bets based on the same emergent, urgent narratives built on flimsy foundations? Do they all look to one another for social proof they are doing the right thing?
If this first set of questions explores attributes of a given opportunity, a second set asks who is investing. For any asset or class of assets, if many novice investors are investing when asset values are fundamentally unknowable, this is reason for concern. Such investors are unlikely to have access to information that would allow them to provide sound reasons to be bullish and are more likely to make decisions based on what others have told them. That is, novice investors are unlikely to understand what is unknowable. Thus, understanding who else is investing and why is critical to making an informed evaluation of whether an asset or class of assets is being traded at unjustifiably inflated prices.
While we hope you find this interpretation of the dot-com bubble intriguing, generalizing from a single convincing story is unwise. There are many problems with making the leap from statements like “entrepreneurs didn’t know how they were going to convert eyeballs into profits” and “there were novices investing in dot-coms” to a causal statement such as “there were novices investing in dot-coms who thought they understood how dot-com entrepreneurs would convert eyeballs into profits, and this was a significant factor in causing the bubble.” This leap requires not only a plausible cause-and-effect argument that links investor type and beliefs as well as the nature of uncertainty to investment decisions and asset prices, but also some “counterfactual” evidence to convince us that the dot-com bubble might have been avoided altogether in the absence of novice investors and the narrative that emerged.
More generally, one strategy to help convince a skeptical reader would be to demonstrate that novice investors were systematically not investing in the companies commercializing early-stage technologies that were not associated with bubbles, and conversely, that novices were active investors in new industries that experienced bubbles. We would then need to demonstrate that when novices were present but there were no compelling narratives, bubbles were less likely to form. To find examples of each of these situations, we would need to sample across a wide range of assets with varying financial histories. This exercise is the intellectual journey of this book.
Our principal methodological challenge is fundamental to the scientific method: identifying causal links requires that we observe instances when the outcome of interest does not happen. For example, imagine that we wanted to breed faster thoroughbreds and so examined the dietary histories of all horses that had won the Triple Crown. Further, imagine we discovered that most Triple Crown winners were found to have received more oats and grains than vegetables. Is this sufficient to change the recommended diet of all racehorses? Hopefully not. It could be that the horses that finished last in every Triple Crown race also received more oats and grains than vegetables. To conclude that diet was an important causal determinant of the outcome of the races, we would need to compare the diets of winning and losing horses, and show that horses that won had different diets from those that lost.18 Similarly, identifying causal factors requires an analysis of assets that were associated with speculative episodes and those that were not associated with speculation at all. Although there are many prior studies that relate the theory of market speculation to the existence of a bubble, we have been unable to identify studies that systematically compare such speculative episodes to historical instances when broad-based market speculation might have occurred but did not.19
To do so, we need a class of assets that appears to be at similar risk of sparking speculative episodes. The category “major technological innovations” meets our requirements. Major technological innovations, as defined in the literature on long waves in economic activity, are interesting and important precisely because they are hypothesized to be economically and socially significant.20 We examine a subset of major technological innovations identified in the long-wave literature so as to observe when bubbles do and do not occur. Then, we relate those observations to, among other things, whether novices were present and whether technological narratives were available that might have aligned investors’ and entrepreneurs’ beliefs in support of speculative activity. In this way we identify robust conditions for the appearance of a bubble.
We analyze fifty-eight major innovations appearing between 1850 and 1970 that may or may not have led to speculative activity. For each, we delve into the history of the innovation and its commercialization—with a particular focus on the uncertainty surrounding how entrepreneurs and businesspeople would make money in the emergent industries. Such uncertainty accompanied many, though not all, new technologies. We then examine the contemporaneous press coverage and historical accounts to understand how entrepreneurs, investors, and the public perceived the market opportunities associated with the innovation. Which types of technology and investment narratives could a given innovation support? We provide the list of technologies in Table A.1 in the Appendix. The table has many fields, which we describe in the forthcoming chapters.
Our interpretation of investment activities would be incomplete without a close examination of the market institutions of the day. Many technology stocks were floated in the early part of the twentieth century when financial market regulation was nonexistent, and trades were literally conducted “on the curb” outside the New York Stock Exchange building in Lower Manhattan. The historical contexts help us understand the level of market access enjoyed by different classes of investors, and understanding the nature of the technology and its related narratives provides windows onto investor composition and entrepreneurial beliefs.
Early on in our study, we discovered important practical barriers to the identification of bubbles associated with the introduction of new technologies. First, there was no comprehensive database of stock market movements that covered the periods of introduction of such profoundly important technological innovations as the telephone or the steel industry. Sometimes, though, we were able to supplement our use of existing databases with indices derived from primary sources. Second, our focus on beliefs and the narratives that string them together required a similar window into public perceptions of the various technologies under study, one that allowed for cross-technology comparisons to find the presence or absence of bubbles, as well as the identification of events that may have coordinated beliefs about the promise (Charles Lindbergh’s successful transatlantic flight) or limitations (the Hindenburg disaster) of a new technology. Understanding these narratives required a careful reading of contemporaneous press accounts. It is doubtful that this exercise would have been possible without the digitization of historical newspapers. Our next step is to clarify precisely our definition of a bubble, then outline when we think bubbles are more likely to occur.
Bubbles, Booms, and Busts
A bubble refers to the rise and fall in asset prices such that prices deviate from “fundamental” or “intrinsic” value. Defining “fundamental” value is hard, so financial economists have tried to tie it to something real, the asset’s future discounted returns. This is easy when considering a bond with a fixed interest rate but much harder to think about when we consider a new, highly uncertain start-up.
But we are getting ahead of ourselves. Simply predicting rises and falls in asset prices—which we call boom and bust episodes—would be sufficient for any practical use. However, such cycles are much more interesting when the price movements fail to reflect underlying intrinsic value; that is, when they are irrational, inspired by “animal spirits” or the “madness of crowds.” Financial economists call such episodes “bubbles,” and so will we.21
Distinguishing between bubbles and mere boom and bust cycles requires a statement about the rationality of traders. This in turn requires some idea of what might have been reasonable to believe at the time trades were made. One has to have a theory of what is reasonable to believe about a future profit stream. The problem is, though, that one can come up with a justification to explain any price as rational. For example, if one has good reason to believe that the $7.7 billion eToys valuation in February 2000 was a reasonable assessment of eToys’ future profits from selling toys on the web, then the eToys episode is properly classified as a boom and bust cycle, not a bubble. In general, many stories are plausible in highly uncertain settings. To quote the famed baseball philosopher Yogi Berra, “It’s tough to make predictions, especially about the future.”22 This prediction challenge has led to claims that even the most excessive price fluctuations, such as those of the dot-com bubble, were not examples of irrational exuberance but measured decisions of thoughtful traders.23 Such arguments rely on an options-based logic that suggests that prices should increase with uncertainty; in this view, high prices reflect the possibility that a given venture might be the next General Electric or Apple while also taking into account the fact that losses are limited—stock prices can’t fall below $0. However, rational theories do not explain why the presence of novice investors increases the likelihood of the phenomenon, nor do such accounts square well with contemporaneous descriptions of bubbles and other market anomalies. They do not incorporate the role of narratives and stories in human decision making. While we will be more precise about these arguments and our definitions in later chapters, we use the term “boom and bust episode” to refer to a substantial increase and subsequent decrease in prices. We label such an episode a “bubble” if we find that the boom and bust occurred at a time with a substantial influx of novice investors and was also accompanied by identifiable narratives.
Causal Factors
What causes technology bubbles? Inevitably, this is the bottom-line question that drives our study, haunts investors in their sleep, and has brought you this far. As noted already, we can offer only probabilistic statements. We identify four principal factors that, taken together, increase the likelihood of a speculative bubble forming around a given technological innovation: the nature and degree of uncertainty surrounding the innovation, the existence of “pure-play” firms whose fortunes are tightly coupled with the commercialization of the innovation, the availability of narratives that coordinate and align beliefs about the likely development of the innovation, and the presence of novice investors to fund those firms. We take up each of these factors in depth in the body of the book but give a brief overview here.
Uncertainty
The arrival of a major technological innovation is often associated with uncertainty about how firms will capture value from the innovation and which firms will profit. The financial economics literature has suggested that bubbles are more likely to occur under greater uncertainty and that speculation will end as this uncertainty is resolved.24 If positive beliefs are both pervasive and, in hindsight, misplaced, then a boom and a bust will follow. In retrospect, this will appear to be speculative.25 Unfortunately, existing research says little about how uncertainty will manifest in the context of new technologies, and if and to what extent institutional and market features will mitigate or exacerbate the effect of uncertainty on the likelihood of a speculative bubble forming.
For example, there might be considerable uncertainty regarding which business model will prove to be an advantageous means to exploit a new technology.26 A business model describes the way businesses will make money selling or using the new technology. It depends on the entire economic system used to deliver value to the end user. Do the best opportunities come from selling cars to consumers or tires to car manufacturers? Although it might appear counterintuitive, when investors have trouble understanding how a new technology will fit into this system, or alternatively, when it is surmised that a new technology might displace extensive portions of a value chain, then this will encourage investment. If there is uncertainty about which part of the value chain will be able to appropriate returns, then we can rest assured that there will be a variety of opinions, and those opinions will be woven into stories justifying investment. Moreover, if firms are replacing greater proportions of a value chain, they may have a better chance of appropriating more value. Different types of investors will get caught in the different webs of stories generated to make sense of each idea about capturing value. This dynamic will push up the entire sector.27 For example, in the case of radio, it was unclear how anyone would make money in broadcasting. In the early 1920s department stores produced broadcasts as a loss leader to attract customers to their store, and the Radio Corporation of America (RCA) began broadcasting as a means to increase demand for its primary product, radio sets. But this also encouraged entry of dozens of independent radio broadcast and receiver producers, and the airwaves were quickly filled with many stand-alone, privately financed radio stations. Contemporaneous observers did not know whether great profits would emerge in broadcasting, radio production, or the production of radio broadcast equipment, although there were opportunities to invest in any of those segments. This variation may have appealed to different investor segments, thereby increasing overall demand for stock in the sector.28
Similarly, electric lighting was demonstrably useful and a sight to behold when all one had experienced was lower quality gas lighting.29 It was first introduced before a metering technology existed and before it was well understood whether electricity should be transmitted using direct or alternating current, or for that matter, whether value would be appropriated by light-bulb producers or electricity suppliers.30 It was also unknown whether electricity would most profitably be sold on a per-light, per-watt, or subscription basis. Different firms and their subsidiaries each pursued different potential solutions (e.g., Brush, Edison, Westinghouse).31
Counterintuitively, knowing who might profit from an innovation might reduce the likelihood of a bubble. Because all bets are tied up in one firm, the bet is more closely aligned with the success of the technology, as opposed to different segment or monetization strategies associated with the new technology.32 There is less room for competing narratives to appeal to different populations and thereby drive up the entire sector. For example, once the US Supreme Court upheld Alexander Graham Bell’s broad patent claims on the invention of the telephone, uncertainty surrounding the fate of other inventors’ claims was reduced. Bell had successfully prevented their entry into the market. Thereafter, the expected value of their ideas and ventures decreased, even if the exact business model that American Telephone and Telegraph (AT&T) would follow had yet to become clear.33 In general, strong intellectual property protection may reduce uncertainty regarding who will profit, even before the precise mode of profit is known.
Uncertainty is necessary for the existence of a boom and bust episode. Without it there are no surprises, and hence neither booms nor busts.34 As we discuss in further detail in Chapter 2, technological innovation is not the only source of uncertainty, but uncertainty is the sine qua non for the formation of a bubble. Uncertainty does not last forever. We expect the likelihood of asset bubbles to wane as appropriate business models are discovered, and it becomes clear who will profit. These periods map closely onto stages in industry evolution that are identified in the strategic management literature.35
Pure Plays
For a bubble to form, pure-play firms—firms tightly coupled to the commercial fate of the technology or innovation—must exist, and investors must be able to buy and sell shares in them. This factor highlights several important features of the landscape that predict the presence or absence of a bubble. First, the existence of pure plays is tied to the degree of uncertainty. Uncertainty is higher when it is not understood whether the skills and capabilities of existing firms will be necessary or useful in the commercialization of a new technology. The presence of pure-play firms indicates that uncertainty may be exploitable by new entrants. Second, pure plays make good stories. Given an interest in, say, electric vehicles, it is more exciting to invest in Tesla than in General Motors, despite the fact that both companies are deeply involved in the electrification of transportation. Conversely, the public and the media are less likely to attend to technology stories that lack a pure-play protagonist. Finally, for a bubble to form, there must be a way for investors to literally buy into the story. This point emerges from our sampling methodology of technologies. Many important technologies were not commercialized by publicly traded companies, or if they were, the companies’ fortunes were broadly diversified. If there are no tradable financial assets that closely track the fortunes of the technology, then there can be no market speculation. Without a pure-play investment opportunity in a given technology, it is simply not possible for a speculative bubble to form for that technology. Simply put: a market must already exist for there to be a market bubble.
Coordination or Alignment of Beliefs Through Narratives
As pointed out in theories of herding and in studies of fads and fashions, beliefs must be sufficiently focused to drive up the value of a class of assets; investors with heterogeneous beliefs must become aware of the opportunity to participate in an emerging market for a new technology. On the one hand, attention must be focused on a particular market. On the other hand, uncertainty is necessary. Bubbles are rarer when attention is focused on a single means of generating returns and more likely when there is uncertainty about how to exploit the new opportunity.
Beliefs are coordinated through stories that circulate in the media and among investors. These stories or narratives piece together different facts, ideas, and guesses about a new technology and its potential profitability. Toby Lenk of eToys told a compelling story that was believable because of the uncertainty surrounding the viability of e-commerce and whether niche players could survive in online retailing. Some ideas and technologies are better subjects of narratives. It was easier to tell a story about human flight than the world’s first synthetic plastic, Bakelite. The degree to which technologies lend themselves to storytelling is an important factor in driving bubbles.
The arc of a narrative is often propelled or stalled by particular actors and events. There are many historical examples of events that appear to have propelled narratives by aligning investor beliefs about the potential profitability of an opportunity. For instance, President James K. Polk, prior to the California gold rush, publicly confirmed the veracity of the rumors of gold in California in his State of the Union address.36 Similarly, Charles Lindbergh’s transatlantic flight was followed by a wave of 127 IPOs of airline and aircraft-related stocks, just like in 1995 the successful Netscape IPO brought increased attention to internet opportunities.37 The Hindenburg disaster halted interest in airships. For other technologies, such as polyester or the laser, neither of which generated a boom and bust cycle, we find no associated coordinating event and no set of plausible entrepreneurial narratives.
Novice Investors
The fourth and final causal factor that contributes to the likelihood of speculation is the presence of novice or unsophisticated investors. Overoptimism or overconfidence may lead to poor buying decisions, thereby increasing demand for risky assets. Certain populations may be especially vulnerable to such biases. This thinking dates back at least to 1841, with Charles MacKay’s Extraordinary Popular Delusions and the Madness of Crowds.38 Contemporary scholars have explored this idea and observed that investors possess different levels of sophistication. Less sophisticated investors, sometimes called “noise traders,” may be overly bullish, and individual traders appear less sophisticated than professional investors.39 We expect that noise traders are especially likely to invest when the technology or its application is something they can understand, even if it is unclear how one might profit from the new technology. For example, in the late 1990s it was evident to the casual observer or investor that the internet was useful, although it was unclear who might profit from its adoption and how.
With this in mind, we argue that potential investors are more likely to buy an asset when they believe that they understand how value will be appropriated. If investors are more likely to invest in something they think they understand than in something they do not, then we suspect that at a minimum, the commercial potential of an innovation, or at least its usefulness, needs to be comprehensible and accessible to the investor. For example, relatively obscure developments in science such as the Nipkow disk in 1885 did little to stimulate the public imagination, despite the fact that the innovation was critical to the eventual development of television. In contrast, the public broadcast of the Metropolitan Opera on the radio in 1922 was accessible to the general investor and may have helped stimulate and align investor beliefs about the commercial prospects of radio.40 Thus, retail-facing innovations may be more likely to grab the attention of a broad set of investors, even when that retail-facing quality is not perfectly correlated with profitability.41 This observation is in line with evidence that individuals tend to invest in assets with which they are familiar.42 We expect (and find) that speculative activity is more likely in innovations or ideas that are familiar and understandable based on common experience. The role of familiarity is exacerbated when the arrival of a retail-facing innovation coincides with an influx of novice or unsophisticated investors.43
The ebb and flow of new investors depends on many factors. Of course, each generation brings new investors to the market. Other factors, such as new investment technology (e.g., the stock ticker, E*Trade) or changes in regulation (e.g., bans on insider trading or the Jumpstart Our Business Startups Act and its influence on crowdfunding in the United States), may increase the influx of novices. To assess the importance of novices, we proceed with direct and indirect measurement. We piece together estimates of the number of households investing across our time period. We then supplement this direct measurement by developing a timeline of innovations and structural changes that increase (or decrease) market access for equity investments. These supply-side innovations lower barriers to entry for investors and reduce transaction costs. To help quantify this across our time periods, we put together a long-term series of the months of labor it takes the average worker to buy the average share traded on the New York Stock Exchange. We put these factors together to identify periods in which the number of possible participants in a financial market increase, thereby allowing us to identify influxes of novice or unsophisticated investors into markets. We term this process “market democratization.”
Moreover, the performance of the market itself will attract or discourage investors. A bull market will attract more novices, and a level of optimism will persist among participants who have yet to experience a bear market. In contrast, a bear market will not only drive investors from the market but also discourage new entrants. The most dramatic of these events is the bull market of the 1920s and the investment desert that prevailed during the Great Depression. We summarize major events in the democratization of investment in Chapter 3.
While the first factor, uncertainty, may lead to rational boom and bust episodes, the fourth factor, the presence of novices, is associated with bubbles: rational models from financial economics do not explain why the presence of novices might be associated with price fluctuations. As we discuss in Chapter 3, these two factors, uncertainty and novice investors, may interact in ways that exacerbate the likelihood of a bubble, because uncertainty exacerbates the liabilities of inexperience in investing.
An Illuminating Example
A single example should never convince us of the importance of these institutional features—there are simply too many other factors that can plausibly explain one event. Nevertheless, such an exposition can illustrate our approach. We develop the four-factor model for two similar cases: the commercialization of Brush electric arc lighting in Cleveland and in London.
Electric lighting, a novel and, to contemporary observers, amazing technology, was demonstrated in Cleveland, Ohio, on April 29, 1879, when Charles F. Brush, backed by Cleveland financier George Stockly, lit up Public Square—then known as Monumental Park—with twelve arc lamps. As reported at the time, Brush’s streetlights turned night into day and were visible for miles. The demonstrations were widely covered in the media and served to coordinate beliefs around the potential of this marvelous new technology.44 A narrative emerged about the inevitability of electrical lighting. The eventually successful Brush Electric Company was capitalized at $3 million. Nevertheless, investors were still unsure how to profitably exploit the innovation. Given the general uncertainty surrounding the technology, and the difficulty investors may have had in assessing the ability of entrepreneurs to exploit it, investors looked for endorsements of prominent businesspeople. Indeed, the Brush Electric Company was funded by what would be known today as business angels. These wealthy investors, mostly Cleveland’s business elite, were connected to Brush through social networks.45 More generally, Brush Electric and its numerous competitors were financed through informal, private equity networks.
The success of the Brush company sparked entry. However, the market for new equity investment in quality firms was limited to these individuals. For example, Brush spin-off Linde was subscribed by “prominent Cleveland businessmen.”46 Not only were investors in these assets relatively sophisticated, at least by the test of using endorsements as signals of underlying quality; they also had strong incentives to make sure that the underlying assets were of long-term value. There was a very illiquid market for shares in early high-technology enterprises in Cleveland, as described by economic historians Naomi Lamoreaux, Margaret Levenstein, and Kenneth Sokoloff:
The wealthy Clevelanders who bought shares in these new high-tech enterprises seem to have been motivated by the returns they expected to earn from owning and holding them rather than the profits they could reap by selling them after an initial run-up in price. Although a few investors cashed out their investments relatively early, the practice seems uncommon. Before the formation of the CSE [Cleveland Stock Exchange] in 1900, the only firms associated with the Brush network for which share prices were quoted in Cleveland papers were Brush Electric itself and the Walker Manufacturing Company. Even after the formation of the exchange, we do not see much trading in equities of concerns associated with the hub. The one major exception, National Carbon, was listed on the exchange from the very beginning, but by that time it was a consolidation of a large number of previously competing firms.47
While it is clear that the most promising opportunities were funded through Cleveland’s angel investor network, it is possible that smaller, individual investors funneled money into ventures of inexperienced entrepreneurs (or worse). The same authors report that there were perhaps forty attempts by fly-by-night artists to raise money in pursuit of dubious electric lighting companies in Cleveland. However, there is little evidence that they raised much money.48 Although public beliefs were aligned in the presence of uncertainty, the structure of the Cleveland investment market limited the influx of new investors and stifled speculative activity.
A remarkably different history can be told about the Anglo-American Brush Company (AABC), founded in London in 1882. This company, established to commercialize Brush’s arc lighting system in Britain, generated a number of “little Brushes,” each receiving territorial exclusivity to establish central stations and supply lighting. Through a political process, monopoly rights were granted to central stations for a period of seven years, which, at the time, was predicted to provide sufficient time to generate a return for investors, although later, in August 1882, this was amended to twenty-one years under pressure from business interests. AABC became part of a larger speculative bubble in electric company assets in the spring of 1882. In the first five months of 1882, British electrical companies registered with authorized capital of £9 million, reflecting investments of £7 million. In mid-May shares of the Anglo-American Brush Electric Light Corporation dropped £600,000 in three days of trading (though they remained above par value).
Why was there a bubble in Britain but not in the United States? While it was clear that lighting was valuable, it was not clear in the 1880s which business model would sustain a lighting company. Would money be made on light bulbs or by selling electricity? (Electricity meters were not yet invented.) To what degree were inexperienced investors interested in this innovation? The historical record is clear that lighting generated interest, if not awe, among contemporary observers. Early entrepreneurs lit up prominent areas of both Cleveland and London (and other cities as well). However, there is reason to suspect that inexperienced investors had much greater market access in London than in Cleveland. The London Stock Exchange (LSE) was a very democratic institution that accommodated smaller, less sophisticated traders. First, commissions on the LSE were lower than on the New York Stock Exchange (NYSE). Second, perhaps more important, trades were settled in London every fortnight. Thus, London traders enjoyed a two-week “float”; they could “buy” for the account what they could not afford and sell short as well. This increased liquidity and allowed for greater speculation. While there was no market in Cleveland, stocks may have been floated in New York on the NYSE. However, in New York trades settled the following day. What’s more, the NYSE had a policy of monopoly that held down the number of securities that were traded.49 By contrast, the London exchange would list any security for which there was a market, and hence traded smaller companies’ shares. Even as late as 1914 the average capitalization of the NYSE was $24.7 million, whereas the average LSE listing was capitalized at one-fifth that amount (£1.03 million, or contemporaneously approximately $5 million). Importantly, the inability to list on the NYSE also made smaller-capitalization stocks less available as collateral for other margin purchases.
Through this example we see that even when the specific asset in question is the same—that is, investors in Cleveland and London were investing in the same underlying technological system—the potential for speculation can be determined by the institutional and organizational context through which investors access the relevant financial market. In this case, the specific features of the LSE supported, perhaps encouraged, speculation. In Cleveland, because shares of Brush and related companies were not readily traded, speculation was certainly harder to engage in, if not impossible.
The London bubble had deleterious effects on the British electric lighting industry and on the British economy more generally.50 In the aftermath of the bubble, and with the help of entrenched interests (gas lighting companies), the British Parliament passed the Electric Light Act of 1885. Not only did this law retard the adoption of the new lighting systems, but perversely, the more “developed” London capital markets set up the darkness exploited by Jack the Ripper. London remained dark well into the 1890s while other worldly cities such as Paris and New York were lit. This episode also deprived British entrepreneurs of valuable opportunities to move down the learning curve with the new technology. This latter cost is quite difficult to quantify.
The fact that public investors in London were able to invest easily in the electric lighting companies itself became part of the story of electric lighting. While journalists in America focused on the magic of the electric light, their British counterparts overlay on this a narrative of investment opportunities. This reporting, in turn, fed the narrative of speculation.
Road Map
In the chapters that follow, we summarize our studies of the history of the commercialization of fifty-eight major technological innovations. This sample—which we describe in greater detail in Chapter 1—contains variation in the features in which we are interested and allows us to generate the causal model we have described here. In the spirit of our previous work with David Miller, our analysis is stochastic.51 Even if all the potential causes are present, a bubble still may not form—and a bubble may form even if few potential causes are present. A successful theory of bubbles will identify factors that, when present, imply that a bubble is more likely to occur. We identify such factors. In no place do we claim or mean to imply that these conditions are sufficient or necessary to generate asset bubbles.
To evaluate whether our ideas have any external validity—that is, to see if the theory applies beyond our initial fifty-eight technology “training sample”—we then test the theory on a different set of thirty more recent technologies that includes the internet, laparoscopic surgery, and liquid crystal displays. Because at this point our framework was fixed, this exercise allowed us to assess how well the framework works outside the historical settings of the initial sample. We then consider whether the theory helps us understand recent events such as the housing crisis and the Great Recession (Chapter 5).
Our work puts the role of narrative at center stage. We cannot understand real economic outcomes without also understanding when the stories that influence decisions emerge and under which conditions they are most likely to be created. Much to the dismay of economists such as the Nobel laureate Robert Shiller, the history of much of economics has been an attempt to assume away the role of stories; they have no space for a rational decision maker.52 However, our analysis suggests that ignoring stories and narratives makes it much harder to understand important economic phenomena such as bubbles.
Independent of our theoretical interpretation, our basic finding—that certain major technological innovations are associated with speculative bubbles and others are not—affirms our methodological approach. This establishes a point of departure for subsequent debates about the possible causes of speculative behavior, regardless of whether one agrees with the specific conditions we describe. To understand the antecedents of bubbles, we must examine when there are and when there are not bubbles. Although we believe that our analysis and interpretation advance theoretical and practical understanding of the causes of bubbles, one might plausibly disagree with these implications, yet still accept the basic empirical framework we set forth. Other explanations of why some but not all major technological innovations lead to speculation are possible.
Notes
1. Robert J. Shiller, Irrational Exuberance (Crown Publishers, 2006). Toys “R” Us declared bankruptcy in March 2018 after failing to successfully compete in a transformed retail space. The failure of Toys “R” Us is attributed to increased competition not from specialized retailers such as eToys, but rather from general retailers such as Walmart and Amazon.
2. Erick Schonfeld, “How Much Are Your Eyeballs Worth? Placing a Value on a Website’s Customers May Be the Best Way to Judge a Net Stock. It’s Not Perfect, but on the Net, What Is?” Fortune, February 21, 2000, http://archive.fortune.com/magazines/fortune/fortune_archive/2000/02/21… index.htm. See also Brett M. Trueman, M. H. Franco Wong, and Xiao-Jun Zhang, “The Eyeballs Have It: Searching for the Value in Internet Stocks,” Journal of Accounting Research 38 (2000): 137–62.
3. Heather Green, “The Great Yuletide Shakeout,” Businessweek, November 1, 1999, 22.
4. Michael Sokolove, “How to Lose $850 Million—and Not Really Care,” New York Times, June 9, 2002; Erin Kelly, “The Last E-Store on the Block: Toby Lenk’s Toy Shop May Be the Best-Run Specialty Store on the Web, Which Raises a Question: If eToys Can’t Make Money Online, Can Anyone?” Fortune, September 18, 2000, http://archive.fortune.com/magazines/fortune/fortune_archive/2000/09/18… index.htm.
5. Kelly, “Last E-Store on the Block.”
6. Kelly, “Last E-Store on the Block”; “eToys to Shut Down Web Site March 8,” CNN, February 26, 2001, http://cnnfn.cnn.com/2001/02/26/companies/etoys/.
7. To see how powerful narratives can be in driving capitalist activity, consider that Amazon.com has sustained a price-to-equity ratio of 200 for more than 20 years, with profit margins hovering near 1%, one-quarter that of grocery stores. Notwithstanding Amazon’s modest profitability, its narrative has been wildly successful, so much so that when we hear the adjective Amazonian, we often cannot distinguish the river from the firm.
8. Ben Eisen, “Nasdaq Tops Inflation-Adjusted High from Dot-Com Boom,” Wall Street Journal, January 17, 2018, https://blogs.wsj.com/moneybeat/2018/01/17/nasdaq-poised-to-top-inflati….
9. Eli Ofek and Matthew Richardson, “Dotcom Mania: The Rise and Fall of Internet Stock Prices,” Journal of Finance 58, no. 3 (2003): 1113–38.
10. The contemporaneous and related telecommunications bust happened at the same time. See Ofek and Richardson, “Dotcom Mania”; S. Greenstein, “The Crash in Competitive Telephony,” IEEE Micro 22, no. 4 (July 2002): 8–9, 88.
11. For a discussion of the British Railway Mania, see Ofek and Richardson, “Dotcom Mania”; Greenstein, “Crash in Competitive Telephony”; Andrew Odlyzko, “Charles Mackay’s Own Extraordinary Popular Delusions and the Railway Mania,” SSRN eLibrary, September 14, 2011, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1927396. For a discussion of market speculation in the 1960s, see Robert Sobel, The Last Bull Market: Wall Street in the 1960s (Norton, 1980).
12. Sobel, Last Bull Market; Ofek and Richardson, “Dotcom Mania.”
13. Ruth Simon, “Margin Investors Learn the Hard Way That Brokers Can Get Tough on Loans,” Wall Street Journal, April 27, 2000, C1.
14. Robin Greenwood and Stefan Nagel, “Inexperienced Investors and Bubbles,” Journal of Financial Economics 93, no. 2 (August 2009): 239–58.
15. Based on author calculations using VentureXpert, June 2008, from Securities Data Company.
16. Brent Goldfarb, Michael D. Pfarrer, and David A. Kirsch, “Searching for Ghosts: Business Survival, Unmeasured Entrepreneurial Activity, and Private Equity Investment in the Dot-Com Era,” 2005, https://doi.org/10.2139/ssrn.825687.
17. It was not true that no one was making money. Like those selling the tools to miners during the gold rush, firms that relied on business models from the old economy did quite well. For example, purveyors such as the web development firm Scient, network infrastructure firm Cisco, Wall Street investment houses that earned a cut from IPOs, and commercial real estate agents all did quite well—at least for a time.
18. Of course, that would not be enough, either. What if owners who feed their horses more oats and grains than vegetables also live in more northern climates, and the colder weather there enhances their horses’ performance? Absent the ability to run a controlled experiment, we can look for other clues. For example, can we measure greater metabolic efficiency in the horses’ muscles and relate this to diet? Does variation in the diet of individual horses change their speed when training? We ask these types of secondary questions in this book as well.
19. Ours is the first to develop a rich historical and institutional narrative to study intermarket variation in the formation of bubbles. Some statistical studies do exploit intermarket variation to identify antecedents: e.g., Gerard Hoberg and Gordon Phillips, “Real and Financial Industry Booms and Busts,” Journal of Finance 65, no. 1 (February 1, 2010): 45–86; and Matthew Rhodes-Kropf, David Robinson, and S. Viswanathan, “Valuation Waves and Merger Activity: The Empirical Evidence,” Journal of Financial Economics 77 (2005): 561–603.
20. Christopher Freeman, Long Waves in the World Economy (F. Pinter, 1984).
21. S. F. LeRoy, “Rational Exuberance,” Journal of Economic Literature 42, no. 3 (2004): 783–804.
22. Fred R. Shapiro, The Yale Book of Quotations (Yale University Press, 2006).
23. Ľuboš Pástor and Pietro Veronesi, “Was There a Nasdaq Bubble in the Late 1990s?,” Journal of Financial Economics 81, no. 1 (July 2006): 61–100.
24. Lubos Pastor and Pietro Veronesi, “Learning in Financial Markets,” Annual Review of Financial Economics 1, no. 1 (2009): 361–81; Andrea Devenow and Ivo Welch, “Rational Herding in Financial Economics,” European Economic Review 403 (1996): 603–15; B. Goldfarb, D. Kirsch, and D. A. Miller, “Was There Too Little Entry During the Dot Com Era?” Journal of Financial Economics 86 (2007): 100–144.
25. Adherents to the efficient markets hypothesis suggest that boom and bust episodes can be explained in ways that are consistent with rational behavior. We consider this idea in Chapter 2.
26. We follow C. Zott and R. Amit, “Business Model Design and the Performance of Entrepreneurial Firms,” Organization Science (2007): 181, http://pubsonline.informs.org/doi/abs/10.1287/orsc.1060.0232, in which a business model is defined as “the content, structure, and governance of transactions designed so as to create value through the exploitation of business opportunities.”
27. Similar phenomena are observed in technology prize competitions, where cash prizes are able to elicit aggregate investments that appear to be above the total prize amounts. Kevin J. Boudreau, Nicola Lacetera, and Karim R. Lakhani, “Incentives and Problem Uncertainty in Innovation Contests: An Empirical Analysis,” Management Science 57, no. 5 (April 1, 2011): 843–63.
28. Gleason Leonard Archer, History of Radio to 1926 (Arno Press, 1971).
29. See Charles Bazerman, The Languages of Edison’s Light (MIT Press, 1999); “The Electric Light,” New York Times, April 22, 1878.
30. “Curious Features of the Electric Lighting Business,” Scientific American 53, no. 15 (October 10, 1885).
31. From the perspective of a financial economist, this uncertainty can be thought of as the variance surrounding the expected returns to the innovation, although in some cases, such factors affect the mean as well.
32. The option value also decreases with the elimination of competitive uncertainty (although this might be counteracted by an increase in the expected value).
33. Christopher Beauchamp, “Who Invented the Telephone? Lawyers, Patents, and the Judgments of History,” Technology and Culture 51, no. 4 (2010): 854–78.
34. Experimental economists have demonstrated bubbles in the lab when there is no uncertainty in fundamental value but there is uncertainty in what others might be willing to pay for an asset. Vernon L. Smith, Gerry L. Suchanek, and Arlington W. Williams, “Bubbles, Crashes, and Endogenous Expectations in Experimental Spot Asset Markets,” Econometrica: Journal of the Econometric Society 56, no. 5 (September 1988): 1119.
35. The resolution of business model uncertainty may be closely related to the emergence of a dominant design, a concept first introduced by J. Utterback and W. Abernathy, in “A Dynamic Model of Product and Process Innovation,” Omega 3 (1975): 638–56.
36. Richard Thomas Stillson, Spreading the Word: A History of Information in the California Gold Rush (University of Nebraska Press, 2006).
37. Mary O’Sullivan, “The Expansion of the US Stock Market, 1885–1930: Historical Facts and Theoretical Fashions,” Enterprise and Society 8, no. 3 (2007): 489–542; Goldfarb, Kirsch, and Miller, “Was There Too Little Entry?”
38. Charles Mackay, Memoirs of Extraordinary Popular Delusions and the Madness of Crowds, 2nd ed., vol. 1 (1841; Office of the National Illustrated Library, 1852).
39. See J. Bradford De Long, Andrei Shleifer, Lawrence H. Summers, and Robert J. Waldmann, “Noise Trader Risk in Financial Markets,” Journal of Political Economy 98 (1990): 703–38. While many models rely on restrictions of short selling to limit the ability of rational traders to sell overvalued stocks, more recent work suggests that some sophisticated traders who possess “strategic IQ” may be better than others at anticipating novice irrationality, then buy overvalued stocks with the expectation of selling them. Sheen S. Levine, Evan P. Apfelbaum, Mark Bernard, Valerie L. Bartelt, Edward J. Zajac, and David Stark, “Ethnic Diversity Deflates Price Bubbles,” Proceedings of the National Academy of Sciences of the United States of America 111, no. 52 (December 30, 2014): 18524–29, examine how the diversity of market participants can attenuate bubbles. See also Brad M. Barber and Terrance Odean, “All That Glitters: The Effect of Attention and News on the Buying Behavior of Individual and Institutional Investors,” Review of Financial Studies 21, no. 2 (April 1, 2008): 785–818.
40. We think of novices broadly: they may be new to investing, or alternatively, they may be new to a particular investment environment or asset class and overestimate similarities between investment environments or asset classes they understand and those they do not. To wit, we do not predict that investors with experiences in a new class of assets that is substantially similar to another in which they are experienced will exhibit investor sentiment. For example, biotechnology investors investing in nanotechnology may behave rationally. Instead, we suggest that the key insight that allows investors to avoid sentimentality is an understanding of what they do not understand about particular assets. Savvy investors understand what they know and what they do not know and make decisions accordingly. Less savvy investors fail to understand what they do not know and are in this sense overconfident. Such investors may mistakenly make analogies between particular asset classes, which may lead to more pricing errors.
41. There are, of course, exceptions to this, as in the case of electronics firms that served the growing military industrial complex. We note in our discussion of the transistor that Sputnik may have been a coordinating event.
42. Gur Huberman, “Familiarity Breeds Investment,” Review of Financial Studies 14, no. 3 (July 1, 2001): 659–80; R. Zeckhauser, “Investing in the Unknown and Unknowable,” Capitalism and Society 1, no. 2 (2006): 5.
43. It may also manifest in heterogeneity in beliefs across investor classes. Different information sets can be attributable to information asymmetry. For example, some investors may have access to insider information before it is available more generally. A strong argument can be made that this was of significant importance in explaining the bubble in the gold market that led to Black Friday on September 24, 1869. Charles Poor Kindleberger and Robert Z. Aliber, Manias, Panics, and Crashes: A History of Financial Crises (John Wiley & Sons, 2005). Others have linked the crash of dot-com stocks to the expiration of lock-up provisions of insiders (Ofek and Richardson, “Dotcom Mania”), although this has been more recently challenged (LeRoy, “Rational Exuberance”). Conceptually, the presence of novice or unsophisticated investors may be thought of as increasing the mean expected return for a given innovation because expected returns of the more naïve investors are biased.
44. Carolyn Marvin, When Old Technologies Were New: Thinking About Electric Communication in the Late Nineteenth Century (Oxford University Press, 1990); Wolfgang Schivelbusch, Disenchanted Night: The Industrialization of Light in the Nineteenth Century (University of California Press, 1995).
45. Naomi R. Lamoreaux, Margaret C. Levenstein, and Kenneth Lee Sokoloff, “Financing Invention During the Second Industrial Revolution: Cleveland, Ohio, 1870–1920,” in Financing Innovation in the United States, 1870 to the Present, ed. Naomi R. Lamoreaux and Kenneth Lee Sokoloff (MIT Press), 50.
46. Lamoreaux, Levenstein, and Sokoloff, “Financing Invention During the Second Industrial Revolution,” 50, 57.
47. Lamoreaux, Levenstein, and Sokoloff, “Financing Invention During the Second Industrial Revolution,” 57. The lack of liquidity for assets in the market for early electric company assets is striking: “Completely missing from the [description of financing of new electric ventures] is any formal role for formal financial institutions in the founding of the original Brush Electric Company or the many startups and spin-offs that came out of [the] Brush cluster. The entrepreneurs who organized and promoted these new ventures secured investment capital largely by relying on personal connections. These could be familial, as when the father of Eugene and Alfred Cowles provided much of the initial capital for the Cowles Electric Smelting and Aluminum Company; they could result from friendships, as when George Stockly agreed to support Brush’s initial work in electrical lighting; or they could be based on the recommendations of men who had established their expertise in the community, as when Brush secured backing for the Linde Air Products Company simply by assuring local businessmen of the merits of the technology. Association with a hub enterprise such as Brush could in and of itself be a means of attracting both attention and funds. Thus, Bentley and Knight, as well as Short, were able to use their very visible association with Brush to raise local capital for their streetcar companies.” Lamoreaux, Levenstein, and Sokoloff, “Financing Invention During the Second Industrial Revolution,” 56–57.
48. Lamoreaux, Levenstein, and Sokoloff, “Financing Invention During the Second Industrial Revolution,” 50.
49. See Ranald C. Michie, “The London and New York Stock Exchanges, 1850–1914,” Journal of Economic History 46, no. 1 (March 1, 1986): 171–87. While there may have been a speculative episode on the Curb or Consolidated exchanges in New York, it was underdeveloped during this period, and we have found no evidence in the historical literature of such an event. These exchanges were outdoor, unregulated markets that traded in less established securities. The Consolidated Exchange folded, and the Curb moved indoors in 1922, becoming the American Stock Exchange in the 1950s. Robert Sobel, The Curbstone Brokers: The Origins of the American Stock Exchange (Beard Books, 2000).
50. Thomas Parke Hughes, “British Electrical Industry Lag: 1882–1888,” Technology and Culture 3 (1962): 27–44.
51. Goldfarb, Kirsch, and Miller, “Was There Too Little Entry?”
52. Robert J. Shiller, “Narrative Economics,” American Economic Review 107, no. 4 (2017): 967–1004.