The collapse of the investment bank Lehman Brothers on Sunday, September 14, 2008, caught almost everyone by surprise. It surprised investors, who dumped stocks and brought the market index down by 500 points on Monday. It surprised policymakers, who rushed to rescue other financial institutions after declaring for months that there would be no government bailouts. It also surprised economic forecasters. Only six weeks before the Lehman bankruptcy, in early August 2008, both the Federal Reserve and professional forecasters predicted continued growth of the U.S. economy. Contrary to that prediction, the U.S. financial system nearly melted down after the Lehman bankruptcy, and the economy slid into a deep recession.
Why was the Lehman crisis such a surprise? After all, fragility had been building up in the financial system for quite some time. In the mid-2000s, the U.S. economy went through a massive housing bubble. As home prices rose, households levered up to buy homes with mortgages. Banks and other financial institutions levered up to hold mortgages and mortgage-backed securities. As the bubble deflated after 2006, the financial system experienced considerable stress, as reflected in runs on financial institutions, followed by bankruptcies, rescues, and mergers. Yet the system and the economy stayed afloat until the fall of 2008, supported by successful interventions by the Federal Reserve aimed to avoid a financial panic. By mid-2008, investors and regulators expected that, despite the deflating housing bubble, the situation was under control. On May 7, 2008, Treasury Secretary Henry Paulson said that “the worst is likely to be behind us.” On June 9, 2008, Fed Chairman Ben Bernanke stated that “the danger that the economy has fallen into a ‘substantial downturn’ appears to have waned.”
The relative quiet before the storm, expressed in both the official and private-sector forecasts of the economy and the speeches of government officials, gives us important clues as to why Lehman was such a surprise. It surely was not the news of Lehman’s financial weakness per se, since the investment bank was in trouble and expected to be sold for several months prior to its September bankruptcy. U.S. banks more generally were making large losses for several months as the housing and mortgage markets deteriorated, and no major economic news surfaced that weekend. Nor can the surprise be attributed to the government’s reiteration of its “no bailout” policy. For if that were the reason for the collapse, the markets would have bounced back as soon as it became clear on Monday that bailouts were back in. In fact, markets bounced around a bit but continued their slide as the financial system deteriorated over the next several weeks, despite all the bailouts.
The evidence about the beliefs of investors and policymakers instead tells us that the news in the Lehman demise was the extreme fragility of the financial system compared with what was previously thought. Despite consistently bad news over the course of 2008, investors and policymakers came to believe that they had dodged the bullet of a major crisis. The pressures building up from home price declines and mortgage defaults were attenuated by the belief that the banks’ exposure was limited and alleviated by effective liquidity support from the Fed. The risks of a major crisis were neglected. The Lehman bankruptcy and the fire sales it ignited showed investors and policymakers that the financial system was more vulnerable, fragile, and interconnected than they previously thought. Their lack of appreciation of extreme downside risks was mistaken. The Lehman bankruptcy had such a huge impact because it triggered a major correction of expectations.
A decade after Lehman’s collapse, economists agree that the underestimation of risks building up in the financial system was an important cause of the financial crisis. In October 2017, the University of Chicago surveyed a panel of leading economists in the United States and Europe on the importance of various factors contributing to the 2008 global financial crisis. The No. 1 contributing factor among the panelists was the “flawed financial sector” in terms of regulation and supervision. But the No. 2 factor among the 12 considered, ranking just below the first in estimated importance, was “underestimation of risks” from financial engineering. The experts seem to agree that the fragility of a highly leveraged financial system exposed to major housing risk was not fully appreciated in the period leading up to the crisis.
These judgments are made with the benefit of hindsight. The world, however, has witnessed an extensive history of financial bubbles, expanding credit, and subsequent crises as the bubbles deflated. Errors in beliefs appear in multiple narratives. Classic studies such as Kindleberger (1978), Minsky (1977), and more recently Reinhart and Rogoff (2009) argue that the failure of investors to accurately assess risks is a common thread of many of these episodes. Rajan (2006) and Taleb (2007) stressed the dangers from low probability risks to financial stability. Even before the Lehman bankruptcy, Gerardi et al. (2008) drew attention to expectation errors in the developing subprime crisis. Since the 2008 crisis, a great deal of new systematic evidence on credit cycles, both for the United States and worldwide, has been assembled, starting with the pioneering work of Greenwood and Hanson (2013). Much of this work points to errors in expectations over the course of the cycle. We take this point of view further and put inaccurate beliefs at the center of the analysis of financial fragility.
Our argument proceeds in three steps. First, we show that survey expectations data are a valid and extremely useful source of information for economic research. Expectations in financial markets tend to be extrapolative rather than rational, and this basic feature needs to be integrated into economic analysis. Second, we provide an empirically motivated and psychologically grounded formal model, called diagnostic expectations, that can be used across a variety of domains. Third, we use this model of expectation formation to account for the central features – including both market outcomes and beliefs – of the 2008 crisis both before and after Lehman and to explain credit cycles and financial fragility more generally. Getting the psychology right allows us to shed light on the conditions under which financial markets are vulnerable to booms and busts.
SURVEY DATA ON EXPECTATIONS
A natural starting point for assessing the significance of financial “instability from beliefs” is to analyze the beliefs themselves. This entails not only directly measuring expectations of market participants and systematically testing whether these beliefs are rational, but also characterizing the type of mistakes that investors make.
This enterprise is feasible because a wealth of available survey data reports the beliefs of investors, corporate managers, households, and professional forecasters. This data offers important insights on whether, in 2008 and in other historical episodes, investors appreciated the risks building up before the crisis or alternatively failed to see the trouble coming. More generally, survey data helps identify regular patterns in beliefs during economic fluctuations, needed to develop better theories of expectation formation and credit cycles.
For the period leading to the 2008 crisis, we have a good deal of data on the expectations of homebuyers about future home price growth, on investor beliefs about the risk of home price declines and mortgage defaults, and on forecasts of economic activity made by both private forecasters and the Federal Reserve. We also have a variety of contemporaneous documents and speeches of policymakers, as well as discussions at the Federal Open Market Committee (FOMC) meetings, which shed light on the beliefs of policymakers. We can then ask directly: What were homebuyers, banks, investors, and policymakers thinking as the events leading up to the crisis unfolded?
The answers to this question cast doubt on the “too big to fail” theory of the crisis, which holds that the banks knew the risks but gambled on bailouts. The expectations of bank executives and employees seem to be very similar to those of other investors. Bankers were optimistic about housing markets and made loans as well as personal home purchases accordingly. There is no evidence that bankers understood the risks better than anybody else.
Beliefs are more in line with the classical analyses of Kindleberger (1978) and Minsky (1977) that emphasize excessive optimism before crises. Homebuyers were unrealistically optimistic about future home price growth. Investors in mortgages and in securities backed by these mortgages, including financial institutions, considered the possibility that home prices might fall but did not fully appreciate how much and what havoc these declines would wreak. And macroeconomic forecasters from both the private sector and the Federal Reserve did not, in forming their expectations, recognize the risks facing the U.S. financial sector and the economy as late as the summer of 2008. The evidence does not suggest that investors or policymakers were totally naive or oblivious to the risks in the financial system. Rather, they did not fully appreciate tail risks until the Lehman collapse laid them bare.
The data on beliefs prior to the Lehman crisis point to two key patterns: the extrapolation of past home price growth into the future, and the neglect of unlikely downside risks. Extrapolation of past home price growth sheds light on the housing bubble. Neglected downside risk explains how the financial system became so leveraged. This levering up of both households and financial institutions was most plausibly supported by the widely shared beliefs that the prices of homes were unlikely to collapse and that financial institutions were protected from bad shocks by diversification and hedging before the 2008 crisis.
Looking at beliefs data also sheds light on financial fragility more broadly, beyond the financial crisis. A great deal of survey data on investor and professional forecaster expectations about not only stock markets, individual stocks, and credit markets, but also the real economy is available and can be examined. We argue that extrapolation of past trends is in fact a common feature of expectations held by investors, corporate managers, and professional forecasters. The neglect of downside risk is present in several documented instances of financial innovation. The kinds of patterns we see in 2008 appear in other financial and economic episodes.
The empirical facts on expectations in the run-up to and during the 2008 financial crisis present a challenge for the standard theory of rational expectations. Nor can naive theories of irrational beliefs explain how extrapolation and neglected downside risk are connected and how they come and go. Adaptive expectations, a theory of mechanical extrapolation of past trends, can explain the growth of the housing bubble but not why the system stayed afloat after the bubble started deflating in 2006 or why a single event such as the failure of Lehman induced such a drastic revision of expectations.
We present one psychologically founded theory of expectation formation, which we call diagnostic expectations. We have developed this theory over the past several years together with Pedro Bordalo and have taken it both theoretically and empirically to a number of different domains with Katherine Coffman, Yueran Ma, and Rafael La Porta. In developing this model of expectations, we are guided by several principal considerations. First, we would like a theory of beliefs to be biologically and psychologically plausible, and in particular based on the evidence on human judgment obtained in experimental data. Next, we would like the same theory to explain evidence in psychological experiments, social judgments individuals make, financial markets, and perhaps other domains. Finally, we would like a theory in which beliefs are forward-looking and a theory that can be testable using survey evidence. The theory is surely not the last word in modeling expectations, but it suggests that one can make some progress in understanding the reality of financial markets.
Our model of expectations builds on the famous representativeness heuristic of human judgment under uncertainty initially proposed by psychologists Daniel Kahneman and Amos Tversky in 1972. According to Kahneman and Tversky (1983), “an attribute is representative of a class if it is very diagnostic, that is, if the relative frequency of this attribute is much higher in that class than in a relevant reference class.” Representativeness entails a judgment error of overestimating the likelihood of representative attributes in a class.
To illustrate, suppose someone is asked to predict the most likely hair color of an Irish person. In several informal surveys we conducted, many people said red. It is absolutely the case that red hair is objectively more common among the Irish than among other humans: 10% of the Irish have red hair, compared to 1% of others. But because red hair is a representative attribute of the Irish, people tend to believe that the Irish are even more likely to have red hair than they actually do. Judgments by representativeness contain a kernel of truth in that they respond to information in the objectively correct direction. However, they do so excessively.
Applied to expectations in macroeconomics and finance, representativeness has some distinctive implications. The kernel of truth principle implies that people tend to overweight future outcomes that become more likely in light of incoming data. Just as they overreact to the news that a person is Irish in estimating the color of their hair, they react to macroeconomic news in the correct direction but excessively. Good macroeconomic news makes good future outcomes more representative, and therefore overweighted, in judgments about future states of the world. The converse is true for bad macroeconomic news. The same principles of belief formation that apply to lab experiments and social judgments translate one-for-one into our model of diagnostic expectations.
Under some conditions, diagnostic expectations tie together extrapolation and neglect of tail risk. News pointing to higher likelihood of economic growth causes high-growth scenarios to be representative and recessions to be unrepresentative, leading investors to both neglect downside risk and to display excess optimism about average conditions. News pointing to reduced volatility renders extreme shocks unrepresentative, leading investors to neglect risk. Diagnostic expectations also generate systematic reversals of optimism and pessimism in the absence of news. When trends in news cool off, no particular outcome is representative and expectations revert toward rationality. If the corrective news is bad enough, the left tail becomes representative and investors display excess pessimism. These movements in beliefs are entirely due to investors’ overreaction to objectively useful information, not to their mechanical extrapolation of the past.
DIAGNOSTIC EXPECTATION AND THE 2008 FINANCIAL CRISIS
Diagnostic expectations provide a useful unifying account of the 2008 crisis. They can serve as a foundation of extrapolative beliefs that characterized the housing bubble, which can be seen as updating and overreacting to repeated good news about home prices and general economic conditions. But they can also account for the neglect of downside risk, due to good news both about economic conditions (which rendered the left tail unrepresentative) and about the safety of financial institutions. More subtly, diagnostic expectations can account for the quiet period between the first tremors in housing and financial markets in the summer of 2007, which the Fed contained so successfully, and the eventual Lehman crisis. Even though the housing bubble was deflating and expectations about economic conditions were revised downward, the perception of tail risks remained dampened due to Fed policies and to the “diversification myth,” an exaggerated faith in the new insurance mechanisms. Diagnostic expectations may thus explain why both Federal Reserve and private-sector forecasts of future economic activity made as late as August 2008 point to a widely shared – and exaggerated – belief that, despite the early tremors, the situation was under control.
The theory also accounts for the extreme reaction to the Lehman bankruptcy, as the tail risks to the financial system came out into the open and market participants reacted. The Lehman bankruptcy revealed that the situation was far from being under control, that financial institutions were highly interconnected, so that systemic risk was much higher than previously expected. As a consequence, the previously neglected left tail became representative, causing beliefs to overweight the black swan of a financial meltdown. The market panic, asset fire sales, runs on financial institutions, mergers to avoid bankruptcy, and of course government rescues can be viewed as reflecting – at least in part – the massive revision of expectations about financial fragility. The Lehman crisis was a crisis of beliefs.
To summarize, we provide a new narrative of the 2008 financial crisis that assigns a central role to beliefs in accounting for both periods of quiet and those of significant volatility. More broadly, we propose a new model of expectations derived from first principles of psychology. Our model of beliefs – diagnostic expectations – can be incorporated into standard macroeconomic analysis. It is testable – and has been tested – using survey expectations data. We believe it not only fits the evidence on the 2008 financial crisis but also provides the foundation for future work on the role of beliefs and their impact on economic outcomes.
Gerardi, Kristopher, Andreas Lehnert, Shane M. Sherlund, and Paul Willen. 2008. “Making Sense of the Subprime Crisis.” Brookings Papers on Economic Activity (Fall 2008): 69-159.
Greenwood, Robin, and Samuel G. Hanson. 2013. “Issuer Quality and Corporate Bond Returns.” Review of Financial Studies 26(6): 1483-525.
Kahneman, Daniel, and Amos Tversky. 1983. “Extensional versus Intuitive Reasoning: The Conjunction Fallacy in Probability Judgment.” Psychological Review 90(4): 293-315.
Kindleberger, Charles P. 1978. Manias, Panics, and Crashes: A History of Financial Crises, 1st ed. New York: Basic Books.
Manski, Charles F. 2004. “Measuring Expectations.” Econometrica 72(5): 1329-76.
Minsky, Hyman P. 1977. “The Financial Instability Hypothesis: An Interpretation of Keynes and an Alternative to ‘Standard’ Theory.” Nebraska Journal of Economics and Business 16(1): 5-16.
Rajan, Raghuram G. 2006. “Has Finance Made the World Riskier?” European Financial Management 12(4): 499-533.
Reinhart, Carmen M., and Kenneth S. Rogoff. 2009. This Time Is Different: Eight Centuries of Financial Folly. Princeton, NJ: Princeton University Press.
Taleb, Nassim N. 2007. The Black Swan: The Impact of the Highly Improbable. New York: Random House.
About the Author
Nicola Gennaioli studies topics at the intersection of psychology, finance, and economics. He obtained a Ph.D. in economics from Harvard University in 2004. Today he is a Professor of Finance at Bocconi University in Milan. He is also a Managing Editor of the Review of Economic Studies, a leading journal in economics, and he holds appointments in various international scientific associations.
Andrei Shleifer is John L. Loeb Professor of Economics at Harvard University. He holds an undergraduate degree from Harvard and a Ph.D. from MIT. Before coming to Harvard in 1991, he taught at Princeton and the University of Chicago’s Booth School of Business. Shleifer has worked in the areas of comparative corporate governance, law and finance, behavioral finance, and institutional economics.