Economists have long been interested in being able to identify stock market bubbles in advance because they are not only associated with significant mispricing in financial markets (in defiance of the efficient markets hypothesis), but the mispricings lead to distortions in allocations of capital (overinvestment). In addition, bubbles are followed by crashes as the consequences of inefficient investment play out.
We can define a stock bubble as a market that booms (rises more than 100% within two years) and then crashes (a drawdown of at least 40% in two years). Stock market bubbles generally follow the same five stages, first identified by American economist Hyman Minsky:
- Displacement: A big change or a series of changes affects how investors think about markets.
- Boom: Prices increase, attracting speculators who drive prices higher as word spreads.
- Euphoria: Investors are driven by excitement rather than rational justification for surging prices.
- Profit taking: The surge in prices ends up being too good to be true and the bubble is pricked.
- Panicked selling: Investors faced with margin calls and plunging values seek to liquidate at any price.
Factors that can contribute to the formation of stock bubbles include:
- Low-interest rates: When interest rates are low, investors seek higher returns, often leading to increased investment in stocks.
- Easy credit availability: Abundant credit can fuel speculation as investors borrow to invest in stocks.
- Economic growth: Periods of strong economic expansion can create optimism and drive up stock prices.
- Technological innovation: The emergence of new technologies can generate excitement and investment, sometimes leading to overvaluation.
- Investor psychology: Herd mentality, fear of missing out (FOMO), and overconfidence can contribute to a bubble.
When a bubble bursts, the consequences can be severe:
- Economic downturn: Sharp declines in stock prices can lead to decreases in consumer spending and business investment, triggering recessions with rising unemployment.
- Bankruptcy: Financial institutions that have invested heavily in the bubble can face bankruptcy.
- Loss of confidence: The public’s trust in financial markets can be eroded.
Unfortunately, there is little evidence that financial economists have been able to identify bubbles in advance. Consider the following from a 2013 NPR interview with Nobel Prize-winning economist Gene Fama.
Eugene F. Fama: The word “bubble” drives me nuts, frankly, because I don’t think there’s anything in the statistical evidence that says anybody can reliably predict when prices go down …
NPR: What would prove it to you that there were bubbles?
Eugene F. Fama: Empirical evidence.
NPR: Such as?
Eugene F. Fama: Well, that you could show me that you can predict when these things turn in some reliable way.
Empirical Evidence
In order to determine if accounting information could ex-ante identify a stock market bubble Salman Arif and Edward Sul, authors of the July 2024 study “Does Accounting Information Identify Bubbles for Fama? Evidence from Accruals” examined industry-level investments in net operating asset accruals and stock returns for 49 countries around the world. They measured investment using changes in net operating asset accruals capturing net investment in both working capital accruals and long-term operating accruals.
They focused their analysis on the industry level, “in line with historical evidence that bubbles are often industry phenomenon.” Using a large sample of countries, they identified run-up episodes in which value-weighted industry stock prices increased over 100% in terms of both raw and net of market returns over the prior two years. Crashes were defined as drawdowns of at least 40% over the following two years. Since accounting data was only available starting in the early 1990’s for non-US countries, they examined run-ups between 1992 and 2020. This resulted in 18 U.S. run-ups and 222 non-U.S. run-ups, for a total of 240 industry run-ups across 49 countries. Their tests focused on univariate predictive return regressions (a statistical model used to predict the future return of a financial asset based on the information contained in a single past variable), sample return predictability, multiple regression tests, predictability of analyst forecast errors, and the economic magnitude associated with the predictability. Here is a summary of their key findings:
Of the 240 total run-ups, they identified 114 crashes—47.5% ended up crashing within the next two years. Of the 18 U.S. run-ups, 10, or roughly 56%, subsequently crashed. China and Hong Kong experienced the most crashes in the international sample with eight crashes each, followed closely by Brazil and India with seven each.
While the average past two-year industry return in any given month was around 24.2% in the full panel, the average return was over 205% in the run-up sample. The run-up sample displayed higher average volatility, one-year changes in volatility and turnover, equity issuance, sales growth, CAPE ratio, convexity of price path (acceleration), and NOA accruals. Run-ups were also associated with younger firms and lower book-to-market ratios.
The change in industry-level NOA accruals was a statistically significant predictor of crashes, with a coefficient of 0.687 and t-statistic of 4.23. A one standard deviation increase in accruals, all else equal, was associated with a 12.4% greater likelihood of a crash in the next two years. Accruals were significantly higher for price run-ups that subsequently crash compared to those that did not—a sharp increase in stock prices at the industry level did not unconditionally predict low returns going forward.
Industry-level NOA accruals were a robust negative predictor of industry stock returns. Run-ups in the lowest tercile of industry accruals experienced returns of 23.8% net of the risk-free rate on average over the following two years, while run-ups in the highest tercile of industry-level accruals experienced returns of -8.1%. The difference of 31.9% was statistically significant. However, industry-level NOA accruals associated with price run-ups negatively forecasted aggregate country-level returns, but industry-level accruals that were not associated with price run-ups did not generically forecast aggregate country-level returns.
Accruals delivered positive out of sample r-squared when predicting each of the post run-up return measures.
Their findings led Arif and Sul to conclude: “Overall, these results suggest that accruals identify bubbles in a statistically robust and economically significant manner.” They added: “The predictive ability of accruals for industry crashes, returns and forecast errors almost quintuples following run-ups compared to the baseline. This indicates that our results are not the product of accruals on average generically predicting future performance. Rather, our findings indicate that the misallocation of capital due to bubble-driven overinvestment has a distinctly negative impact on future asset prices and corporate fundamentals.”
Turning to providing the explanation for overinvestment predicting bubbles, Arif and Sul noted: “Historical bubble accounts suggest that under the overinvestment explanation, managers are more likely to overinvest when sentiment is buoyant, earnings expectations are inflated and financing easy to obtain. Consistent with this, we find a positive contemporaneous correlation between accruals and two investor sentiment proxies: the Baker, Wurgler and Yuan (2012) country-level sentiment index as well as the Dichev (2007) measure of investors’ net equity market fund flows computed at the country-industry level.” They also found: “Higher accruals portend greater earnings shortfalls relative to analysts’ EPS expectations.”
Investor Takeaways
Arif and Sul’s findings are consistent with an overinvestment channel—corporate investment rises when investor sentiment around run-ups is more exuberant, yet such periods tend to be followed by price crashes and disappointing corporate fundamentals. Thus, they have provided Fama with his quest for empirical evidence—financial statement analysis can be used to detect and predict important sources of capital market inefficiencies at the industry and market level with NOA accruals identifying bubbles and providing a leading signal of downturns in industry and aggregate-level returns. With the run up in the prices of many stocks associated with artificial intelligence Arif and Sul’s findings provide a warning. Arif and Sul also found that several other measures were significant predictors of a crash: “Volatility, Volatility1yrChange, IndustryAge, AgeTilt, PercentIssuers, BooktoMarket, Acceleration and CAPE.” Forewarned is forearmed.
Larry Swedroe is the author or co-author of 18 books on investing, including his latest, Enrich Your Future: The Keys to Successful Investing
