VoxEU Column Economic history Financial Markets International Finance

Rumors and runs in opaque markets: Evidence from the Panic of 1907

The story of the run-up to the Global Crisis is, unfortunately, not an entirely new one. This column argues that regulators would do well to read up on the ‘Panic of 1907’. What quelled rumours and panicky behaviour back then still applies – maintaining market liquidity through measures that encourage transparency.

Opacity may be a major source of market freezes. This phenomenon arose in dramatic fashion in the 2007-8 Global Financial Crisis, when interbank markets suddenly froze, causing a major withdrawal of global liquidity.

Much of the post-Crisis concern from policymakers was focused on both financial instruments that were frequently labelled ‘complex’ and ‘exotic’ and on trading that took place in opaque markets and institutions, outside the purview of regulatory bodies (otherwise known as the ‘shadow banking system’, see Gorton and Metrick 2012). Fundamentally, market participants could not fully assess the value of these assets or the risks to their valuations.

The Panic of 1907 and the Global Crisis

Interestingly, it turns out that most of the component parts of this most recent episode are not new. Indeed, the Panic of 1907 in the US displays many parallels with the 2007-8 Crisis, despite the fact that the institutional environment was quite different: financial markets in 1907 were yet to be regulated by the federal government; the Federal Reserve System was only founded in 1913; and statutory regulation of stock trading and corporate reporting came even later. The world of 1907 was largely characterised by spontaneous actions, private initiative and self-regulation.

This all begs the question of whether early markets were more or less resilient with respect to economic or financial crises.

Since the Panic of 1907 has been characterised in the literature as a financial crisis driven purely by rumours rather than fundamentals (Moen and Tallman 1990, Frydman et al. 2015), it lends itself perfectly to a rigorous analysis of the role of information costs. While many previous studies have examined the Panic at the aggregate level, we are able to offer a much more nuanced ‘microstructure’ view of the unfolding crisis by exploiting a new database for all stocks traded on the New York Stock Exchange from 1905 through to 1910. For every trading day during this six-year period, we hand-collected opening, closing, high and low transactions prices, market-makers’ closing bid and ask quotations, and total trading volumes for all stocks traded at the New York Stock Exchange as well as from its primary competitor, the Consolidated Exchange.1 We compute several measures of liquidity and then, using a methodology developed by Huang and Stoll (1997) and refined by Gehrig and Haas (2014), we decompose bid-ask spreads into their main transaction cost components: information costs (or adverse selection costs), inventory holding costs, and operating costs (cum market power).  Our results demonstrate the crucial role of information opacity in turning a relatively minor hiccup into a market-wide liquidity freeze.

Development of the 1907 Panic

The Panic of 1907 began on 16 October 1907 with the failure of the brokerage firm of Otto Heinze. Heinze's brokerage house was forced to close when he attempted to corner shares of the United Copper Company and pull a classic short squeeze. The manipulations caused wild swings in the price of United Copper, but the price ultimately plummeted and forced Otto into financial ruin. United Copper was partly owned by Otto's brother, the notorious copper magnate, FA Augustus Heinze, and was traded in an ‘on the curb’ market, not on the New York Stock Exchange. The Heinze failure set off rumours that certain financial institutions had financed the failed short squeeze and therefore held unpayable debts from Otto Heinze. Much of the lending among these New York City banks rested on the collateral of securities.   Thus, with the stock market already several months into a decline (see Figure 1), and the fear of margin calls, the market was particularly vulnerable to a run. As the graph of stock prices shows, mining sector stocks were particularly volatile and most hit by the falling prices in the run-up to the crisis.

Figure 1. Average daily stock prices, 1905-1910

It is interesting to track in more detail the progression of the rumours through the financial system, as it underscores how little it takes to cause a big problem. In this case, Augustus Heinze was the key link in the rumour chain. He had moved – just a few months prior – to Manhattan and taken an active interest in banking and finance. He became president of the Mercantile Bank and took on directorships at several other banks and trust companies. Thus, as rumours spread about counterparties to his brother Otto's brokerage firm, depositors ran on Mercantile National and on the trust companies with known ties to Heinze – first and foremost, the Knickerbocker Trust Company, with $69 million in assets (Tallman and Moen 1990). After the closure of the Knickerbocker Trust Company on Tuesday 22 October, depositors rapidly began withdrawing from other trust companies. In an era in which investors learned price information by traveling to or phoning their brokers – who, in turn, relied on a stream of information printed onto ticker tape arriving via telegraph – the only way to learn news in real time was to appear in person. Given that disclosure requirements and transparency were low, traders faced a continual threat of informational contagion.  

Figure 2 reproduces the now-famous photograph in Harper's Weekly following the peak of the panic, and gives an impression of what that ‘price discovery’ process looked like.

Figure 2. The run on Wall Street, October 1907 (Harper’s Weekly 1907)

Note: Image reproduced under a Creative Commons licence.

Information as the driver of illiquidity

Clearly, information asymmetry sets the stage for panic, and we can observe the fallout in measures of market liquidity. We find that daily relative quoted spreads increased six-fold over the peak of the crisis, and rose significantly in advance of the most acute period of crisis – starting already by September 1907. The heightened illiquidity lasted until March 1908, several months after the crisis ended. At the same time, trading volume declined steadily and remained depressed well into 1908.

Figure 3.  Daily cross-sectional median relative effective spreads, 1905-1910

Moreover, we find that transaction costs increased more for smaller and less liquid stocks and for companies in the more opaque market segments (e.g. mining and manufacturing).  Illiquidity also disproportionately hit the so-called ‘unlisted department’ of the New York Stock Exchange. As part of its strategy vis-à-vis its main competitor – the Consolidated – the New York Stock Exchange had formed this ‘department’ to allow trading in the stocks of smaller or riskier firms that did not meet the Stock Exchange’s stringent listing requirements or that did not want to go through the vetting process involved in listing and preferred instead to keep their accounting data private. As we expect, these unlisted stocks generally suffered more from higher informational costs, and these information costs were especially elevated when rumours were most active in the last quarter of 1907. It also took longer for adverse selection risk to decrease in unlisted stocks compared to listed stocks. This implies that investments in listed companies that had to publish accounting information indeed served as a hedge against adverse selection risk and especially so in times of heightened uncertainty.

The decomposition of spreads into their adverse selection, inventory holding and order processing components establishes that all three of them played a significant role in increasing market illiquidity during the Panic. Arguably, however, adverse selection costs were the largest component, which is of little surprise given the informationally more opaque environment in the early 20th century.

Since liquidity has been established as a priced factor in stock returns (Acharya and Pedersen 2005), we extend their work and ask whether the components of liquidity costs themselves are also priced in the market. Indeed, our asset-pricing analysis verifies that each factor of our bid-ask decomposition was independently priced, and hence carried a corresponding (illiquidity) premium already in the early markets. Hence, not surprisingly, liquidity concerns always have constituted fundamental characteristics of securities markets.

The upshot of the Panic of 1907

The Panic of 1907 marked the beginning of the end of unregulated capital markets and weak central monetary authority in the US. While private initiatives – the concerted effort organised by John Pierpont Morgan – ultimately resolved the crisis, it provided central banking advocates the ammunition they needed to push through the Federal Reserve Act, and in the meantime the provision of emergency currency via the Aldrich-Vreeland Act (which would come into play in the summer of 1914). The episode prompted the famous Money Trust hearings in Congress that led to the Clayton Antitrust Act, as well as a state-level investigation in New York that ultimately led to the disbanding of the NYSE's ’unlisted' department. These regulatory steps laid the foundation for the more far-reaching interventions as the US Securities and Exchange Commission that emerged much later.

This long-ago episode elucidates how opaque markets and institutions allow rumours to cause runs and can freeze liquidity in short order. Our microstructure analysis of this historical crisis reminds us of three fundamental points:

  • Transparency is the lynchpin of properly functioning markets;
  • Liquidity is the measure of a market’s health status;
  • Illiquidity in one minor market segment – however minor -- can readily infect other, if not all, market segments.

Policy should focus on maintaining market liquidity through measures that encourage transparency.


Acharya, V and L Pedersen (2005), “Asset Pricing with Liquidity Risk”, Journal of Financial Economics 77(2): 375-410.

Fohlin, C (2014), “A New Database of Transactions and Quotes in the NYSE, 1900-25 with Linkage to CRSP”, JHU mimeo.

Fohlin, C, T Gehrig, and M Haas (2015), “Rumors and Runs in Opaque Markets: Evidence from the Panic of 1907”, CEPR Discussion Paper 10497, London.

Frydman, C, E Hilt and L Y Zhou (2015), “Economic Effects of Runs on Early 'Shadow Banks': Trust Companies and the Impact of the Panic of 1907”, Journal of Political Economy, forthcoming.

Gehrig, T, C Fohlin (2006), "Trading Costs in Early Securities Markets: The Case of the Berlin Stock Exchange 1880-1910", Review of Finance 10: 585-610.

Gehrig, T and M Haas (2014), “Lehman Brothers: Did Markets Know?”, ECGI-DP, 424/2014, Brussels.

Gorton, G and A Metrick (2012), “Securitized banking and the run on repo”, Journal of Financial Economics 104(3): 425-451.

Huang, R and H Stoll (1997), “The Components of the Bid-ask Spread: A General Approach”, Review of Financial Studies 10(4): 995-1034.

Tallman, E W and J R Moen (1990), “Lessons from the Panic of 1907”, Federal Reserve Bank of Atlanta Economic Review.


1 See Fohlin (2014) for more detail on the larger data collection project.

2,205 Reads