VoxEU Column Financial Markets Global crisis

Price discovery during anomalous market trading: The Lehman Brothers case

During normal operations, price discovery is an important feature of decentralised market trading. But the process can be distorted when markets are under great stress, such as during the run up to the collapse of Lehman Brothers in 2008. This column uses trading data from the days leading up to and following the collapse to show that price discovery at US stock exchanges remained remarkably efficient, even at the height of the turmoil.

Price discovery is the hallmark of the benefits derived from decentralised trading in a market economy. When markets operate smoothly, the flow of information about traded goods or securities miraculously gets translated into prices reflecting this information almost instantly (Hayek 1945, Fama 1970). But how do markets perform in periods of ultimate stress? Does orderly price discovery still work? The final weeks prior to the Lehman Brothers’ failure, and the weeks after, present such a period of ultimate stress.

In joint work, my co-author and I analyse price discovery in US stock markets during this dramatic period, both in the Lehman’s stocks as well as in the stocks of its large US peers in commercial and investment banking (Gehrig and Haas 2016).

The Lehman insolvency

Troubles for US investment banks started well before the start of the year 2008 and continued throughout for all of them. All their business models relied heavily on short term funding in money markets and since the drying up of the repo market in late 2007, funding costs exceeded revenues from their long-term assets. Accordingly, throughout 2008 all investment banks were accumulating negative earnings, and thus shedding equity value (Figure 1).

Figure 1 Relative evolution of stock prices of the US investment banks in 2008

Bear Stearns was the first to leave the market, deplete of capital in March 2008 and ultimately taken over by JP Morgan. Bear Stearns was one of investment banks with the lowest capital-to-asset ratios in January 2008, at 4%. Lehman Brothers and Merrill Lynch were close in terms of capital-to-asset ratios in early 2008 – they were to follow suit in terms of depleting their equity given their loss-making business models. At that time, Morgan Stanley held a capital-to-asset ratio of about 6% and Goldman Sachs 8%, but were also losing capital.

Because of its loss-making business model, Lehman Brothers desperately needed to raise capital during the worst possible times. Nevertheless, they managed to raise 4 billion US dollars in April 2008 and another $5 billion in June 2008. On 10 September 2008, Lehman Brothers had to disclose their warnings of an enormous pending loss of $3.9 billion for the third quarter of 2008. This was the largest ever quarterly loss in Lehman’s and US banking history. The CEO, Richard Fuld, also announced hasty emergency operations, but after a failed purchase attempt by the British bank Barclay’s to take over Lehman’s assets through the weekend after 12 September (the purchase was stopped by the British supervising FSA), at the beginning of business on Monday 15 September Lehman Brothers had to file for protection under Chapter 11. Within days about 25,000 employees lost their jobs. Parts of Lehman’s assets were taken over on 17 September  by Barclays Bank and on 22 September by Nomura Holdings.

Lehman’s failure sent shockwaves throughout the financial system – not only in the US but worldwide, hugely amplifying the liquidity crisis that had already been triggered a year earlier in September 2007 (Gorton and Metrick, 2009). As such, this failure is widely considered the climax of that Great Depression. (In historical comparison, Lehman’s failure resembles the role of the failed corner and the ensuing run on Knickerbocker Trust in the Panic of 1907; see Fohlin et al. 2015).

How did these hectic developments affect the performance of US stock markets? The drying-up of repo markets was the key driver of the losses incurred by investment banks’ business models. Did this dry-up also affect stock markets – and if so, how?

In Gehrig and Haas (2016), my co-author and I answer this question by analysing price discovery in Lehman stocks in the run-up to the failure on 15 September, and in the stocks of their peers – i.e. commercial banks and investment banks around and after Lehman’s failure.

Price discovery for Lehman Brothers stock

Daily price and volume data suggest that a lot of trading had already taken place on 9 September, one day before the public announcement on 10 September (Figure 2).  This could imply that markets knew the information before it was publicly announced, or perhaps points to the existence of (privately) informed trading prior to the announcement. Is the risk of informed trading priced in the market?

Figure 2 Daily stock price and trading volume of Lehman Brothers stock traded in NMS

Note: The green line is the date of the public announcement of 3rd quarter losses.

Based on real time transactions by date, we reach a slightly different conclusion. While relative bid-ask spreads remain low, fluctuating between 0.3 and 0.5% (Figure 3), the composition of the spread components did change only after the public announcement (Figure 4). 

Figure 3 Relative bid-ask spreads in Lehman Brothers stock traded at NYSE

Note: The red line is the date of the public announcement of 3rd quarter losses.

Following Huang and Stoll (1997), we decompose bid-ask spreads into an information and non-information component, in order to find whether information risk was priced differently in the run-up to the crises. Figure 4 documents that information risk did indeed shoot up on 10 September, to become the main driver of spreads after the public announcement. However, there is only weak indication of a relative increase in market perception of informed trading prior to 10 September. Certainly, information risk was not priced higher before Monday, 8 September.

Figure 4 Decomposition of the bid ask spread of Lehman Brothers stock traded in NMS

Note: The information component is in blue. In red is inventory holding costs, and in green are operating costs.

Based on real time data and grouping them by hour, we find strong indications of exceedingly intense trading on 9 September, just one day before the public loss announcement (Figure 5). After the public announcement, trading activity calms down before it heats up a bit on 11 September, and calms down once more on Friday, 12September, which should become the last public trading day for Lehman stocks.

Figure 5 Number of trades in Lehman Brothers stock traded at NYSE

Note: The red line is the date of the public announcement of 3rd quarter losses.

Interestingly, the size of trades did increase significantly only after the public announcement (Figure 6), while trading costs (bid-ask spreads) essentially remained un-elevated (see again Figure 3).

Figure 6 Trade size in Lehman Brothers stocks traded at NYSE

Note: The red line is the date of the public announcement of 3rd quarter losses.

Despite the large increases in trading volume, transaction prices remained efficient to an extraordinary degree. Apparently, panic did not arise at the height of the troubles at Lehman Brothers. This evidence sets a stark contrast to the US Panic in October 1907 (Fohlin et al. 2015). Throughout the last two weeks of trading of Lehman Brothers stocks, we find that trades impacted transaction prices for fewer than ten subsequent trades, which implies an information based price impact of typically less than five seconds. During the most turbulent days, price efficiency even increased and information based price impacts are discernible only for roughly half a second. Figure 7 shows the average price impact on 10 September, the day of the public announcement. This reflects a remarkable degree of efficiency of price discovery even in a period of extreme stress.

Figure 7 Efficiency of price discovery

Price discovery in the stocks of the largest US banks

With the remarkable information efficiency in pricing the stocks of the troubled Lehman Brothers, it comes as little surprise that we do not find any significant disorderly developments in the pricing of its main competitors and peers. Applying the same spread decomposition methodology, we do find a general and significant increase in bid-ask spreads after the Lehman failure but virtually no cross-market spillovers of information risk before 15 September. Overall information efficiency also remains high after the Lehman failure.

Figure 8 Daily average stock price and trading volumes of largest commercial (Bank of America, Citigroup, JP Morgan Chase, Wells Fargo) and investment banks (Goldman Sachs, Merrill Lynch, Morgan Stanley) in NMS

Note: The green line is the date of the public announcement of 3rd quarter losses.


Overall, price discovery at US stock exchanges remained remarkably efficient even at the height of turmoil during the Great Recession. Even though there is no discernible leakage of information in pricing, there is clear evidence that a revaluation of the stock of Lehman Brothers did occur at least the very day before the public disclosure of what was going to be the final and lethal losses. Trading on 8 and 9 September is characterised by an extraordinary number of standard trades at constant spreads with little price impact individually, but with a clear aggregate impact and strong revaluation of the stock prior to the public disclosure. Price adjustments continue at high pace after the disclosure, but trade size increases while the number of trades recedes to even below normal levels.


Fama, E F (1970), “Efficient Capital Markets, A Review of Theory and Empirical Work”, Journal of Finance, 25(S): 383–417

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

Gehrig, T, and M Haas (2016), “Anomalous Trading Prior to Lehman’s Failure”, CEPR Discussion Paper 11194

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

Hayek, F A (1945), “The Use of Knowledge in Society”, The American Economic Review, 519-530

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

2,835 Reads