Crowd Disasters or:
How and Why Things Can Get Totally Out of Control
Question: Professor Helbing, we are currently enjoying the European Soccer Championship and are already looking forward to the Olympic Games 2012. Are our stadia safe?
Answer: As far as I know, the authorities have carefully prepared for these events with international workshops discussing the lessons of previous crowd disasters such as the one in Brussel’s Heysel Stadium. But when large masses of people gather in one place, there is always an element of risk.
Q: You have just completed an empirical study of the crowd disaster at the Love Parade in Duisburg, Germany, on July 24, 2010, where 21 people died and more than 500 were injured. Who is to blame?
A: Our study analyses how things can get totally out of control, but it does not address the issue of legal responsibilities. When something goes wrong, people often search for a scapegoat, i.e. they tend to personalize things. Of course, we do not claim that no individual or institutional mistakes were made, but our study focuses on the systemic failure that led to the crowd disaster. We investigate why and how crowd disasters can happen, even though nobody wants to harm anybody and everybody wants to avoid them.
Q: How is this possible?
A: Let us start with another example. On May 6, 2010, the financial markets were shaken by a so-called ‘flash crash’. Within just 20 minutes, many solid assets turned into penny stocks. 600 billion dollars of stock values evaporated. The ownership structure of some companies changed within minutes. In the beginning, people thought it was the result of a criminal act.
Q: And what was the reason?
A: It turned out to be an interaction effect of automated trading transactions performed by computer programs. A large sell transaction triggered other computer programs to sell as well. This created cascading effects, which human traders could not analyze and respond to quickly enough. In fact, it took many months to analyze what was really going on.
Q: Just as for the crowd disaster at the Love Parade in Duisburg, which shocked the public two years ago. In the first moment, some people thought the fatalities resulted from a bomb attack. Then, it was claimed that people’s behavior got out of control: they were storming the festival area by overcoming barriers, climbing light poles and using a narrow staircase. It was also claimed that some of these people fell down on the crowd, thereby causing the fatalities.
A: All of this did not cause the disaster. There is the common misconception that crowd disasters result from a psychological state of panic that causes a stampede of people. But that would blame the victims for the disaster. This absurdity shows that we still have totally misleading ideas of the causes of crowd disasters, and that is also why they happen time and again.
Q: So what has caused the crowd disaster in Duisburg?
A: We found the same as in our analysis of previous crowd disasters during the Hajj, the Muslim pilgrimage. When the density becomes too high, body contacts are unavoidable. These transmit forces from one body to others. People perceive this as a pushy behavior. But what makes things worse is that the forces add up and vary strongly, in unpredictable ways. A phenomenon called ‘crowd turbulence’ or ‘crowd quake’ occurs, which makes it very difficult to stay on one’s feet. When people stumble and fall to the ground, a ‘domino effect’ sets in. More and more people fall on top of others. Those on the ground have difficulties to breathe. Eventually, they suffocate.
Q: So, this dynamics cannot be controlled, not even by a large number of security or police forces?
A: Exactly, when many people accumulate in one place and the density reaches a certain level, the situation can get out of control sooner or later. We know a similar phenomenon from vehicle traffic. Imagine a group of vehicles driving in a circle. When the density in the circle is high enough, traffic flow will break down and drivers will be stopped by ‘phantom traffic jams’ even when they are well skilled and trying hard to avoid this. If you would ask the drivers what happened, they would say: “There was a stupid driver in front of me who did not know how to drive.”
Q: And this is not the reason?
A: No, it is the instability of traffic flows, which causes the breakdown. If the density reaches a certain level, then any small variation in the vehicle flow will be amplified. A slightly delayed response of a driver due to the reaction time will require a slightly stronger braking of the following vehicle, and the next one will brake even stronger. This chain reaction finally causes drivers to be stopped, although nobody wants this to happen.
Q: This is surprising.
A: Yes, it is hard to understand. But we find similar kinds of dynamics in many complex systems. When many individuals adaptively respond to each other, this may cause a seamless self-organization, as Adam Smith assumed it for the economy with his paradigm of the ‘invisible hand’. However, if we drive a system towards a ‘tipping point’, a totally different system behavior may unexpectedly result.
Q: Such as?
A: Crowd turbulence rather than smooth and well-organized flows of people, for example, or the implosion of a bubble in the stock market, social conflicts, political revolutions, or wars. In many of these cases, nobody wants these things to happen, but when a systemic instability occurs, the situation will get out of control sooner or later. This applies even if everybody would have the best intentions, would be well skilled, perfectly equipped, and would try hard to avoid bad things happening. Of course, we are not claiming that these conditions are always fulfilled. Obviously, they are often not, and this makes things even worse.
Q: Does this mean we cannot do anything about systemic instabilities?
A: Not really. Systemic failures may be understood as institutional or procedural failures. They happen if institutions or procedures are not suitably set up. But if interactions cause the problem, we must change the interactions!
Q: What specifically does this mean for the organization of mass events?
A: One should avoid dense accumulations of many people, particularly where they are supposed to move. For this, one needs to have a suitable location, an adequate preparation, including contingency plans for all kinds of problems, a quick response to early warning signs, and last but not least good information and communication. One must also recognize the ‘laws of the crowd’. People are not robots. They have psychological, social, physiological and physical needs. For example, one should not require them to wait long without food, water, toilets, information and entertainment.
Q: That means, one should not stop the flow of people?
A: Stopping the flow often makes the situation worse. In many cases, it is better to organize the flows in ways that keep people moving, even if only slowly. Detours are a possible solution. But it is also important to avoid crossing flows and counter-flows. These can cause serious obstructions and situations that may get out of control.
Q: As it happened during the Love Parade…
A: Yes, problems resulted from the fact that arriving and leaving visitor flows were using the same paths and could not be separately controlled.
Q: So, how did the overcrowded situation actually occur?
A: The causal interdependencies were quite complex, so one must be aware that everything one can say in a few sentences is an oversimplification. It seems that the festival area and the ramp leading to it did not provide enough capacity for all visitors who wanted to visit the Love Parade. Therefore, everything depended on effective crowd control. But studying the course of events in detail revealed many surprises.
Q: Such as?
A: For example, jams of people did not build up at the narrowest part of the ramp, where one would have expected this, but at the upper end, where people were trying to enter the festival area. The ‘floats’ (i.e. the music trucks) were not effective enough in pulling people with them ‘like a magnet’. It seems that, when the density on the festival area reached a certain level, it obstructed the movement of the trucks, and these obstructed the inflow of people.
Q: And what happened then?
A: From 15:31 on, when the ramp got crowded, people overcame fences and used the slopes on both sides of the ramp to get up to the festival area. But this was reasonable. If they had not done so, the whole ramp and the tunnels would have been clogged very quickly. The situation would have escalated much earlier.
Q: What was the crowd management doing?
A: They were trying to control the inflow. Moreover, already between 14:30 and 15:15, they asked the police for help. This was around the time, when the police shift changed. Hence, the new shift was immediately and unexpectedly confronted with a difficult situation. The police tried to control the flows with cordon tactics, but it did not work out as expected. The cordons turned out to be ineffective and had to be given up. The result was a large accumulation of people on the ramp, and the ramp could not be evacuated, because arriving and leaving visitors were blocking each other. The people in the crowd had no overview of the situation, and the communication with the crowd was not effective enough to manage the flows.
Q: People were trapped on the ramp and densely packed…
A: Yes, exactly. Approximately from 16:20 on, people were trying to escape the ramp by climbing light poles, traffic signs, the container of the crowd manager, and a narrow staircase. In the beginning, this might have appeared like a crowd that was trying to ‘storm’ the festival area by all possible means. But in fact it was a sign that a life-threatening situation was emerging. At about 16:35, crowd turbulence occurred and caused a domino effect around 16:45, so that many people were piled up on top of each other. All of this can be reconstructed from the videos that the visitors of the Love Parade have uploaded to the Web.
Q: You mention that these publicly available materials are opening an era of citizen science and public investigation.
A: Yes, when we analyzed the crowd disasters that occurred during the Hajj in the past, this was based on exclusive video materials. However, for the Love Parade disaster, all evidence needed to get a picture of the causes of the crowd disaster is accessible to the public. This has helped us enormously in our study, which wants to create a better awareness of the origins of crowd disasters and their advance warning signs. We hope that our study helps to organize safer mass events in the future, including the exciting soccer matches to come.
- D. Helbing, P. Mukerji, Crowd Disasters as Systemic Failures: Analysis of the Love Parade Disaster. EPJ Data Science 2012, 1:7 (published on June 25, 2012).
- For more information, please visit http://www.soms.ethz.ch and http://loveparadevideos.heroku.com and http://blogs.ethz.ch/crowd/
- M. Moussaïd, D. Helbing, and G. Theraulaz (2011) How simple rules determine pedestrian behavior and crowd disasters. PNAS 108 (17) 6884-6888.
- A. Johansson, D. Helbing, H. Z. A-Abideen, and S. Al-Bosta (2008) From crowd dynamics to crowd safety: A video-based analysis. Advances in Complex Systems 11(4), 497-527.
- D. Helbing, A. Johansson, and H. Z. Al-Abideen (2007) The dynamics of crowd disasters: An empirical study. Physical Review E 75, 046109. See this link for a complementary webpage on how the Hajj was made safer.
- D. Helbing, I. Farkas, and T. Vicsek (2000) Simulating dynamical features of escape panic. Nature 407, 487-490.
- D. Helbing, A. Johansson (2010) Pedestrian, Crowd and Evacuation Dynamics. Encyclopedia of Complexity and Systems Science 16, 6476-6495.
- Y. Sugiyama et al. (2008) Traffic jams without bottlenecks – Evidence for the physical mechanism of the formation of a jam, New J. Phys. 10 033001 doi:10.1088/1367-2630/10/3/033001
- D. Helbing and A. Johansson (2010) Cooperation, norms, and revolutions: A unified game-theoretical approach. PLoS ONE 5(10), e12530.
- D. Helbing, H. Ammoser, and C. Kühnert (2005) Disasters as extreme events and the importance of network interactions for disaster response management. Pages 319-348. in: S. Albeverio, V. Jentsch, and H. Kantz (eds.) Extreme Events in Nature and Society (Springer, Berlin).
- K. Peters, L. Buzna, and D. Helbing (2008) Modelling of cascading effects and efficient response to disaster spreading in complex networks. Int. J. Critical Infrastructures 4(1/2), 46-62.
- D. Helbing (2010) Systemic Risks in Society and Economics. International Risk Governance Council (irgc).
Note that aggressive and seriously misleading comments may be removed. The responsibility for the comments is with those who post them. The site editor may not agree with the opinions posted to this blog.