The world of PRA models: when certainty does not exist, frequency and probability is the best we can do

Risk & Insurance, April 14, 2003 by Roger Crombie

Probabilistic risk assessment (PRA) is the use of modeling applied to technology systems. The technique relies on the application of logic, such as fault-tree or event-tree analysis, or component failure databases, to arrive at a quantitative assessment of the potential for systemic failure. The analysis is expressed as some function of the probability of the failure for each of the components of the system. PRA integrates various analytical techniques to assess the potential for failure and to help find ways to reduce risk.

PRA relies on assumptions for each individual risk assessment and requires that the procedure for combining the probabilities of failures to arrive at an overall assessment of failure has been correctly calculated.

The value of this approach is limited by errors of observation and measurement. Since failures in precision can occur for each component of an overall system, and since each probability for error is multiplied by the probability for error in subsequent components, the overall assessment of failure can be highly inaccurate. PRA is simply an analytical tool, rather than a magic wand.

Risk assessment models can run both deterministic as well as probabilistic assessments. Deterministic risk assessment involves defining an event, such as a catastrophe or an earthquake, and computing the likely damage associated with that event. PRA computes the potential damage for different events, while accounting for the probability of each event.

A fundamental principle of PRA techniques is that risk due to natural catastrophes is location dependent, and that it can be assessed within an acceptable range of uncertainty, if reliable historical and location specific data are available.

Risk assessment of natural catastrophes has two components: hazard and vulnerability. The hazard is a measure of the physical intensity of the peril (earthquake, wind, surge, etc.) at a particular location and the associated probabilities of these intensities. Hazard is location dependent. For instance, a location surrounded by seismic faults and a weak surface geology has a higher hazard potential than a location far away from faults and with strong surface geology. Vulnerability is a measure of the damage caused to man-made objects at that location. Man-made structures respond to different perils in different ways, depending on design, structure and construction methods.

PRA is becoming more popular. Why? A single example will suffice. The clean-up cost after the 1979 accident at Three Mile Island was about $1 billion, but only $300 million of property insurance was available at the time to General Public Utilities (GPU). The $300 million compensated GPU for a fraction of their clean-up costs. PRA is now used by such agencies to formulate a more complete picture of the total risk.

An integrated picture allows managers to evaluate policies necessary to cover an acceptable level of risk, and to provide a comprehensive basis for supporting that decision. Risk assessment provides a means of prioritizing resources, to ensure the most efficient use of available capital.

Regulatory agencies are increasingly recognising the potential benefits of the use of risk assessment techniques in solving environmental problems.

Engineers use PRA to assess potential damage to pressure vessels and piping system failures. Both the U.S. National Academy of Science and the National Response Team (which addresses "technical guidance for hazards analysis") use PRA techniques. In pollution prevention and control, the use of PRA is often stipulated. Developers use PRA to make decisions on the feasibility of projects. The nuclear power industry uses PRA to determine the risk to the public from nuclear power plants.

Meteorological departments use tropical cyclone threat maps to improve the accuracy of tropical storm warnings. PRA creates hazard maps to show earthquakes, landslides, floods or fires for cities, districts or countries. They tend to be drawn on the basis of zip codes. These hazard maps, easy to understand, are helpful to the population as they can serve to identify high-risk areas. Planners, developers and insurance companies also find them useful.

No one has made more widespread use of PRA techniques than the insurance and reinsurance industry. Both apply the models to low-probability, high-consequence events. In catastrophe insurance and reinsurance circles today, PRA is a standard tool. Most companies report using PRA models to look at natural (rather than man-made) events, such as hurricanes, earthquakes, tornados and winds, all events where history, seismic conditions and a welter of data can produce a more accurate picture.

PRA is widely used in catastrophe risk securitization in the form of cat bonds. They are then sold to investors in the capital markets. The bonds can increase the amount of capital available for catastrophic risk, and alter the pricing of risk.

EQECAT has been a leader in the development of this market. Through use of PRA models, earthquake and hurricane risk assessments are performed to aid an issuer in determining how much risk to transfer and at what price. Working with the investment bank, EQECAT assists in assessing alternative loss triggers and associated probabilities to meet investor and issuer requirements.