Moles and Lichenstein have both carried out studies highlighting some of the problems in risk assessment resulting from perception. Peter Moles argues that larger events are more likely to be perceived than smaller onesa. This implies that a risk assessor would ignore low probability and low consequence events, yet we do not know if a low probability event is low consequence until it occursa. A study by Lichenstein (1978) shows how an assessor’s perception may affect their ability to judge probabilitya. His study suggests people have an inability to assess probability accurately. He points out how dramatic and infrequent events tend to be overestimated, whilst those more common and infrequent events tend to be underestimated.
Having examined how underwriters’ awareness and perception hinder their ability to assess risk, let’s look more closely at the methods used by underwriters in risk assessment – and in particular how probability is obtained. Where at all possible underwriters will seek to ascertain a probability based on the frequency of an event’s occurrence in the past. Alternatively they can obtain a probability through subjective expert opinion, yet this requires the availability of willing and proficient experts. Furthermore the experts’ opinion will depend on their knowledge and understanding of the context which will be influenced by their perceptionb. Therefore the argument is that, provided the data is available, a frequentist probability based on historical evidence is more accurateb. Yet frequentist probabilities are nonetheless fraught with problems, including their reliance on potentially unavailable and misleading data.
A major problem for underwriters using frequentist probabilities to assess risk is that such probabilities are based on data collected in samples. These samples may not be a true reflection or representation of the wider population and the derived implications from the sample data won’t necessarily correlate with the risk being assessedb. There is also the issue of whether the sample itself is accurate. The method in which the data was collected and recorded may have weaknessesb. Death certificates for example have been criticised for giving vague or incorrect statements as to the causes of death. Even where gross assumptions are made from huge surveys such as consensus data there are concerns that the data is misleading.
An additional problem is that in using historic data as the bases for their assessment models, underwriters are assuming that past histories are good predictors of the futureb. Take for example, life assurers, who predict future death rates based on past history and life tables. Yet children have a higher chance of reaching 100 years old than their parents, moreover past histories fail to take account of new and emerging diseases such as HIV/AIDSc. The frequency of events can and will change, looking at another context, consider weather storms, the frequency of these may increase or decrease as a result of climate changeb. Yet we cannot predict climate change and the effect it will have with any degree of accuracy, so adding to the difficulties of risk assessment. Society and the environment is constantly evolving, the underwriter’s assessment models’ ability to take this into consideration is questionable. Whatever controls they attempt to use so as to allow for changes will be subjective and therefore potentially false.
This brings into light a further problem in risk assessment, which is associated with the general belief that scientific theory is an authority and correct. Wynne (1992) questions the validity of such authoritya. Scientific theory’s ability to know who is likely to be affected and the implications, both short and long term, is in doubt. Scientific assumptions may hold within known context but not where the conditions of the theory are not satisfieda. The problem for underwriters is that in some circumstances, such as nuclear accidents, it may not be possible to tests the scientific theory.
Another problem with the models used in risk assessment is that there is a need to attenuate information. Let’s again consider risk assessment for the purposes of life assurance, here a number of factors will be taken into consideration including age, gender and other variables such as medical history and lifestyle. The assessors may seek information from a number of sources such as:b
a). Proportionality mortality studies
b). Studies of deaths among workers employed at time of death
c). Case-Control studies
d). Simple prospective studies
e). Studies of spatial distributions
All of these studies provide relevant data and, one could argue, equally irrelevant data, because to obtain the results assumptions were made and factors left outb. In addition the model cannot take too many variables into consideration because the subgroups will become too small to provide reliable estimatesb, again necessitating the assessors own judgement, based on awareness and perception, as to which factors to omit. The assessor will therefore need to attenuate the information including that which he feels is most relevantb. However important factors may be ignored, as only those that ‘fit’ with the assessor’s “world view” will be recognised and accepted. Furthermore the assessor may not have access to certain variables either because they cannot or could not collect data or because previous investigators did not deem them to be relevant and so did not record data on their basisb. This again reiterates the problems associated with awareness and perception in risk assessment.
I have discussed many of the problems underwriters face in assessing risk with particular reference to obtaining an accurate probability, and many of these factors are equally problematic when attempting to ascertain the likely impact of a realised risk. However ascertaining probability is relatively simple in comparison to assessing the likely consequences of its occurrenceb. The frequency of an event’s occurrence in general remains fairly constant yet its possible impact must be continually revised. The sums of money underwriters are obliged to pay out will continually change as the environmental, economic and societal conditions change. For example new drugs and increased healthcare costs would have substantial affects on the cost to life assurers should a risk be realised. Interest and exchange rate changes also substantially affect underwriters’ potential outlays. This makes assessment a continual and extremely complex process. In accepting to take on a risk they not only have to attempt to predict the likely chance of it occurring but also the constantly evolving potential impact.
Some forms of insurance have a fixed maximum impact should a risk be realised. For example with life insurance a contract is drawn up, whereby the underwriter agrees to pay out a maximum sum should a contingency arise. This gives the assessors a clear quantification as to the maximum impact of a risk, however for accurate risk assessment the underwriter will seek to assess the likely possible consequences that may be within their maximum liability. Other forms of insurance do not have the luxury of knowing the maximum possible loss they could face should a risk be realised. Nine insurance companies were put out of business as a result of Hurricane Andrewc, this was because of their inability to assess the maximum possible impact the risk posed and prepare accordingly.
The difficulty for underwriters is being able to put an upper limit on the loss or risk they facec. Let us consider fire insurance. Insurance firms may face claims for fitments and accessories within a building, they may also face even higher claims for damage to the actual fabric of the building. The risk assessment for one building would be fairly simple, however if the fire were to spread the claims could become substantial. The hypothetical graph below illustrates the pointc. The shaded area represents the actual claims, and the other lines represent the predictions based on the models used by
underwriters.
Household Fire Insurance Claims
Graph taken from Ansell, 2002, University of Edinburgh, Risk Management Course, Lecture 6 notes
The circled upper tail shows where high impact events are not accounted for in the models used. However it is virtually impossible to predict the extreme situations and how fat the tail should bec. Asbestosis is a prime example of this. Underwriters will be unaware of all the possible consequences. In a medical context underwriters cannot avoid the fact that they are unaware of new diseases that may emerge, and therefore cannot even speculate their potential impact.
As a final point let’s consider problems associated with risk assessment that are specific to insurance firms. These include the fact that insurance is a moral hazardc. Underwriters must take account for that fact that the insured person or party may take less care as a result of being insured. Moreover insurance firms tend to suffer from adverse selectionc, whereby the people most likely to seek insurance are those who have experienced or are likely to experience large losses. It is factors such as these that further complicated the risk assessment process in insurance.
In conclusion the problems associated with risk assessment in insurance stem from the necessity to construct models, which by their very nature cannot truly represent reality. It is argued that there very basis on probability and consequence is too basic. Furthermore the way in which probability and consequence should be combined to give an overall assessment is questionable, many argue that expected loss is over simplistic. The models basis on probability and consequence, as has been highlighted in this essay, is subject to erroneous data, subjectivity and human fallibility.
(Word Count = 2041)
References
Adams, J, Risk, University College Press.
Ansell, (2002), Risk Management Course - Lecture Notes, School of Management, University of Edinburgh
Ansell & Wharton, (1992), Risk – Analysis, Assessment and Management, John Wiley and Sons, Chichester
Huczynski, AA, and Buchanan, DA, (1991). Organizational Behaviour, Prentice-
Hall, Hemel Hempstead.
- Monday 13/1/03 1700-1800 (UK Time)
- Monday 13/1/03 1700-1800 (UK Time)
a Take from Ansell, (2002), Risk Management Course - Lecture 2 Notes, School of Management, University of Edinburgh
b Take from Ansell, (2002), Risk Management Course - Lecture 3 Notes, School of Management, University of Edinburgh
b Take from Ansell, (2002), Risk Management Course - Lecture 3 Notes, School of Management, University of Edinburgh
a Take from Ansell, (2002), Risk Management Course - Lecture 2 Notes, School of Management, University of Edinburgh
b a Take from Ansell, (2002), Risk Management Course - Lecture 3 Notes, School of Management, University of Edinburgh
b Take from Ansell, (2002), Risk Management Course - Lecture 3 Notes, School of Management, University of Edinburgh
c Take from Ansell, (2002), Risk Management Course - Lecture 6 Notes, School of Management, University of Edinburgh
c Take from Ansell, (2002), Risk Management Course - Lecture 6 Notes, School of Management, University of Edinburgh