I am not sure what I add will enhance your knowledge but can only highlight my own personal confusion with some of the terms or rather the ways in which you have used the terminology that I, and possibly some others, usually associate with risk management.

I have always thought of risk management as an exercise that is specific to one particular product. The proposal that this exercise can be based on a model is not in itself a problem as one can capture the specific details relating to the product under question and incorporate those into the model. The problem, for me at least, begins when you wish to consider a range of products that are different in nature. The specific details that are required for a useful risk management model or program are then lost or diluted as you attempt to “aggregate” them in order to build your model. You may well end up with a model but what would be the practical use of it.

For example you consider policy and regulation as “products” and assume a willingness on the part of the consumers to buy into it. Similarly in the draft paper, you refer to taxation as a product and point out the willingness of the citizens to pay it. Your generalisation is obviously omitting the very little fact that non-payment of taxes by a citizen gives rise to sever penalties which makes me wonder how you can assume willingness for tax payment in the first place.

In the draft paper you refer to engineers having lost the sight of unseen parameters (for example the change in driving behaviour after the introduction of airbag or seat belt) but when commenting on the failure of supersonic passenger planes you seem to also not pay enough attention to the parameters that may not be directly part of the supersonic flight scenario and may have contributed to its failure as a popular mode of transport. For example, the concord flew between London and NY but to get on board for someone who lived just on the outskirts of London would have taken twice as long as the flight itself. The teleconferencing technologies may have also contributed to this without even being considered as a parameter by analysts.

On the whole I believe any risk model needs to be as specific as it can be in order to cater for and accurately capture all the risks. On the other hand a risk model that does not capture all the risks is in itself a risk.

Hamid

Hamid, thank you for your comments.  Yes I have encountered that kind or risk management.  It is a very engineering-centric viewpoint.  Engineers believe if they calculate from the details, everything is considered and their answer is correct.

In an economic model, to a first approximation everything boils down to money.  All details are given an economic weight, a dollar value.  Even with lives are lost, they boil down to economic payments to relatives.  The utility of getting somewhere faster or in more comfort, etc., has a monetary value.  All features, regardless of how "cool", have a monetary value.  So these features can be weighed against the cost of crashes.  The engineering model cannot close this loop and cannot quantify risk compensation, but this model can. 

There is a tremendous negative reaction to my model in the "system engineering" community.  Did you see the anecdote about the fire safety blankets?  This system engineer thought he had saved lives by designing a better fire safety blanket.  The very next year statistics proved all he had done was give fire fighters the confidence to get too close to the fire and remain there too long and get themselves killed.  The engineer is not thinking logically in this case, but is emotionally attached to his tools and methods and unwilling to think he might be inadvertently causing harm.

Another important insight is the economic costs of policy thrashing.  There is no way to get this insight from an engineering model.

Robert, it is very nice to get a response from you and I feel slightly under pressure to clarify that I do agree with you that using a monetary value for various aspects allows a more coherent approach and permits a more global calculation to be performed. In fact being a bioengineer, many years working in the health service, my risk management computations did on many occasions require attributing monetary value to an injury or an irreversible injury or death. In all such cases I used data from the health service or the litigation authorities to attribute a monetary value to a harm such as death or injury. That makes the exercise work and yield results and I advocate that myself.

My comments were mostly aimed at the aggregations that may occur in order to bring together different facets of the subject under investigation. It is these aggregations (or generalizations) that dilute the risk computations.

I have to say that I am not familiar with the term “risk compensation” but my guess is that it might be the same as what I would call “risk mitigation”. If so then are you assuming that risk mitigation is something that happens in some natural way or is somehow forced through some kind of interaction (e.g. market forces) in order to achieve some equilibrium? If so then you are trying to use some type of optimization in risk management.

Although risk management seem to always aim to minimise risk (thereby making itself a good candidate for an optimized model) it is not always the option that presents the lower risk that is selected. This is because “risk management” is not the aim but merely one of the means by which we control how we achieve our aim. Take the example of a simple surgical scalpel which obviously presents the risk of injury to the tissue with which it comes into contact. To mitigate the risk would be to put the scalpel in its holder and avoid any tissue contact. However, the surgery cannot proceed with this option!

Using your own draft paper as example, on page 4, you draw similarity between policy and taxation and some product like a washing machine and state that citizens must be willing to pay taxes associated with regulations...

I would be very interested to know how you have worked out that citizens are willing to pay taxes or be penalised by regulations that are usually drawn up by individuals whose own lives are usually not touched by those very incentives. The fact that governments fall (at least in democratic countries that use the ballot box) because of their taxation policies alone is evidence that citizens are not willing to pay those taxes. Moreover, that very fact shows that the policymakers did not attempt to mitigate the risk of (being thrown out of office) by lowering the taxes or giving in to what the voters wanted.

Going back to what I said in my first post, a risk model needs to be specific so that it captures all the risks as I believe any unknown risk is an unacceptable risk.

I would also like to add what I, as a bioengineer would do with the captured risks is to use them to achieve my engineering aims. As for the fire safety blanket, I am not familiar with the details but given the severity of any potential harm I would have thought adequate risk management supported by experimental data would have yielded a safer product.

Once again to ensure any confusion is clarified, as an engineer I do use economic impact as a way of harmonising the units of various parameters during risk management and that is very much an engineering approach.

Hamid, risk compensation is the opposite of risk mitigation, I suggest just Google it.  I give a one paragraph summary of it in the paper.  It is a strong empirical finding but mine is the first quantitative model that explains it.

It sounds like you are doing a good job incorporating unexpected effects in your models.  Keep it up.  Look at how you value and handle testing and defect rates in your computations and tell me if it is similar or different than my crash rate model.

Making products and processes more nearly error free corresponds to lowering the defect ratio and is a very good thing.  This is one way risk analysis and mitigation CAN be practiced.  It USED to be practiced that way where I work, but has become diluted.

Making it less costly to add new valuable features to a product has an unfortunate and undesired effect on the crash rate.  The additional utility invites greater usage and offsets greater failure rate costs.  This is the sticking point of many "system engineering" professionals. 

Making it less costly to find errors or defects has different effects depending on what one does with it.  If it is used only to reduce costs, with testing only to the same defect ratio, then the reduce costs measured against the utility of the product invite the company to continue adding complexity (features) until the total costs including possible increased failure rates again balance the (now larger) revenues.  Companies cannot leave anything on the table.

If one tests to a lower defect ratio, even if some cost savings are realized, then the defect ratio multiplies the effect of testing costs in the model.  This is what most people do not understand, that when lowering costs you have to test to higher levels of quality.

The Japanese understood this in the 1970s and 1980s.  They came up with something called six-sigma methodology which American companies thought was absurd.  They don't think that anymore.  Is this something I should mention in the paper?  Maybe in the conclusion?

Excellent strategy in your responses but i still have not worked out how tax payers are willing to pay taxes and how policymakers can overlook the risks, become unpopular and get themselves voted out of office, the majority of them anyway.

Interesting opinion. When I finish divorce process and arrange other care about me and my kids I may think about it more deeply..

Michal look forward to hearing from you.  Hamid, yes the idea of extending the model to political processes is something I came up with only recently and obviously it doesn't work as reliably as the corporate model, so that in particular is a part I'd like comments on. 

Governments do get overturned if taxes are too high.  Essentially King John was forced to agree to the Magna Charta for this reason (a copy of the thing actually came to Houston, first time out of England).  The next year John got it annulled, which I did not know, by the pope to whom he had sworn allegiance.  But then John died and his son's advisers agreed to a toned down version of it.

That political processes and also war might be describable as economic processes is something I started thinking about last fall, after writing my equity premium book, and noticing the twists and turns the Arab spring was taking.  The thing was started because a street vendor in Tunisia was unable to make ends meet, as he had been hassled by the police to pay too many fees and fines.  He said "I just want to work" and set himself on fire, and the rest is, as they say, "history."  If there were a high level of employment in the Arab world, likely ISIS would not find recruits.  Perhaps I should mention this in my background?  I only give corporate examples at present.

I have uploaded an UPDATE to the paper with changes suggested by the above comments, including discussion of taxation and governmental fees' role in the process of governmental change with examples of changes in totalitarian governments from medieval England to the Arab Spring.  Also including a new section discussing the relationship of this model to Six Sigma.  Amazingly I found that the old Deming-Japan process (not called six sigma) was not entirely a statistical process but had abstract equations comparing total cost and total work much like in the model.

Article Optimizing Innovation and Calamity