TWP asks: "What weight to give to each?" Surprisingly, it might be easier to answer that question that to try to assign probabilities of occurrence. I say "easier" because there is a well-used technique, developed by Saaty and Saaty in the 1970s (I think) which is particularly suited to divining RELATIVE importance factors between several choices. And, it's relative importance that would seem to be relevant here. In this instance, "choice" means choice of threat to mitigate, and so "choice" = "threat." Humans are particularly good at choosing between one of two things, and much worse at choosing between a large number of alternatives. This is the problem that was tackled by Saaty and Saaty.
The method is very simple, but it helps greatly if you use a spreadsheet.
Create a square matrix (equal number of rows and columns). Label each row with the threat (Monetary Policy, Earthquake, Flood, Storm, Civil Unrest, War) from top to bottom. Label each column from left to right with the same threats, and in the same order.
Put a "1" in each diagonal square.
Then, start working on the first row; the one labeled "Monetary Policy." In the first row, first column, you will have already put a "1." Next, go to column 2, the one labeled "Earthquake." Consider which you believe to be the bigger threat, by any way that you want to define "bigger." Then, how certain are you that your choice is correct? Score your belief on a scale of 1 to 5 (no zeroes!). If you think that both threats are of equal importance to you, put a "1" in that square on the grid. If you think that Monetary Policy is a much 'bigger' threat to you than Earthquake, then put a "5" in the square. If you think, by contrast, that Earthquake is a much bigger threat to you than Monetary Policy, put 1/5 into the grid square. Use numbers between 2 and 5 to set intermediate levels of confidence. For example, "2" means "with a whole lot of work, I could probably prove to myself that one choice is better than the other," "3" means "I wouldn't have much work to do to prove that my choice is correct, but some people would probably argue that I got it wrong, and they would probably have a rationale for their beliefs," "4" means "I can easily substantiate that my choice is correct, but there is still some doubt that others could come up with a contradiction," and "5" means "my choice is axiomatic or nearly so."
Work your way across the matrix in the same fashion, giving scores to each grid square. Remember, if the threat in the row is more important than the one in the column, use a score of 1 to 5. If the threat in the column is more important than the one in the row, then use a score of 1 to 1/5 (i.e., 1, 1/2, 1/3, 1/4 or 1/5).
Now, go to the second row. But, don't start at column 1. Rather, start at the first column after the grid square with the "1" in it (the diagonal); that is, row 3. Then, continue to the end of the row as before.
Work your way down and complete the last row. When you are done, in addition to the "1" put into the diagonal, you will have filled out all the grid squares on the upper left half of the matrix. For example, if you had a total of six threats, you would have put six "1s" in the diagonal grid squares plus you would have filled out all the other 15 grid squares to the right of the diagonal.
Here is where it gets tricky in that you need to fill out the remaining blank grid squares with the reciprocals of what you put into the grid squares across the diagonal.
The reciprocal is simply =1/x, where "x" = the value in the reciprocal grid square.
What is the reciprocal grid square? It's the grid square that has the row and column indices swapped or inverted from the grid square you are currently filling in. For example, if you are working on filling in grid square row 3, column 2, then x is the value you will find by looking in grid square row 2, column 3 -- which will be a value you filled in earlier. Let's say grid square row 2, column 3 contains "3," then in the blank grid square at row 3, column 2 you would enter "=1/3"
Fill out the remaining 15 blank grid squares in this way.
The hard work is almost done!
In each row, compute the geometric sum:
For each row in the matrix, multiply all the numbers together and take the nth root of the result, where n= number of columns in the matrix.
The result is the relative weight of each choice, denoted by the label on each row.
You can see the result more easily if you normalize by dividing each resulting row value by the sum of all row values, so that the total will add to 1.00, but that's just a niceity.
The attachment has a worked example (in .pdf). I added "EMP" to the list of threats, which makes a 7x7 matrix.
The Saaty & Saaty method was developed to match, as simply as possible, the way human psychology seems to work. The trick is forcing the evaluator to consider only pairwise choices at any one time. Another breakthrough is the recognition that animal brains tend to accumulate evidence as geometric, rather than linear, sums and work in relative space (logarithmic space), not cartesian space.
If you want more background, I think the original paper is readily available from the IEEE.
The Saaty & Saaty method has been used for technology development risk management for two generations.
15 Replies
1303 Views