Decisions I

Risk and

Decision Strategies

The Classical Textbook Decision Process

1. Identify the problem

2. Specify objectives and decision criteria

3. Identify alternatives

4. Analyze and compare alternatives

5. Select the best alternative

6. Implement

7. Monitor results Step 8: admit and correct mistakes step zero: determine that there really is a problem

The First Complication--There are Different Decision Environments--

Certainty--the outcome of several courses of action are known for sure. We need only pick the best alternative.

Risk--We can calculate the outcomes, but we only know probabilities they will occur

Uncertainty--We can calculate outcomes, but have no idea about relative probabilities. This is also called G.O.K.

God Only Knows

Decision Strategies under Uncertainty

$M, NPV

demand low moderate high

Small 10 10 10

Medium 7 12 12

Large -4 2 16

Result of preferred choice (numeric)

StatisticianPessimistoptimist

What does the optimist see as the most likely result from each strategy? How about the pessimist? And the statistician?

If each wants the best outcome from what they see as most likely, which would they choose,S, M, or L?

Fill in the results for each strategy, then click the corresponding check button to see if the responses are correct.

Not knowing the probability of success for specific products, should we build small, medium or large? Depends on what you expect from each strategy. The optimist assumes the best would happen in each case, the pessimist expects the worst. Our marketing folks and engineers came up with these projections for NPV for each combination of plant size and demands. Fill in the numbers, then click the check button.

Is this a surprising result?With this set of numbers--

The optimist (Maximax strategy) would build a large plant The pessimist (Maximin strategy) would build a small plant The statistician (LaPlace strategy) would build a medium size plant

But is this always what these strategies would give as results?

Consider another feasible situation:suppose having a small plant with a high demand market puts you at a disadvantage because competitors are enticed into the market, and you suffer low credibility as a supplier. Thus you have the ironic result of a lower return because demand is higher. There could be a similar result for a medium size plant. Perhaps even with low demand, a large plant could give economies.

Decision Strategies under UncertaintyA different set of numbers gives a different result

$M, NPV

demand low moderate high

Small 10 10 6

Medium 10 14 8

Large 10 14 20

Result of preferred choice (numeric)

Pessimistoptimist

What does the optimist see as the most likely result from each strategy? How about the pessimist? And the statistician?

If each wants the best outcome from what they see as most likely, which would they choose,S, M, or L?

Fill in the results for each strategy, then click the corresponding check button to see if the responses are correct.

Not knowing the probability of success for specific products, should we build small, medium or large? Depends on what you expect from each strategy. The optimist assumes the best would happen in each case, the pessimist expects the worst. Our marketing folks and engineers came up with these projections for NPV for each combination of plant size and demands.

check Pesscheck Opt

What a surprising result!With this set of numbers-

The optimist would build a large plant The pessimist would also build a large plant!

Pessimists and optimists both want the best result available. They differ in their views of what is available.

Is it better to be an optimist or a pessimist?

Attitude affects outcome. In each situation, Is it better to be an optimist or a pessimist?

Situation:

1.Working hard at this research may give a breakthrough.2.People at the party will be interesting & friendly.3.The concert will be fun.4.I will continue to live a few more years.5.The chemical reaction won't be violent. 6.The airplane may have enough fuel.7.The prisoners will not get violent.8.The river would never flood.9.I might win at the gambling boats.10.I might win the Reader's Digest sweepstakes.

Optimist . . . Pessimist

Is there a pattern to your responses? Does it relate to severity of the risk?

Most People are not “Optimists” or “Pessimists”Rather, they make decisions to avoid perceived Personal Risk

The Rule of Thumb is

“Never Risk more than you can afford to lose”

Thus, real-life decision makers only want to know three things:

1. What’s the probability this will turn out okay?

2. If it goes bad, how bad could it get?

3. What can I do to affect the outcome?

Real life decision makers consider the potential for blame and punishment (Fear is a great motivator.)

This Strategy is called “Minimax Regret” because the selection is made by considering the worst thing that could happen in each alternative and avoiding those that have the worst possible consequences For each possible state in the future consider how much

criticism you could get from your boss. This is called “regret”. Consider the worst thing that could happen for each alternative. Avoid alternatives that have large potential regrets Pick the alternative with the lowest worst regret.

For example, let’s look at another case, and look at potential for regrets (blame) for each state of demand, depending on which size plant we built.

Decision Strategies under Uncertainty—Minimax Regret

$M, NPV

demand low moderate high

Small 10 10 10

Medium 7 12 12

Large -4 2 16

Not knowing the probability of success for specific products, should we build small, medium or large? For each possible future demand state, fill in how much criticism you would get for having built each size plant. For example, If in the future demand is low, a small plant would give the best possible return, so there would be 0 regret. On the other hand, if you had a medium-size plant, your boss might criticize you for only making $7 M instead of $10 M. The regret would be $3M. After filling in the 3 values for each demand, click the button to check your calculations for each alternative, then click the button for the one you would choose based on this approach.

Regrets

selectMax Regret

check mod check hicheck Low chk max

small

medium

large

Decision Strategies under Risk--Calculation of expected Monetary Values for the Small Medium and large strategies

EMV

Suppose we knew ahead of time what the demand would be in a specific case? Could we do better than consistently building medium?

Marketing people reviewed history and estimated Probabilities of the levels of demand for this type of product. We can’t tell for each individual case what will happen, but knowing probabilities allows us to project an average or expected return for a consistent strategy of building small, medium, or large size plants. The expected return is a weighted average of the returns under different demands or “states of nature” . This weighted average is called an “Expected Monetary Value” or EMV, and the goal in this case is to pick the strategy with the Maximum EMV

Max EMV

$M, NPV

demand low moderate high P( ) 0.3 0.5 0.2

Small 10 10 10

Medium 7 12 12

Large -4 2 16

check EMV

Decision Strategies under Risk--Calculation of expected Monetary Value When we have perfect information ahead of time

Max EMV, when we have risk, or only know overall probabilities gives the best result for the large size plant in this case. Having perfect information ahead of time for each individual decision would allow us to know with certainty what the best size plant would be. This allows us to pursue a specific strategy rather than a consistent general strategy. If We knew ahead of time that demand would be low, what size plant would we build and how much would we make? How about if we knew, this time, time would be moderate? What would we do in high demand?

Expected Value of Perfect Information (EVPI) is the difference between EMV with the information(certainty) and EMV with only probabilities (risk)

EVPI =

$M, NPV

demand low moderate high EMV P( ) 0.3 0.5 0.2

Small 10 10 10 10

Medium 7 12 12 10.5

Large -4 2 16 3.0

check EMVcCheck EVPI