Causes of Poor Decisions

Transcription

1 Lecture 7: Decision Analysis Decision process Decision tree analysis The Decision Process Specify objectives and the criteria for making a choice Develop alternatives Analyze and compare alternatives Select the best alternative Implement the chosen alternative Monitor the results 1 2 Causes of Poor Decisions i Bounded Rationality The limitations caused by costs, human abilities, time, technology, and availability of information Sub-optimization Different units each attempting to reach a solution that is optimum for itself Decision Environments Certainty t - Environment in which h relevant parameters have known values Risk - Environment in which certain future events have probable bl outcomes Uncertainty t - Environment tin which hiti is impossible ibl to assess the likelihood of various future events 3 4

2 Decision Making under Uncertainty Maximin i Choose the alternative ti with the best of the worst possible payoffs Maximax - Choose the alternative with the best possible payoff Laplace - Choose the alternative with the best average payoff of any of the alternatives Decision Making under Risk Maximin i Choose the alternative ti with the best of the worst possible payoffs Maximax - Choose the alternative with the best possible payoff EMV - Choose the alternative with the best expected monetary value of any of the alternatives 5 6 List of Factors Possible future conditions or states t of nature Likelihood or probability of each future condition Decision alternatives Outcome or payoff for each alternative under every future condition Decision criterion An Example A firm must decide whether to construct a small or large stamping plant. A consultant s report indicates a 0.2 probability that demand will be low, and a 0.8 probability that the demand will be high. If the firm builds a small facility and the demand turns out to be low, the net present value will be million. If demand turns out to be high, the firm can either subcontract and realize a net present value of million or expand greatly to realize a net present value of million. If the firm builds a large facility and the demand is low, the net present value will be million, whereas high h demand d will result in a net present value of million. 7 8

3 Constructing a Decision Tree Decision Tree Pick an evaluation date and units List the decisions and random events, including all decisions and events which affect the final outcome all decisions that may affect the outcomes of future events or alternatives of future decisions Work through the possible decisions and random events chronologically Fill in final outcomes and payoffs, and probabilities of random events if possible Decision Point Chance Event 9 10 Applying EMV Criterion Step 1. Starting from the top of the tree, find a node whose immediate successors are all either end points or have already been examined. If the node is a decision node, look at the EMVs of all successors and choose the one with the highest EMV. If the node is a chance node, then multiply the probability of each branch with the EMV (or final outcome) of that branch, and sum over the branches. Step 2. Go back to Step 1 until every node has been examined. Expected Monetary Value (EMV) Optimal course of action: Build small and, if demand high, expand

4 Different Decision Criteria Criterion Maximin Optimal Decision Small Maximax Large EMV Small 13 Using Decision Tree to Evaluate Capacity Alternatives Sales growth at Hackers Computer Store over the past few of years has been good. The owner is considering three courses of action: A: Move to new location B: Expand the store C: Do nothing now; consider expansion one year later The demand growth may be strong or weak. The probabilities are estimated as 0.55 and The owner also estimates the cost of each alternative and the revenue under the differing probable demand levels. 14 Decision Tree Analysis $765,000 $365,000 $863,000 $413,000 $843,000 $850,000 How would You Answer? Does an optimal decision always lead to an outcome more desirable than a non-optimal decision? Should money and effort be spent to investigate which alternative would actually happen? What should be done if the probabilities biliti of the states t of nature are known but only imprecisely? $525,

5 Expected Value of Perfect Information EVPI -- the difference between the expected payoff under certainty and the expected payoff under risk Expected value of Expected payoff Expected payoff perfect information = under certainty under risk _ EVPI - An Example Without information 56.4 With perfect information EVPI = 56.4 = Sensitivity Analysis Sensitivity Analysis 42P L +48(1-P L ) =48-6P L EMV Small plant, 48-6P L P L (-20)P L +60 (1-P L ) = 60-80P L -20 The two lines cross at P L = 0.16, i.e., small plant is better if P L >