2008S_153.4_Preferences.s

2008S_153.4_Preferences.s - MS&E 153: Intro. to...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: MS&E 153: Intro. to Decision Making in Organizations Modeling and Measuring Preferences Sam Holtzman Copyright © 2002-8 Samuel Holtzman. All rights reserved. We want to achieve clarity of action based on decision quality. Meaningful, Reliable Information Clear, Genuine Preferences Sound Creative, Doable Reasoning Alternatives 100% Useful Frame Copyright © 2002-8 Samuel Holtzman. All rights reserved. Commitment to Action 1 of 23 1 Modeling and Measuring Preferences ● Preferences ● Modeling ● Focusing Attention ● Measuring Copyright © 2002-8 Samuel Holtzman. All rights reserved. 2 of 23 How were the preference numbers at the right of each tree branch derived? Plant Size Demand High Large .4 .6 Low High Medium .4 .6 Small Copyright © 2002-8 Samuel Holtzman. All rights reserved. Low Profits US$1,000 million – US$100 million US$900 million US$300 million US$500 million 3 of 23 Expressing preferences requires introspection and can be emotionally difficult. Which of the following two (deterministic) alternatives would you chose? A. $100 million profit 50% of the company laid off. B. $10 million profit No lay-offs. Copyright © 2002-8 Samuel Holtzman. All rights reserved. 4 of 23 Some preference models balance several value sources; they don’t just compute NPV over cash flows. Job Security Profitability Copyright © 2002-8 Samuel Holtzman. All rights reserved. Environmental Consequences Share Price 5 of 23 For unfamiliar, emotional decisions, mathematical clarity yields understanding and calm. ● When thrown into an unfamiliar situation, individuals almost always have difficulty developing and expressing preferences. ● In unfamiliar situations, people need to become familiar with possible prospects. – Explicitness makes challenging prospects easier to handle. ● Exploring each prospect and its personal implications can greatly enhance preference formation. ● By modeling preferences mathematically, we can guide clients to better understand their preferences and their implications for action. ● A preference model is best used as a tool to better understand key preference trade-offs. Copyright © 2002-8 Samuel Holtzman. All rights reserved. 6 of 23 How would you grade MS&E 153? Achievement 1 Individual -1 0 1 Team -1 Effort Copyright © 2002-8 Samuel Holtzman. All rights reserved. 7 of 23 Modeling and Measuring Preferences ● Preferences ● Modeling ● Focusing Attention ● Measuring Copyright © 2002-8 Samuel Holtzman. All rights reserved. 8 of 23 First, compose a pilot preference model. ● A preference model maps deterministic outcome scenarios into a scalar numeraire. ● Many pilot models can be described on the “back of an envelope” (BoE) – The simpler the model, the better, as long as it captures the essence of the trade-offs material to the decision. ● In many cases, it is best to create the pilot model it in an hour or so following a heart-to-heart conversation with the decision maker. ● Make sure you capture both the upside and the downside. – Beware of the decision-maker’s bias. – Champions only see the up-side, administrators only see the down-side ● Get feedback from the decision maker, and revise the BoE model as needed. – Keep the revised pilot model simple. Copyright © 2002-8 Samuel Holtzman. All rights reserved. 9 of 23 A preference model is typically implemented as a spreadsheet. ● A preference model yields a scalar measure of the decision-maker’s preferences for each possible scenario of uncertainties and decisions. ● Except for the simplest cases, preference models have two parts: – A composition model translates resolved decisions and uncertainties into primary preference measures. - E.g., profit, level of employment, environmental effect(s) - E.g., lifespan, lifestyle, comfort, self esteem, dignity, and wealth – A value trade-off function trades off primary preference measures and yields a single scalar numeraire. - NPV, worth numeraire ● Most value trade-off functions are simple – They a single primary preference measure (e.g., NPV). Copyright © 2002-8 Samuel Holtzman. All rights reserved. 10 of 23 Preference models typically have two parts: a composition model and a value trade-off function. Preference Model Composition Model Submodel 1 Uncertainty 1 Uncertainty 2 Submodel 3 Uncertainty 3 Submodel 2 • • • Overall Numeraire Value Trade-off Function Submodel 4 Uncertainty N Submodel 5 1 2 M on sion • • • ion i i is is ec ec ec D D D Copyright © 2002-8 Samuel Holtzman. All rights reserved. 11 of 23 Preference-model spreadsheets are most useful when they are organized by regions. Output: Scenario Value (copied from bottom) Parameters: Locally defined, numeric or nominal They do not change during a set of spreadsheet evaluations Can be set manually or by the probabilistic model 1 Calculations: Scalar and matrix operations on spreadsheet inputs and locally defined parameters. Parametric streams: Arrays calculated from inputs, parameters, calculation results, and other parametric streams Typically used to represent such time sequences as cash flows, symptomatic states, and intermediate time-indexed results Value Trade-off Function Composition Model Inputs: Decisions 1 Uncertainties 1 Value trade-off function: Scalar scenario value calculated from inputs, parameters, calculations, and parametric streams 1 Copyright © 2002-8 Samuel Holtzman. All rights reserved. Typically done by defining inputs as labeled spreadsheet cells. 12 of 23 7 Only three value trade-off function forms are allowed. ● Additive W=Ax+By+Cz ● Multiplicative W=Kxyz ● Sum of products W = M1 ( A1 x1 + B1 y1 + C1 z1) x M2 (A2 x2 + B2 y2 + C2 z2) E ● Where: W is a value trade-of function (or worth numeraire) A, B, C, K, M, and N are constant coefficients x, y, and z are outcomes over which the decision maker has primary preferences E is a constant elasticity coefficient Copyright © 2002-8 Samuel Holtzman. All rights reserved. 13 of 23 Operational and strategic models differ greatly and serve very different purposes. ● Operational – Large – Detailed ● Strategic – Small – Incisive Copyright © 2002-8 Samuel Holtzman. All rights reserved. 14 of 23 Modeling and Measuring Preferences ● Preferences ● Modeling ● Focusing Attention ● Measuring Copyright © 2002-8 Samuel Holtzman. All rights reserved. 15 of 23 Range sensitivity analysis quickly focuses attention on the preference model’s most sensitive inputs. Range 90th% 50th% 10th% Sensitivity Bar $264 $175 $58 ⇐ ⇐ ⇐ $400 Extreme Case Reference Case Extreme Case $500 $600 $700 $563 $264 $175 NPV $58 ● A range is obtained for each preference model input, yielding a first view of the uncertainties affecting the decision at hand. – A range is usually a three-branch tree, comprising a reference case and two opposite extreme cases. – For numerical distinctions, the extreme cases are chosen as the distribution’s 10th and 90th percentiles, and the distribution’s 50th percentile is chosen as the reference case. – To avoid creating an anchor, a range’s extremes are assessed first, followed by the reference case. (Hence, numerical ranges are referred to as 10-90-50 ranges.) ● Each range is mapped into a sensitivity bar. – The preference model input being considered is set to each range value, and the resulting numeraire is recorded. – All other ranges are set to their reference-case value. – The sensitivity bar reflects the highest, lowest, and reference numeraire values. Copyright © 2002-8 Samuel Holtzman. All rights reserved. 16 of 23 A tornado diagram displays all sensitivity bars in decreasing-swing order. Net Present Value –400 –300 –200 –100 Peak Market Size [$MM/year] 0 100 Average Market Price [$000/load] Normalized Labor Costs [$MM/year] Available Nearby Land [acres] Effluent Cleaning Costs [$MM/year] National Distribution Costs [$MM/year] Raw Material Costs [$MM/year] 400 500 2,300 780 Medium Low High 565 83 34 115 480 540 370 120 20 400 210 165 280 35 50 Average Gas Price at Peak Market Size [$/bbl] Process Royalties [$MM/year] 300 1,500 1,100 Consumer Environmental Sensitivity [nominal] 200 Swing 42 100 4 11 60 58 27 43 45 480 370 440 20 1,080 1,320 18 1,530 A sensitivity bar’s swing is the difference between its maximum and its minimum. 5 Reference Case: $125 million Copyright © 2002-8 Samuel Holtzman. All rights reserved. 17 of 23 1 A range’s normalized squared swing approximates* the corresponding uncertainty’s contribution to risk. Net Present Value –400 –300 –200 –100 0 100 200 300 400 500 Swing Swing 2 Swing 2 Σ Swings 2 Cumulative Normalized Swing 2 1,500 2,300 780 608,400 51.78% 51.78% Medium High 565 319,225 27.17% 78.95% 400 160,000 13.62% 92.56% 210 44,100 3.75% 96.32% 165 27,225 2.32% 98.63% 100 10,000 0.85% 99.49% 1,100 Low 83 34 115 480 540 370 120 280 35 50 20 42 4 11 60 3,600 0.31% 99.79% 58 27 43 45 2,025 0.17% 99.96% 480 370 440 20 400 0.03% 100.00% 1,080 1,320 5 25 0.00% 100.00% 18 * Risk ~ variance; exact only when all uncertainties are mutually irrelevant and the preference model is additive. 1,530 $125 million Copyright © 2002-8 Samuel Holtzman. All rights reserved. 18 of 23 We can reliably fixate the distinctions below a threshold and aleate only those above the threshold. Net Present Value –400 –300 –200 –100 Peak Market Size [$MM/year] 0 100 Average Market Price [$000/load] Normalized Labor Costs [$MM/year] Available Nearby Land [acres] Effluent Cleaning Costs [$MM/year] National Distribution Costs [$MM/year] Raw Material Costs [$MM/year] 400 500 2,300 51.78% Medium Low High 78.95% 83 34 115 480 540 370 120 20 92.56% 96.32% 98.63% 280 35 50 Average Gas Price at Peak Market Size [$/bbl] Process Royalties [$MM/year] 300 1,500 1,100 Consumer Environmental Sensitivity [nominal] 200 Cumulative Normalized Swing 2 42 99.49% 4 11 99.79% 58 27 43 99.96% 480 370 440 100.00% 1,080 1,320 100.00% 18 1,530 $125 million Copyright © 2002-8 Samuel Holtzman. All rights reserved. 19 of 23 Modeling and Measuring Preferences ● Preferences ● Modeling ● Focusing Attention ● Measuring Copyright © 2002-8 Samuel Holtzman. All rights reserved. 20 of 23 The walk-away wizard helps you quickly measure a decisionmaker’s preferences over “intangible” prospects. 1. Help your client visualize the “intangible” prospect. 2. State that a wizard has entered the room who, for a fee, will make the desired outcome happen. 3. By asking for (ridiculously) low and high wizard fees determine that the desired outcome has value, but that the value is limited. 4. Ask for the most your client would pay the wizard (low anchor fee) 5. Increase the amount several times until you find a value that your client would not pay (initial indifference fee). 6. Now the fun starts … Copyright © 2002-8 Samuel Holtzman. All rights reserved. 21 of 23 The walk-away wizard conversation switches to an ultimatum mode. 7. Summarize the prospect and once again help your client visualize it. 8. Tell your client that the wizard will walk away (hence the technique’s name) unless your client pays 50% more than the initial indifference fee.) 9. Keep asking ultimatum questions with higher (or possibly lower) fees until you find your client’s “breaking” fee – just before the point at which your client would allow the wizard to walk away. 10. That breaking fee is your client’s value for the intangible prospect Copyright © 2002-8 Samuel Holtzman. All rights reserved. 22 of 23 The walk-away wizard assessment method has valuable characteristics. ● The ultimatum nature of the walk-away wizard method prevents negotiating behavior, which can significantly distort expressions of value. ● You may see dramatic valuation changes as your client’s perception of each scenario in question improves. – It is not uncommon for the breaking indifference fee to be over 10-100 times higher than the initial indifference fee. ● The walk-away wizard method works in a broad range of situations involving “intangibles”. – Experience shows that it works well for measuring people’s willingness to pay related to: consumer goods, medical interventions, negotiating, or bidding. Copyright © 2002-8 Samuel Holtzman. All rights reserved. 23 of 23 MS&E 153: Intro. to Decision Making in Organizations Modeling and Measuring Preferences Sam Holtzman Copyright © 2002-8 Samuel Holtzman. All rights reserved. ...
View Full Document

This note was uploaded on 06/16/2010 for the course MS&E 153 taught by Professor Ronhoward during the Summer '08 term at Stanford.

Ask a homework question - tutors are online