Introduction to Decision Analysis for Public Policy Decision-Making

British Decision to Develop Radar During WWII Introduction to Decision Analysis for Public Policy Decision-Making

Ross D. Shachter Management Science and Engineering Stanford University

1 © Ross D. Shachter MS&E 290, Public Policy Analysis 2

British Decision to Develop Radar During WWII The Serenity Prayer

Choice between Infrared and Radar to detect God grant me

German bombers  the serenity to accept Churchill protégé advocated the things I cannot change, Uncertainties infrared, which would have  the courage to change been ineffective the things I can, and Decisions Consequence: infrared  the wisdom to know the difference. Decision Analysis would not have prevented a German invasion Reinhold Niebuhr (1892-1971) Right decision by chance Inspired by Friedrich Oetinger and Boethius.

Action

Actional Thought Thought Bayesian Perspective

Some of our actions merit thought and reflection. We measure our beliefs about thosee uncertain Not all of our thoughts are focused on possibilities we care about using probabilities. interventions we can make in our world. Those probabilities are based on our current state Decision analysis lets us think clearly about a of information, &, from life experience, expert complex problem so that we can choose those opinions, and observation. actions that yield our preferred prospect. (There are no “objective” probabilities.) The quality of a decision is determined by the The possibilities are distinctions we impose on the process we use, not by the eventual outcome. world. We define them with enough clarity that A good decision can lead to a bad outcome. the uncertainty arises from our lack of knowledge, rather than from ambiguity.

1 Decision Analysis for Public Decisions Uncertainties Decisions Policy and Prospects Prospect

 Individual Decision Maker  A decision is an irrevocable allocation of resources,

 Identify single “rational” decision maker for public decision, such completely under the decision maker’s control.

as President, Governor, etc.  Doing nothing can be a decision.  Many stakeholders are affected by public policy decisions and  The decision to act or intervene in the world can have process should engage them multiple possible effects.  Normative Approach  A prospect is the future the decision maker faces with a  How we would want choices made rather than how choices might otherwise be made in a political process particular set of choices. It can be a deal, made up of multiple possible prospects.  Adds transparency and defensibility to a public process  The decision maker should choose those actions and interventions that yield his/her preferred prospect.

Decision Basis $20 Auction

 To build a model, we  What can you do? We will conduct an auction address the following Alternatives questions:  The highest bidder will receive a $20 bill  What is your real  Who makes the final problem/opportunity?  The top TWO bidders have to pay commitment? Frame  I reserve the right to back out of the deal Decision Maker  Do we have a process?  What do you want? Logical Reasoning Preferences Decision Analysis!  What do you know?  And afterwards? Information Commitment to Action

The Normative Rules of Jane’s Party Problem Actional Thought ? Outdoors  Jane wants to decide where & Porch Lead to a logically consistent normative to have her party but she is Indoors not prepared to compare analysis for a rational decision maker: the prospects of each  1. Given your observations and choices, you Party location. Weather can recognize each possibility that you care  She recognizes that the Location about and you can assign a probability to it. party value depends on the Party  2. You can think about the prospect of each of weather. Value  She consider the uncertain those possibilities, and you can identify a best possibilities for the prospect b and a worst prospect w. (These do Sun Weather to be Sun or Rain. & not have to be unique.) Rain

2 Rules of Actional Thought Normative Analysis of (Continued) Jane’s Party Problem

0.4  3. For any prospect x, there is a number u(x), Sun & 0.6 for any party location 0 ≤ u(x) ≤ 1, such that: u(x) b Rain x ~ w 1-u(x) Party Outdoors Porch Indoors  4. You prefer prospect x to prospect y if and Values only if u(x) > u(y). Sun $100 $90 $40  5. If you are indifferent between prospects x b and y then you are indifferent between two deals if one can be formed from the other by Rain $0 $20 $50 w substituting x for y. For Jane, u(x) = x/$100 for any prospect value x.

Normative Analysis of Value of Clairvoyance

Jane’s Party Problem (Continued)  Suppose Jane knows the weather (clairvoyance) before she chooses the party location: 0.4 $40 Sun $100 >> Outdoors $100 Outdoors 0.6 0.4 $100 Rain $0 Sun Porch $90 0.4 $70 Indoors $48 $48 Sun $90 & $40 & >> Porch 0.6 Rain $20 Outdoors $0 0.4 0.6 Rain $50 $46 Sun $40 Porch $20 Indoors 0.6 Rain $50 >> Indoors $50  To determine the prospect value of Jane’s decision  The prospect value of Jane’s decision opportunity with opportunity, we must determine clairvoyance is $70. (It cannot decrease!)

 her prospect value for each party location  Thus, she is willing to pay up to $22 = 70 - 48 for clairvoyance.

 her preferred choice for party location  This is the most she would pay for any information.

Sensitivity Analysis Decision Hierarchy $100 Frame determines the level at which this Policy

Value decision is made Optimal Strategy  Assuming levels above Indoors Porch  Setting assumptions for Prospect Outdoors the levels below Tactics It is impractical to $0 relate each decision to Or to relate it to the 0.0 Probability of Sun 1.0 fundamental values detailed prospects

3 US/Soviet Decisions during Decision Analysis Cycle Cuban Missile Crisis

Continue to improve the modeled basis -- Both sides had “facts” that uncertainties, alternatives, preferences -- were not correct and did not until there is clarity of thought about action. understand each other Consequence: closest we Formulate Analyze Recognize ever came to a nuclear war Basis Model Decision Right decisions and good Opportunity Appraise process, but wrong Basis information Take Action © Ross D. Shachter MS&E 290, Public Policy Analysis 19 © Ross D. Shachter MS&E 290, Public Policy Analysis 20

Decision  Determined by the decision process, not by the eventual outcome.  Appropriate Frame: solving the Important Extensions Quality correct problem;  Creative Alternatives: taking full advantage of the decision Value Models within Information Models Commitment to Action opportunity; the DA Framework  Representing  Information: accurate modeling of relationships in possibilities, relationships,  Multiple objectives probabilities, and observations; graphical models Logic  Risk aversion  Values: capturing the important  Probability assessment features of the decision maker's  Time value of money preferences;  Using experimental • capital expenditures  Logical Reasoning: applying clear

• deferred benefits evidence thinking to the entire decision Preferences

Alternatives Information situation; and  Multiple decision  Commitment to Action: makers Frame implementation of the decision- maker's insights.

In Closing

A good decision can lead to a bad outcome. The quality of information available at the time of the actual decision is critical. You are always better off if you have

 More alternatives or

 More information