Decision Theory Lecture 4 Decision Theory – the foundation of modern economics • Individual decision making – under Certainty • • • Choice functions Revelead preference and ordinal utility theory Operations Research, Management Science – under Risk • • • • Expected Utility Theory (objective probabilities) Bayesian decision theory Prospect Theory and other behavioral theories Subjective Expected Utility (subjective probabilities) – under Uncertainty • • • Interactive decision making – – – – • Decision rules Uncertainty aversion models Non-cooperative game theory Cooperative game theory Matching Bargaining Group decision making (Social choice theory) – Group decisions (Arrow, Maskin, etc.) – Voting theory – Welfare functions • Individual decision making – under Certainty • Choice functions Choice Choice function Weak axiom of revealed preference (WARP) Exemplary choice functions • Pick the cheapest (e.g. public tenders) • Pick the second cheapest (wine for a party) • Maximize the IRR (investment projects) • Pick whoever gets majority of votes (Talent shows on TV) • … 4 good 2. Choice functions – some intuition (1) Out of the gray set, A was chosen (a unique choice) B Out of the blue set, B was chosen (a unique choice) Do we find these choices confusing? (when considered collectively) A good 1. 5 Choice functions – some intuition (2) good 2. Out of the gray set, A was chosen (a unique choice) Out of the blue set, B was chosen (a unique choice) Do we find these choices confusing? (when considered collectively) B A good 1. 6 Choice functions – some intuition (3) good 2. Out of the gray set, A was chosen (a unique choice) Out of the blue set, B was chosen (a unique choice) Do we find these choices confusing? (when considered collectively) A B good 1. 7 Good 2. Choice functions – some intuition (4) Out of the gray set, A was chosen (a unique choice) A Out of the blue set, B was chosen (a unique choice) Out of the golden set, C was chosen (a unique choice) B Do we find these choices confusing? (when considered collectively) C Good 1. 8 Homework 1. Can we, using only linear budget constraints, construct such an example for two goods, that there is a „consistency problem” when considering more than two alternatives, and no problem when considering only each two alternatives separately? 2. And when considering three goods? 9 Choice functions – a formal definition • Notation: X B 2X , B C :B B set of decision alternatives available menus (non-empty subsets of X) choice function, working for every menu • (Technical) properties: C (B ) always a choice C ( B) B out of a menu • If C(B) contains a single element this is the choice • If more elements these are possible choices (not simultaneously, the decision maker picks one in the way which is not described here) 10 An exercise • Let X={a,b,c}, B=2X • Write down the following choice functions: – C1: always a (if possible), if not – it doesn’t matter – C2: always the first one in the alphabetical order – C3: whatever but not the last one in the alphabetical order (unless there is just one alternative available) – C4: second first alphabetically (unless there is just one alternative) – C5: disregard c (if technically it is possible), and if you do disregard c, also disregard b (if technically possible) 11 The solution B C1(B) C2(B) C3(B) C4(B) C5(B) {a} {a} {a} {a} {a} {a} {b} {b} {b} {b} {b} {b} {c} {c} {c} {c} {c} {c} {a,b} {a} {a} {a} {b} {a,b} {a,c} {a} {a} {a} {c} {a} {b,c} {b,c} {b} {b} {c} {b} {a,b,c} {a} {a} {a,b} {b} {a} 12 The solution B C1(B) C2(B) C3(B) C4(B) C5(B) {a} {a} {a} {a} {a} {a} {b} {b} {b} {b} {b} {b} {c} {c} {c} {c} {c} {c} {a,b} {a} {a} {a} {b} {a,b} {a,c} {a} {a} {a} {c} {a} {b,c} {b,c} {b} {b} {c} {b} {a,b,c} {a} {a} {a,b} {b} {a} 13 Desirable properties • Sometimes an internal consistency is postulated • Why so? – positive approach – non-consistent will go bankrupt – normative – in order not to go bankrupt • We’ll discuss the following: – – – – weak axiom of revealed preferences a property b property g property 14 WARP – weak axiom of revealed preferences Definition (WARP): A pair(B,C()) satisfies WARP, if the following holds: if for some B from B, s.t. x,yB, we have xC(B), than for every B’ from B, s.t. x,yB’, if yC(B’), then xC(B’). Intuitively: if x was shown to be at least as willingly picked as y (for a menu B), then for every menu B’ containing x,y, if y is picked, so does x have to be. 15 good 2. WARP – an intuition Out of the gray set, A was chosen (a unique choice) B Out of the blue set, B was chosen (a unique choice) Do we find these choices confusing? (when considered collectively) A good 1. 16 WARP – an intuition good 2. Out of the gray set, A was chosen (a unique choice) Out of the blue set, B was chosen (a unique choice) Do we find these choices confusing? (when considered collectively) B A good 1. 17 WARP – an intuition good 2. Out of the gray set, A was chosen (a unique choice) Out of the blue set, B was chosen (a unique choice) Do we find these choices confusing? (when considered collectively) A B good 1. 18 An exercise • Check which functions C1-C5 do not fulfill WARP, prove by giving exemplary menus B C1(B) C2(B) C3(B) C4(B) C5(B) {a} {a} {a} {a} {a} {a} {b} {b} {b} {b} {b} {b} {c} {c} {c} {c} {c} {c} {a,b} {a} {a} {a} {b} {a,b} {a,c} {a} {a} {a} {c} {a} {b,c} {b,c} {b} {b} {c} {b} {a,b,c} {a} {a} {a,b} {b} {a} 19 The solution • C1 – fulfils • C2 – fulfils • C3 – doesn’t! b picked from {a,b,c} and not from {a,b} • C4 – doesn’t! b picked from {a,b,c} and not from {b,c} • C5 – doesn’t! b picked from {a,b} and not from {a,b,c}, while a picked 20 a property (Chernoff property) Definition (a property): Assume B=2X. C() meets a, if the following holds: if for some B out of B we have xC(B), then for every B’B, s.t. xB’, we have xC(B’). Intuitively: if x picked from menu B, then shall be picked from each smaller menu B’ (if present in it). 21 a property differently • If something not picked from menu B’, shan’t be picked from a bigger one: B' B B'\C ( B' ) B \ C ( B) • If we add to B1 some new alternatives B2, then the choice will either not change, or something out of new alternatives should be picked B1 , B2 B : C( B1 B2 ) C( B1 ) B2 22 Homework Prove that the previous definitions are equivalent 23 An exercise – check the a property for C1-C5 B C1(B) C2(B) C3(B) C4(B) C5(B) {a} {a} {a} {a} {a} {a} {b} {b} {b} {b} {b} {b} {c} {c} {c} {c} {c} {c} {a,b} {a} {a} {a} {b} {a,b} {a,c} {a} {a} {a} {c} {a} {b,c} {b,c} {b} {b} {c} {b} {a,b,c} {a} {a} {a,b} {b} {a} WARP yes yes no no no a yes yes no no yes 24 b property Conclusion for the previous exercise – a and WARP differ (let’s look for other properties) Definition (b property): Take B=2X. C() meets b property, if the following holds: if form some B’ in B we have x,yC(B’), than for each B, B’B, we have xC(B) yC(B). Intuitively: if x and y are picked in a menu B’, then their status is equal in every greater menu B. 25 An exercise – check b property for C1-C5 B C1(B) C2(B) C3(B) C4(B) C5(B) {a} {b} {a} {b} {a} {b} {a} {b} {a} {b} {a} {b} {c} {a,b} {a,c} {c} {a} {a} {c} {a} {a} {c} {a} {a} {c} {b} {c} {c} {a,b} {a} {b,c} {a,b,c} {b,c} {a} {b} {a} {b} {a,b} {c} {b} {b} {a} WARP yes yes no no no a yes yes no no yes b yes yes yes yes no 26 g property Definition (g property): Assume B=2X. C() meets g, if the following holds: if for every menu Bi out of a family of menus we have xC(Bi), then for B=Bi we have xC(B). Intuitively: if x is picked in every menu (in a family of menus), than it is also picked in a joint menu 27 An exercise – check g property for C1-C5 B C1(B) C2(B) C3(B) C4(B) C5(B) {a} {a} {a} {a} {a} {a} {b} {b} {b} {b} {b} {b} {c} {c} {c} {c} {c} {c} {a,b} {a} {a} {a} {b} {a,b} {a,c} {a} {a} {a} {c} {a} {b,c} {b,c} {b} {b} {c} {b} {a,b,c} {a} {a} {a,b} {b} {a} WARP yes yes no no no a yes yes no no yes b yes yes yes yes no g yes yes yes no no 28 The complete solution B C1(B) C2(B) C3(B) C4(B) C5(B) {a} {a} {a} {a} {a} {a} {b} {b} {b} {b} {b} {b} {c} {c} {c} {c} {c} {c} {a,b} {a} {a} {a} {b} {a,b} {a,c} {a} {a} {a} {c} {a} {b,c} {b,c} {b} {b} {c} {b} {a,b,c} {a} {a} {a,b} {b} {a} WARP yes yes no no no a yes yes no no yes b yes yes yes yes no g yes yes yes no 29 no Properties and manipulation • Assume C1-C5 can be used in a public tender (a,b,c denote offers) • Take C3({a,b})={a}, C3({b,c})={b}, C3({a,b,c})={a,b} – different choice for a complete problem (b may be selected), – different when short listing – … pairise comparisons also change the outcome – b „better than” c, a „better than” b, hence a – putting c on the table impacts the chocie (favours b – possible alliance) 30 An exercise • Public tender • Alternatives – offers described by: price and time to deliver (quality is constant) • Rule #1: – minimize the expression a pricei + b timei (for some weights a>0, b>0 determined irrespectively of set of offers) • Rule #2: – calculated the minimal price (MP) and minimal time (MT) for all offers (assume MP>0 and MT>0) – minimize the expression pricei/MP + timei/MT • Which rule do you like? 31 The solution • Rule #1 – meets’em all: WARP, a, b, g (intuitively – the evaluation does not depend on the menu, will be formalized later) 32 The solution • Rule #2 – doesn’t meet a single one • Take B={x,y,z}, x=(4,4), y=(1,9), z=(16,1) – what will be selected? • Try to find some modifications in order to show how a, b, g are broken 33 Summing up • Different views on decision making – choice and choice functions – preferences – utility function • We can judge not only alternatives, but also choice rules – not meeting some properties yields a risk of being manipulated – different properties, not all of them equivalent 34 Materials • Compulsory: – A. MasColell, M. Whinston, J. Green Microeconomic Theory, Oxford University Press, 1995, rozdz. 1 • Supplementary: – A. Sen, Choice Functions and Revealed Preference, The Review of Economic Studies, 1971, 38(3), s. 307-317 35