Discussion of Optimal Taxation of Behavioral Agents by Emmanuel Farhi and Xavier Gabaix

Transcription

1 Discussion of Optimal Taxation of Behavioral Agents by Emmanuel Farhi and Xavier Gabaix Ali Shourideh CMU May , BFI Conference on Taxation and Fiscal Policy

2 BOUNDED RATIONALITY AND TAXES People make choices not consistent with the canonical models in PF: Chetty, Looney and Kroft (2009): They fail to account for sales taxes Lacetera, Pope and Sydnor (2012): buyers of used cars look at the left digit of odometer even when other information is available Etc. etc. How does this change the basic lessons of optimal taxation? Inverse elasticity rule, Production efficiency, etc.?!2

3 THIS PAPER!3

4 THIS PAPER Distinguish choice from welfare: decision utility vs. experience utility!3

5 THIS PAPER Distinguish choice from welfare: decision utility vs. experience utility!3

6 THIS PAPER Distinguish choice from welfare: decision utility vs. experience utility!3

7 THIS PAPER Distinguish choice from welfare: decision utility vs. experience utility Use Gabaix (2014) s framework to revisit the classic results in Public Finance!3

8 THIS PAPER Distinguish choice from welfare: decision utility vs. experience utility Use Gabaix (2014) s framework to revisit the classic results in Public Finance Inverse Elasticity Rule: holds but inattention increases taxes!3

9 THIS PAPER Distinguish choice from welfare: decision utility vs. experience utility Use Gabaix (2014) s framework to revisit the classic results in Public Finance Inverse Elasticity Rule: holds but inattention increases taxes Production Efficiency and irrelevance: it does not hold!3

10 THIS PAPER Distinguish choice from welfare: decision utility vs. experience utility Use Gabaix (2014) s framework to revisit the classic results in Public Finance Inverse Elasticity Rule: holds but inattention increases taxes Production Efficiency and irrelevance: it does not hold Many others: Mirrleesian and Pigouvian taxes, etc.!3

11 BEHAVIORAL VS NON-BEHAVIORAL PF Non-behavioral PF: Individual behavior of i: xi ( ) 2 arg max ui (x) x2bci ( ) Data: revealed preferences argument gives us ui (x) Choose social welfare function W i u (x) i2i Optimal Taxation Problem: Z i i max W u (x( )) + x ( )di!4

12 BEHAVIORAL VS NON-BEHAVIORAL PF Behavioral PF a la Farhi-Gabaix: Individual behavior of i: xi ( ) 2 arg max uid (x) x2bci ( ) i Data: revealed preferences argument gives us ud (x) Choose experience utility uie (x) Choose social welfare function Optimal Taxation problem: max W i W ue (x) uie (x( )) i2i Z + xi ( )di!5

13 BEHAVIORAL VS NON-BEHAVIORAL PF Behavioral PF a la Farhi-Gabaix: Individual behavior of i: xi ( ) 2 arg max uid (x) x2bci ( ) i Data: revealed preferences argument gives us ud (x) Really irrelevant here. So we can just directly i Choose social welfare function W ue (x) estimate choice i2i Choose experience utility uie (x) Optimal Taxation problem: max W uie (x( )) Z + xi ( )di!5

14 BEHAVIORAL VS NON-BEHAVIORAL PF Behavioral PF a la Farhi-Gabaix: Individual behavior of i: xi ( ) 2 arg max uid (x) x2bci ( ) i Data: revealed preferences argument gives us ud (x) Really irrelevant here. So we can just directly i Choose social welfare function W ue (x) estimate choice i2i Choose experience utility uie (x) Optimal Taxation problem: max W uie (x( )) Z + xi ( )di!5

15 THIS PAPER S APPROACH Use Gabaix (2014) to choose uie (x) and uid (x) Gabaix s sparsity model - an example: Agents perceive after tax prices to be pi (1 + mi i ) Decision utility: Maximize u(x)given perceived after tax prices Experience utility: Maximize u(x)given true after tax prices Interprets mi as perception; assume a cost function that leads to sparse choice, i.e., many zeros!6

16 THIS PAPER S APPROACH Misperception, i.e, mi 6= 1, change elasticities When mi < 1, typically households are less responsive to tax changes Difference wedge i between ue (x) and i ud (x) creates behavioral!7

17 MY DISCUSSION Decision vs. experience utility Focus on misperception as limited ability to process information and show it could increase people s responses to taxes and reduce optimal taxes!8

18 DECISION VS EXPERIENCED UTILITY!9

19 DECISION VS EXPERIENCED UTILITY Main Question: Why do we have to take this stance?!9

20 DECISION VS EXPERIENCED UTILITY Main Question: Why do we have to take this stance? Simon s approach: bounded rationality is!9

21 DECISION VS EXPERIENCED UTILITY Main Question: Why do we have to take this stance? Simon s approach: bounded rationality is rational choice that takes into account the cognitive limitations of the decision-maker - limitations of both knowledge and computational capacity.!9

22 DECISION VS EXPERIENCED UTILITY Main Question: Why do we have to take this stance? Simon s approach: bounded rationality is rational choice that takes into account the cognitive limitations of the decision-maker - limitations of both knowledge and computational capacity. Alternative: use behavioral theories that accommodate welfare:!9

23 DECISION VS EXPERIENCED UTILITY Main Question: Why do we have to take this stance? Simon s approach: bounded rationality is rational choice that takes into account the cognitive limitations of the decision-maker - limitations of both knowledge and computational capacity. Alternative: use behavioral theories that accommodate welfare: Sims, Caplin and Dean, Gabaix, etc.: Rational Inattention!9

24 DECISION VS EXPERIENCED UTILITY Main Question: Why do we have to take this stance? Simon s approach: bounded rationality is rational choice that takes into account the cognitive limitations of the decision-maker - limitations of both knowledge and computational capacity. Alternative: use behavioral theories that accommodate welfare: Sims, Caplin and Dean, Gabaix, etc.: Rational Inattention Gul and Pessendorfer: Preferences over sets of available choices!9

25 BOUNDED RATIONALITY AND WELFARE Without welfare, it is hard to think about where is the objective of the optimal taxation problem coming from In rational models: political institutions can explain where social welfare functions come from Here: people vote with their experience utility but make decision with decision utility!!10

26 RATIONAL INATTENTION AND INFORMATION Alternative model of behavior: Sims, Caplin and Dean, Matejka and McKay, etc. Untaxed Numeraire Suppose we have two goods: c0, c1 Utility function c / c1 1-1/ Price of both goods normalized to 1. Pay tax on good 1!11

27 RATIONAL INATTENTION AND INFORMATION!12

28 RATIONAL INATTENTION AND INFORMATION Households do not know taxes and want to learn it but have to pay attention cost!12

29 RATIONAL INATTENTION AND INFORMATION Households do not know taxes and want to learn it but have to pay attention cost Prior about taxes U[0, ]!12

30 RATIONAL INATTENTION AND INFORMATION Households do not know taxes and want to learn it but have to pay attention cost Prior about taxes U[0, ] Information choice (s )!12

31 RATIONAL INATTENTION AND INFORMATION Households do not know taxes and want to learn it but have to pay attention cost Prior about taxes U[0, ] Information choice (s ) Cost of information K( )!12

32 RATIONAL INATTENTION AND INFORMATION Households do not know taxes and want to learn it but have to pay attention cost Prior about taxes U[0, ] Information choice (s ) Cost of information K( ) It is higher the more precise is the information!12

33 RATIONAL INATTENTION AND INFORMATION Households do not know taxes and want to learn it but have to pay attention cost Prior about taxes U[0, ] Information choice (s ) Cost of information K( ) It is higher the more precise is the information!12

34 RATIONAL INATTENTION AND INFORMATION Households do not know taxes and want to learn it but have to pay attention cost Prior about taxes U[0, ] Information choice (s ) Important: independent of actual tax chosen by the government! Cost of information K( ) It is higher the more precise is the information!12

35 RATIONAL INATTENTION AND INFORMATION Households do not know taxes and want to learn it but have to pay attention cost Prior about taxes U[0, ] Information choice (s ) Important: independent of actual tax chosen by the government! Cost of information K( ) It is higher the more precise is the information!12

36 RATIONAL INATTENTION AND INFORMATION Households do not know taxes and want to learn it but have to pay attention cost Prior about taxes U[0, ] Information choice (s ) Cost of information K( ) Important: independent of actual tax chosen by the government! Sims: Shannon entropy It is higher the more precise is the information!12

37 RATIONAL INATTENTION AND INFORMATION Consumption choice conditional on a signal Z (s ) 1-1/ V(s; ) = max [c0 ( ) + c1 /(1-1/ ) R d (s )d c1,c0 ( ) subject to c0 ( ) + c1 = w Information choice ZZ max V(s; ) (s )dsd - K( )!13

38 INFORMATION STRUCTURES Full information - standard public finance 1 s= (s ) = 0 otherwise Partition information - not fully attentive (s ) = 1 0 s = si, 2 [ i-1, i ] otherwise!14

39 GOVERNMENT Z Z Objective max V(s; ) (s )ds + c1 (s) (s )ds Since actual tax does not affect information cost, only effect of tax on the realized signal and consumption matters!15

40 OPTIMAL TAXES Proposition. Suppose (s ) is partition and is fine enough. Then Sims Ramsey < FG <!16

41 OPTIMAL TAXES Why? With partition: never optimal to choose inside an interval Tax base with partition attention: (1 + - Tax base under Ramsey: (1 + )- Tax base under FG: (1 + mi ) - ) - - Also consumption of good 1 is too high; behavioral wedge!17

42 CONCLUSION Very important first step towards understanding of optimal taxes away from full rationality and attention More work needed to understand the precise effect of the nature of bounded rationality on optimal taxes!18