Outline: Crowding out 1. Field data (Kingma, Hungerman, Andreoni and Petrie) 2. Lab data a) Andreoni b) Bolton and Katok c) Limits of the impure altruism model (Ribar and Wilhelm) d) Testing motives 3. Field experiments on motives (DellaVigna, List and Malmendier) Econ 2230: Public Economics Lecture 7: Motives for giving (cont.) 2: Crowding out in the lab (Andreoni, 93) 2: Crowding out in the lab (Andreoni, 93) “The Public Good Crowding-Out Hypothesis.” Examine crowding out in a public good game with interior NE Two ppyrimary treatments: No tax Tax NE? 1 2: Crowding out in the lab (Andreoni, 93) 2: Crowding out in the lab (Andreoni, 93) Force a 2 unit tax Results: Average contribution including the tax: Equivalent to no tax – except first two columns and four first rows No tax: 2.78 removed Tax: 3.35 Find crowd out of 71.5 percent Looking at last rounds of the five games crowd out 84 percent Incomplete crowd-out. Subjects who are taxed give more than those who are not taxed – giving for others not a perfect substitute for giving by self. Subjects are concerned about their private contribution to the public good Bolton and Katok (1998) Bolton and Katok (1998) “An experimental test of the crowding out hypothesis: The nature of beneficent behavior” Purely altruistic donors: Andreoni’s crowd out analysis based on the assumption that Decrease contribution by 3 if contribution > 3 in Game 2 individuals aim to maximize own payoff. If altruistic and concerned Decrease contribution to zero otherwise for the payoff of others then many contribution levels may be consistent with complete crowd-out Game 1: (πD, πR) = (15, 5) avg. giving $3.48 Game 2: (πD, πR) = (18, 2) avg. giving $2.49 Compare two dictator games Game 1: (πD, πR) = (15, 5) Accountingg$g for those who contributed less than $3 this generates the Game 2: (πD, πR)=(182)) = (18, 2) complete crowd-out prediction that giving should be $1.83 in $15/$5. Thus crowd-out is (3.48-2.49)/(3.48-1.83) = 60% Complete crowd-out predictions? 2 Ribar and Wilhelm (2002) Limits of pure and impure altruism: n → ∞ Revisit impure altruism Pure altruism: G* = q(w‐T+G‐i) Max U(xi, G, gi) s.t. xi + gi = w ‐ T Andreoni, 1988: Limn → ∞ dG*/dG‐i = Limn → ∞ q1 = 0 Max U(w‐T‐G+G‐i,G, G‐G‐i) Complete crowd out G* = q(w‐TGT+G‐i, G‐i) dG*= q (dw‐dT+dG ) + q dG 1 ‐i 2 ‐i Impure altruism: G* = q(w‐T+G‐i, G‐i) Impure altruism model embraced because it results in incomplete crowd out Ribar and Wilhelm, 2002, examine what happens as n → ∞ Pure altruism: q1 > 0 and q2 = 0 Limn → ∞ dG*/dG‐i = Limn → ∞ q1 + q2 = ? Pure warm glow: q1 + q2 = 1 R&W show that when U(., ., .) is concave and the warm glow component of Impure altruism: q1 + q2 < 1 utility is strictly operative at all levels of G, then Limn → ∞ G‐i → ∞ and Limn → ∞ q1 + q = 1 Giving by others not seen as perfect substitute: 2 Interpretation? dG*/dG‐i = q1 + q2 > dG*/dw = q1 Incomplete crowd out when fully funded (dt=dG‐i): dG*/dG‐i |dG‐i=dT = q2 > 0 incomplete crowd‐out Limits of pure and impure altruism: n → ∞ Limits of pure and impure altruism: n → ∞ Pure altruism: G* = q(w‐T+G‐i) Pure altruism results in complete crowd out Andreoni, 1988: Limn → ∞ dG*/dG‐i = Limn → ∞ q1 = 0 Impure altruism results in no crowd out Complete crowd out Thoughts? Impure altruism: G* = q(w‐T+G‐i, G‐i) Impure altruism model embraced because it results in incomplete crowd out Ribar and Wilhelm, 2002, examine what happens as n → ∞ Limn → ∞ dG*/dG‐i = Limn → ∞ q1 + q2 = ? R&W show that when U(., ., .) is concave and the warm glow component of utility is strictly operative at all levels of G, then Limn → ∞ G‐i → ∞ and Limn → ∞ q1 + q2 = 1 Interpretation? Altruism is diminishing as a motive at the margin: UG → 0. As a consequence, the first order condition of the impure altruism model converges to the first order condition of a pure warm glow model (‐ Ux + Ug = 0 ) 3 Ribar and Wilhelm, 2002 Ribar and Wilhelm, 2002 Implication for identifying motives for giving: Implication for identifying motives for giving: Impure altruism: crowd out sensitive to contribution of others Impure altruism: crowd out sensitive to contribution of others Not sufficient to look at one crowd out measure. Degree of crowd out Not sufficient to look at one crowd out measure. Degree of crowd out ddthllthihGdepends on the level at which G-i idThtGis measured. The greater G-i the less ddthllthihGdepends on the level at which G-i idThtGis measured. The greater G-i the less weight put on altruism weight put on altruism Previous tests of motives focused solely on the predictions of pure Previous tests of motives focused solely on the predictions of pure altruism altruism A direct test of impure altruism: increase G-i from a low level to a high A direct test of impure altruism: increase G-i from a low level to a high Low High Low High level and test H0: dG*/dG-i ≤ dG*/dG-i level and test H0: dG*/dG-i ≤ dG*/dG-i Increases in contributions by others shifts marginal motive from concern for securing the public good to concern for warm glow Individual response to changes in giving by others should be large when initial contribution by others is low, and small when initial contribution by others is large Vesterlund, Wilhelm and Xie (2008) Charitable Public Good: Books for Children “Why Do People Give? Testing Pure and Impure Altruism” Created a new public good through collaboration with the American Objective of paper: Red Cross in Southwestern Pennsylvania Identify private benefit of giving in impure altruism model In the event of a fire in SW PA they help the affected families find Examine multiple crowd-out measures temporary shelter, provide them with clothing, a meal, and give them Test comparative static predictions of both pure and impure essential toiletries altruism model Prior to study no items given to the children affected by the fire Pair each participant with a child (1‐12 years old) whose family home has suffered extensive fire damage Environment that mirrors that of the theory: Each participant informed of an initial amount to be given to the child, A real charitable public good (no VCM) and given an endowment which may be allocated between the child Provision determined by sum of an initial contribution and and him/herself partici pant cont rib uti on ( no ex is ting c har ity ) Money given towards the child will be spent on books, which will be Multiple individual public goods of varying sizes distributed by the American Red Cross, immediately after the child has been affected by a severe fire (Few boundary decisions) “Each participant in this study is paired with a different child. Only you have the opportunity to contribute books to the child, neither the American Red Cross nor any other donors provide these books to the child.” 4 Budgets Test 1: Income Effect 1. Child $4 Self $40 1. Child $4 Self $40 2. Child $10 Self $40 2. Child $10 Self $40 3. Child $28 Self $40 3. Child $28 Self $40 4. Child $34 Self $40 4. Child $34 Self $40 5. Child $4 Self $46 5. Child $4 Self $46 6. Child $28 Self $46 6. Child $28 Self $46 Change in giving in response to $6 increase in income What aspects of our models can be tested with these budgets? Normal public good: Increase in giving Normal private good: Increase smaller than $6 Test 2: Balanced Budget Increases in G‐i Test 3: Increasing G‐i 1. Child $4 Self $40 1. Child $4 Self $40 2. Child $10 Self $40 2. Child $10 Self $40 3. Child $28 Self $40 3. Child $28 Self $40 4. Child $34 Self $40 4. Child $34 Self $40 5. Child $4 Self $46 5. Child $4 Self $46 6. Child $28 Self $46 6. Child $28 Self $46 Balanced budget crowd out: Response to a $6 tax contributed to the charity (budget 2 vs. 5 and 4 vs. 6) Pure Altruism U(x,G):$6 decrease Impure Altruism U(xGgx,G,g): Less than $6 decrease 5 Test 3: Increasing G‐i Test 3: Increasing G‐i 1. Child $4 Self $40 1. Child $4 Self $40 2. Child $10 Self $40 2. Child $10 Self $40 3. Child $28 Self $40 3. Child $28 Self $40 4. Child $34 Self $40 4. Child $34 Self $40 5. Child $4 Self $46 5. Child $4 Self $46 6. Child $28 Self $46 6. Child $28 Self $46 Pure Warm‐Glow: U(x,g) gi does not respond to changes in initial giving Test 3: Increasing G‐i Test 3: Increasing G‐i 1. Child $4 Self $40 1. Child $4 Self $40 2. Child $10 Self $40 2. Child $10 Self $40 3. Child $28 Self $40 3. Child $28 Self $40 4. Child $34 Self $40 4. Child $34 Self $40 5. Child $4 Self $46 5. Child $4 Self $46 6. Child $28 Self $46 6. Child $28 Self $46 Pure Warm‐Glow: U(x,g) Pure Warm‐Glow: U(x,g) gi does not respond to changes in initial giving gi does not respond to changes in initial giving Pure Altruism: U(x,G) Pure Altruism: U(x,G) 6 Test 3: Increasing G‐i Increasing G‐i: Pure Altruist q1 increasing (luxury): gi diminished sensitivity to G‐i 1.