The Burden of Past Promises

Transcription

1 The Brden of Past Promises Jin Li, Niko Matoschek, & Michael Powell Kellogg School of Management

2 A Good Relationship Takes Time Common view: relationships are bilt on trst, and trst develops over time (Sobel, 1985; Datta, 1996; Ghosh and Ray, 1996; Kranton, 1996; Watson, 1999; Watson, 2002; Eeckot, 2006; Fjiwara-Greve and Ohno-Fjiwara, 2009; Halac, Forthcoming) Once trst is established, can motivate cooperation throgh promises regarding ftre actions Time is therefore the friend of a good relationship

3 The Brden of Past Promises... bt eventally, the ftre becomes the present Yesterday s promises become today s obligations Time can be the foe of a good relationship

4 The Jobs Bank The Jobs Bank is a legacy of the early 1980s, when then-chairman Roger B. Smith was embarking on a strategy to atomate GM s North American Factories. The Wall Street Jornal, Janary 7 th, 2006

5 The Jobs Bank In a recent interview, UAW President Ron Gettelfinger [...] said the Jobs Bank originally was a company proposal, aimed at convincing UAW leaders not to oppose new technology. `The idea was, Yo help s get prodctive and we ll bring work in to occpy the displaced workers, Mr. Gettelfinger said. The Wall Street Jornal, Janary 7 th, 2006

6 The Jobs Bank In a recent interview, UAW President Ron Gettelfinger [...] said the Jobs Bank originally was a company proposal, aimed at convincing UAW leaders not to oppose new technology. `The idea was, Yo help s get prodctive and we ll bring work in to occpy the displaced workers, Mr. Gettelfinger said. Bt that decision came back to hant the company in later years as it began to embrace Toyota s methods of car making [...]. Bt the Jobs Bank never got redefined. Instead, after fighting a series of costly strikes with the UAW in the mid-1990s, GM management conclded it was better to bild a harmonios relationship than provoke fights. The Wall Street Jornal, Janary 7 th, 2006

7 Prpose of the Paper Show that transition of promises into legacy costs is a natral featre of optimally managed relationships Examine implications for evoltion of firms

8 Model Sketch Infinitely repeated delegation game a la Armstrong and Vickers

9 Limited Transfers in Organizations A striking characteristic of work life is that one cannot reward individals in cash for some things, bt can compensate them in other ways. (Prendergast and Stole, 1999) a significant nmber of [payments within firms] are in the form of policy commitments (Cyert and March, 1963)

10 Model Sketch Infinitely repeated delegation game a la Armstrong and Vickers No transfers

11 Model Sketch Infinitely repeated delegation game a la Armstrong and Vickers No transfers Characterize the optimal relational contract: PPE that maximizes principal s expected payoff

12 Main Reslts Principal initially delegates with the nderstanding that Agent will choose Principal s preferred project if possible Agent rewarded and pnished w/changes in contination payoffs Two thresholds: if contination payoff crosses pper threshold, entrenchment. If contination payoff crosses lower threshold, permanent centralization or exit Rewards and pnishments are permanent

13 Main Implication 1: Inertia and Decline One of the most consistent patterns in bsiness is the failre of leading companies to stay at the top of their indstries when technologies or markets change (Bower and Christensen, 1996) Inertia of established firms is the reslt of commitments that allowed these firms to adapt when they were still yong. Firm performance declines over time

14 Main Implication 2: PPDs Organizational policies/procedres tend to be derived from the early history of the organization (Stinchcombe, 1965; Hannan and Freeman, 1977 and to be derived (or at least crystallized ot of) specific noteworthy events in the early history of the organization (Schein, 1983) (Kreps, 1996, p. 577) Same starting point and mltiple steady states: long-rn performance is history-dependent Differences in performance linked to differences in organization

15 Agenda The Model The PPE Payoff Set The Optimal Relational Contract Pblic Information Conclsions

16 Model Sketch One principal and one agent Risk-netral No transfers Repeated trst game Only agent knows what projects are available Common discont factor δ

17 Timing P Players A t = 1,2,3, t t e, e 0,1 chosen 1: P and A simltaneosly decide whether to enter the relationship. e = 1 if enter and e = 0 if not. j = P, A.

18 Timing P Players A t = 1,2,3, t t e, e 0,1 P chooses chosen d 0,1. If d = 0, k = S 2: P decides whether to delegate decision making to A. If P does not delegate, then P chooses safe project, k = S.

19 Timing P Players A t = 1,2,3, t t e, e 0,1 P chooses K A, A, P chosen d 0,1. chosen. K = A, P If d = 0, k = S with prob p <. (A s private info) 3: Natre determines which projects are available. A s project always available. P s project available with prob p <. Only A knows which projects are available.

20 Timing P Players A t = 1,2,3, t t e, e 0,1 P chooses K A, A, P If d = 1, A chosen d 0,1. chosen. K = A, P chooses k K If d = 0, k = S with prob p <. (A s private info) 4: If P has delegated, A chooses an available project.

21 Timing P Players A t = 1,2,3, t t e, e 0,1 P chooses K A, A, P If d = 1, A Payoffs chosen d 0,1. chosen. K = A, P chooses k K Π k, U k If d = 0, k = S with prob p <. (A s private info) 5: Stage payoffs are realized

22 Timing P Players A t = 1,2,3, t t P chooses K A, A, P If d = 1, A Payoffs d 0,1. chosen. K = A, P chooses k K Π k, U k If d = 0, k = S with prob p <. (A s private info) e, e 0,1 chosen x pblicly observed 6: The otcome, x, of a pblic randomization device is commonly observed.

23 Stage-Game Payoffs P Players A t = 1,2,3, t t P chooses K A, A, P If d = 1, A Payoffs d 0,1. chosen. K = A, P chooses k K Π k, U k If d = 0, k = S with prob p <. (A s private info) e, e 0,1 chosen x pblicly observed

24 Stage-Game Payoffs Π U

25 Stage-Game Payoffs Exit: Π E = 0 U E = 0 Π U

26 Stage-Game Payoffs Exit: Π E = 0 U E = 0 Π No Delegation/Safe Project: Π S = a U S = a U

27 Stage-Game Payoffs Exit: Π E = 0 U E = 0 No Delegation/Safe Project: Π S = a U S = a Principal s Preferred Project: Π P = B U P = b Π (b, B) Parameter Restrictions: 1. B > a > b U

28 Stage-Game Payoffs Exit: Π E = 0 U E = 0 No Delegation/Safe Project: Π S = a U S = a Π (b, B) Principal s Preferred Project: Π P = B U P = b Agent s Preferred Project: Π A = b U A = B Parameter Restrictions: 1. B > a > b U (B, b)

29 Stage-Game Payoffs Exit: Π E = 0 U E = 0 Π (b, B) No Delegation/Safe Project: Π S = a U S = a p b, B + (1 p)(b, b) Principal s Preferred Project: Π P = B U P = b Agent s Preferred Project: Π A = b U A = B Parameter Restrictions: 1. B > a > b 2. pb + 1 p b > a U (B, b)

30 Stage-Game Payoffs P Players A t = 1,2,3, t t P chooses K A, A, P If d = 1, A d 0,1. chosen. K = A, P chooses k K If d = 0, k = S with prob p <. (A s private info) e, e 0,1 chosen x pblicly observed B > a > b pb + 1 p b > a

31 Repeated Game P Players A t = 1,2,3, t t P chooses K A, A, P If d = 1, A d 0,1. chosen. K = A, P chooses k K If d = 0, k = S with prob p <. (A s private info) e, e 0,1 chosen x pblicly observed = 1 δ E δ e, e, Π k = 1 δ E δ e, e, U k B > a > b pb + 1 p b > a

32 Soltion Concept (Pre Strategy) Perfect Pblic Eqilibrim Optimal relational contract: PPE that maximizes P s average payoff Goal: characterize the dynamics of the optimal relational contract

33 Agenda The Model The PPE Payoff Set The Optimal Relational Contract Pblic Information Conclsions

34 PPE Payoff Set Abre-Pearce-Stacchetti: characterizes PPE payoff set

35 PPE Payoff Set

36 PPE Payoff Set 1. PPE set ℇ is convex and compact ℇ

37 PPE Payoff Set 1. PPE set ℇ is convex and compact 2. ℇ is convex hll of its frontier: = max, ℇ ℇ

38 PPE Payoff Set 1. PPE set ℇ is convex and compact 2. ℇ is convex hll of its frontier: = max, ℇ ℇ

39 PPE Payoff Set 1. PPE set ℇ is convex and compact 2. ℇ is convex hll of its frontier: = max, ℇ ℇ

40 PPE Payoff Set 1. PPE set ℇ is convex and compact 2. ℇ is convex hll of its frontier: = max, ℇ 3. is self-generating ℇ

41 PPE Payoff Set Abre-Pearce-Stacchetti: characterizes PPE payoff set Can focs exclsively on frontier

42 Actions? Abre-Pearce-Stacchetti: characterizes PPE payoff set Can focs exclsively on frontier Also want to characterize actions taken at each point on the frontier Any eqbm payoff pair on frontier either generated by pre actions or by randomization b/t two eqbm payoff pairs generated by pre actions

43 For Classes of Actions in Eqilibrim Centralization C both enter, P does not delegate Cooperative Delegation D both enter, P delegates, A chooses k = P whenever possible Uncooperative Delegation D both enter, P delegates, A always chooses k = A Exit E neither enter

44 For Classes of Actions in Eqilibrim Centralization C both enter, P does not delegate If spported at,, then: Cooperative Delegation D both enter, P delegates, A chooses k = P whenever possible Uncooperative Delegation D both enter, P delegates, A always chooses k = A Exit E neither enter

45 For Classes of Actions in Eqilibrim Centralization C both enter, P does not delegate If spported at, = 1 δ a + δ, then: Cooperative Delegation D both enter, P delegates, A chooses k = P whenever possible Uncooperative Delegation D both enter, P delegates, A always chooses k = A Exit E neither enter

46 For Classes of Actions in Eqilibrim Centralization C both enter, P does not delegate If spported at, = 1 δ a δ, then: Cooperative Delegation D both enter, P delegates, A chooses k = P whenever possible Uncooperative Delegation D both enter, P delegates, A always chooses k = A Exit E neither enter

47 For Classes of Actions in Eqilibrim Centralization C both enter, P does not delegate If spported at, = 1 δ a δ, then: Cooperative Delegation D both enter, P delegates, A chooses k = P whenever possible Uncooperative Delegation D both enter, P delegates, A always chooses k = A Exit E neither enter

48 For Classes of Actions in Eqilibrim Centralization C both enter, P does not delegate If spported at, = 1 δ a δ, then: Cooperative Delegation D both enter, P delegates, A chooses k = P whenever possible Uncooperative Delegation D There mst be two contination vales, and satisfying: both enter, P delegates, 1 δ b + δ 1 δ B + δ A always chooses k = A and Exit E = p 1 δ b + δ + 1 p 1 δ B + δ neither enter

49 For Classes of Actions in Eqilibrim Centralization C both enter, P does not delegate If spported at, = 1 δ a δ, then: Cooperative Delegation D both enter, P delegates, A chooses k = P whenever possible Uncooperative Delegation D There mst be two contination vales, and satisfying: both enter, P delegates, 1 δ b + δ = 1 δ B + δ A always chooses k = A and Exit E = p 1 δ b + δ + 1 p 1 δ B + δ neither enter

50 For Classes of Actions in Eqilibrim Centralization C both enter, P does not delegate If spported at, = 1 δ a δ, then: Cooperative Delegation D both enter, P delegates, A chooses k = P whenever possible Uncooperative Delegation D There mst be two contination vales, and satisfying: both enter, P delegates, 1 δ b + δ = 1 δ B + δ A always chooses k = A and 1 δ b 1 δ B Exit E = = δ δ neither enter

51 For Classes of Actions in Eqilibrim Centralization C both enter, P does not delegate If spported at, = 1 δ a δ, then: Cooperative Delegation D both enter, P delegates, A chooses k = P whenever possible = = 1 δ b δ 1 δ B δ Uncooperative Delegation D both enter, P delegates, A always chooses k = A Exit E neither enter

52 For Classes of Actions in Eqilibrim Centralization C both enter, P does not delegate If spported at, = 1 δ a δ, then: Cooperative Delegation D both enter, P delegates, A chooses k = P whenever possible = = 1 δ b δ 1 δ B δ Uncooperative Delegation D both enter, P delegates, A always chooses k = A = 1 δ B δ Exit E neither enter = δ

53 Agenda The Model The PPE Payoff Set The Optimal Relational Contract Pblic Information Conclsions

54 The PPE Payoff Frontier (b, B) p b, B + (1 p)(b, b) (B, b)

55 The PPE Payoff Frontier (b, B) p b, B + (1 p)(b, b) (B, b)

56 Actions Spporting the Frontier (b, B) p b, B + (1 p)(b, b) (B, b) E C D D

57 Optimal Relational Contract (B, b)

58 Optimal Relational Contract Period 1: D (B, b)

59 Optimal Relational Contract Period 1: D, P chosen (B, b)

60 Optimal Relational Contract Period 1: D, P chosen Period 2: D (B, b)

61 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen (B, b)

62 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen (B, b)

63 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen Randomization (B, b)

64 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen Randomization (B, b)

65 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen Randomization Period 3: D (B, b)

66 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen Randomization Period 3: D, A chosen (B, b)

67 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen Randomization Period 3: D, A chosen Period 4: D, A chosen Entrenchment (B, b)

68 Optimal Relational Contract Period 1: D (B, b)

69 Optimal Relational Contract Period 1: D, P chosen (B, b)

70 Optimal Relational Contract Period 1: D, P chosen Period 2: D (B, b)

71 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen (B, b)

72 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D (B, b)

73 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen (B, b)

74 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D (B, b)

75 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen (B, b)

76 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen (B, b)

77 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization (B, b)

78 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization (B, b)

79 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization Period 5: C, S chosen (B, b)

80 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization Period 5: C, S chosen Period 6: C, S chosen (B, b) Permanent Centralization

81 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization (B, b)

82 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization (B, b)

83 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization Period 5: D (B, b)

84 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization Period 5: D, A chosen (B, b)

85 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization Period 5: D, A chosen (B, b)

86 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization Period 5: D, A chosen Randomization (B, b)

87 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization Period 5: D, A chosen Randomization (B, b)

88 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization Period 5: D, A chosen Randomization Period 6: E (B, b)

89 Optimal Relational Contract Period 1: D, P chosen Period 2: D, A chosen Period 3: D, A chosen Period 4: D, A chosen Randomization Period 5: D, A chosen Randomization Period 6: E Period 7: E (B, b) Exit

90 Optimal Relational Contract Properties Start off with cooperative delegation If reward (ncooperative delegation), reward forever If pnish (permanent centralization or exit), pnish forever Mltiple steady states. Will eventally reach one of them and stay.

91 Implications Firm performance declines over time Initially flexible firms develop inertia and stop adapting to the private information of the agent Same starting point and mltiple steady states: long-rn performance is history-dependent Differences in performance linked to differences in organization

92 Agenda The Model The PPE Payoff Set The Optimal Relational Contract Pblic Information Conclsions

93 Frontier of Baseline Model (b, B) p b, B + (1 p)(b, b) (B, b)

94 New Project Becomes Available (b, B) In each period, with probability q, new centralized project becomes available permanently p b, B + (1 p)(b, b) (, ) (B, b)

95 New Project Becomes Available (b, B) In each period, with probability q, new centralized project becomes available permanently (, ) p b, B + (1 p)(b, b) Mst characterize two frontiers: post-opportnity and pre-opportnity First characterize ; then characterize, taking into accont (B, b)

96 New Project Becomes Available (b, B) In each period, with probability q, new centralized project becomes available permanently (, ) p b, B + (1 p)(b, b) Mst characterize two frontiers: post-opportnity and pre-opportnity First characterize 2 ; then characterize, taking into accont (B, b)

97 New Project Becomes Available (b, B) In each period, with probability q, new centralized project becomes available permanently (, ) p b, B + (1 p)(b, b) Mst characterize two frontiers: post-opportnity and pre-opportnity First characterize 2 ; then characterize, taking into accont (B, b)

98 New Project Becomes Available (b, B) In each period, with probability q, new centralized project becomes available permanently (, ) p b, B + (1 p)(b, b) Mst characterize two frontiers: post-opportnity and pre-opportnity First characterize ; then characterize 1, taking 2 into accont (B, b)

99 New Project Becomes Available (b, B) (, ) p b, B + (1 p)(b, b) (B, b) In each period, with probability q, new centralized project becomes available permanently Mst characterize two frontiers: post-opportnity and pre-opportnity First characterize ; then characterize 1, taking 2 into accont

100 New Project Becomes Available (b, B) (, ) p b, B + (1 p)(b, b) In each period, with probability q, new centralized project becomes available permanently Mst characterize two frontiers: post-opportnity and pre-opportnity First characterize ; then characterize 1, taking 2 into accont Each point on specifies an action and contination payoffs if opportnity navailable and if available (B, b)

101 Optimal Relational Contract Period 1: D (, ) (B, b)

102 Optimal Relational Contract Period 1: D, P chosen (, ) (B, b)

103 Optimal Relational Contract Period 1: D, P chosen (, ) (B, b)

104 Optimal Relational Contract Period 1: D, P chosen Period 2: D (, ) (B, b)

105 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen (, ) (B, b)

106 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen (, ) (B, b)

107 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen Period 3: D (, ) (B, b)

108 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen Period 3: D, P chosen (, ) (B, b)

109 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen Period 3: D, P chosen Period 4: D (, ) (B, b)

110 Optimal Relational Contract Period 1: D, P chosen Period 2: D, P chosen Period 3: D, P chosen Period 4: D, A chosen (, ) (B, b)

111 Optimal Relational Contract (, ) (B, b) Period 1: D, P chosen Period 2: D, P chosen Period 3: D, P chosen Period 4: D, A chosen Period 5: D, A chosen Period 6: D, A chosen Entrenchment

112 Reslts from Pblic Opportnity Case The breadth of promises is also an important design variable Broad promises may lead to rigidity with respect to pblic information Older firms less likely to adapt to new pblic information

113 Agenda The Model The PPE Payoff Set The Optimal Relational Contract Pblic Information Conclsions

114 Conclsion Corporate cltres do change over time, and modeling the endogenos evoltion of trst and incentives to invest in it wold be a fascinating avene for ftre research. (Bloom, Sadn, and Van Reenen, 2012) View that good relationships take time sggests that trst increases over time, and discretion is positively related to trst. Or model sggests that trst (cooperative delegation) decreases over time, while discretion may move in the opposite direction Interaction between trst and discretion and their evoltion depend on whether there is ncertainty abot employees types or abot their actions

115 EXTRA SLIDES

116 The PPE Payoff Frontier = max,, q,,, st,,, q = 1 & q,,, = if q = 1, 1 δ a = δ = 1 δ a + δ 1 δ b 1 δ B if q = 1, = = δ δ = p 1 δ B + δ + 1 p 1 δ b + δ if q = 1, 1 δ B = δ = 1 δ b + δ if q = 1, = δ = δ

117 New Project Becomes Available (b, B) (, ) p b, B + (1 p)(b, b) (B, b) Region 1: Mix b/t N and C Region 3: CD Actions spporting these PPE frontier payoffs Region 2: Mix b/t C and CD Region 4: Mix b/t UD and CD

118 New Project Becomes Available (b, B) (, ) p b, B + (1 p)(b, b) (B, b) Region 1: Mix b/t N and C Region 3: CD Actions spporting these PPE frontier payoffs Region 2: Mix b/t C and CD Region 4: Mix b/t UD and CD