0 views

Uploaded by Javier

Common-value all-pay auctions with asymmetric information

- Mixed Strategy
- A Theory of Social Decision Making
- Problems Repeated and Bayesian game.pdf
- Game Th. TalkNotes
- Lecture Note 4 Extensive-Form Games
- 05715450
- Equivalence Relations_a.ppt
- Behavioral Game Theory.pdf
- Martin Galas
- principio_del_consensualismo.pdf
- ManualUsuario K XCT300i
- Master 5
- Nociones de lógica.pdf
- Mundo interconectado
- Ch_03
- K-XCT_300i_201303271114_MR.SK60BA-V1_
- Stoch Proc Sim
- Modelo_Hedonico_Rosen.pdf
- jeas_0118_6679.pdf
- Modelo Hedonico Rosen

You are on page 1of 24

DOI 10.1007/s00182-015-0524-4

ORIGINAL PAPER

**Common-value all-pay auctions with asymmetric
**

information

**Ezra Einy1 · Mridu Prabal Goswami2 ·
**

Ori Haimanko1 · Ram Orzach3 · Aner Sela1

**Accepted: 20 December 2015 / Published online: 7 January 2016
**

© Springer-Verlag Berlin Heidelberg 2016

**Abstract We study two-player common-value all-pay auctions in which the players
**

have ex-ante asymmetric information represented by finite connected partitions of the

set of states of nature. Our focus is on a family of such auctions in which no player

has an information advantage over his opponent. We find sufficient conditions for the

existence of equilibrium with monotone strategies, and show that such an equilibrium

is unique. We further show that the ex-ante distribution of equilibrium effort is the

same for every player (and hence the players’ expected efforts are equal), although

their expected payoffs are different and they do not have the same ex-ante probability

of winning.

Keywords Common-value all-pay auctions · Asymmetric information

JEL Classification C72 · D44 · D82

**Einy, Haimanko and Sela gratefully acknowledge the support of the Israel Science Foundation Grant
**

648/13.

B Ori Haimanko

orih@bgu.ac.il

Ezra Einy

einy@bgu.ac.il

Ram Orzach

orzach@oakland.edu

Aner Sela

anersela@bgu.ac.il

**1 Department of Economics, Ben-Gurion University of the Negev, Beersheba 84105, Israel
**

2 Economic Research Unit, Indian Statistical Institute, 203 B.T. Road, Kolkata 700108, India

3 Department of Economics, Oakland University, Rochester, MI 48309, USA

123

80 E. Einy et al.

1 Introduction

**All-pay auctions are used in diverse areas of economics, such as lobbying in organi-
**

zations, R&D races, political contests, promotions in labor markets, trade wars, and

biological wars of attrition. In an all-pay auction each player submits a “bid” (i.e.,

exerts effort) and the player with the highest bid wins the contest. However, regardless

of who the winner is, each player bears the cost of his bid. All-pay auctions have been

studied when players have either private or common values of winning.1 In this paper

we focus on common-value all-pay auctions and consider contests with two ex-ante

asymmetrically informed players where, in each state of nature, the value of winning

is identical for both players. We assume that each player’s information is represented

by a finite partition of the set of states of nature, but these partitions are different for

the two players.2 When a state of nature is chosen, each player learns which element

of partition contains the realized common-value, but the players do not necessarily

know the exact value of winning the contest.3

In our model of asymmetric information we assume that the information partitions

of both player are connected with respect to the value of winning the contest (see Einy

et al. 2002 and Forges and Orzach 2011). This means that if a player’s information

partition does not enable him to distinguish between two possible states of nature

with v1 ≤ v2 as the corresponding values of winning, then he cannot distinguish

between any states of nature where the values are in the range [v1 , v2 ]. Connectedness

is plausible in environments where the information of a player allows him to put upper

and lower bounds on the actual value of winning, without ruling out any outcome

within these bounds.

In studying common-value all-pay auctions with asymmetric information, one par-

ticular case of interest is when one player has an information advantage (henceforth,

IA) over another, which means that his information partition is finer than that of his

opponent. This assumption simplifies the analysis when there are only two players,

since it can usually be reduced to the postulate that one player is completely informed

about the state of nature, while the other player has no information about it. Significant

results have been obtained for the IA case. For example, Siegel (2014) has shown4 the

existence of a unique equilibrium in mixed strategies in a two-player common value

auction with IA, in which the ex-ante distribution of effort is the same for both players,

as is the ex-ante probability of winning the contest.

In this paper we study two-player common-value all-pay auctions in which, except

in the extreme states of nature (corresponding to the lowest and the highest possible

1 To mention just a few works, all-pay auctions have been considered by, e.g., Hillman and Riley (1989),

Baye et al. (1993, 1996), Amman and Leininger (1996), Krishna and Morgan (1997), Che and Gale (1998),

Moldovanu and Sela (2001, 2006), Siegel (2009), and Moldovanu et al. (2012).

2 This partition representation is equivalent to the more common Harsanyi-type formulation of Bayesian

games.

3 This framework has been used in several works to analyze common-value second-price auctions [see

Einy et al. (2002), Forges and Orzach (2011), and Abraham et al. (2012)], and common-value first-price

auctions (see Malueg and Orzach 2009, 2012).

4 See Section 2 of the online appendix to his work.

123

the ex ante distributions of both players’ efforts are the same (like in the IA case). Siegel (2014) proves existence and uniqueness of a monotone equilibrium under a strong assumption (M) of joint monotonicity of conditional densities and valuations. Our other results pertain to the properties of the equilibrium with monotone strate- gies. In a study of asymmetric two-player all- pay auctions with interdependent discrete valuations. His assumption M presupposes that the common prior distribution of type profiles has full support. more favorable signals never lead to lower bids). We show that although the players have asymmetric strategies that yield different expected payoffs as well as different chances to win the contest (unlike in the IA case). The emergence of monotone mixed strategy equilibria in non-IA common-value all-pay auctions is not unique to our set-up. In Sect.5 We first construct a “candidate” for an equilibrium.6 We also show that a player’s expected payoff monotonically increases with the values of winning conditional on an element of his partition. and hence in par- ticular that a player cannot rule out any specific type of his rival given his information. In Rentschler and Turocy’s (2015) examples of non-monotonicity of equilibrium strategies. and sometimes there are only non-monotone equilibria.e. It turns out that there can exist at most one equilibrium with monotone strategies. which is natural as by assumption no player has an inherent advantage over his rival in our model. some types of the other players are neces- sarily ruled out. in the IA case Siegel (2014) has shown that the uninformed player has a lower equilibrium payoff—he gets zero in expectation. In contrast. in which both players’ strategies are monotone (i. Failure of M has been known to be related to non-monotonicity of equilibria—Rentschler and Turocy (2015) show that equilibria with non-monotone strategies may exist. Thus. while the expected payoff of his informed rival is positive. In Sect. 7 Among other conditions. since our players’ information endowments are such that given one’s type (information set).Common-value all-pay auctions with asymmetric information 81 values of winning). our assumption of connected information partitions plays a role in the emergence of equilibrium with monotone strategies. 6 The expected payoff of a player may be higher or lower than that of his opponent. 3 we give a numerical example that demonstrates how to find an equilibrium in our model. the information endowments are non-connected. and hence. neither player has an IA over his opponent. In our model M is inapplicable. 5 we explicitly describe a “candidate” for equilibrium with monotone mixed strategies. if such an equilibrium exists. and shows how these assumptions enable a reduction to information structures with singleton and doubleton information sets. in the common-value set-up with symmetric and highly correlated types. higher/lower val- ues in the information set of a player lead to a higher/lower conditional expected payoff. 2 we present the general model. it must be our “candidate”. the asymmetry of information between the players manifests itself not in unequal expected efforts. present a 5 Malueg and Orzach (2012) studied the first-price auction with information partitions of this type. Section 4 presents our assumptions on the information structure in the auction. but in the allocation of payoffs between the play- ers. 123 . In Sect. Thus. and then find simple sufficient conditions for the “candidate” to be a true equilibrium.. Our work identi- fies an important instance where monotone equilibria exist despite that M does not hold.7 The paper is organized as follows.

Slightly abusing notation. i. 6 we analyze the players’ expected efforts and payoffs in the unique equilibrium with monotone strategies. In Sect. and thus are ex-ante asymmetric. ·) and y wherever appropriate. i. Fi (·. x2 }.. If player i plays a pure strategy given π . x) is constant on every element of i ). 2 ). ·) is a cumulative distribution function (c. such that for every ω ∈ . x) = −xi . Proofs of equilibrium existence and uniqueness are in the Appendix. if xi < max{x1 . but all the players bear the cost of their effort. A function v : → R++ represents the common value of winning the contest. i.f. Section 7 concludes. we will identify between Fi (π.l. whenever convenient. Given a mixed strategy profile F = (F1 . Thus our attention will be given to mixed strategy equilibria. ∞ ∞ E i (F) ≡ E u i (·. A mixed strategy of player i is a function Fi : × R+ → [0. denote by E i (F) the expected payoff of player i when players use that strategy profile. and then he chooses an effort xi ∈ R+ . In all-pay auctions. 1 . where m(x) denotes the number of players who exert the highest effort.e. The utility (payoff) of player i ∈ N is given by the function u i : × R2+ → R defined as follows: m(x) v(ω) − x i . Einy et al.e. if ω ∈ is realized then the value of winning is v(ω) for every player. x2 }. x1 )d F2 (·.) on R+ . 0 0 123 . x). if the distribution represented by Fi (π. ·) is supported on some y ∈ R+ . Each player i is informed of the element πi (ω) of i which contains ω (thus. and for all x ∈ R+ .. xi = max{x1 . (x1 . for any π ∈ i we will denote the constant value of Fi (·. 1 if u i (ω. v. F2 ). 2} of two players who compete in an all-pay auction where the player with the highest effort wins the contest. x2 )) d F1 (·. The players will have different information partitions.o. x2 )}|.. A common-value all-pay auction starts when nature chooses a state ω from according to the distribution p. x2 ) . m(x) = |{i ∈ N : xi = max(x1 . namely. Fi (·. A two-player common-value all-pay auction with differential information is fully described by and identified with the collection G = ((. p(ω) > 0 for every ω ∈ ). x) is a i - measurable function (that is. 1]. The uncertainty in our model is described by a finite set of states of nature and a probability distribution p over which is the common prior belief about the realized state of nature (w. p). Fi (ω.e. 2 The general model Consider the set N = {1. sufficient condition for it being an equilibrium and claim that an equilibrium with monotone strategies is unique.82 E.d. πi (ω) constitutes the information set of player i at ω).g. The private information of each player i ∈ N is described by a partition i of . there is usually no equilibrium in pure strategies. x) on π by Fi (π.

⎩ 1. Consider a two-player common-value all-pay auction with three states of nature. if x > 21 and ⎧ ⎪ ⎪ 0. state ωi occurs with probability pi = 13 and the value for the prize in ωi is v(ωi ) = i. x) = x. For i = 1. F) will denote the conditional expected payoff of player i given his information set π .f. It can be easily verified that the corresponding common-value all-pay auction does not have an equilibrium in pure strategies. given by ⎧ ⎨ 0. if x < 0.f. the following inequality holds: E i (F ) ≥ E i Fi . and every mixed strategy Fi of that player. if 0 ≤ x ≤ 21 . ⎩ 1. there does exist a mixed strategy equilibrium. F2 ) of mixed strategies constitutes a (Bayesian Nash) equilib- rium in the common-value all-pay auction G if for every player i. 3. if 21 ≤ x ≤ 1. ∗ F2 ({ω3 }. if 1 ≤ x ≤ 25 . 0 0 A profile F = (F1 . ω3 }}. if x > 1 and ⎧ ⎨ 0. F1 ({ω2 . ω3 }. 3 Example We begin with a simple example of the behavior of players in our model. ω2 }.. given by ⎧ ⎨ 0. ∞ ∞ E i (π. ⎩ 1. if x < 1. 1 if x < 21 . x2 ) | π . i. if 1 < x ≤ 25 . if x < 0. 123 . if 0 ≤ x ≤ 1. However. player 1’s mixed strategy F1∗ is a state-dependent c.e. if x > 25 . Player 1’s information partition is 1 = {{ω1 }. 2. (x1 . In this equilibrium. E i (π.d. while player 2’s information partition is 2 = {{ω1 . F) ≡ E u i (·. x2 ))d F1 (·. Player 2’s mixed strategy F2∗ is a state-dependent c. {ω2 .d. F1∗ ({ω1 }. ω2 }. x1 )d F2 (·. if x > 25 . x) = x 21 ⎪ ⎪ 3 + 6. ⎨ ∗ x− .Common-value all-pay auctions with asymmetric information 83 For π ∈ i . ⎩ 1. x) = 23 x − 23 . F2∗ ({ω1 . where −i denotes i’s rival. x) = 2x. {ω3 }}. F−i .

2 2 3 3 and when 1 exerts effort x ∈ [ 21 . x. when payer 1 uses mixed strategy F1∗ . 21 ]. 25 ] given {ω2 . ω2 } and on [1. Similarly. 25 ] is a best response of player 1 to F2∗ conditional on {ω2 . 21 ] is a best response of player 1 to F2∗ conditional on {ω1 }. Thus the mixed strategy F1∗ (supported on [0. F2 is a mixed strategy equilibrium. 2 2 2 and when 1 exerts effort x ∈ [0. F1∗ . x = + · 2 · x − − x = 0. It is easy to see that efforts above 21 would result in a non-positive conditional expected payoff to player 1. 25 ] given {ω3 }) is an (unconditional) ∗ ∗ ∗ ∗ best response of player 2 to F1 . 1]. player 2’s expected payoff condi- tional on the event {ω1 . F1∗ . 1 E 1 {ω2 . x = · 1 · 2x − x = 0. F2∗ = · 2 + · 3 · x− − x = 0. Furthermore. player 1’s expected payoff conditional on the event {ω1 } is E 1 {ω1 }. In order to see that the above strategies are an equilibrium. 3 6 2 Any effort above 25 or below 1 would lead to a conditional expected payoff smaller than 21 . The ex-ante distrib- utions of each player’s equilibrium effort are identical—indeed. ω3 }) is an (unconditional) best response of player 1 to F2∗ . x. the pair F = F1 . 1 1 2 2 E 1 {ω2 . F1∗ . Einy et al. ω3 } is as follows: when 1 exerts effort x ∈ (1. conditional on the event {ω3 }. 1] given {ω1 . 1 E 2 {ω1 . Thus the mixed strategy F2∗ (supported on [0. Thus. F2∗ = · 2 · x − x = 0. ω2 } payoff is as follows: when 2 exerts effort x ∈ ( 21 . 2 It is easy to see that any effort above 1 would result in a negative expected payoff. and thus any effort in [ 21 . any effort in [0. x). ω3 }. ω3 }. 2 It is easy to see that exerting efforts below 21 or above 25 would lead to non-positive conditional expected payoffs. 21 ]. 25 ]. F2∗ = 1 · x − x = 0 for any effort x ∈ [0. ω3 }. x) = 3 · + −x = . 21 ] given {ω1 } and on [ 21 . 1]. x. the ex-ante probability 123 . ω2 } . 25 ] is x 1 1 E 2 ({ω3 } . player 1’s expected payoff conditional on the event {ω2 . Hence. 1 1 1 E 2 {ω1 . the expected payoff of player 2 when he exerts effort x ∈ [1. ω2 } .84 E. Also. note that when player 2 uses mixed strategy F2∗ .

l. the states’ indices reflect the (weak) ranking of the values. ω2 . .g.Common-value all-pay auctions with asymmetric information 85 to exert an effort smaller than or equal to x is the same for both players and is given by ⎧ ⎪ ⎪ 0. if x > 25 . In the next two sections we will lay out conditions on the information structure that will be sufficient for exis- tence and uniqueness of a mixed-strategy equilibrium (with monotone strategies) in a general two-player common-value all-pay auction without information advantage.o. when the value is not known with precision. if x < 0. F2∗ ) = 0. . (1) Thus. for n > 1 . 1 1 1 E 2 (F1∗ .4722. ⎩ 9 9 1. 24 and 24 respectively. ⎨ 2 x. F2∗ ) = · = . if 0 ≤ x ≤ 1. where w. . ωn }. ∗ F (x) = 2x 3 ⎪ ⎪ + 4 . 4 Connected and overlapping information partitions Let us write = {ω1 . if 1 < x ≤ 25 . the ex-ante probabilities of players 1 and 2 to win are distinct. 123 . player 2 has a similar monopoly on knowledge at ω3. each player knows that the value is above (if he is player 1) or below (if he is player 2) the value in the full knowledge state. . In particular. The expected payoff of the players in F ∗ are E 1 (F1∗ . 0 < v (ω1 ) ≤ v (ω2 ) ≤ · · · ≤ v (ωn ) . the expected equilibrium efforts of both players are identical and equal to 2. 3 2 6 The ex-ante probability of player 2 to win is given by 1 1 2 x 1 P2 = 2ds 1d x + 1d x 3 0 0 1 2 1 x 1 + 1ds 1d x 3 1 2 1 2 5 1 2 x 1 2 1 + ds dx + 1ds 3 1 1 3 3 1 2 13 = 24 11 13 Thus. Note that in this example no player has a general information advantage over another: while player 1 is the only player with the full knowledge of the value at ω1 . Also.

ω j } and π2 (ω1 ) = {ω1 . Malueg and Orzach (2012) showed that the environment where the players partitions are connected and overlapping can be viewed in the following simpler but equivalent form: Proposition 1 Every common-value all-pay auction with information partitions i . i = 1.. for any π ∈ 1 and π ∈ 2 the following holds: / π ⇒ π π .8 However. i. (2002).. 2. after observing their own signals neither player has an information advantage over his opponent. ω1 . .e. the same conclusion applies to π1 (ωn ) and π2 (ωn ). if ωk . that |1 |. Assumption 1 Each partition i . the possibility that the two players have identical information sets in some states of nature. . which are connected. ωn belong to the same information set for one of the players (in which case i is a singleton for that player i). and that an equality π1 (ω1 ) = π2 (ω1 ) would easily contradict Assumption 2. . The notion of connectedness. i.e. |2 | > 1. . and non-trivial. correspondingly). then also ω j ∈ π .86 E. is also ruled out by Assumption 2. ωl ∈ π for k ≤ l and j is such that v (ωk ) ≤ v ω j ≤ v (ωl ). overlapping. as their information sets are incomparable. Hence either π1 (ω1 ) π2 (ω1 ) (2) or π2 (ω1 ) π1 (ω1 ) (3) must hold. Indeed. since we combine Assumptions 1 and 2. as can be easily observed. 123 . ω1 . except in the extreme states of nature ω1 . . Assumption 2 The (distinct) partitions 1 and 2 are overlapping.e. 2. the existence of information advantage is not ruled out globally—at each of the extreme states of nature one player is necessarily better informed than the other. We shall therefore additionally assume that partitions 1 and 2 are non-trivial. It means that. . . in that it contains only consecutive elements of (there are no “holes”).e. is connected with respect to the common value function v. i = 1. Assumption 2 would lose its bite if the extreme states ω1 . Einy et al. k ≥ 1 by Assumption 1. i. ωk } for some j. ωn (containing the lowest and the highest values of winning. for every information set π ∈ i of player i. implies that each information set of a player is an “interval”. i. is strategically equiv- alent (up to a permutation of the players’ names) to a common-value all-pay auction with information partitions given either by 8 Ties in information.. note that π1 (ω1 ) = {ω1 . that was introduced in Einy et al. ωn ∈ The notion of overlapping partitions was introduced by Malueg and Orzach (2012).. The following two assumptions will be made on the information partitions in the auction. ωn ∈ / π ⇒ π π. .

9 that π1 (ω1 ) π2 (ω1 ) (otherwise swap the players’ names). 5}. {ωn }}. {6. Malueg and Orzach (2012) provided other 9 To be precise. 6}. denote = {1. Define a probability distribution p on by p ({ω }) = p(ω ) for every ω ∈ . . . As there is an natural payoff-preserving correspondence between mixed strategy profiles in the original auction and the induced auction (with the rule that player i chooses the same effort distribution on the induced information set π ∈ i as on the original π ∈ i ).g. {4. . 123 . {ωn−2 . ωn }}. a modification of the example given by Malueg and Orzach (2012). . {ω2 . 3} and = {4. for instance. since 1 and 2 are non-trivial. satisfies Assumption 2 and therefore this environment is strategically equivalent (conditional on a realization of either or ) to a common-value all-pay auctions with informa- tion partitions given by either (4) or (5). Then index the elements of in accordance with their rank. there is a natural enumeration of the elements of = {ω1 . ω5 }. {ω3 . 2}. (4) for an odd n ≥ 3. ω2 . However.Common-value all-pay auctions with asymmetric information 87 1 = {{ω1 }. ω3 }. . It can be shown that some environments not satisfying Assumptions 1 and 2 can be nonetheless reduced to strategically equivalent environments that do sat- isfy them. . . {ω4 . . Consider. . 3}. ωn }. . . {ω2 . The recipe to obtain the equivalent information structure is straightforward. . Let = 1 ∨ 2 be the coarsest partition of that refines both 1 and 2 . 2 = {{ω1 . {ωn−1 . the strategic equivalence follows. . . ω2 }. 1 = {{1}. 7. 7}. . Since both 1 and 2 are connected. . will be viewed as the new set of states of nature. ω5 }. Assumption imply that 1 and 2 and Definition (6) further 2 are described either by (4) (if is odd) or by (5) (if is even). (6) For each ω ∈ set the value of winning v ω as the conditional expectation E[v | ω ]. It obviously does not satisfy Assumption 2. {7}}. 6. (5) for an even n ≥ 4. {ωn }}. 2 = {{1. . {3}. Note that each of and . 5. and let vi = i for every i = 1. . together with the partitions induced by 1 and 2 . π1 ω1 = {ω1 }. {5. . 2. ω2 }. {ω4 . (2) and (3) imply that ≥ 3. {2. . or 1 = {{ω1 }. {4}. An outline of the proof. ω4 }. {ωn−1 .l. ω4 }. 7}. . . regardless of the realized state it is commonly known whether the state is in or . induced by the numbering of the elements of . Let = {1. . . clearly. 7}}. consider the partition i of that consists of the sets π = ω ∈ | ω ⊂ π for every π ∈ i . ωn }}. for every i ∈ N . Assume w. consider the following well-defined ranking of the elements of : let ω < ω if and only if there are indices k < l such that ωk ∈ ω and ωl ∈ ω . {ω3 . ω3 }. and. Clearly.o. . . . ωn−1 }. Then. . 2. 2 = {{ω1 .

{ωn }}. if k = 0.88 E. for every k = 0. {ωn−1 . and. {ω2 . We shall henceforth adopt (4). . ωn } for an odd n ≥ 3. that = {ω1 . as in the other case the formulae and the analysis are entirely analogous. let pk pk+1 p k. .1 Preliminaries As has been claimed in Proposition 1. and that 1 = {{ω1 }. (8) To write the singleton or doubleton information sets of the players more compactly. .n = p 0. correspondingly. . .1 ≡ 0. Furthermore. . . Accordingly. accordingly. . 2 = {{ω1 . following (1). n let i(k) designate the (unique) player for whom π k ∈ i(k) . of ωk+1 . Einy et al. n − 1. . . . ω3 }. {ωn−2 . set p n.g. ωn }}. if k = n. n. . for every k = 1. and 2 = {π k | 1 ≤ k ≤ n is odd}. . . ω2 }. . . . 0 < v1 ≤ v2 ≤ · · · ≤ vn . and thus. . p n+1. ωn−1 }. n − 1. . . ⎩ {ωn }. ω2 . . ω4 }. Also denote vk = v(ωk ) and pk = p(ωk ) > 0 (7) for every k = 1. Hence i(0) = 1. Assume. . . i(n) = 2.k+1 ≡ and p k+1. define ⎧ ⎨ {ωk . {ω3 . i(1) = 2. . given the occurrence of the event π k . examples depicting environments that can be transformed into the setting analyzed in this paper. that the information partitions 1 and 2 are given by either (4) or (5). 1 = {π k | 0 ≤ k ≤ n − 1 is even}. . . . . Also. .l. .o. .n+1 = p 1. 123 . 1.k ≡ pk + pk+1 pk + pk+1 be the conditional probabilities of the state ωk . ω5 }. Additionally. π k ≡ {ω1 }.0 ≡ 1. ωk+1 }. it can be assumed w. 2. . if k = 1. i(2) = 1. 5 Equilibrium analysis 5. {ω4 .

xn−1 ) = 0 and F2∗ (π n . xk = p k.2 v1 .k+2 (10) ⎪ ⎪ x+ p . of a continuous probability distribution supported on [x0 . strictly increasing on xk−1 . if xn−1 ≤ x ≤ xn . continuous. x) x−xn−1 = p n. xk+1 = 1 and Fi(k) π k . · is well defined.k−1 v Fi(k) πk. x1 ]. (11) ⎩ 1. con- ∗ ∗ ∗ tinuous.d. j+1 v j . Finally. Note that the function F1∗ π 0 . given π k ∈ i(k) . Let x0 = 0. Thus. player 1’s mixed strategy is ⎧ ⎨ 0. ⎪ ⎩ p k+1. if x > x1 . j=1 Given π 0 ∈ 1 .n−1 vn . strictly increasing on [x0 . 123 .d. . Fi(k) πk. ·) is a c. if xk < x ≤ xk+1 . if 0 ≤ x ≤ x1 . xn . n set k xk ≡ p j. of a continuous probability distribution supported on xk−1 . Fi(k) π . x = p1. The function F2∗ (π n . strictly increasing on xn−1 . F1∗ π 0 .k+1 . ·) is well defined. ∗ p k. x (9) ⎩ 1. F2∗ ) that will be an equilibrium of the all-pay auction under some conditions. of a probability distribution supported on xn−1 . if x < 0. · is well defined. xn ) = 1. if x > xn . 2 . player 2’s mixed strategy is ⎧ ⎨ 0. ∗ k The function Fi(k) π .Common-value all-pay auctions with asymmetric information 89 5.f. F1∗ π 0 . · is a c. j−1 p j. x1 ]. 5. F2∗ (π n . x0 = 0 and F1∗ π 0 . F2∗ ) to be an equilibrium in the common-value all pay auction a simple condition on the values of winning suffices—each value vk+1 must be sufficiently above its immediate predecessor vk . if x < xn−1 . F2∗ (π n . Thus. .k+2 vk+1 1. x = 0. the mixed strategy of player i(k) is ⎧ ⎪ 0. x1 = 1. For k = 1. con- tinuous. xn . · is a c. if x > xk+1 . n − 1. Fi(k) π k .f. F1∗ π 0 . ⎨ if xk−1 ≤ x ≤ xk .d.f. ∗ k k−1 Thus. given π n ∈ 2 . . x = k vk+1 −xk+1 k+1.2 Description of equilibrium strategies We now define a mixed strategy profile (F1∗ . . if x < xk−1 . ⎪ ⎪ x−xk−1 . . xk+1 . F2∗ (π n . . .3 Equilibrium existence and uniqueness In order for the pair (F1∗ . 2. xk+1 . and for every k = 1.

the inequality p n. · is supported on [x k−1 . 2. 12 We additionally use the fact that. It follows from (16). . x k+1 ] if 1 ≤ k ≤ n − 1. F1∗ π 0 . j+1 p j+1. π l ∈ i for k < l. (10). and (11) in the previous section]. when v1 = v2 = · · · = vn (or. Proof See the Appendix. Note that Proposition 3 imposes no condition on the auction other than those spec- ified in Sect. 123 . . . . (17). by definition. y ∈ Al . then F = F ∗ . . 11 This can be seen using (13) in the next section. under (F1∗ . Mixed strategy Fi of player i is monotone if his efforts are (weakly) increasing with the index of his information set (that indicates a higher value of winning). . ·) is supported on Ak .k+2 vk+1 − vk ≥ 0 for every k = 1. and hence. The proof of Proposition 2 (given in the Appendix) provides an explicit formula for expected equilibrium payoffs of the players conditional on each event π k .11 which cannot occur in equilibrium. .1. Einy et al. i. · is supported on [xn−1 . . n − 2. whenever π k . (13) j=1 10 Note that (12) can be assumed to hold only for i = 1. F ∗ = p j. (10). xn ]. F ∗ k i(k) π . and both F1 and F2 are monotone. and F2∗ π n . F2∗ ) the expected payoffs to both players are negative. Note that both F1∗ and F2∗ are monotone. F2∗ ) need not be an equilibrium if (12) does not hold. and (18) there that12 k−1 E i(k) π k .e. j+2 v j+1 − v j . n − 1. The pair (F1∗ . F2∗ ) is not an equilibrium of the auction. F2 ) is an equilibrium of the common-value all-pay auction G. · is supported on [x0 . 6 Other results In this section we derive some comparative results about the players’ expected payoffs and efforts in the equilibrium F ∗ [described in (9). 2. F2∗ ) described in (9). . (12) Then the strategy profile F ∗ = (F1∗ . As for i = n − 1. x ≤ y for every x ∈ Ak . It is easy to see that the pair (F1∗ . It turns out that (F1∗ . . whenever (F1∗ . . and (11) is a mixed strategy equilibrium of the common-value all-pay auction G. F2∗ ) is of particular interest for the following reason. Fi is monotone if for every π k ∈ i there is a set Ak ⊂ R+ such that the distribution corresponding to Fi (π k . Proposition 2 Suppose that p k+1. more generally.10 Proof See the Appendix.n+1 vn − vn−1 = vn − vn−1 ≥ 0 holds trivially by our assumption that vi ≥ vi−1 . and.90 E. x1 ]. the set Ak is to the left of the set Al . 1 < i ≤ n. n. no equilibrium with monotone strategies exists in our framework. F2∗ ) is the only possible equilibrium with monotone strategies: Proposition 3 If F = (F1 . Put precisely. 5. for k = 0. .. For instance. when the values are very close to each other).

k−1 vk j=1 13 As in the proof of Proposition 2. 3. F ∗ ) ≤ E 1 (π 2 . F ∗ ) ≤ E 2 (π 3 .e. . x j=1 k−1 x − xk−1 = p j + ( pk + pk+1 ) p k. n − 1.. n.Common-value all-pay auctions with asymmetric information 91 when the sum in (13) is defined as 0 if k ≤ 1. . F ∗ − E i(k−1) π k−1 . This yields the following immediate result that compares the players’ conditional expected payoffs. conditional on π k ∈ i . 4.13 E i(k) π k . Then the ex-ante probability that player i(k) exerts an effort smaller than or equal to x is ∗ k−1 ∗ F i(k) (x) = p j + ( pk + pk+1 )Fi(k) πk. . . . in particular. . F ∗ ≥ E 2 π2 (ωk ). F ∗ ) ≤ · · · ≤ E 1 (π n−1 . that (2) the expected payoff of each player i. both players have the same ex-ante distribution of effort.k+1 vk − vk−1 ≥ 0. have different information partitions). we use the convention that when k = 1. .k p k. v k−1 = v0 is defined as 0. It follows. (1) for every k = 1. F ∗ ) and E 2 (π 1 . i. . . Proposition 5 In equilibrium F ∗ . i k i i k 123 . . is increasing in k: E 1 (π 0 . F ∗ ≤ E 2 π2 (ωk ). . the expected efforts of both players are the same. for every even k = 2. n − 114 E 1 π1 (ωk ). n E 1 π1 (ωk ). F ∗ ). Proof Let xk−1 ≤ x ≤ xk .. F ∗ . F ∗ and for every odd k = 1. F ∗ ) ≤ · · · ≤ E 2 (π n . . . Proposition 4 Under condition (12). . (3) no player is dominant with respect to the expected payoff. The next result shows that although the players are asymmetrically informed (i. their ex-ante distributions of equilibrium effort are identical. 14 Recall that π (ω ) ∈ denotes the element of that contains ω . F ∗ = p k−1.e. Furthermore. . In particular. for k = 2. .

In asymmetric information models. common value all-pay auctions may have only non-monotone equilibria (with overlapping sup- ports of equilibrium strategy distributions corresponding to different types of the same player). probabilities of winning. we showed in Sect. 7 Concluding remarks We presented sufficient conditions for the existence and uniqueness of a monotone equilibrium in a model of asymmetric all-pay auctions where no player has an infor- mation advantage over his opponent. as was exhibited by Rentschler and Turocy (2015).k+1 vk − xk = p j + ( pk−1 + pk ) p k. 2013a). It is worth noting that although the players have the same ex-ante distribution of effort. and since obviously ∗ ∗ F 1 (x) = F 2 (x) = 1 for x > xn . and also sufficient for the existence of a monotone equilibrium when the different information endowments can be ranked (see Einy et al. xn ]. although no player has an information advantage over his opponent. (15) k−1 With the convention that k−2 j=1 p j = j=1 p j = 0 when k < 2 and that p0 = pn+1 = 0. 3 that their ex-ante probabilities to win the contest need not be the same. and thus the expected efforts of all players are equal. exerts an effort smaller than or equal to x is then ∗ k−2 ∗ F −i(k) (x) = p j + ( pk−1 + pk )Fi(k−1) π k−1 . x j=1 k−2 x + p k. ∗ ∗ Since we have shown that F 1 (x) = F 2 (x) for every x ∈ [0. differences in players’ information usually result in different equilibrium strategies. we show that equilibrium efforts have the same distribution for all players. In our model. and expected pay- offs. (14) pk vk j=1 The ex-ante probability that i(k)’s rival. (14) and (15) also hold for k = 1 and k = n. the ex-ante distributions of effort in equilibrium F ∗ are identical for both players. k−1 ( pk + pk+1 )( pk + pk−1 )(x − xk−1 ) = pj + .92 E. player −i(k) = i(k − 1).k+1 vk j=1 k−1 ( pk + pk+1 )( pk + pk−1 )(x − xk−1 ) = pj + pk vk j=1 ∗ = F i(k) (x). We also show that the conditional expected 123 . This assumption is natural in many contexts. Without connectedness. Einy et al. An important assumption underlying our results is that the information endowment of each player is connected with respect to the value of winning the contest.

. . .k+1 vk − xk . j+2 v j+1 − v j ) is non-negative by assumption (12). 3. xk ]. Then (17) applies also when k = 0 (in which case i(k) = 1). (iii) indifferent between all efforts in [xn−1 . x − x x + p k. (16) If 1 ≤ k ≤ n − 1 and i exerts effort x ∈ [xk . if 1 ≤ k ≤ n − 1 and i = i(k). . player i is: (i) indifferent between all efforts in [xk−1 . x. 2012)]. xn ] given π n (if i = 2). then ∗ ∗ E i(k) π k . From our results we can conclude that in common-value all- pay auctions the players’ information does not affect the ratio of the players’ expected efforts. . i. and i(k) exerts effort x ∈ [xk−1 .k+1 vk Fi(k−1) π k−1 . for k = 2. It can be shown by induction on k that.k vk+1 −x p k+1. . Fi(k−1) = p k. then ∗ ∗ E i(k) π k . j+1 p j+1. conditional on the event π k . xk+1 ] given the event π k . j+1 ( p j+1.k+1 vk − xk = p j. 2013b) and (Warneryd 2003. (ii) indifferent between all efforts in [x0 . Equalities (16) and (17) establish the following fact: Fact 1 When player i’s opponent uses the strategy from the profile F ∗ . (17) holds for every 0 ≤ k ≤ n − 1. Appendix Proof of Proposition 2 For any k = 0. and assume that his rival uses the strategy from the profile F ∗ . equality (18) 123 . The expected payoff of player i(k). is given as follows. k−1 p k.k+1 vk + p k+1.k vk+1 Fi(k+1) π k+1 .k+1 vk −x p k.e.k+1 vk − xk .k+1 vk + p k+1.k+1 vk = p k. xk+1 ]. .k vk+1 = p k. n. . n consider the player i(k) for whom π k ∈ i(k) . x − x x − xk = p k.k+1 vk − xk = p k. (18) j=1 The expression in (18) is non-negative as every summand in k−1 j=1 p j. j+2 v j+1 − v j ≥ 0. If 1 ≤ k ≤ n. (17) Now set v0 = 0. When k = 0 or k = 1. x1 ] given π 0 (if i = 1). x. Fi(k+1) = p k.Common-value all-pay auctions with asymmetric information 93 payoffs of the players increase with the expected value of winning (as conveyed by the revealed information). It is interesting to note that this property does not necessarily hold for other forms of contests such as Tullock contests with or without the assumption of an infor- mation advantage [see (Einy et al.

(19) The inequality in (19) holds since.k+3 vk+2 −xk+2 ≤ p k.k+1 vk + p k+1.94 E. player i (weakly) prefers effort xk to any effort above xmin(k+1. and the last equality in (19) holds by (17). Given the event π k . y. F2∗ . F1∗ ≤ E 2 π n .k vk+1 − xk+1 p k+2. Fi(k+1) ≤ E i(k) π k . Since obviously.k+3 vk+2 > p k+1.k−2 vk−1 123 .k vk+1 Fi(k+1) π k+1 .k − xk+1 p vk+1 ∗ = p k. y. xn−1 .k vk+1 k+1. Fi(k−1) = p k.k+1 vk − xk = E i(k) π k .k vk+1 −y p k+2. if y > xn . Now consider 2 ≤ k ≤ n and the player i(k). E 2 π n . given the event π k . E 1 π n−1 . y − y y − xk−2 = p k. p k+2. xk+2 . (21) and (19) establish the following: Fact 3 When player i’s rival uses the strategy from the profile F ∗ .n) .k+3 vk+2 − xk+2 = p k.k+1 vk + p k+1.k+3 vk+2 xk+1 − xk = p k. F1∗ (21) Then (20).k+1 vk + p k+1. Next consider 0 ≤ k ≤ n − 2 and the player i(k). if y ∈ [xk+1 .k+1 vk −y p k−1. if y ∈ [xk−2 . F2∗ ≤ E 1 π n−1 . ∗ ∗ E i(k) π k . Einy et al. if 0 ≤ k ≤ n and i = i(k).k vk+1 . y.k+1 vk Fi(k−1) π k−1 . y − y y + p k+2. It then follows from (16) and (18). Notice that. if y > xk+2 and k ≤ n − 2. given the event π k . xn . y. Fi(k+1) . Fi(k+1) = p k. (17) that: Fact 2 The conditional expected payoffs of player i considered in (16) and (17) are non-negative for the corresponding efforts. F2∗ = E 1 π n−1 . xn . xk . xk+2 ] then ∗ ∗ E i(k) π k . remains meaningful if the sum is defined as 0. y. xk−1 ] then ∗ ∗ E i(k) π k . Fi(k+1) (20) and.k+1 vk + p k+1.k+3 vk+2 xk+1 + p k+2. by (12).

player i(k) can improve his expected payoff by shifting mass from a to an effort slightly below a. F2 ) is an equilibrium. Thus. for some k. μ Fi(k) (π k .. ·) is a compact set.k+1 vk − xk−1 p k−1.k−2 vk−1 ∗ = p k. conditional on π k .f. the c. and that both strategies are monotone. and μ Fi(k+1) (π k+1 . Note also that when k ≥ 2 and 0 ≤ y ≤ xk−2 . ·) the support of that distribution (i. Fi(k−1) . conditional on π k . a is an atom of μ Fi(k) (π k .·) the probability distribution (measure) with a c.·) ((a −δ.f. Lemma 2 For each k = 1. there exists δ > 0 such that μ Fi(k−1) (π k−1 . and by supp Fi (π k . conditional on the event π k any effort in the support of Fi∗ π k .e. 3. . ∗ E i(k) π k . if 2 ≤ k ≤ n and i = i(k). ·) = [xk−1 . Lemma 1 For any k = 1. a]) = 0 if k ≤ n − 1.·) has no atoms in (0. Fi(k−1) ≤ 0. (22) The inequality in (22) holds since by (12) p k. Fi (π k . This contradicts the assumption that.e. F2 ) = (F1∗ . .·) ((a − δ. (23) Then (22).k+1 vk ≥ p k−1. 2. y.d. x + ε) > 0). ∞). F2∗ ). . supp Fi(k) (π k . each supp Fi (π k . Facts 1. Furthermore. . a]) = 0 if k ≥ 1. Since equilibrium strategies’ supports are necessarily bounded (as the marginal cost of effort is positive and fixed at 1). Proof of Proposition 3 Assume that F = (F1 . (24) 123 . i. xk . Thus Fi∗ is also an unconditional best response for each player i. · is a best response of player i against the mixed strategy of his rival in the profile F ∗ . ∞) such that for any ε > 0. for every 1 ≤ k ≤ n − 1. Proof of Lemma 1 Suppose that. 2. ·) is continuous on (0. which means that F ∗ is indeed an equilibrium of G. ·) is discontinuous at some a > 0. (23). xk+1 ]. Fi(k) (π k . n. player i (weakly) prefers effort xk to any effort below xk−1 given the event π k . . a is chosen by i in equilibrium with positive probability. . and Fact 2 lead to the following: Fact 4 When player i’s rival uses the strategy from the profile F ∗ . ·). denote by μ Fi (π k .. Fi(k) (π k . ·) .k−2 vk−1 . ∞). Hence all conditional expected payoffs of each player are continuous at every positive effort (chosen as that player’s pure strategy).e. provided his rival sticks to his strategy in F. For any i = 1. Then x0 ≤ x1 ≤ · · · ≤ xn . and 4 show that for any 0 ≤ k ≤ n with i = i(k). and let x0 ≡ 0. n. We will show that (F1 . and the last equality in (22) holds by (16).) . the closure of set of all x ∈ [0.Common-value all-pay auctions with asymmetric information 95 xk−1 − xk−2 ≤ p k. 2 and k satisfying i(k) = i. μ Fi (π k . denote xk ≡ max supp Fi(k−1) (π k−1 . Since i(k)’s rival best-responds to Fi(k) . i.d.k+1 vk − xk = E i(k) π k . .·) ((x − ε. .

Proof of Claim 1 By definition x1 ∈ supp F1 (π 0 . Case 2: x + δ ∈ (supp F2 (π 1 . F1 . and w. This implies the existence of δ > 0 such that [x. x + )) = 0. supp F1 (π 0 . Claim 2 If x1 > 0. x + δ − δ1 . F2 ) . The following claim is needed to establish the equality (24) for k = 1. In particular. x +δ ∈ / supp F2 (π 1 . we conclude that x + δ ∈ / supp F1 (π 0 . a contradiction. F2 ) > E 1 (π 0 . xn ]. and so E 2 (π 1 . x1 ]. (26) Proof of Lemma 2 We start by showing (25). As F2 is an equilibrium strategy. ·))c . x1 ) ∩ (supp F1 (π 0 . Now suppose by the way of con- tradiction that there exists x ∈ [0. Since contradiction arises in both cases. Proof of Claim 2 Since [x0 . we can assume that x + δ ∈ supp F1 (π 0 . Given the monotonicity of F1 and the first part of Lemma 1. in contradiction to the assumption that x ∈ [x0 . ∞) ∩ (supp F2 (π 1 . ·). it can be shown that [x1 . Case 1: x + δ ∈ supp F2 (π 1 . ·))c . Using the first part of Lemma 1. Arguing as in Case 1 above. x + δ) ⊆ [0. By recycling the arguments used to prove Claim 1 above. Hence. x1 ] ⊂ supp F2 (π 1 . ·))c . x + δ + δ1 ) of x + δ. x1 ) ∩ (supp F2 (π 1 . μ F2 (π 1 . x +δ).l. ·) ⊂ [x0 . x + δ. ·). ·). F2 ) > E 1 (π 0 . Einy et al. Suppose by the way of contradiction that there exists x ∈ (0. it suffices to show that (x0 . By the second part of Lemma 1. x1 ) ⊂ supp F2 (π 1 .g.96 E. it follows that μ F1 (π 0 . 1 123 . ·).·) ((x − . By the monotonicity of F2 and the first part of Lemma 1. ·). by the second part of Lemma 1. x) > E 2 (π 1 . z. By definition. x1 ) ∩ (supp F1 (π 0 . the claim is established. x) > E 2 (π 1 .·) ([x. ·). x1 ≤ x2 . x + ) of x. x1 ] = supp F1 (π 0 . Given the monotonicity of F2 and the first part of Lemma 1. [x0 . a contradiction. this means that under efforts x + δ − δ1 and x + δ player 1 has the same expected value of winning (conditional on π 0 ). x] ⊂ (0. hence E 1 (π 0 . player 1 has the same expected value of winning (conditional on π 0 ) under efforts x − δ and x . ·).o. As F1 is an equilibrium strategy. ·) = [x0 . additionally. E 1 (π 0 . ·))c . ·))c . x − δ. x1 )∩(supp F2 (π 1 . In conjunction with Claim 2 this implies15 [x0 . x1 ] ⊂ supp F1 (π 0 . μ F1 (π 0 . x2 ] ⊂ supp F2 (π 1 . ·) = [xn−1 .·) ((x + δ − δ1 . Now two possible cases emerge. x2 ] ⊂ 15 This is regardless of whether x > 0 or = 0. ·). F1 . z) for any z in some open neighborhood (x + δ − δ1 . In this case there exists δ1 > 0 such that [x + δ − δ1 . under efforts x and x + δ player 2 has the same expected value of winning (conditional on π 1 ). F1 . x1 ] (25) and supp F2 (π n . x + δ]) = 0. x + δ + δ1 )) = 0. and so Claim 1 implies (25). F1 . supp F1 (π 0 . x1 ] is assumed to have a non-empty interior. ·). E 2 (π 1 . x + δ] ⊂ (0. Claim 1 [x0 . ·). ·))c . This implies the existence of δ > 0 such that [x −δ. F2 ) for any z in some open neighborhood (x − . and hence.

if xn−1 < xn . x1 . obviously. ·) is continuous at 0. assume that (27). Fi(k−1) = E i(k) π k . and E i(k) (π k . But. xk ) p k.k+1 p k−1. Now.2 [1 − F1 (π 0 . 0. j+1 p p j=1 p 0. and Fi(k) (π k . xn ] ⊂ supp F2 (π n . ·) is continuous at 0. xk−1 .Common-value all-pay auctions with asymmetric information 97 supp F2 (π 1 . (27) holds for k = 1. F2 ) = 0. ·.−1 F1 (π 0 . Since the reverse inclusion holds by the definition of x2 . then for every 1 ≤ k ≤ n. j−1 xk = p v j 1 − j−1. 0). 0)].k p k−2. E 2 (π n . F1 . Since xk−1 . j−2 1 − j−2. x) < E 2 (π n . given 1 < k ≤ n. using additionally the assumption that F2 (π 1 . p p p p (28) Proof of Lemma 3 The proof is by induction on k. i. xk−1 > 0. ε→0+ Thus x1 = p 1. 123 . ·). and hence the inclusion holds as equality. Lemma 3 If x1 > 0 and F2 (π 1 . F1 . The analogue of Claim 2 would also show that [xn−1 . x1 = lim E 2 π 1 . Fi(k−1) = p k. · . 0)].2 v1 F1 (π 0 . ·).k−3 1−· · · 1 − 0. Fi(k−1) ) is continuous on (0. x1 ] ⊂ supp F2 (π 1 . ∞). F2 ) = E 1 (π 0 . F1 .1 × 1 − 0. v1 F2 (π 1 . F1 . 0) · · · . 2 can be used to iteratively extend the proof in order to establish (24) for k > 1.k−1 1− k−1. ·) is continuous on (0. x n ) for every x > xn . ∞). since [0.2 v1 [1 − F1 (π 0 . and hence F2 (π 1 .k−1 p 0.1 = k. F1 . 0) · · · (27) p (with the convention that p 0. Similarly. ε = p 1. x1 ) − x1 = E 1 (π 0 . k p j−1. xk−1 − xk−1 . (28) have been shown to hold for k − 1.k+1 vk − xk = E i(k) π k . The analogues of Claims 1.−1 F1 (π 0 . which establishes (28) for k = 1. establishing (26).1 = p 1.k−2 1− k−2. we have established the equality (24 ) for k = 1. and E 2 (π 1 . Notice that. x1 > 0. xk . j−3 1 − · · · j. p k.e. p 1.0 = 1).2 v1 − x1 = E 2 π 1 . and to show that x2 ≤ x3 ≤ · · · ≤ xn . xk ⊂ supp Fi(k) π k .k+1 vk Fi(k−1) π k−1 . j p j−2. ·). x1 ) = p 1.−1 = p 0.

·. p p p j=1 (29) where the last two equalities hold by the induction hypothesis on (27) and (28). 0) · · · = p n.e. p n−1. Fi(k) = E i(k−1) π k−1 . then F1 (π 0 . Corollary 4 If x1 > 0 and F2 (π 1 . (31) j=1 and Fi(k) π k . xn ) = 1. xk−1 . Einy et al. 0) · · · 0 .1 = k. p n−1.1 = p j. ·) is also continuous at 0).k−1 vk Fi(k) π k .k vk−1 + p k. by (28) in Lemma 3 taken for k = n.k−2 1 − k−2. Since xk−1 . and hence Fi(k) π k . Fi(k) = p k−1. j p j−2. ∞).n−2 p n−2.k+1 vk 1 − k−1. and E i(k−1) (π k−1 . j−3 1 − · · · 1 − 0.k−1 p 0.n−3 p 0.k−2 1 − k−2.98 E.k p k−2. (32) Proof of Corollary 4 Since F2 (π n . xk 1 = (xk − xk−1 ) p k. j−1 p j.1 = p k.n−1 .n−1 p 0. by Lemma 3. F1 (π 0 . xk−1 + xk−1 p k−1.k+1 vk 1 − Fi(k−1) π k−1 . k xk = p j.k+1 . xk .n p n−2. ·) is continuous at 0.k p k−2.1 1− 1 − 1 − · · · 1 − F1 (π 0 . j+1 v j .k−3 1 − · · · 1 − 0.k−1 1− k−1.k+1 p k−1. p k−1. j−1 p 0. Fi(k) ) is continuous on (0.−1 F1 (π 0 . j+1 v j 1 − j−1.k−3 1 − · · · 1 − 0. Hence. This completes the induction step for (27). · . xk ⊂ supp Fi(k−1) π k−1 .k vk−1 − xk−1 .−1 F1 (π . xk = p k. j−2 1 − j−2. xk − xk = E i(k−1) π k−1 . 0) · · · . 0) = 0 (i. xk−1 > 0. p p p p (30) This completes the induction step for (28) and finishes the proof. for every 1 ≤ k ≤ n..−1 123 .−1 F1 (π 0 .k−1 p 0.k−1 vk p k. 0) · · · + xk−1 p p p k p j−1. and hence xk = p k.

since [0. ε ε→0+ = p 1. F1 . (37) p n. p 1. ∞). ·) is continuous at 0 then x1 > 0. Then [0. Similarly.2 v1 + p 2.k+1 vk 1 − k−1. This yields 1 2. x2 > 0. xn = (xn − xn−1 ) . Since E 1 (π 2 . F1 . x2 = E 1 π 2 . 0 = 0. the first equalities in (29) and (30) in the proof of Lemma 3 hold in the current context as well. 0 . (33) p For k > 2. ·). ·) is continuous on (0. (35) and (33) that 1 x3 − x2 = v3 1 − 2. F2 . F2 . 0) . ∞). (36) p vk−1 and also that 1 (1 = ) F2 π n . 0) p 123 . p 2. and they imply that xk − xk−1 (34) =p vk 1 − Fi(k−1) π k−1 . x2 ) = 2. Proof of Lemma 5 Assume to the contrary that F2 (π 1 .3 + p 2.1 F1 (π 2 . F1 .1 F1 (π 2 . x2 − x2 = E 2 π 1 . ·) is continuous at 0 (and hence x2 > 0) but x1 = 0.1 v2 F1 π 2 . Thus x2 = p 2.1 v2 F1 π 2 . x2 ] ⊂ supp F1 (π 2 .3 F1 (π 2 . x2 ] = supp F2 (π 1 .3 v2 − x2 = E 1 π 2 .1 p + p 2. 0) = 0. xk−1 k.Common-value all-pay auctions with asymmetric information 99 which can only occur when F1 (π 0 .k−2 (xk−1 − xk−2 ) . ·). Substituting F1 (π 0 .k+1 (35) 1 =p k. x2 = lim E 2 π 1 .3 v2 . 0) = 0 into (27) and (28) yields (31) and (32). Lemma 5 If F2 (π 1 . F2 .1 p 2.n−1 vn It follows from (34).2 v1 + p 2. and E 2 (π 1 . ·) is continuous on [0.

· is continuous at 0. From (37) we have xn − xn−1 = p n.1 p 2. ∞). x1 > 0. x1 > 0. since [0.2 + F2 π 1 . ∞). by induction based on (34)–(36).2 = p 2. v1 F2 π 1 . 0 = 0.k+1 vk 1 − k−1. F1 .3 + p 2. if n = 3. ·) is continuous on [0.2 v1 .1 p 2. · is continuous at 0. F2 ) is continuous on (0.n−3 1 − · · · 1 p p 3. Thus x1 = p 1. ·) and E 2 (π 1 . · is continuous at 0.2 v1 − x1 = E 2 π 1 . ·. (42) p vk−1 123 .n p n−2.2 1 − 2. the left-hand side is negative while the right-hand side is non-negative. since [0. F2 = lim E 1 π 0 . and E 1 (π 0 .1 F1 (π 2 .1 F1 (π 2 .4 p 1 − 3. the first equalities in (29) and (30) in the proof of Lemma 3 hold in the current context as well. p Equivalently. However.1 F1 (π 2 . Einy et al. and they imply that xk − xk−1 (40) =p k.100 E. ·).3 p 1.3 + p 2. p 1. which is equal to the expression in (38) only if 1 1 − 2. 0 . x1 = E 2 π 1 .n−1 vn .2 . 0 1 Lemma 6 If F1 π . 0). x1 .k−2 (xk−1 − xk−2 ) . that p n−1. x1 − x1 = E 1 π 0 . (39) For k > 1. Proof of Lemma 6 Since F1 π 0 .k+1 vk 1 − Fi(k−1) π k−1 . − p 2.n−2 1 − n−2. 0) = p 3. x1 ] = supp F1 (π 0 . F2 = v1 F2 π 1 . F1 . 0) ··· (38) p p if n > 3. Notice that. ε. F1 . And. a contradiction. xk−1 (41) 1 = p k. x1 = p 1.n−1 xn − xn−1 = vn 1 − n−1. ε→0+ and hence F2 π 1 . then F2 π . 0 . x1 ] ⊂ supp F2 (π 1 . and.

n−1 xn − xn−1 = vn 1 − n−1.n−2 1 − n−2. Accordingly. (41 ) and (39) that x2 − x1 = p 2. Fi(k) = p k−1. · and E i(k+1) (π k+1 . Since.1 p − F2 (π . xk+1 : p k+1. Fi(k) = p k+1. Fi(k) (π k . x. 0) = 0. x. as has been shown. ·) is continuous at 0. Thus F2 (π 1 . · .k−1 vk Fi(k) π k . · cannot be discontinuous at 0.k vk−1 + p k. by Lemma 6. since otherwise player 1 would have had a profitable deviation from F1 . xk (if k > 1) and on the interval xk . F2∗ )). consisting of a shift of mass from 0 to an effort slightly above zero (conditional on π0 ). 0) . · can have a discontinuity at zero.k+2 vk+1 − xk+1 = E i(k+1) π k+1 . then as xk−1 . then as xk . ∞). xn = (xn − xn−1 ) . 0 Proof of Proposition 3 Note first that at most one of the functions F1 π . 0) · · · 2.3 v2 p 2.Common-value all-pay auctions with asymmetric information 101 and also that 1 (1 = ) F2 π n .1 − F2 (π 1 .3 − 2. Fi(k) ) is continuous on [0. if 1 < k ≤ n. that p n−1. also xn − xn−1 = p n. xk : p k−1.n−1 vn It follows from (40). xk ⊂ supp Fi(k−1) π k−1 . and the sequences x k k=1 . if 0 ≤ k < n. · must be continuous at 0. a contradiction. · and E i(k−1) (π k−1 . xk ) k=1 are determined by (31) and (32) in a unique fashion (and are equal to the corresponding parameters for the strategy profile (F1∗ . Note now that.k vk−1 − xk−1 . xk+1 . xk−1 .1 1 . It follows that each Fi(k) π k . · and F2 π 1 . and.k+2 vk+1 Fi(k) π k . Fi(k) = E i(k+1) π k+1 . x − x = E i(k−1) π k−1 . (44) p From (43). Note that F2 π 1 .n−1 vn . xk+1 (if k < n). ∞). · is continuous at 0 and therefore so is F2 π 1 . which is equal to the expression in (44) only if F2 (π 1 . x − x. Then by Lemma 5 and Corollary 4 n n F1 π 0 . Fi(k) π k .n−3 1 − · · · 1 p p p 2. Fi(k) = E i(k−1) π k−1 . · is linear on the interval xk−1 . since then F1 π 0 . xk+1 ⊂ supp Fi(k+1) π k+1 . by induction based on (40)–(42).n p n−2. · is a continuous 123 . (43) p n. ·. for every x ∈ xk−1 . · is also continuous at 0. Similarly. F2 π 1 . Fi(k) ) is continuous on [0. ·. for every x ∈ xk .

J Math Econ 37:247–258 Einy E. J Math Econ 47:760–767 Hillman A. Turocy T (2015) All-pay auctions with interdependent valuations: the highly competitive case. Moreno D. lemons. Haimanko O. Econometrica 77(1):71–92 Siegel R (2014) Asymmetric contests with interdependent valuations. Orzach R (2012) Equilibrium and revenue in a family of common-value first-price auctions with differential information. 0 = 0. Kovenock D. Shi X (2012) Carrots and sticks: prizes and punishments in contests.102 E. Morgan J (1997) An analysis of the war of attrition and the all-pay auction. and cookies: designing auction markets with dispersed information. F2∗ ). Haimanko O. Sela A (2006) Contest architecture. Leininger W (1996) Asymmetric all-pay auctions with incomplete information: the two-player case. xk+1 = 1 if k < n (and 0 n F1 π . F2 ) = (F1∗ .s Fi(k) (π k . J Econ Theory 126(1):70–97 Moldovanu B. CEPR Discussion Paper No. Thus (F1 .f. ·) k=1 in a unique fashion. Am Econ Rev 91(3):542–558 Moldovanu B. xk ) k=1 determine the n c. Shitovitz B (2013b) Tullock contests with asymmetric information. Working paper Forges F. de Vries CG (1993) Rigging the lobbying process: an application of the all-pay auction. Sela A (2002) Dominance solvability of second-price auctions with differential information. Econ Inq 50(2):453–462 Rentschler L. Econ Theory 8:291–305 Che Y-K. Athey S. Games Econ Behav 14:1–18 Baye MR. Econ Polit 1:17–40 Krishna V. function. F2 (π n . Available at SSRN: http://ssrn. Am Econ Rev 83:289–294 Baye M. Orzach R. Kovenock D. Sela A (2001) The optimal allocation of prizes in contests. Grubb M (2012) Peaches. Riley J (1989) Politically contestable rents and transfers. Gale I (1998) Caps on political lobbying. J Econ Theory 110:121–136 Warneryd K (2012) Multi-player contests with asymmetric information.d. References Abraham I. Harvard University Amman E. J Econ Theory 72(2):343–362 Malueg D. DP9315 Einy E. de Vries C (1996) The all-pay auction with complete information. Orzach R. Am Econ Rev 88(3):643–651 Einy E. Econ Lett 105:177–180 Malueg D. J Econ Theory (Forthcoming in) Warneryd K (2003) Information in conflicts. Working paper. with Fi(k) π k . Sela A. xk−1 = 0 if k > 1 and Fi(k) π k . Orzach R (2011) Core-stable rings in second price auctions with common values. Econ Theory 51:277–287 123 . Int J Game Theory 41(2):219–254 Moldovanu B. Sela A. xn ) = 1). Babaiof M. Orzach R (2009) Revenue comparison in common-value auctions: two examples. Sela A (2013a) Common-value all-pay auctions with asymmetric infor- mation. Haimanko O. the values of Fi(k) (π k . Einy et al.com/abstract=2446789 Siegel R (2009) All-pay contests.

- Mixed StrategyUploaded byendu wesen
- A Theory of Social Decision MakingUploaded byYoyokKurniawan
- Problems Repeated and Bayesian game.pdfUploaded byvictorginer8
- Game Th. TalkNotesUploaded byBj Kr
- Lecture Note 4 Extensive-Form GamesUploaded byKashi Kafle
- 05715450Uploaded bygeekmasta
- Equivalence Relations_a.pptUploaded byKaleem Chaudhary
- Behavioral Game Theory.pdfUploaded byAnonymous 7UBIVlKM

- Martin GalasUploaded byJavier
- principio_del_consensualismo.pdfUploaded byMarco Antonio Rodriguez
- ManualUsuario K XCT300iUploaded byJavier
- Master 5Uploaded byJavier
- Nociones de lógica.pdfUploaded bytambooks
- Mundo interconectadoUploaded byJavier
- Ch_03Uploaded byJavier
- K-XCT_300i_201303271114_MR.SK60BA-V1_Uploaded byJavier
- Stoch Proc SimUploaded byJavier
- Modelo_Hedonico_Rosen.pdfUploaded byJavier
- jeas_0118_6679.pdfUploaded byJavier
- Modelo Hedonico RosenUploaded byJavier
- 47722418-CAPITULO+6-Fundamentos+microeconomicosUploaded byJavier
- hoja1-cefUploaded byJavier
- Tabla de DerivadasUploaded byAdam Vang
- top_ mod_ent_binUploaded byJavier
- TEMA 4. LOS CICLOS ECONÓMICOS REALES (1).docUploaded byMary Turpo Pancca
- Master4 14 RUploaded byJavier
- Master4 14 RUploaded byJavier
- Simulacion y Finanzas-CompletoUploaded byJavier
- 47722416-CAPITULO+5-Teorias+keynesianasUploaded byJavier
- MintUploaded byJavier
- EDEstocasticas_LuisRincon.pdfUploaded byJavier

- Power seriesUploaded bysondhani
- pnas01660-0010bUploaded bysid_senadheera
- Robust Control ToolboxUploaded byEmre Sarıyıldız
- FemUploaded byapu
- Lecture Note on StatisticsUploaded byHo Thai Anh
- GFASUploaded bynagy_andor_csongor
- Cushing-An Introduction to Structured Population DynamicsUploaded byMarluci Galindo
- Matrices JonyUploaded byJony Rolaivers Mosquera
- c101contentUploaded byuserraj
- basten form 63Uploaded byapi-287205585
- Speciale1Uploaded byzappy14
- 04359193Uploaded byRyan Dutt
- Delhi_30-1-(1, 2, 3) FinalUploaded byjgkrishnan
- Chapter 12Uploaded byAyman Al Kafrawy
- Arnold, V. I. - On Teaching MathematicsUploaded byTathagata Ghosh
- Principles of DiscretizationUploaded bymanmohancq
- 1.2. Adding and Subtracting Real NumbersUploaded byKzea Jae
- Metrics and distanceUploaded byRedwan Itlearning
- 3exp Functions and Graphs 1Uploaded byJohn Goh
- homework1-1Uploaded bys4368
- Diff CalUploaded byrex tanong
- 9709_s03_qp_3Uploaded bySong Lim
- Baez, Higher-Dimensional Algebra and Topological Quantum Field TheoryUploaded byΣωτήρης Ντελής
- The singular value decomposition.Uploaded byDundi Nath
- A problem with one variable.docxUploaded byAmanda Patterson
- Estimation TheoryUploaded byMurash javaid
- The Method of Lagrange MultipliersUploaded bySon Hoang
- Convergence Catalan Series 4Uploaded bymark_villarino7158
- Detailed Lesson Plan-dspUploaded byemssiet
- C3 Solomon EUploaded byJosh Zngr