Tadelis (2001) - The Market For Reputations As An Incentive Mechanism

Article
Has bibtex key
Has article title
Has year	2001
In journal
In volume
In number
Has pages
Has publisher
© edegan.com, 2016

• This page is referenced in BPP Field Exam Papers

Reference(s)

Tadelis, S. (2001) "The Market for Reputations as an Incentive Mechanism," Journal of Political Economy 110(4):854-882 pdf

Abstract

Reputational career concerns provide incentives for short-lived agents to work hard, but it is well known that these incentives disappear as an agent reaches retirement. This paper investigates the effects of a market for firm reputations on the life cycle incentives of firm owners to exert effort. A dynamic general equilibrium model with moral hazard and adverse selection generates two main results. First, incentives of young and old agents are quantitatively equal, implying that incentives are "ageless" with a market for reputations. Second, good reputations cannot act as effective sorting devices: in equilibrium, more able agents cannot outbid lesser ones in the market for good reputations. In addition, welfare analysis shows that social surplus can fall if clients observe trade in firm reputations.

Summary

The model seperates identities from entities (which have names), with the later having reputations. There is a market for names and two types of agent: good and opportunistic. The agents are short-lived but the transferability of the names creates a long-lived entity seperate from the identity of the agents. There is no seperating equilibria where only good or bad types buy names, but there are reputation dynamics - reputations increase after good performance and decrease after bad performance.

The Model

The model has the full base assumptions:

• Buyers and sellers are risk neutral
• Types of sellers are indistinguishable and are active for two periods with overlapping generations
• Sellers can choose a new name, keep their name or buy a name
• Only names have track records
• There is a competitive market of buyers so that prices are bid up to their expected surplus
• There is no discounting

In addition there are the following driving assumptions:

• There are no contingent contracts
• Shifts of name ownership are not observable
• Name changes are costless
• With probability [math]\epsilon \ge 0\,[/math] a seller can not change his name (note that this is a technical assumption to rule out some 'bad' equilibria from costless name changes)

The model uses a rational expectations equilibrium.

There are two types of seller: [math]G\,[/math] (good) exist in proportion [math]\gamma\,[/math] and succeed with probability [math]P_G \in (0,1)\,[/math], and [math]O\,[/math] (opportunistic) exist in proportion [math]1-\gamma\,[/math] and success with probability [math]eP_G\,[/math] where [math]e \in [0,1]\,[/math] and have a cost of effort [math]c(e)\,[/math] that satisfies Inada conditions.

The sequence of events is:

1. New agents chooses or buys his name
2. Buyer pays upfront
3. Opportunistic type chooses effort
4. Success or failure realized
5. Retiring agent can sell his name/continuing agent can change their name

Benchmark: No Reputations Markets

Here we assume that names can not be traded and restrict attention to two periods with three generations. Generation 0 dies at the end of period 1, generation 1 lives for both periods, and generation 2 is born half way through and lives for period 2.

In period 1 there is one wage as everyone has a new name. In period 2 there are three wages depending on the [math]h\,[/math] (histories) observed which are either [math]S\,[/math] (success), [math]F\,[/math] (failure), or [math]N\,[/math] (new name). [math]F\,[/math] histories are irrelevant as sellers who failed are better off costly changing their name.

[math]O\,[/math] types will exert zero effort in the second period (as it is costly and contracts are made in advance) so [math]w_{2}(h)\,[/math] only depends on clients' beliefs about the likelihood of [math]h\,[/math] belonging to a [math]G\,[/math] type.

Expected period 2 utility of an [math]O\,[/math] type is:

[math]\mathbb{E}u_{O} =w_{1}+eP_{G}w_{2}(S)+(1-eP_{G})w_{2}(N)-c(e) =w_{1}+eP_{G}\Delta w+w_{2}(N)-c(e)\,[/math]

where [math]\Delta w\equiv w_{2}(S)-w_{2}(N)\,[/math]

The equilibrium is characterized by the tuple [math]\left\langle w_{1},w_{2}(S),w_{2}(N),e\right\rangle\,[/math] such that [math]e\,[/math] is a best response given [math]w_{2}(S)\,[/math] and [math]w_{2}(N)\,[/math], and all wages are correct given rational expectations about [math]e\,[/math]

Equilibrium beliefs imply:

[math]\Pr \{G|S\}=\frac{\gamma P_{G}}{\gamma P_{G}+(1-\gamma )eP_{G}}\,[/math]

[math]\Pr \{G|N\} =\frac{\gamma (1-P_{G})+\gamma }{\gamma (1-P_{G})+(1-\gamma)(1-eP_{G})+1} =\frac{2\gamma -\gamma P_{G}}{2-P_{G}[\gamma +(1-\gamma )e]}\,[/math]

Equilibrium wages are:

[math]w_{2}(h)=\Pr \{G|h\}\cdot P_{G}\,[/math]

Fixing (for a moment) [math]\Delta w\,[/math] the equilibrium [math]e\,[/math] solves the FOC:

[math]P_{G}\Delta w=c^{\prime}(e)\,[/math]

However, optimal effort must solve:

[math]P_{G} \ge c^{\prime }(e)\,[/math]

The actual wage differential depends on the amount of effort that the [math]O\,[/math] types exert. If all [math]O\,[/math] types exert no effort (i.e. are bad) or full effort (i.e. are good) then we have:

[math]\Delta w_{B}=\frac{2(1-\gamma )P_{G}}{2-\gamma P_{G}} \mbox{ and } \Delta w_{G}=0\,[/math]

This leads to:

Proposition 1: There is a unique equilibrium of the two-period model with no reputation markets. If [math]P_{G}\Delta w_{B} \le c^{\prime }(0)\,[/math] then [math]e=0\,[/math] in equilibrium. If [math]P_{G}\Delta w_{B}\gt c^{\prime }(0)\,[/math] then [math]e\in (0,1)\,[/math] in equilibrium.

Essentially if the opportunistic types exert no effort then the wage differential is the highest, but this provides them with an incentive to work in period 1.

The Market for Reputations

We now let retiring sellings from generation 0 to sell their names. Only successful names will be traded - and in fact [math]S\,[/math] names will be traded in all equilibria. If no names were traded it is because they are worthless. But the supply of [math]S\,[/math] names is positive, so it must be the case that [math]w_{2}(S)\leq w_{2}(N)\,[/math]. But then [math]O\,[/math] types would exert [math]e=0\,[/math] in the first period and [math]Pr \{G|S\}=P_{G}\,[/math]. However, [math]1-\gamma \gt 0\,[/math] implies that there are some [math]O\,[/math] types and [math]\Pr \{S|N\}\lt P_{G}\,[/math], which in turn means that [math]w_{2}(S)\gt w_{2}(N)\,[/math], which is a contradiction.

To get an equilibrium the model assumes an arbitrage condition. The value of an [math]S\,[/math] name is:

[math]v(S)=w_{2}(S)-w_{2}(N)\,[/math]

This naturally leads all period 1 [math]O\,[/math] types to exert the same effort, as they have identical incentives, whether or not they bought a name.

We now need to establish the correct beliefs by buyers about effort in period 1 (again - [math]O\,[/math] types will choose zero effort in period 2):

[math]w_{1}=\left[ \gamma +(1-\gamma )e\right] P_{G}\,[/math]

and

[math]w_{2}(h)=\Pr \{S|h\}=\Pr\{G|h\}\cdot P_{G}\,[/math]

Let [math]\mu\,[/math] denote the proportion of [math]G\,[/math] types who buy [math]S\,[/math] names in [math]t=2\,[/math], and [math]\rho\,[/math] the proportion of [math]O\,[/math] types. Then an equilibrium is specified as a tuple [math]\left\langle \mu ,\rho ,w_{1},w_{2}(S),w_{2}(F),w_{2}(N),v(S),e\right\rangle\,[/math], with the beliefs about [math]\mu ,\rho and e\,[/math] pinning it down. The equilibrium must satisfy the (supply equals demand) market clearing condition:

[math]\gamma P_{G}+(1-\gamma )eP_{G}=\mu \gamma +\rho (1-\gamma)\,[/math]

The beliefs must therefore be (after some algebra and plugging in market clearing):

[math]\Pr \{G|S\}=\frac{\gamma P_{G}+\gamma \mu }{2\gamma P_{G}+2(1-\gamma )eP_{G}}\,[/math]

[math]\Pr \{G|N\} =\frac{2\gamma -\gamma P_{G}-\mu \gamma }{2-2\gamma P_{G}-2eP_{G}(1-\gamma)}\,[/math]

Proposition 4: There exist [math]\underline{\mu }\lt \overline{\mu }\,[/math] so that [math](\mu ,\rho ,e)\,[/math] is an equilibrium if and only if the following three conditions hold:

1. [math]\mu \in [\underline{\mu },\overline{\mu}]\,[/math]
2. [math](\mu ,\rho ,e)\,[/math] satisfy market clearing
3. [math]c^{\prime }(e)=\Delta wP_{G}\,[/math]

As long as the price for names reflects the wage differential that the name generates, sellers will be indifferent. At [math]\underline{\mu}\,[/math] either the price of an [math]S\,[/math] name is sero, or there are too few good types in the second period so that even when all [math]S\,[/math] names are bought by bad types this is still better than having no history. In the interval prices for the [math]S\,[/math] names are positive, and at the upper bound all good new types are buying the names without violating market clearing.

In order to perform some welfare comparisons between the benchmark (which would result if trade could be made observable, say by regulation) and this model, we need to select an equilibrium from each. To do this, we need to restrict the multiple equilibria in this model - this is done by assuming that name buyers are in proportion to their population. Therefore:

[math]\mu ^{*}=\rho ^{*}=\gamma P_{G}+(1-\gamma )eP_{G}\,[/math]

Again we now consider when all [math]O\,[/math] types are good [math](e=1)\,[/math] or bad ([math]e=0)\,[/math] to get:

[math]\Delta w_{B}(\mu ^{*})=\frac{P_{G}(1-\gamma )}{2(1-\gamma P_{G})} \mbox{ and }\Delta w_{G}(\mu ^{*})=0\,[/math]

The market for names is bad (in the welfare sense) if:

[math]\Delta w_{B}(\mu^{*})\lt \Delta w_{B}\,[/math]

Which is true when:

[math]P_{G}\lt \frac{2}{3\gamma}\,[/math]

Longer Horizon and Sorting

The paper goes on to discuss longer horizons and sorting. The derivations will not be discussed here but the following propositions are provided:

Proposition 6: Names with a history consisting of only one success and no failures must be traded in all equilibria.

Proposition 7: For the infinite horizon model there is no equilibrium in which only [math]S\,[/math] names are traded, and these are bought only by good types and by opportunistic types who choose to be good.

This sorting result is important. Suppose it were true, then it would affect beliefs - Observing an [math]S\,[/math] name would lead to the inference that it is held by a good (or good opportunistic) type. Then an [math]S\,[/math] name would command a premium, and [math]O\,[/math] types would value it higher because they face a less attractive future without it.