Grossman Helpman (2001) - Special Interest Politics Chapters 4 And 5

• This page is referenced in BPP Field Exam Papers

Reference(s)

Grossman, Gene and Elhanan Helpman (2001), "Special Interest Politics", Chapters 4 and 5, MIT Press pdf

These chapters reference:

• Crawford, Vincent P. and Joel Sobel (1982), "Strategic Information Transmission", Econometrica, Vol. 50, No. 6 (Nov.), pp. 1431-1451 pdf

Abstract

No abstract available - these are book chapters.

Summary

These chapters cover:

• One Lobby
  • Two states of the world
  • Three states of the world
  • Continuous states of the world
  • Welfare considerations
• Two Lobbies
  • Like Bias
  • Opposite Bias
  • Multidimensional Information
• More General Lobbying

The Model(s)

In general there is a policy maker who choses a policy to implement [math]p\,[/math], based on some facts about the state of the world [math]\theta\,[/math]. There is also a Special Interest Group (SIG) who observes the facts about the state of the world but has a bias [math]\delta \gt 0\,[/math]. The utility functions of both types of players are inverse quadratic (i.e. quadratic loss functions).

Utility of the policy maker:

[math]G(p,\theta) = -(p - \theta)^2\,[/math]

Utility of the SIG:

[math]U(p,\theta) = -(p - \theta - \delta)^2\,[/math]

The policy maker sets [math]p=\theta\,[/math] when the SIG reveals the true state of the world and [math]p=\mathbb{E}\tilde{\theta}\,[/math] otherwise.

One Lobby - Two States

There are two states of the world:

[math]\theta_H \gt \theta_L\,[/math]

Given that the bias is positive the SIG never has any incentive to misrepresent the high state, that is a report of [math]\theta_H\,[/math] will be trusted by the policy maker who will implement [math]p=\theta_H\,[/math].

Supposing that the true state is [math]\theta_L\,[/math], then we can use the distance from ideal points to see for what levels of bias the SIG will truely report this. Specifically the sig prefers to truthfully report if:

[math](\theta_L + \delta) - \theta_L \le \theta_H - (\theta_L + \delta)\,[/math]

[math]\therefore \delta \le \frac{\theta_H - \theta_L}{2}\,[/math]

When the bias satisfies this criteria there is informative lobbying and we can have full relevation, however this is not the sole equilibrium. Suppose the policy maker distrusted the SIG, then the policy maker would implement:

[math]p=\frac{(\theta_L+\theta_H)}{2}\,[/math]

Knowing this the SIG has no incentive to report truthfully - this is the Babbling Equilibrium and it always exists.

However, there is an equilibrium refinement by Farrel where the SIG makes a speech essentially saying "I have no incentive to lie". The SIG prefers [math]\theta_L\,[/math] to the policy maker's distrust implementation if is closer to his ideal point of [math](\theta_L+\delta)\,[/math]:

[math](\theta_L+\delta) - \theta_L \le \frac{(\theta_L+\theta_H)}{2} - (\theta_L+\delta)\,[/math]

[math]\therefore \delta \lt \frac{(\theta_H- \theta_L)}{4}\,[/math] provides the criteria for not trusting the SIG.

One Lobby - Three States

Again there is no incentive for the SIG to misreport [math]\theta_H\,[/math] when it is the true state. When [math]\theta_M\,[/math] is the true state we can perform the bias restriction calculation as before to get:

[math]\delta \le \frac{\theta_H-\theta_M}{2}\,[/math]

If we were just distinguishing between all three states this would hold for [math]\theta_L\,[/math] too:

[math]\delta \le \frac{\theta_M-\theta_L}{2}\,[/math]

However, the SIG now has the option of reporting [math]\theta_L\,[/math] or not [math]\theta_L\,[/math], in the latter case expecting the policy maker to implement [math]\frac{(\theta_H-\theta_M)}{2}\,[/math]. In this case of partial information transmission we have to recalculate the bias restriction for when the real state of the world is [math]\theta_L\,[/math] (For [math]\theta_M\,[/math] the bias restriction is less binding). Here the SIG's ideal point is [math]\theta_L+\delta\,[/math] so the SIG will not report falsely iff:

[math](\theta_L+\delta) - \theta_L \ge \frac{(\theta_H + \theta_M)}{2} - (\theta_L + \delta)\,[/math]

[math]\therefore \delta \le \frac{\theta_H - \theta_M}{4} + {\theta_M - \theta_L}{2}\,[/math]

This allows for an equilibrium with partial transmission of information - this can be possible when the bias conditions for full transmission are violated.

One Lobby - Continuous States

As the number of states grows it becomes harder to get full relevation - the bias must be less than half of the distance between two states and this tends to zero as the number of states tends towards infinity. However, we can partition a continuous range of information and report credible with range the true state of the world falls in.

Suppose that:

[math]\tilde{\theta} \sim U[\theta_{min},\theta_{max}]\,[/math]

The SIG then reports [math]\theta_1\,[/math] if [math]\theta_{min} \le \theta \le \theta_1\,[/math] and so forth. The policy maker getting such a report implements [math]\frac{(\theta_{min} + \theta_1)}{2}\,[/math]. The temptation to lie is greatest when [math]\theta\,[/math] is on the boundary of [math]\theta_1\,[/math], and [math]\theta_2\,[/math] is the greatest temptation (all other higher values greater exagerate the outcome and will overshoot the bliss point). Credibility therefore requires:

[math]\frac{(\theta_1 + \theta_2)}{2} - (\theta_1 + \delta) \ge (\theta_1 + \delta) - \frac{(\theta_{min} + \theta_1)}{2}\,[/math]

This solves to:

[math]\theta_2 \ge 2\theta_1 +4\delta - \theta_{min}\,[/math]

If the truth is in [math]\theta_2\,[/math], then the SIG would prefer to report this rather than [math]\theta_1\,[/math] iif:

[math](\theta_2 + \delta) - \frac{(\theta_1 + \theta_2)}{2} \le (\theta_1 + \delta) - \frac{(\theta_{min} + \theta_1)}{2}\,[/math]

This solves to:

[math]\theta_2 \le 2\theta_1 +4\delta - \theta_{min}\,[/math]

Putting these two together we have that:

[math]\theta_2 = 2\theta_1 +4\delta - \theta_{min}\,[/math]

Or more generally:

[math]\theta_j = 2\theta_{j-1} +4\delta - \theta_{j-2}\,[/math]

There are two boundaries where [math]\theta_n = \theta_{max}\,[/math] and [math]\theta_0 = \theta_{min}\,[/math], which can be used together to solve the general solutions for [math]\theta_j\,[/math]:

[math]\theta_j = \frac{j}{n}\theta_{max} +\frac{n-j}{n}\theta_{min} -j(n-j)\delta\,[/math]

Considering that [math]\theta_1 \gt \theta_{min}\,[/math] it must be the case that:

[math]2n(n-1)\delta \lt \theta_{max} - \theta_{min}\,[/math]