Decentralization and ideology

Decentralization and ideology (DOI: https://doi.org/10.1017/psrm.2025.13) was written by Anna M. Wilke, Georgiy Syunyaev, and Michael Ting in 2026. It was published in Political Science Research and Methods (vol. 14).

Decentralization is often justified on principles:

• U.S. states as 'laboratories of democracy'

• adapt laws to local circumstances
• etc

The authors demonstrate that decentralization can also be explained ideologically. They develop a game theoretic model.

• two periods: t in {1,2}

• four players:

  • two (prospective) executives j in {I, O}

    • Incumbent of first period is I.

    • Preferred policy positions as z_j. Assume that z_O = -z_I.

  • two local governments i in {1,2}

    • Preferred policy positions as y_i. Assume that y₂ = -y₁. Let y = [y₁, 1₂].

• There are two states for a locality i: centralization and decentralization. Centralization is separable across localities, and may be reset between periods, so the centralization profile is notated as c_i,t. Let c_t = [c_1,t, c_2,t].

• Under decentralization (c_i,t = 0), in each period, policies are chosen by the local governments in each period as x_i,t. Let x_t = [x_1,t, x_2,t].

• Under centralization (c_i,t = 1), in each period, policies are chosen by the executive.

• Between periods there is an election.

  • I remains executive in period 2 with probability p; O becomes executive with probability (1-p).

  • Furthermore, let q be the probability that centralization can be reset. Perhaps the probability that the executive is weak in period 2.

Clearly under decentralization the local governments will select x_i,t = y_i. And clearly in a 1 period game the executive will select full centralization. (It follows that in two periods, the second period executive will select full centralization, too.) The policy that would be selected is not necessarily obvious however.

• Local governments care about policies in their own locality.

  • -(x_i,t - y_i)²

• Local governments also care about policies in the other locality. Let gamma be the weight given to own locality policies and 1-gamma be the weight for other locality policies.

  • Assume gamma is not zero.

  • For locality 1, altogether: U₁(x_t | y₁) = -gamma(x_1,t - y₁)² + -(1-gamma)(x_2,t - y₁)².

• Executives care about policies in all localities equally.

  • -Sigma_i(x_i,t - z_j)².

• Executives also care about local welfare. Let omega be the weight given to policies and 1-omega be the weight for welfare.

  • Assume omega is not zero.

  • Altogether U_j(x_t | y, z_i) = -omega Sigma_i(x_i,t - z_j)² + (1-omega) Sigma_i U_i(x_t|y_i).

• Altogether, executives do not necessarily select x_i,t = z_i.

Continuing in the 1 period game case:

• Polarization is measured as |z_I| (since symmetry is assumed).

• Under low polarization, O prefers full centralization; under high polarization, full decentralization.

• By making simplifying assumptions about the preference directions of localities, the authors demonstrate that I and O prefer centralizing 'enemy' localities over 'ally' localities.

Now in the 2 period game case:

• As noted above, the second period executive will select full centralization.

• If I is uncertain about re-election, they may select decentralization.

• Under low polarization, full centralization is still selected.

• Under moderate polarization, optimal strategy depends on rigidity (q). With high rigidity, an 'ally' locality can remain decentralized even if O wins the election.

  • Under high polarization, it may even be preferable to select full decentralization, i.e. give up control over 'enemy' locality policy.
  • By contrast, with low rigidity, 'locking in' decentralization for one or both localities doesn't make sense.

CategoryRicottone CategoryReadingNotes