The Complexity of Strategic Behavior in Primary Elections

We study the computational complexity of strategic behaviour in primary elections. Unlike direct voting systems, primaries introduce a multi-stage process in which voters first influence intra-party nominees before a general election determines the final winner. While previous work has evaluated primaries via welfare distortion, we instead examine their game-theoretic properties. We formalise a model of primaries under first-past-the-post with fixed tie-breaking and analyse voters’ strategic behaviour. We show that determining whether a pure Nash equilibrium exists is $Σ_2^{\mathbf P}$-complete, computing a best response is NP-complete, and deciding the existence of subgame-perfect equilibria in sequential primaries is PSPACE-complete. These results reveal that primaries fundamentally increase the computational difficulty of strategic reasoning, situating them as a rich source of complexity-theoretic challenges within computational social choice.

💡 Research Summary

The paper investigates the algorithmic difficulty of strategic voting in primary elections, a two‑stage process in which parties first hold intra‑party primaries and then the winners compete in a general election. The authors formalise a model based on first‑past‑the‑post (FPTP) voting with a fixed tie‑breaking order. Each voter’s strategy consists of a primary ballot for every party (or abstention) and a general‑election decision function that maps every possible profile of party nominees (the “finalist” tuple) to a chosen candidate or to abstention. Participation costs are incurred for each primary and for the general election, and a voter’s utility is the utility from the eventual general‑election winner minus all incurred costs.

Two central decision problems are defined. Best‑Response asks, given a full strategy profile of all other voters, to compute a strategy for a distinguished voter that maximises her utility. Equilibrium asks whether there exists a pure Nash equilibrium (PNE) – a profile where no voter can improve her utility by deviating. The analysis first treats the simplest timing where all primaries are simultaneous, and later extends to a sequential setting where primaries occur one after another.

For the simultaneous case the authors show:

• Best‑Response is NP‑complete. The search space grows as the product over parties of (|A_k|+1) possible primary votes, which is exponential in the number of parties. A reduction from SAT demonstrates NP‑hardness, while verification of a candidate response is polynomial, establishing NP‑completeness.
• Existence of a PNE is Σ₂^P‑complete. By encoding the problem as an ∃∀ quantified Boolean formula, the authors map the existence of a profile where each voter’s best response condition holds to the Σ₂^P class. This lifts the complexity one level above the NP‑complete equilibrium existence results known for direct (single‑stage) elections.

The paper then studies sequential primaries, where the outcome of each party’s primary can affect later parties’ primary strategies. In this richer extensive‑form game, the authors consider the existence of a subgame‑perfect equilibrium (SPE). They prove that deciding SPE existence is PSPACE‑complete by a reduction from the Quantified Boolean Formula (QBF) problem. The construction treats each party’s primary result as a Boolean variable and forces players to make optimal moves in a game tree that mirrors the quantifier alternation of a QBF instance.

Technical contributions include:

• A careful treatment of the representation of the general‑election decision function g_i, distinguishing between explicit tables (size Θ(m^p)) and succinct encodings (e.g., preference lists) that allow O(1) queries. The complexity results hold under both representations, though the input size differs.
• An explicit enumeration algorithm for best‑response that runs in O(n p + (m+1) p (m p + n τ)) time, where τ is the query time for g_i, showing polynomial‑time solvability when p and τ are constants.
• Proofs that the multi‑stage nature of primaries directly corresponds to additional quantifier alternations, thereby explaining why equilibrium existence climbs from NP to Σ₂^P and why SPE existence reaches PSPACE.

The discussion highlights the practical implications: primary elections increase the cognitive and computational burden on voters, potentially deterring strategic manipulation but also making the system more opaque. Policy designers can use these results to weigh the trade‑offs between openness (e.g., open vs. closed primaries) and the tractability of strategic reasoning. The authors suggest future work on restricted settings (single‑primary participation, bounded numbers of candidates, specific cost structures) where the complexity might drop, as well as empirical studies to measure how real voters cope with the identified computational challenges.

In summary, the paper establishes that introducing a primary stage transforms the strategic landscape of elections, elevating the computational complexity of best‑response computation, Nash‑equilibrium existence, and subgame‑perfect equilibrium existence to NP‑complete, Σ₂^P‑complete, and PSPACE‑complete respectively. This positions primary elections as a fertile source of complexity‑theoretic questions within computational social choice and algorithmic game theory.