Kawai and Watanabe (2013) Motivation Understanding how voters choose is key to understanding electoral outcomes Study Background 2005 legislative elections in Japan → 480 representatives elected, one per district Focus on 159 districts that satisfy the following : • Elect representative with first-past-the-post (FPTP) rule • Have 3 or 4 candidates running that are part of a subset of the 4 major parties • District has at least 2 municipalities • No missing data Theory Voting Models Expressive Voting → Vote for preferred party • Voter i has to choose among K parties or candidates • Derives utility 𝑢𝑖𝑘 from candidate k being elected • i’s vote is determined by max 𝑘 𝜖 𝐾 𝑓(𝑢𝑖𝑘 , 𝑝𝑘 ) Strategic Voting → Voting decision takes into account pivotality • Voter i holds prior beliefs over each parties’ chance at winning, 𝑝𝑘 • i’s vote is determined by max 𝑘 𝜖 𝐾 𝑓(𝑢𝑖𝑘 , 𝑝𝑘 ) or • Voter i holds prior beliefs over when her vote will be pivotal, µ • Vote for candidate that yields higher utility weighted by the scenarios where the voter is pivotal Literature • Voters consistently overestimate their likelihood of being pivotal → experience increases the voters’ accuracy • 30% to 50% of voters act strategically • Older and richer voters vote more strategically Identification • Model where there are strategic and sincere voters • Voters hold beliefs about when their vote could be pivotal • Combine with data on legislative elections in Japan • Exploit variation within electoral district to estimate % of strategic and sincere voters Identification • We do not observe the distribution of preferences or voters’ beliefs • Identification by exploiting cross-municipality variation within district Example • Three candidates → cA (liberal), cB (moderate) and cC (conservative) • Two districts → D1 mostly conservative and D2 mostly liberal • Take a municipality from each district → 𝑚11 (liberal) and 𝑚12 (liberal) • If there are no strategic voters ⇒ Similar vote shares in 𝑚11 and 𝑚12 • If there are strategic voters ⇒ More votes for moderate candidate in 𝑚11 Empirical Design Model • K ≥ 3 candidates • M municipalities with an electoral district • Finite number of voters ∑𝑀 𝑚=1 𝑁𝑚 < ∞ Voter n derives utility 𝑈𝑛𝑘 from candidate k being elected : 𝑈𝑛𝑘 = 𝑢(𝒙𝑛 , 𝒛𝑘 ) + 𝜉𝑘𝑚 + 𝜀𝑛𝑘 • 𝒙𝑛 → set of voter characteristics • 𝒛𝑘 → set of candidate characteristics • 𝜉𝑘𝑚 → candidate-municipality shock • 𝜀𝑛𝑘 → independent and identically distributed random variables (iid) shock Voters Sincere voters → vote for k if and only if 𝑈𝑛𝑘 > 𝑈𝑛𝑙 Strategic voters → chooses vote considering only the events in which she is pivotal • 𝑻𝑛 = {𝑇𝑛,𝑘𝑙 }𝑘𝑙 are voter n’s beliefs that there will be a tie or one candidate will be only 1 vote behind the other between any two candidates • Expected utility from voting for candidate k is • 𝑈 ̅𝑛𝑘 (𝑻𝑛 ) = 1 ∑𝑙 𝜖 {1,…,𝐾} 𝑇𝑛,𝑘𝑙 (𝑈𝑛𝑘 − 𝑈𝑛𝑙 ) 2 • Vote for k if and only if 𝑈 ̅𝑛𝑘 (𝑻𝑛 ) ≥ 𝑈 ̅𝑛𝑙 (𝑻𝑛 ) 𝛼𝑚 probability of voter n being strategic in municipal m 1 − 𝛼𝑚 probability of voter n being sincere in municipal m Assumptions • Beliefs are common to all voters within an electoral district • A voter is more likely to be pivotal with parties that have a higher vote share • Voters of both types always vote for the candidate that maximizes their utility and not someone else Estimation • Assume the following functional form for the voter’s utility 𝑈𝑛𝑚𝑘 = −∥ 𝜃 𝐼𝐷 𝒙𝑛 − 𝜃 𝑃𝑂𝑆 𝒛𝑃𝑂𝑆 𝑘 ∥ +𝜃 𝑄𝐿𝑇𝑌 𝒛𝑄𝐿𝑇𝑌 𝑘𝑚 + 𝜉𝑘𝑚 + 𝜀𝑛𝑘 𝒙𝑛 → years of education, income level & binary variable indicating if n is 65 or older 𝒛𝑃𝑂𝑆 𝑘 → candidate’s party and economic ideology 𝑄𝐿𝑇𝑌 𝒛𝑘𝑚 → candidate’s political experience and whether m is her hometown 𝜃 𝑃𝑂𝑆 → ideological position of the voter 𝜃 𝑄𝐿𝑇𝑌 → non-ideological component of the voter’s preferences • Estimate preference parameters θ • Construct counterfactuals when changing voter characteristics → sincere voting under plurality rule Data Voting data at the municipality level Candidate characteristics from the Asahi-Todai Elite Survey (ATES) Demographic characteristics at the municipality level Electoral forecasts at the district level from 2 newspapers Key Findings Very high share of strategic voters (compared to previous findings) • Authors argue that is because previous papers only focused on misaligned voting (strategic voters that do not vote for their preferred candidate) Large proportion of strategic voters: 63.4% to 84.9% Interpretation Results would change if all voters would vote sincerely and plurality rule adopted However, this paper does not consider turnout