Governator vs. Hunter and Aggregator: A simulation of party competition with vote-seeking and office-seeking rules

Abstract

The policy positions parties choose are central to both attracting voters and forming coalition governments. How then should parties choose positions to best represent voters? Laver and Sergenti show that in an agent-based model with boundedly rational actors a decision rule (Aggregator) that takes the mean policy position of its supporters is the best rule to achieve high congruence between voter preferences and party positions. But this result only pertains to representation by the legislature, not representation by the government. To evaluate this we add a coalition formation procedure with boundedly rational parties to the Laver and Sergenti model of party competition. We also add two new decision rules that are sensitive to government formation outcomes rather than voter positions. We develop two simulations: a single-rule one in which parties with the same rule compete and an evolutionary simulation in which parties with different rules compete. In these simulations we analyze party behavior under a large number of different parameters that describe real-world variance in political parties' motives and party system characteristics. Our most important conclusion is that Aggregators also produce the best match between government policy and voter preferences. Moreover, even though citizens often frown upon politicians' interest in the prestige and rents that come with winning political office (office pay-offs), we find that citizens actually receive better representation by the government if politicians are motivated by these office pay-offs in contrast to politicians with ideological motivations (policy pay-offs). Finally, we show that while more parties are linked to better political representation, how parties choose policy positions affects political representation as well. Overall, we conclude that to understand variation in the quality of political representation scholars should look beyond electoral systems and take into account variation in party behavior as well.

Fig 1. The effect of number of parties.

Note: Based on corresponding data mined OLS regressions with α at .5, discount factor at .5, and ideal point dispersion factor at 1. Grey shaded areas are 95% confidence intervals. A = Aggregator, S = Sticker, H = Hunter, G = Governator, SG = Satisficing Governator.

Fig 2. Effect of degree of policy-motivation.

Note: Based on corresponding data mined OLS regressions with 5 parties, standard ideal point variance, and discount factor at .5. Grey shaded areas are 95% confidence intervals. A = Aggregator, S = Sticker, H = Hunter, G = Governator, SG = Satisficing Governator. The x-axis shows the degree of policy-seeking of a party, where 0 is fully office-motivated and 1 is fully policy-motivated.

Fig 3. Effect of dispersion in parties’ ideal positions.

Note: Based on corresponding data mined OLS regressions with 5 parties and discount factor at .5. Grey shaded areas are 95% confidence intervals. Columns represent different levels of policy-motivation. A = Aggregator, S = Sticker, H = Hunter, G = Governator, SG = Satisficing Governator.

Fig 4. Boxplot of rule shares in simulated party systems.

Fig 5. Predicted eccentricity for rule shares compared to sticker by number of parties.

Note: Based on corresponding data mined OLS regressions with all parameters at mean value, and other rule shares kept at 0. Grey shaded areas are 95% confidence intervals. Columns represent the number of parties in the party system. A = Aggregator, S = Sticker, H = Hunter, G = Governator, SG = Satisficing Governator.

Fig 6. Predicted party system misery for rule shares compared to sticker by number of parties.

Note: Based on corresponding data mined OLS regressions with all parameters at mean value, and other rule shares kept at 0. Grey shaded areas are 95% confidence intervals. Columns represent the number of parties in the party system. A = Aggregator, S = Sticker, H = Hunter, G = Governator, SG = Satisficing Governator.

Fig 7. Predicted government misery for rule shares compared to sticker by number of parties.

Note: Based on corresponding data mined OLS regressions with all parameters at mean value, and other rule shares kept at 0. Grey shaded areas are 95% confidence intervals. Columns represent the number of parties in the party system. A = Aggregator, S = Sticker, H = Hunter, G = Governator, SG = Satisficing Governator.