Political Parties, Candidate Selection, and Quality of Government

In a recent paper, Cruz and Keefer (2010) study the role of programmatic parties (i.e., parties with a well defined ideology) on the effectiveness of countries to implement public sector reforms.

I discuss these arguments in more detail in Section 4.

In particular the relation between partisan support and primary adoption would be negative if improving selection is the main reason of using primaries, and positive instead if primaries are used to increase political competition and candidates’ pre-electoral incentives.

There is, of course, a large literature examining the effect of nomination procedures on other outcomes such as information dispersion and acquisition (Meirowitz 2005), legislators’ ability to compromise (Alvarez and Sinclair 2012), or party polarization and loyalty (Ansolabehere, Hirano, and Snyder 2007; Ansolabehere et al. 2010). Other papers focus on the sequentiality associated to the use of primaries, specially in the U.S. case. For example, Knight and Schiff (2010) explore the effect of sequential voting on social learning, while Klumpp and Polborn (2006) develop a model to examine the effect of primaries on the effectiveness of campaign spending.

For some recent evidence on the effect of democracy on growth and well-being see Barro (1996), Rodrik and Wacziarg (2005), Papaioannou and Siourounis (2008) and Kudamatsu (2012). The effect of electoral rules and form of government is thoroughly studied in Persson and Tabellini (2003) and Persson and Trebbi (2003).

For a more comprehensive discussion of the role of domestic political institutions on policies in Latin America, see Scartascini et al. (2008).

See Carey and Polga-Hecimovich (2006) for a more detailed discussion of primaries in Latin America.

The political science literature refers to this feature of selection methods as internal democracy or inclusiveness (Hazan and Rahat 2006).

The discussion of possible channels in this section is not exhaustive. I focus the attention on the main arguments discussed in the political economy literature.

This would depend of whether we use a standard Downsian electoral competition model or a citizen-candidate approach.

This period corresponds to the re-introduction of democratic elections in many countries, after failed military dictatorships in the 1960s and 1970s.

Hall and Jones (1999) use the average of these three indicators plus risk of expropriation and an index of government repudiation of contracts. Data on these indices are, however, available only until 1997. This reduces the sample size by half. Nonetheless, including this information for the period when it is available produces similar results.

The survey question is: “How much trust do you have in the President?” There are four possible answers: a lot, some, a little, or none.

Trust in president has been collected since 1997, while president’s approval has been collected only since 2002.

Table 8 in Appendix B. replicates the baseline results using annual data and clustering the errors by president’s term. This alternative procedure produces similar results.

The quinquennium indicators are associated to the year of the presidential election. I consider the following periods: 1980–1984, 1985–1989, 1990–1994, and so on. The last period is 2005–2008. The results are robust to the exclusion of this time trend.

As previously mentioned, a less demanding specification uses annual data instead of aggregated at term level. The results, shown in Table 8, are similar.

Interestingly, mandatory primaries seems to be negatively (though not significantly) correlated to quality of government. A possible explanation is that primaries are more democratic and competitive – and hence contribute more to quality of government—when they are voluntarily adopted instead of required by electoral legislation.

Note that higher values of the indicator of corruption in government represent a perception of less corruption.

I thank an anonymous referee for pointing out this issue.

When using observations aggregated at term level, the use of lags and forward reduces dramatically the number of observations.

The full regression results are available in Table 9 in Appendix B.

This is consistent with the effects lasting only during the term of a primary-nominated president. In the sample, the average duration of a president’s term is 4.2 years.

Including fixed effects produces positive point estimates, but the results are statistically insignificant (see Table 10 in Appendix B).

The results are similar using primary any other.

Note, however, that a similar result could be due to other reasons, such as overreporting of primaries when they are mandatory.

I focus on these two mechanisms because they have been explicitly incorporated in models of endogenous primary adoption, see for example Castanheira, Crutzen, and Sahuguet (2010) and Serra (2011). There are, however, other benefits associated to primaries that may explain their adoption. For example, Kemahlioglu, Weitz-Shapiro, and Hirano (2009) argue that primaries allow parties to solve internal conflict. In a related work, Hortala-Vallve and Mueller (2010) develop a model wherein parties adopt primaries to avoid costly party splits. Meirowitz (2005) presents a model in which primaries allow voters an early chance to reveal their preferences, while Jackson, Mathevet, and Mattes (2007) argue that primaries allow a party to credibly commit to more centrist policies.

To see the intuition, consider a case when candidates are of heterogeneous quality and their effort is fixed. In this scenario, only political selection matters, and parties may adopt primaries if they help them pick up the best politician and attract undecided voters. This benefit of primaries declines with partisan support, since there are fewer voters to attract. Hence, larger partisan support will reduce primary adoption. In contrast, if candidates are homogeneous but their effort levels are not fixed, only incentives would matter. In this case, if the cost of effort is borne by the candidate, then candidates’ optimal effort would be lower than the party’s. Primaries are beneficial to the party since they allow it to extract the maximum level of effort from candidates who are competing for the party nomination. This benefit increases with partisan support (i.e., less political competition) since in that case candidates have even lower incentives to exert effort.

See Table 11 in Appendix B for the full first-stage results. Also note that weak instruments are less of an issue in this case since the system is just identified, and hence 2SLS estimates are median unbiased.

Exploiting within party variation and a larger sample, Aragón (2013) finds a similar positive relation between partisan support and primary adoption.