The majority runoff rule

Since 1965, France has utilized the majority runoff rule to elect the president. Under such as rule, a minimum of a majority is required to win in the first round. If no one candidate receives at least a majority in that first round, then the top two finishers advance to a runoff election, where one of two candidates is guaranteed to obtain a majority. Today, majority runoff rules are common “in African countries under former French control,” in Latin American countries “in the most recent democratic period (starting with Ecuador in 1978), and in post-Communist regimes in Central and Eastern Europe since the 1990s” ([15]: 38).

As O’Neill explains, the majority runoff rule “serve[s] the seemingly noble purpose of electing a candidate with majority support, and many governments have enacted runoff voting in response to candidates winning with less than a majority of the vote” ([16]: 353). Shugart and Taagepera ([17]: 324) add that “[t]he appeal of majority runoff stems largely from a desire to avoid…the ‘Allende syndrome’: the election of a rather radical president by a narrow plurality of the vote, leading to a military dictatorship.” In Chile’s case, Allende only won roughly 36% of the vote. His unpopular presidency was then cut short when General Pinochet launched a successful coup d’état. To avoid another Allende (and the coup that followed), today Chile elects its president using the majority runoff rule. Interestingly, as part of a guided transition back to democracy, in some cases it is the “incumbent military rulers” that have introduced “majority runoff in the fear that a large but minority extreme party could take over power in an open election” ([18]: 92).

One of the downsides of the majority runoff rule is that “[h]olding a second election is expensive, the campaign season is longer, and voters must take the time to go to the polls a second time” ([16]: 352). In order to afford the longer campaign season, candidates have been know to resort to extreme measures. For instance, after failing to win a majority in the first round of the 1994 Colombian presidential election, “Ernesto Samper was accused …of having accepted money from the Cali drug cartel to finance his” runoff campaign ([19]: 158). Although he was ultimately acquitted, the scandal succeeded in damaging the legitimacy of his administration.

The plurality runoff rule

According to Pérez-Liñán ([20]: 144), for most majority runoff elections, “the president could have been elected by plurality rule without altering the final outcome.” As a result, some states have simultaneously sought to avoid the perils of plurality rule, while also avoiding the costs associated with the majority runoff rule. This has led some states to utilized the “plurality runoff” rule to elect the president ([21]: 103). Under such a rule, a qualifying plurality (such as 40%) is required to win in the first round. If no one candidate receives at least that qualifying plurality in the first round, then the top two finishers advance to a runoff election. To date, plurality runoff systems are only used in Latin America, although their popularity is beginning to spread, as some scholars argue that using a “40 per cent threshold has been an unalloyed success” ([22]: 133). Currently, Argentina, Bolivia, Costa Rica, and Ecuador all use a 40% threshold in some form. Nicaragua used to use a variation of a 40% threshold, but has since reverted back to plurality rule. Next I’ll review Argentina’s adoption of this system, as it is illustrative of the reform process many states underwent when adopting the plurality runoff rule.

Argentina’s 1853 constitution was “[l]ike many other Latin American constitutions during the nineteenth century…inspired by the precedent of the American constitution,” and as such, utilized an electoral college to indirectly election the president ([23]: 110). In 1994, Argentina amended their constitution to eliminate their electoral college and introduce a plurality runoff system. As Kamińksi ([24]: 84) notes, during times of reform, “the selection of voting procedures” is often the result of “a bargaining process between two” conflicting parties. In Argentina’s case, the plurality runoff rule was “thus created as a compromise between the plurality rule preferred by the ascending but not dominant PJ [Partido Justicialista or Justicialist Party] and the majority formula preferred by the declining UCR [Unión Cívica Radical or Radical Civic Union]” ([25]: 430). According to Negretto ([23]: 114), “[i]nitially, the UCR demanded a threshold of 50 per cent, while the PJ stuck to a floor of no more than 40 per cent…negotiators from both parties agreed to ‘split the difference’” and came up with a novel solution: a candidate needs a qualifying plurality of at least 45% to avoid a runoff, or they can also avoid a runoff by netting 40% of the vote and having at least a 10-point lead over the second-place candidate. Many scholars have praised Argentina’s plurality runoff rule; Tanaka ([26]: 123), for instance, has suggested that his native country, Peru, should consider adopting a similar rule for electing the president. However, one aspect of runoff rules that few scholars have considered is the relationship between runoff rounds and government respect for human rights [27].

Runoff rounds and human rights

Do democratic elections increase or decrease the likelihood that governments respect human rights? The answer to this question may seem intuitive, but the literature has thus far been mixed. On one side of the debate, scholars argue that elections empower a population vis-à-vis its government, which decreases the likelihood that a government will resort to repressing its own people [28–33]. This body of literature argues that by creating an “opportunity for citizen involvement in the political process,” elections “provide a motive for elected representatives to be [held] accountable to their constituents” ([34]: 205). Through elections, citizens “confront the controllers and supporters of sets of social arrangements that determine patterns of access to resources, services, status and power” ([35]: 6). Mill ([36]: 404) argues that “human beings are only secure from evil at the hands of others, in proportion as they have the power of being, and are, self-protecting.” One of the best ways to ensure that citizens are, indeed, ‘self-protecting’ is to given them in say in who presides over the nation by allowing them to “express a choice among the alternatives” ([37]: 71). As Steiner ([38]: 102) explains, “[b]y periodically subjecting elected officials to the approval of the electorate, [elections] help to arrest governmental violations of widely valued rights.” Thus, “[e]lections thereby serve a vital protective function” by “establish[ing] boundaries for governmental action” ([38]: 101–102).

On the other side of this debate, however, some scholars argue that elections increase political polarization by challenging the established order and threatening the political system, which can lead to an increase in human rights violations [39–44]. From this point of view, governing elites sometimes strategically violate human rights in an attempt to maintain stability and control [45–57]. This body of scholarship notes that “[m]any of the world’s longest-standing consolidated democracies, including France, the United Kingdom and the United States, have experienced periods of election violence” ([58]: 27). Indeed, while democracies are generally more inclined to protect human rights than non-democracies, even in democracies repression still occurs [59–68]. Hafner-Burton, Hyde, and Jablonski ([58]: 2) note that for many democracies, “[g]overnment-sponsored election violence—events in which incumbent leaders and ruling party agents employ or threaten violence against the political opposition or potential voters before, during or after elections—is common”.

Why are some democratic elections seemingly associated with repression, while others are not? I argue some of this variation can be explained by closely examining and comparing different types of elections. For instance, Richards ([69]: 648) initially found “that the presence of national elections, either executive or legislative, has no effect on government respect for human rights.” Interestingly, a few years later Richards and his co-author, Gelleny, re-examined the impact of national elections on government respect for human rights; this time, however, instead of lumping all national elections together (as had previously been common within the literature [70]), Richards and Gelleny disaggregated national elections as being either presidential elections (occurring within presidential systems) or lower-house legislative elections (occurring within parliamentary systems). In this reexamination, “lower-house national legislative elections were found to be associated with greater government respect for human rights, while presidential elections were associated with less respect for human rights” ([71]: 505).

By disaggregating ‘national elections’ as being either presidential elections in presidential systems or lower-house legislative elections in parliamentary systems, Richards and Gelleny [71] noticed that some elections appear to promote government respect for human rights, while others do not, which is something that Richards [69] was unable to see when he initially lumped all national elections together; such efforts at disaggregation could be taken further. For instance, Richards and Gelleny [71] lumped all presidential elections together, as have some subsequent scholars [58]. However, as I have alluded to above, not all presidential elections are the same; some occur over only one round, while those that utilize some variation of a runoff rule can potentially advance to a runoff round. As such, in a recent paper, I reexamined the impact of democratic presidential elections on government respect for human rights, but instead of lumping all presidential elections together (as Richards and Gelleny [71] did), I disaggregated all presidents as being elected under either plurality rule or a runoff rule. In the end, I found that “[i]n comparison to when the president is elected using plurality rule, when the president is elected using a runoff rule, that state is more likely to be associated with high government respect for human rights” ([14]: 3). Once again, however, I believe that such disaggregation could be taken further.

As is evident by my discussion above comparing the majority runoff rule versus the plurality runoff rule, not all runoff rules are the same. When comparing majority runoff and plurality runoff rules, Negretto ([23]: 116) notes that one of the main differences can “be found in the different incentives they provide for coalition-making among parties.” According to Riker ([72]: 33), candidates want to “create coalitions just as large as they believe will ensure winning and no larger.” This is because the more power a candidate negotiates away in order to get elected, the less power they will retain after the election. While the majority runoff rule “induce[s] different opposition parties to coalesce after the first round” in order to win in the runoff round, in plurality runoff systems, “opposing parties may have an incentive to coalesce before the election…to reach the minimum share” and therefore avoid a runoff round ([23]: 116). In other words, presidents elected after two rounds likely have had to broaden their coalitions more widely than those able to avoid a runoff round. I argue that this difference has human rights repercussions, as prior research has shown that “when presidents are able (or forced) to cobble together broad-based coalitions to win…their administrations are less likely (and less able) to violate human rights” ([73]: 1). To provide an illustrate on how this can play out, Freudenreich ([74]: 80) explains that in presidential systems, “the partisan composition of cabinets is largely predetermined by the bargaining and the competition before and during presidential elections.” Pérez-Liñán ([20]: 132) adds that being forced to advance to a runoff round encourages “the formation of inclusive electoral alliances before the second round,” which often forms “the basis for coalition governments.” Parallel to this, recent research has found “cabinets comprised of a higher percentage of individuals from parties other than that of the president to be associated with greater government respect for human rights” ([73]: 1). This is because when “ministers of an opposing party…find themselves in cabinet-level positions,” they “effectively wield a veto because if their president proposes a new policy (or the continuation of an old policy) that they disagree with, they can make the implementation of said policy much more onerous” ([73]: 2).

In sum, presidents elected after only one round likely have had to negotiate away less power, and thus have more power to potentially violate human rights, while presidents elected after two rounds likely had to broadly their winning coalition, oftentimes by negotiating away powerful cabinet-level positions, thus making it harder for such administrations to repress without consequence. As such, with this paper I seek to extend my previous finding that a runoff rule promotes government respect for human rights by focusing entirely on those countries that utilize a runoff rule, and disaggregating among those presidents elected after two rounds versus those elected after only one round. This leads me to my hypothesis:

I argue that country-years where the president has been elected after two rounds are more likely to be associated with high government respect for human rights, in comparison to country-years where the president has been elected after only one round (that could have advanced to a runoff round, but did not).

Sample

As my just stated hypothesis suggestions, this article seeks to ascertain whether presidents elected after a runoff round versus those elected after only one round (that could have advanced to a runoff round, but did not) are more likely to be associated with high government respect for human rights. Ultimately, the purpose of this article is to provide a follow-up to a previous article[14] where I argue that a runoff rule produces presidents more likely to be associated with high government respect for human rights, in comparison to those elected under plurality rule. In order to build upon my previous empirical findings, my sample for this follow-up article is modeled off that which was used in my previous study—i.e. presidential democracies with ‘democracy’ defined using Cheibub, Gandhi, and Vreeland’s Democracy versus Dictatorship (DD) dataset [75]. Although some human rights studies have opted to use either the Freedom House or Polity measures to identify which regimes are democratic, this is controversial as both Freedom House and Polity base their classification—in part—on how regimes respect human rights; therefore, using either measure would partially control for my outcome variable [76]. Poe and Tate ([59]: 856), argue that democracy “must be defined in terms that allow independent operationalization of the concept,” and in light of this advice, I (and many others [14, 27, 73, 76–84]) have opted to use Cheibub, Gandhi, and Vreeland’s [75] typology, as their DD dataset defines democracy in a way that does not incorporate state human rights practices. Per the DD dataset, a regime is considered to be a ‘democracy’ when the president is elected, the legislature is elected, there is more than one party competing in elections, and an alternation under identical electoral rules has taken place ([75]: 69).

Dependent variables

For my primary dependent variable, I follow my previous article [14] in utilizing the Cingranelli-Richards (CIRI) Physical Integrity Rights Index [85], which is an additive nine-point index of four ordinal indicators of government respect for physical integrity rights: the rights of all human beings to be protected from torture, extrajudical killing, disappearance, and political imprisonment. CIRI scores ranges from ‘0’ (no respect for any of the four physical integrity rights) to ‘8’ (full respect for all of them). Whereas my previous article [14] only examined data up through 2011, thanks to the CIRIGHTS Data Project [86], CIRI scores are now available up through 2017.

Unlike my previous article [14], for this follow-up article I also utilize an alternative dependent variable: the Political Terror Scale (PTS), which—like CIRI—is coded using data from the US Department of State’s Country Reports on Human Rights Practices and Amnesty International’s Annual Report [87]. While CIRI scores are determined by individually evaluating instances of torture, extrajudical killing, disappearance, and political imprisonment (then adding together all four constituent scores), PTS scores are determined by collectively evaluating the range of the population effected by instances of torture, extrajudical killing, disappearance, and political imprisonment. States are designated a level ranging from ‘1’ to ‘5’: ‘1’ indicates that the state is under a secure rule of law, people are not imprisoned for their views, torture is rare or exceptional, and political murders are extremely rare; ‘2’ indicates that there is a limited amount of imprisonment for nonviolent political activity, torture is exceptional, and political murder is rare; ‘3’ indicates that there is extensive political imprisonment, and political murders are common; ‘4’ indicates that disappearances, torture, and political murders are all common, though state terror only affects those who interest themselves in politics; finally, ‘5’, which indicates that state-sanctioned repression has been extended to the whole population, and state leaders place no limits on the means or thoroughness with which they pursue personal or ideological goals [88].

While CIRI scores seeks to reflect “actual government practices” ([89]: 406), PTS scores aims to reflect “the ‘range’ of violence committed” ([88]: 368). Despite these differences, however, “PTS and CIRI essentially measure the same thing” ([90]: 88). As such, scholars that employ one of these indices as their dependent variable often report analogous estimations using the alternate index as a robustness check [27, 91–93]. In order to aid in comparability to CIRI, I have followed the literature’s trend by inverting PTS scores, such that higher scores now indicate greater government respect for physical integrity rights. Note that all summary statistics are presented in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary statistics.

https://doi.org/10.1371/journal.pone.0243094.t001

Independent variables

In order to test my hypothesis, I have constructed the following independent variable: president elected after two rounds (vs. one round). To construct this variable, I consulted version 3.0 of Bormann and Golder’s Democratic Electoral Systems Around the World dataset [94, 95]. For each country-year, I looked up the most recent democratic election which brought the current president to power. If that election advanced to a runoff round, my president elected after two rounds (vs. one round) variable was coded as ‘1’. Otherwise, if the election that brought the president to power concluded after only one round, my variable was coded a ‘0’. As was the case for my previous article [14], all country-years where the president was not directly elected were omitted from my analysis. Note that for country-years that are election years, I have coded that year based on which president (i.e. either the outgoing or incoming) presided over the majority of that year.

Beyond my president elected after two rounds (vs. one round) variable, I include control variables based on those that were used in my previous article [14]; these variables take into account executive constraints, the level of civil conflict, population size, gross domestic product (GDP) per capita, and finally the previous year’s level of government repression. Since Poe and Tate’s defining study [59] (and as a result of later extensions of that study [64, 68]), it has become common practice within the human rights literature to include the types of control variables that I just mentioned. Indeed, within the literature it has now become common to refer to a model that includes these types of control variables as simply an “‘off-the-shelf’ Poe and Tate model” ([96]: 663); with this article, I seek to add to this legacy. At this point, I will discuss some of the specifics of my particular ‘Poe and Tate model’.

To begin, while many Poe and Tate models have included Polity IV’s XCONST variable [97] as its ‘executive constraints’ control [58, 71, 98], this “measure was designed to estimate executive constraints across all countries,” and as such, “it does not fully capture the variation within specifically presidential democracies” ([99]: 8). As such, I follow my previous article [14], as well as other recent articles [81, 99], in using Doyle and Elgie’s ([100]: 734) quantification of “the constitutional power of presidents,” which I refer to as presidential power. Values range from ‘0’ to ‘1’ with higher values indicating more powerful presidents.

Next, Poe and Tate ([59]: 859) note that “regimes are more coercive when they are involved in civil conflict.” As such, I follow my previous article [14] in including a measure of civil conflict, which is drawn from version 19.1 of the UCDP/PRIO Armed Conflict Dataset [101, 102]. This variable is coded as ‘0’ for each country-year with less than 25 battle-related deaths, ‘1’ for each country-year where there were between 25 and 999 battle-related deaths, and finally ‘2’ for each country-year where there were more than 999 battle-related deaths.

Continuing on, Poe and Tate models typically include measures for population size and GDP per capita, the former of which seems to be negatively associated with high government respect for human rights, and the latter of which seems to be positively associate with high government respect for human rights. My population size and GDP per capita measures both come from the World Bank [103], and as is common practice, both have been logged to correct their distributional nature.

Finally, Poe and Tate models often include a lagged dependent variable, as government repression in one year seems to influence government repression in subsequent years. However, “[b]ecause the CIRI [and PTS] variables are non-linear, a simple lagged dependent variable is less appropriate because it does not efficiently model the autoregressive trend in the data” ([104]: 12). As such, I follow Hafner-Burton [105]: 615) (and others [27, 68, 73, 99, 106]) in including a series of binary variables representing each level of the corresponding lagged dependent variable “to account for dependence across the categories of the dependent variable over time”.

Results

In Table 2, you can see two ordered probit models that estimate CIRI scores and PTS scores in presidential democracies. Following the human rights literature [14, 27, 73, 81, 84, 99, 107–109], both of these models have robust standard errors clustered by country to address heteroscedasticity and the pooled nature of the data. As detailed above, these two regressions are modeled off the regressions I used in an earlier article [14], so as to build upon previous empirical findings; however, note that instead of examining all presidential democracies, to test my hypothesis, I have limited my sample to democracies that utilize either a majority or plurality runoff rule. Also note that unlike my previous article [14], for the analysis in this article, I utilize PTS scores as a check for robustness. Finally, note that whereas my previous article [14] only examined data from 1990 to 2011, this article examines data from 1990 to 2017. The additional 6 years are a result of recently released updates to both human rights datasets.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Ordered probit estimates of CIRI scores and PTS scores in presidential democracies, 1990-2017.

https://doi.org/10.1371/journal.pone.0243094.t002

Recall that my hypothesis is that country-years where the president had been elected after two rounds are more likely to be associated with high government respect for human rights, in comparison to country-years where the president had been elected after only one round (that could have advanced to a runoff round, but did not). As you can see in Table 2, my president elected after two rounds (vs. one round) variable is found to be positively associated with both high CIRI scores and high PTS scores; these relationships are statistically significant at least at the 95% level. Consistent with previous human rights scholarship, across both models all control variables are statistically significant and have signs pointing in the expected direction. For instance, presidential power, civil conflict, and (logged) population size are all found to be negatively associated with both high CIRI scores and high PTS scores, while (logged) GDP per capita and all of the binary variables for each level of the corresponding lagged dependent variable are found to be positively associated with both high CIRI scores and high PTS scores.

While the results reported in Table 2 are promising, the substantive effects of these models can be better illustrated through an analysis of predicted probabilities. As such, in Table 3, I report predicted probabilities (and 95% confidence intervals) for CIRI scores and PTS scores when the president is elected after two rounds versus one round. Note that these probabilities were estimated using the Clarify software package [110] and are all based upon each control variables’ mean (or mode in the case of categorical variables) for all country-years in my dataset; these values can be seen in the bottom third of Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. The percent change in predicted probabilities of CIRI scores and PTS scores when the president is elected after two rounds vs. one round.

https://doi.org/10.1371/journal.pone.0243094.t003

Starting with CIRI scores—which are reported in the top-half of Table 3—you can see that the probability of a CIRI score of ‘8’ (i.e. the highest possible score for government respect for human rights) is 0.093 when the president is elected after one round. In comparison, note that the probability of a CIRI score of ‘8’ is 0.145 when the president is elected after two rounds. Note that the difference in the probability going from 0.093 to 0.145 is 0.052, which—as you can see—is statistically significant at least at the the 95% level, given that the corresponding 95% confidence interval (which is in brackets) does not overlap with zero. Substantively, increasing a 0.093 probability by 0.052 is roughly a 56% increase. This means that for a given year, the ‘average’ state in my dataset (i.e. a state whose parameters match the mean/mode of my control variables, which can be see in the bottom third of Table 1) is roughly 56% more likely to be at the highest level of government respect for human rights when the president has been elected after two rounds versus one round.

Continuing on, you can see that a CIRI score of 7 corresponds with a positive percent change (i.e. roughly 11%), while lower scores CIRI scores (such as ‘4’ through ‘6’) all correspond with negative percent changes. Together, this suggest that a state is more likely to be at higher levels of government respect for human rights and less likely to be at lower levels of government respect for human rights when the president has been elected after two rounds versus one round. In sum, consistent with my hypothesis, presidents elected after two rounds appear to be better at protecting human rights than presidents elected after one round (that could have advanced to a runoff round, but did not). Moving to the bottom half of Table 3, you seen that this trend is robust to an alternate indicator of human rights, i.e. PTS scores.