Sorry it took me so long: Latin America and rapid governments’ response to COVID-19

Wen facing the COVID-19 pandemic, what was key to governments’ response velocity throughout Latin America? The region had more information on what to do to prevent the disease from spreading itself and social isolation was the most recommended measure to avoid contamination. Still, Latin American countries varied greatly on how fast they adopted strict social isolation measures. We deploy an explanatory work on which institutional designs collaborates with higher delay in governments’ adoption of these measures. Among the institutional variables considered, we find that our variable of interest (delay) correlates strongly and positively with democracy, negatively with concentration of power, and positively with GDP per capita. These might suggest that autocrats faced less institutional and moral constraints to act, while democratic leaders dealing with pluralism and accountability faced higher costs to implement such measures. Due to the small sample, we next investigate some countries’ experience looking for examples for the found correlations.


Introduction
Latin America had a rare comparative advantage in learning how other regions dealt with the novel coronavirus before the pandemic hit the continent hard. But we should ask ourselves if Latin American leaders managed to take advantage of this fortune. Some countries in the region took up to 20 days between the first COVID-19 confirmed case in the country and the adoption of strict social isolation measures, while others took less than a week to do so. Why is that? What was fundamental to a quick government response in adopting the kind of measures recommended by the World Health Organization (WHO)?
In everyday politics, time is already one of the main obstacles faced by the government in policy implementation. During a pandemic, this variable is even more acute. Early action was among the most given advice on how to face the virus, even before its arrival in Latin America (Cheong, 2020;Pueyol, 2020;Wei, 2020). The reason is simple and easy to follow: the novel coronavirus has the ability to rapidly spread in a sustained manner across the population (CDC, 2020), but there is a gap between the true number of cases and the officially recorded cases. After the first case is confirmed, there is a possibility that the virus is already out there in many unaware citizens. Rapid action from governments even when there wasn't a great count of cases could reduce the spread of the virus. Although the spread of the disease followed a similar path in different countries, containment policies had a great impact on slowing the disease (Baldwin, Di Mauro, 2020). Pueyol (2020) brings evidence from China that social distance is the best way to reduce the number of cases since it does not allow people to interact with each other. In South Korea, the president once said that there was no reason to restrain daily routines due to the fear of the virus, and shortly after came to the large-scaled group infections outbreak in a couple of cities in the central part of the country (Cheong, 2020). According to a World Bank's (2020) study, contention measures to deal with COVID-19's spread in the population were more effective when earlier applied in 25 middle-income countries. This evidence is for both strict measures and loose ones. So, to act rapidly, not mattering how, was key to slow down the virus.
We understand that in some cases leaders lacked the will to act, adopting a discourse that reduced the virus mortality potential or even embracing negationism. This was the case for Nicaragua's dictator Daniel Ortega, Brazil's far-right president Jair Bolsonaro and Mexico's leftist president Manuel Lopez Obrador. But mostly, we understand that political institutions operate an important role in allowing rapid government's response in some countries, while not in others. In an emergency or crisis like the coronavirus' pandemic, the capacity and incentives for the government to adopt the necessary measures rest on some institutional features like concentration of power, plurality, and accountability. This is an exploratory work that seeks to identify what kind of institutional arrangements are associated with delay in Latin American governments' adoption of strict social isolation measures, hence we do not have a testable causal hypothesis 4 . We have a few reasons to focus our effort on Latin America. First, we believe this is a region that was lucky in having more time and information on how to deal with the virus, and therefore we can presume that all considered countries in the analysis had the same amount of information to make their choices to act or not. Second, when we focus the analysis in one region, we control for regional specificities, not only at the time of arrival of the virus but mainly constitutional similarities (Corrales, 2018). When we compare political institutions across Latin America, we hope to compare different degrees of similar institutions, most of the time. And third, Latin America is a region where you can find, according to V-Dem's Regimes of the Worlds classification, liberal democracies (Uruguay and Chile), electoral democracies (Argentina, Brazil, Colombia, Mexico, etc) and electoral autocracies (Honduras, Nicaragua, Bolivia, and Venezuela), a miscellaneous of regimes that add diversity to our comparations. In the same way, we have also different kinds of populists dealing with the virus in the region, from left to right, from elected in free and fair elections to term-limits-expanded-reelection ones. If populists are distinguished to be less pragmatic, they can be identified as outlier's performance in a task to compare institutional designs.
We have a few assumptions to address beforehand. We consider that when facing the novel coronavirus' pandemic, the adoption of social isolation measures is the optimal outcome of every responsible leader, guided by science and technical orientation. Some leaders waited longer to act in this sense, even having the institutional capability to do so since the adoption of such measures is unpopular and harmful to the economy. Although some countries had more to lose than others, in terms of economy's size, we should consider that all presidents in this analysis had some concern on the side effects of this measures, and the variance in the concern, which collaborates in the delay's variance, should rest upon political institutions' abilities to reward or punish government performance. That said, since they roughly prime the same outcome and face the same dilemmas, we can compare governments' delay and address this variance to different incentives and containment drawn by political institutions.
To this end, this work is structured as follows. In the first section, we go to the literature on government's response and efficiency to natural disasters and political crisis, emergencies that most resembles COVID-19 urgency, and on public goods provision looking for clues on which institutional designs can provide more efficient responses. Next, we present the data and the analysis we ran to address the relationship between governments' delay and some political institutions. In the third and last section, we build on the mechanisms that might strengthen and illustrate the found correlations.

Political Institutions during an emergency
In order to assess which institutional designs collaborate or embarrass an agile government response, we should go to the literature looking for what has been stated for government action during emergencies. We understand this by accomplishing that the kind of challenges that governments faced when hit by the pandemic was similar to a catastrophe situation, a "big, complex and both difficult and different" (Handmer, Dovers, 2007, 45). Lack of time, a high degree of uncertainty, and the capacity to act properly are constraints faced by decisionmakers both in corona times (Furman, 2020) and catastrophe times (Handmer, Dovers, 2007;Fischhendler et al., 2012), the literature suggests. But since those decision-makers operate in an institutional framework, the handling of the crisis can only be as good as the institutional design allows it to be (Boin et al., 2005;Handmer, Dovers, 2007;Fischhendler et a, 2012).
An important institutional variable that we must first consider is regime type. Democracies and autocracies have different incentives and constraints to deal with crises, given the different types of bargaining between leaders and citizens (Desai et al., 2009). It is demonstrated elsewhere that concerning the time they took to act, autocracies and democracies did not differ significantly, for a global sample (Lins et al., 2020). This do not mean, thought, that there aren't different mechanisms operating whitin these regime, as Pulejo and Querubin (2020) demonstrated that a closer election was associated with milder forms of lockdown. Simmilarly, Frey et al. (2020) found that autocracies impose stronger lockdowns since they don't have to worry much with electoral consequences of this policies. Empirical evidence suggests, however, that in general democracies can deal better with crises than autocracies (Rodrik, 1999).
Also, the public policy process differs. In democracies, the process is accountable to different constituencies, can be corrected through elections, and goes through several veto players. On the other hand, autocracies have an opaque process that only responds to the needs of an elite. Those features affect the speed of the process: slowing down in democratic countries and seeping up in the autocratic ones (Li, 2010;Alon et al., 2020).
Another major determinant for rapid action in abnormal situations is collaborative governance (Berteli et al., 2018). A higher coordination level between agencies, authorities from different government spheres, and involved actors in general, such as private sector and non-governmental organizations (NGOs) can be a great strategy to deal with problems (Ansell, Gash et al., 2008). But the stakes are high, and information is scarce (Handmer, Dovers, 2007), so to coordinate the efforts in a similar way is a particular institutional challenge. As Fischhendler et al. (2012) point, decentralization both increase private sector involvement in the emergency response and can enhance efficiency, since local authorities are closer to the disaster and could enjoy some autonomy to make decisions and take actions (Gopalakrishnan, Okada, 2007). However, decentralization can also lead to a delay in changing the status quo or even preventing such a change (Daehler, 2014).
Although this discussion is legitim, for the COVID-19 pandemic this was a no dilemma. Centralization and enhancement of the national government were key in most of the countries (Stephens, 2020;Edgell et al. 2020). This is due to the nature of the problem the virus presents to countries: a widespread borderless problem, a whole nation problem, or a whole-society problem (Handmer, Dovers, 2007). In this kind of situation, when the crisis hit areas that expand through multiple administrative jurisdictions, responsibility tends to shift to the major authority (Boin et al., 2005). There is an almost consensual understanding in this literature that every catastrophe is unique in a series of ways since disasters are abnormal by definition, policy and institutional responses will vary considering not only the government regular orientation but also the given challenge that the specific tragedy presents to authorities (Boin et al., 2005;Handmer, Dovers, 2007;Fischhendler et al., 2012).
Although decentralizing institutions exist to deal with territorial and socio-cultural differences (Oates, 1999), coordination between governments may be necessary to reach a common goal in some policies (Bakvis, Brown, 2010). Therefore, during the COVID-19 pandemic, coordination between national and subnational levels was crucial for efficient response. This places the pandemic in a different kind of situation when compared to a geographically located catastrophe that is better managed by decentralization, where local authorities or expert agencies lead the handling of the crisis (Boin et al., 2005). We can consider that an autonomous bureaucratic body that controls its own budget has a greater capacity to act (Fischhendler et al.,2012), and if this autonomy also means that there is no need to additional approval to implement its substantive decisions, this autonomous body is more likely to produce responses, as it was drawn from the catastrophe literature. However, we should also pinpoint that evidence from public goods provisions' literature suggests that decentralization is not linked to government efficiency (Treisman, 2000(Treisman, , 2002. Therefore, ultimately, federalism levels, subnational government's autonomy, and alignment between national and local authorities should become investigation variables for us. Also, the discussion presented in the catastrophe literature between a state or market-centered handling of the crisis (Handmer, Dovers, 2007) was completely absent in this pandemic. When facing the novel coronavirus pandemic, governments worldwide usually took the measures to asset a wartime situation in order to properly deal with it. Concentration of power and decision making, expansion of budget and expenditures, Ian Rebouças Batista, Amanda Domingos e Rodrigo Lins and bringing the state back in the scene (Stephens, 2020) were the consequences of this wartime method. We should investigate, then, in what ways the degree of centralization of decision-making correlates with the government's delay.
The ability to reach a national decision certainly depends on the necessity that a government has to reach a consensus before making a decision. Therefore, levels of democracy and a multiplicity of veto points might be an issue (Tsebelis, 2002). We have evidence from financial crisis literature that veto players and close-by elections are a complicating factor for sharp political action (Ha, Kang, 2015), in a way that puts pluralism and coordination dilemmas at the core of fast decision-making variables.
At the same time, in the universalism and public good providing literature, we found that a more institutionalized political party system can help authorities to overcome veto points and coordinate action toward public goods and welfare policies (Rasmussen, Knutsen, 2019). The mechanism is that linkages between society and political parties should make the latter more accountable to the former's demands (Kitschelt, 2000;Samuels, Zucco, 2015). Similarly, we also know that the intensity of electoral competition influences social policy reforms since strong opposition creates incentives to universalistic and public goods provision (Prible, 2013). This way, we should also explore if political parties' institutionalization is somehow linked to faster decision-making concerning strict social isolation measures.
In the liberal world, such policies are not only linked to public provision willingness of authorities or institutions' capacity but are subject to deeper concerns in policy-making calculus. Strict social isolation measures are unpopular and possess a harmful side-effect on the economy, as it was highlighted publicly by some presidents 5 . The situation is indeed complex: politicians should keep the population at their homes, curbing their liberties and hurting the economy in order to fight an invisible enemy. In an electoral regime, all kinds of arguments should populate public opinion trying to take electoral advantage supporting or criticizing the studied-to-be-adopted measures. The greater the pluralism, the further away from the consensus. We should, finally, examine the relationship between pluralism and delay in Latin American governments' actions to fight the novel coronavirus.

Data and analysis
In this section, we present our delay variable and the institutional variables we used to run the correlations 6 .
First, some remarks concerning the creation of our delay variable. We identified in Oxford's database on COVID-19 cases and deaths (Roser et al., 2020) the date of the first confirmed coronavirus case in each country considered in the database. Then, we identified when the government adopted social isolation measures at least at the second level of stringency, according to Oxford's COVID-19 Government Response Tracker database (Hale et al., 2020) 7 . The difference between those days is our delay variable. In Table 1 you can find the summary of this variable, both for the world and for Latin American countries. Now, we wanted to demonstrate that the adoption of social isolation measures is a matter of learning. In other words, the countries that were hit first took longer to adopt such policies because they did not know exactly what to do. After a few weeks and months, social isolation started to be higlighted as the best way to avoid contamination and spread of the virus. That way, the countries where COVID-19 arrived later could adopt these measures in less time. Considering this, we can divide the world sample by when COVID-19 arrived at the countries. We created a categorical variable wave after the identification of for how long since the countries have had the first confirmed case of the novel coronavirus. Those who had confirmed cases first are in the wave "1" category, first wave countries; and those who had confirmed  cases later on are in wave "2" and "3", second and third waves countries. Figure 1 8 shows the distribution of the variable that counts how many days since the first confirmed case 9 , and the red lines indicates the waves categories.
With the countries divided into waves, we can plot the difference of the means of the delay between each group (Figure 2), considering only countries that at some point in time adopted what we are considering here as strict social isolation measure (dropping the cases that simply did not act). It is evident now that countries who were hit by the pandemic later knew what to do and were faster in adopting strict social isolation measures. We ran an analysis of variance (ANOVA) for these variables and the results show that these differences in means are statistically significant, as we can see in Table 2. Now we can focus on Latin American countries, considering that they all belong to the second wave group who knew better what they should do to avoid the COVID-19 spread (compared to wave 1 countries, at least). Figure 3 presents the delay for each Latin American country considered -Nicaragua is not in the Figure since it did not adopt such measures.
The first variable to be addressed here is the regime type, using V-Dem's Regimes of the World measure, a categorical variable that considers that a regime can be a closed autocracy (0), an electoral autocracy (1), an electoral democracy (2), or a liberal democracy (3). In our 16 Latin American countries sample, we have two liberal democracies, eleven electoral democracies, and three electoral autocracies. As the boxplot presented in Figure  4 suggest the electoral autocracies (Venezuela, Honduras and  Table 2: ANOVA between delay and wave Source: Authors. 8 This graph is the histogram for the distribution of the "start" variable, which indicates how many days the country is living with confirmed cases of COVID-19. At the same time, this variable allows us to identify the day when the first case was confirmed in a particular country. Countries with similar values of "start" had the first confirmed case of the disease on the same day. On the x-axis in Figure 1 are all the 136 days we consider, and at the y-axis the number of countries with the first COVID-19 case confirmed on each day. This way, each bar in the graph presents the number of countries with the first case confirmed on that particular day. And this is why we can divide the x-axis into three waves. 9 Our threshold is June 30. Countries who confirmed COVID-19 cases after this date are not in the sample. From now on, we followed Osborne and Overbay (2004) demonstrations and suggestion to transform outliers of the delay variable in the sample for a more honest estimate of population parameters. We transformed the outliers values by the mean, 18,19 days, (see also Barnett and Lewis, 1994), what means we transformed two delay values, Brazil's 69 days and Dominan Republic's 47 days, when estimating for the coorelations below.
So, when plotting the correlation between delay and polyarchy, we found another suggestion that democracy, pluralism, and diversity in the society might be linked to delay in the governments' response. We used V-Dem's polyarchy index and found a statistically significant (p-value < 0.01) 0.656 positive correlation with delay ( Figure 5 10 ), and that for social science could be considered as "strong" (Cohen, 1988). Two might be the reasons for this relation. A democracy, per definition, has a higher amount of veto players and institutional constrains that permeate the struggle to adopt and implement any policy. It is costlier to adopt policies in terms of coordination. We should also consider the dilemmas that elected politicians confronted when deciding whether to adopt or not social isolation. Of course, it is easier to implement a social movement restriction policy in a country where freedom is not institutionally imbricated. Another thing is to suggest such a measure, that will profoundly hurt the economy, in an electoral regime, where the president and governors might be punished by the popular vote for such consequences. There is, then, also a moral cost implicating in delay.
To check if concentration of power gives another perspective to the democracy and delay relation, we run a correlation between delay and V-Dem's presidentialism index, which ultimately measures concentration of political power in the hands of one individual ( Figure 6). In Latin America, presidentialism is a constant, so it is a privileged region to access this association. We found a negative correlation of -0.579, statistically significant (p-value < 0.01). So, we are primed to consider that more veto players, more deliberation and more horizontal accountability might have delayed governments' adoption of the most recommended measures to fight COVID-19. We also run correlations between delay and division of power index (division between central and subnational governments) from V-Dem's data set and a variable for party alignment between the national and the subnational governments, collected by ourselves 11 . These results were not statistically significant nor strong, but their signal  10 When looking at the Figures 5, 6 and 7, please note that Brazil's and Dominican Republic's delay score correspond to the adjusted value (18,19), not the real delay of these countries, since we used these adjusted values for the correlations tests. 11 Alignment between sub-national (state or provinces) governments and the national government in the beginning of 2020. This variable was measured as a ratio of how many subnational governments are aligned with the president out of the total number of subnational entities. By alignment we understood governor's partisanship to parties from the national ruling coalition. For those countries without elected state or province governors, the regional administrator is appointed by the president and therefore this variable score 1.0 in such cases. In the replication material of this work you can find the values for this and all other variables considered here in our data set -with the sources we consulted for construction of variables such as alignment and majority. was as expected, positive (0.25) for the federalism index (more federalism correlates with more delay since federalism means another sphere of action that is prone to coordination problems) and negative (-0.29) for alignment (more alignment correlates with less delay since the coordination problem between this spheres is somehow diminished).
At this point, we tested the correlation between delay and GDP per capita, trying to assess how the economy was con-sidered as a determinant in the strategic calculus operated by the goverments when deciding to act or not. This might operate in two possible ways: as a moral dilemma, if one considers that the size of the economy and the richness of a population were correlated with more delay, suggesting that those dilemmas were higher when the stakes were higher; or as an electoral concern, signifying that the leader himself had more to loose if social isolation was adopted and the economy was damaged. We found a positive, statistically significant (p < 0.01), and strong correlation of 0.663 (Figure 7). It seems that the richer the country, the higher the delay.
We ultimately checked if political parties' institutionalization correlates with our delay variable. We used V-Dem's party institutionalization index, wich measures several institucionalization components such as level of organization, links with the society, supporters within the electorate and legislative discipline. As more institutionalized political parties demonstrate a higher level of commitments to society (Rasmussen, Knutsen, 2019), an institutionalized party system should offer a greater level of authorities' responsiveness to a pandemic. The result was weak and not statistically significant, but the signal was positive (0.41).
As we all know the mantra, correlation it is not causation. If we want to stand our suggestions in a steadier ground, we should employ other methods to address causation. With a small/medium sample, this is not an easy task. In the next section, we dive into some countries of our sample looking for examples of those statistical correlations in the national experiences. We still won't be able to address causation, but we hope to nourish the previous exposed correlations and strength these suggestions.

Cases
In this session, we will proceed with the case studies in three groups: democracy and its correlation with delay, autocracy and its correlation with quick response, and populism and denial. We selected Latin America's most emblematic cases of each group to provide a cross-national approach.

Chile, Uruguay, and Dominican Republic
From what we got from the data a high polyarchy index seems to be positively correlated with delay in government action for social isolation. We suggested that this might be the case as a democracy features more constraints to action, more veto players in the decision-making process, requirement of coordination within the governing coalition (sometimes even with the opposition), and politicians subject to electoral accountability. These elements might collaborate with the delay to adopt social isolation measures, a policy that it is not easy to make considering the economics' costs attached to it. In Latin America, Chile and Uruguay are the region's only two liberal democracies, scoring in the polyarchy index, 0.773 and 0.858, respectively. Chile took 20 days after the first case was confirmed to adopt a strict social isolation policy, while Uruguay took 18. The mean for the region is 18.19 days.
The beginning of 2020 brought mixed feelings to Chilean democracy. The country experienced massive social protests in the last months of 2019 that resulted in a scheduled referendum on the necessity to implement a new constitution, expected to happen in April. The highly contested and unpopular president, Sebastian Piñera, together with the Congress and the Supreme Court, are leading the Constitutional Convention that postponed the referendum to October. Chilean political establishment and institutions are severely hurt after last year's protests, with a 2% trust in the political parties and a 3% trust in the Congress, a public opinion survey revealed 12 . In this scenario, with the pandemic arrival, Piñera had to lead a nation that did not recognize him as a leader anymore. Only 12% of the population approved his administration at the end of February 13 .
The first confirmed case in Chile was on March 4th, while the social isolation measure was only implemented on March 25th following a State of Constitutional Exception decreed by the government on March 18th. This kind of reaction was common across democracies from all over the world. Great powers are therefore requested to deal with great and urgent challenges, indicating that democracies usually do not provide leaders with the appropriate tools to act effectively in such situations. Piñera's capacity to act was priorly condemned not only by popular antipathy but also to institutional constraints such as being a minority leader in the Congress. Any policy proposed in mid-March would have to be discussed in the prism of the calculus that balances social isolation and economic consequences. The minister of health, Jaime Mañalice, was one of the government's voice that publicly disclaimed the "many adverse effects" of social isolation 14 . Since we also found a positive correlation between delay and economy's size, it is appropriate to address that Chile is the first Latin American country in terms of GDP per capita. We should, therefore, conclude that the virus got Chilean government thinking if it worth the shot, to stop the country and somehow cool down protests that were once again striking against the state at the cost of the economy. Eventually, they acted. Better later than never.
Two weeks after president Lacalle Pou took office in Uruguay, the country had its first coronavirus case confirmed, on March 15 th . The left coalition Frente Amplio ruled the country for 15 years and Pou was elected under an economic and security platform. This shift of power between opposite sides was celebrated as being as democratic as it could be. Such transition, however, was not the most efficient. Fifteen days in office and the president had not yet appointed all his staff when he had to deal with the pandemic. Reports on the transition process describe some difficulties within the elected coalition concerning negotiations for positions, unfamiliarity with ministerial structure, and delay in some designation process that needed Senate approval 15 . Such negotiations and bureaucratic processes for nominations are attributes from democratic institutions that you cannot found in more authoritarian regimes. It is democracy's way to discuss as it is democracy's way to put positions under horizontal accountability.
In Uruguay, 14% of the population is over 65 years old 16 . An aged society like this was particularly worried about COV-ID-19 consequences, considering that elderly people are in the 12 "Desafios del processo constituyente chileno". Retrieved from: https://www.celag.org/desafios-del-proceso-constituyente-chileno/. Visited in July 1st, 2020 13 "Aprobacion de Piñera briones". Retrieved from: https://www.cnnchile.com/pais/cadem-febrero-2020-aprobacion-pinera-briones_20200224/. Visited in July 1st, 2020 14 " El ministro de salud de Chile rechaza la cuarentena por el coronavírus, 'el destino és que toda la población mundial se infecte, a menos que exista uma vacuna' ." Retrieved from: https://www.infobae.com/america/america-latina/2020/04/15/el-ministro-de-salud-de-chile-rechaza-la-cuarentenapor-el-coronavirus-el-destino-es-que-toda-la-poblacion-mundial-se-infecte-a-menos-que-exista-una-vacuna/. Visited in July 1st, 2020 15 "Entre el ajuste y el vírus: primer més de Lacalle". Retrieved from: https://www.celag.org/entre-el-ajuste-y-el-virus-primer-mes-de-lacalle/. Visited in July 1st, 2020 16 "Sociedade envejedica busca hacerle frente a pandemia". Retrieved from: https://www.clarin.com/mundo/coronavirus-uruguay-sociedad-envejecidabusca-hacerle-frente-pandemia_0_HL7uFmUjN.html. Visited in July 1st, 2020 risk group of the disease. Uruguay's doctor's Union suggested a total quarantine of the population on March 17th 17 . Besides that, the first official policies did not consider strict stay at home requirements. Schools were suspended, agglomeration was prohibited, and those who could were designated to telework. The government took those measures in the first days of the outbreak in the country, most of the time coordinated with the opposition 18 . Strict social isolation was only adopted on April 2 nd , still not as strict as other countries. The great debate during the delay in adopting these policies was whether it would be a total and mandatory or partial and volunteered quarantine, considering the economic costs. These two visions were defended, each of them, by government and opposition 19 , with the government much more reluctant to adopt strict isolation. We cannot say the government was paralyzed until the adoption of such a policy, this is not our suggestion. What it seems to us is that consideration and coordination, two elements more likely to be found in democracies than autocracies, takes time. Uruguayan case illustrates that even with opposite political forces agreeing in the macro policies, a lot still has to be discussed before a complex measure is adopted.
Although it is not a role model of Latin American democracy, Dominican Republic experience has to be shortly addressed here (polyarchy index, 0.598). It took 47 days to the government to adopt a strict social isolation measure after the first confirmed case of COVID-19 in the country. The president had the majority in the Legislative (his party has 76% of the seats in the Senate 20 ) and the sub-national governments are not elected but appointed by the Executive. One could say that he was a strong president with institutional support to act. Why did he take so long?
Probably because 2020 is an electoral year in Dominican Republic, with presidential elections initially scheduled for May 15 and postponed to July 5 21 . Landman and Splendore (2020) list several possibilities for the pandemic to make elections unfeasible and, therefore, to be a risk to democratic rule itself, including low turnout, term's expansion, and compromised steps in the electoral cycle, as voter's registration processes and campaigns. But a close-by election also affects the government's behavior, particularly those concerning economic performance (Drazen, 2000). Since voters end up evaluating the government only by the last year of economic performance, as a cognitive short-cut (Healy, Lenz, 2013), it becomes urgent to the incumbent to deliver a safe economic policy considering his prospects to the election. Pulejo and Querubin (2020) already found some evidence that incumbents who have a coming election implement less strict policies dealing with the pandemic. In these cases, we have to considerer that the dilemma between isolation and economy comes with different weights with a close-by election. The president does not want to be punished in the ballot by the economic downturns caused by the quarantines or curfews. In the Dominican Republic case, the government first adopted flawless isolation measures, not paralyzing industrial and agricultural sectors for example 22 . Electoral accountability emphasized the dilemma that governments faced, therefore consisting of another democratic attribute that collaborated with delay to adopt isolation measures.

Venezuela and El Salvador
On the other face of the coin, Venezuela, the lowest polyarchy index of the region, 0.229, was the only country to be ahead of the disease, in the sense that it adopted strict social isolation measures two days before the first confirmed case of COVID-19 in the country. Nicolás Maduro, the Bolivarian president, announced on March 13 a State of Alert, increasing his powers even more than regular, since Venezuela has already the second most powerful president in the region, according to the presidentialism index, only slightly behind Nicaragua's dictator Daniel Ortega -Venezuela has a 0.963 index and Nicaragua 0.976. We found a somehow moderated and negative correlation between this index and delay. According to V-Dem's codebook, this index measures the extent to which the president is free from constraints by other political actors and institutions. In practice, it means that the more power the president concentrates, the easier it is to implement the complex and debatable social isolation policies.
As an electoral autocracy, Venezuela might have an institutional facede that resembles a liberal democracy, but its institutions have been deturped in favor of the ruling elite (Corrales, 2020). In this scenario institutions of representation are just another kind of domination's institutions. Subnational governors, for example, can serve to delegate power to agents aligned with the government than to effectively share power (Schedler, 2013). In 2020, 19 of the 23 state governors in Venezuela were aligned to Maduro. This is a context were unconformity is unlikely. Thus, Frey et al. (2020) demonstrated that autocratic regimes impose more stringent lockdowns during the COVID-19 pandemic, indicating that autocracies overcome democracies' institutional barriers and dilemmas, while not necessarily resulting in efficient achievement of policies objectives.
El Salvador did not start the year 2020 considered as an autocracy. In fact, it is considered an electoral democracy, with a 0.631 polyarchy index and 0.298 presidentialism index, both for the year 2019. Less than a year in power, though, Nayib Bukele started an executive aggrandizement in 2020, invading the Congress accompanied by the military in February 24 and repeatedly defying the Supreme Court 25 . Bermeo (2016) appoints that to weaken constraints on the executive power and to strategically harass and manipulate opposition are the most common tools to erode democratic regimes in this century. Coppedge (2019) similarly suggests that the power that an autocratic executive enhance comes at the expense of courts and legislatures. On March 19, El Salvador had the first confirmed cases of COVID-19. On March 21, Bukele announced a State of Emergency enhancing his powers and adopting strict social isolation measures. What follows suggests that Bukele did not have the most democratic of the intentions.
The military was on the streets and containment centers were installed to arrest those accused of gathering and disrespect the curfew. These centers were accused to violate human rights, without basic hygiene conditions nor adequate care considering the pandemic. When the Supreme Court ordered the release of those prisoners without coronavirus symptoms, Bukele did not obey and claimed to be "the man in charge" after the opposition suggested that the pandemic measures should be widely discussed 26 . Bukele is an example of an autocrat that took advantage of the pandemic to erode democratic institutions. During a state of emergency, those institutions can be eroded while keeping constitutional legitimacy, the favorite strategy of the 21 st century autocrat. In fact, democracies are 75% more prone to erode while in a declared state of emergency (Luhrman, Rooney, 2020). During the COVID-19 pandemic, while the social isolation measures were the most recommended and suggested measures to avoid the spread of the virus, as it was highlight earlier, these policies could be used to violate fundamental rights and to severely hurt democracy (Edgell et. al, 2020). From the evidence discussed here, we have reasons to believe that this is the case for Bukele, an autocrat in the making. With enhanced powers and malice, he acted fast, in part because he could, in part because he needed and wanted to.

Negationists (Brazil, Mexico and Nicaragua)
Three Latin American leaders have been notorious deniers of the gravity imposed by the novel coronavirus. Presidents of Brazil, Mexico and Nicaragua tried to soften the impact that CO-VID-19 could have in the public health system and, ultimately, in the diseases' death toll. Perhaps the best illustration of this reasoning is the classification made by Jair Bolsonaro, president of 23 "Venezuela necessita una tregua política para lidiar con el coronavírus". Retrieved from: https://www.nytimes.com/es/2020/03/23/espanol/opinion/ venezuela-maduro-guaido-coronavirus.html. Visited in June 28, 2020. 24 "Presidente de El Salvador invade o Congresso com militares amparado por direito divino". Retrieved from: https://brasil.elpais.com/ internacional /2020-02-10/presidente-de-el-salvador-invade-o-congresso-com-militares-amparado-por-direito-divino.html. Visited in June 28, 2020 25 "El Salvador presidente defies supreme court". Retrieved from: https://www.hrw.org/news/2020/04/17/el-salvador-president-defies-supreme-court. Visited in June 28, 2020. 26 "Presidente de El Savador usa quarentana para radicalizar autoritarismo". Retrieved from: https://www1. folha.uol.com.br/mundo/2020/04/ presidente-de-el-salvador-usa-quarentena-para-radicalizar-autoritarismo-no-pais.shtml. Visited in June 28, 2020  Brazil when he stated that the COVID-19 was only a gripezinha ("minor flu") 27 . The three countries also have one more feature in common: their presidents' scores are high in a populist index. Figure 8 shows the values for all presidencies of the three countries, from 1998 to 2019: The data comes from the Global Populism Dataset, created by Hawkins and coauthors (2019). This dataset codes each president based on textual analysis (speeches, manifestos, political documents, etc.) and takes an ideational approach of populism. This means that every populist discourse will have "at least a discourse in which the putative will of the common people is in conflict with a conspiring elite" (HAWKINS et al, 2019, p. 2). In other words: Political leaders are positioned in a populist scale based on how much they stress the conflict between the common people and a corrupt elite.
While there is no necessarily link between populism and negationism, all cases of populist leaders in Latin America flirtedat least temporarily -with the negation of the pandemics' gravidity. But the the populist's behavior can be constrained or echoed depending on the political institutions in place. In countries where the concentration of power by the president is higher, the impact of negationism might be also stronger. On the other hand, where party institutionalization is strong and political power is shared with subnational entities, populists should be easily constrained.
The values from the Global Poulism Dataset range from 0 to 2, with the following cut-points: below 0.5, the leader is not populist; from 0.5 to 0.9, the politician is somewhat populist; from 1 to 1.49, the leader is populist; and above 1.5 it is considered very populist. As we can see from the figure, Bolsonaro is the only president from Brazil to be considered at least somewhat populist. The same can be said about Mexico's Obrador. And all three presidencies of Ortega are among the top-five, with his first presidency (2007)(2008)(2009)(2010)(2011)(2012) the most populist for the three countries for the period 28 .
Not surprisingly, the three countries are among those that waited the most to take action against COVID-19. Brazil is the country with the highest delay in the region: 69 days. Mexico is the fourth, with a delay of 29 days. Both are well over the mean for Latin America (18.19 days). But the most worrisome case is that of Nicaragua: until the consolidation of the dataset used in this work, the country had not taken a strict isolation policy. And according to a press release from early June, the president said that the country never will adopt such measures 29 . The link between populist leader and weak response has already been shown by Kavakli (2020). The main argument has two features. First, populists usually distrust elites, and scientists are no exception. Policy suggestions made by the scientific community fall in deaf ears. The second feature is what the author calls "economic conservatism". Populist leaders tend to balance the economic effects of the policy being implemented and underreact in the early days of the pandemic. However, once the diseases become more apparent, they may overcompensate (Kavakli, 2020).
All three leaders have, at some point, made use of dilemmas opposing the reaction to the novel coronavirus and the state of the economy or even civil liberties. In the case of Brazil, Bolsonaro made use of both. The Brazilian federalist arrangement tries to associate level of coordination and autonomy, which historically varies between states and specific policies (Arretche, 2004). During the novel coronavirus pandemic, governors and mayors had autonomy to lead efforts in fighting the disease. But this dynamic took on a new chapter: the conflict between subnational governments and national executive. Despite the 1988 Federal Constitution's tradition of cooperation, Bolsonaro ended up implementing a very centralized kind of federalism: (i) reinforcing the division of powers, at the same time as decreasing federal funding and transfers for inequality reduction; (ii) concentrating decisions that affect subnational units; and (iii) creating a constant struggle with governors and mayors, stirring his constituency against local executive governments. (Abrucio et al, 2020).
Bolsonaro also stated that "the cure cannot be worse than the disease itself" 32 . The argument is that keeping people at home would kill poorer people that usually work in the informal sector and do not have job security. Also, the slowdown of the economy would generate mass unemployment, thus creating a big problem to be solved post-pandemics. This was also the argument made by Obrador when reopening the economy on June 2nd. He stated that it was necessary to "normalize productive, economic, social, and cultural activities" 33 .
Another point in common for the three deniers of the region is the strong attack on the media. Common adjectives are "alarmist" and "irresponsible". When in a press conference in early June, Obrador was asked about the record-breaking number of 1,092 registered deaths in the previous 24 hours. In response, the Mexican president lashed out at the media, arguing that the numbers did not mean that 1,092 persons died, but only that their deaths were registered in that interval 34 . A similar approach was made by Bolsonaro. In the Brazilian case, however, the government officially tried to change the way figures related to the pandemic were publicized. From June 5th to June 9th, the government only released the number of confirmed deaths. The Supreme Court needed to give an order for the government to resume the release of numbers related to registered deaths 35 .
So far, we have seen that all three presidents had similar behavior -at least in the early days of the pandemic. And although ideologically very distant, Bolsonaro and Obrador have some similarities. Their presidencies began only one day apart (December 31 st 2018 for Obrador; January 1 st 2019 for Bolsonaro). That means that during the pandemic both of them were far from facing voters in the ballot box for reelection 36 . Both of them also adopted the strategy of minimizing the novel coronavirus. But their divergences are also relevant. Soon after shooting a video having breakfast in a restaurant to prove that the danger was minimal 37 , Obrador recognized the problem.
The institutional barriers they face are also different. While Obrador was elected with an overwhelming majority in the legislature, Bolsonaro is what can be called a minority president. In the 2018 election, his support group could not win the majority of seats. But it gets even worse: Bolsonaro had a strong fallout with the president of his party and abandoned it. Now, he is trying to create his own political association 38 .
Politically speaking, Nicaragua is a different kind of animal. It can be classified straightforwardly as an authoritarian country, or electoral autocracy according to the Regimes of the World classification. If not de jure, at least de facto. To understand the personality of the Nicaraguan president, he went missing for 34 days at the beginning of the pandemic outbreak 39 . The country not only did not adopt strict measures of social distancing but is also accused of hiding the real figures. According to the Pan American Health Organization (PAHO), the country is not informing the number of tests conducted.
Daniel Ortega is in his third consecutive term as president, fourth in total. He is governing with the assistance of his wife, Rosario Murillo, elected as his vice-president. Since 2018 the country is facing a political crisis, with 326 citizens killed in conflicts 40 . This standoff between government and opposition began in April 2018, when Ortega proposed a reform in the country's pension system. The opposition took the streets to protest and was firmly restrained by the police. This led the opposition to call for resignation of the president. Since them, negotiations started but it did not have success 41 .
We believe that a note on these three presidents' popularity is also relevant. In the case of Bolsonaro, the gap between those who disapprove and approve the president first widened in the early months of the pandemic. In April, 38% of respondents in a national survey disapproved the president; this number went up to 44% by June. The number for those who approve his presidency is stable: went from 33% to 32% 42 . However, this figures started to change from June onwards. According to a Datafolha poll conducted in mid-September, those who approve the government had a five percentage point increase, reaching 37%. An even greater change was seen in the disapproval rate: a ten percentage point decrease, going from 44% to 34%. The poll suggests a correlation between those changes and the distribution of a R$ 600 emergency aid. From those who collected the emergency aid, 42% approves the government of Bolsonaro. Among those who were not recipients of the emergency aid, this figure is 36% 43 . The value of such aid will be cut in half for the last for months of its vigency, starting in September. This could lead to a new shift in the sentiments of Brazilians toward president Bolsonaro.
The downfall of Obrador is still more sensitive. Two years ago, when he got elected president of Mexico, he had an approval rate of 77%. In April 2020, his approval was 68%. The latest poll shows approval of 56% 44 . Still a well-rated president, but it seems that COVID-19 is melting his numbers. In Nicaragua, there is a constant fear of reprisal. This could skew the poll. But the results showed that Ortega has the support of only 24% of the population 45 .

Conclusion
In this paper, we were concerned about exploring possible political institutions constraints to the Latin American governments' adoption of strict social isolation measures during the COVID-19 pandemic. We brought from the catastrophe, political crisis, and public goods provision literature some variables that are usually addressed for government efficiency discussions. Limited by our small sample, we used descriptive statistics to investigate which institutions were correlated with delay, and then we dived in some countries' experiences to further analyze the found relationships.
Our findings can be summarized as follows: (i) for the world sample, countries who experienced the novel coronavirus later had more information on how to deal with the pandemic and therefore took less time to adopt strict social isolation measures; (ii) democracies present some institutional constrains to presidential action that might have influenced governments' delay, including horizontal and electoral accountability; (iii) autocrats generally can act more freely, without checks and balances to their actions or worries with electoral costs, and therefore could adopt the complex social isolation measures more rapidly; (iv) the size of the economy can illustrate what was on stake during government's policy study and considerations; (v) populists clearly took their time before taking action. Following their anti-science rhetoric, populist leaders were less willing to sacrifice the economy to combat COVID-19.
One cannot abstract from this paper the idea that democracies are the worst regimes that misconducted the pandemic policies. Democracies, during pandemics or not, are still the regimes that consider civil liberties the most. So, democratic leaders think too much, reflect too much, consider too much. This is not necessarily a problem. Politics is complex and urge responsibilities.