I. INTRODUCTION

Why is it relevant to identify the circumstances under which countries drastically change their foreign policies? First, despite their brevity, foreign policy shifts (FPSs) have historical importance because their results and effects are long-term, as they are generally followed by long periods of stability. Also, FPSs have a significant impact on international relations, as they have a disruptive effect on the balance of power and/or result in alliance changes. Finally, FPSs in Latin America have historically represented the alternation between pro and anti-US foreign policies, with a direct impact on the hegemon's hemispheric interests, in some cases leading to a change of posture towards the region.

This article aims to identify historical conjunctures that led Latin American and Caribbean countries to restructure their foreign policies, conceptualized here as FPS), which are defined as the most extreme foreign policy change (FPC) processes. The analysis seeks to identify common domestic and international conditions that led leaders to implement FPSs.

Few studies have systematically investigated the reasons for FPC. ^{Volgy and Kenski (1976}, ¹⁹⁸²) developed a quantitative analysis regarding cases of “distance change” – a conceptual cousin of FPC. ^{Huxsoll (2003)} made the most significant effort to empirically study the effects of domestic variables on FPC processes, despite not investigating how these processes worked, and only seeking to identify differences in the way in which variables impact the degree of FPC. ^{Mattes, Leeds, and Carroll (2015)} showed that, in autocracies, as leaders aiming to stay in office tend to pursue foreign policies that favored their core societal support groups, changes in the domestic sources of support for leaders had a direct impact in United National General Assembly (UNGA) voting patterns.

Most qualitative studies of the phenomenon analyze decision-making dynamics to investigate FPC in unique case studies, with the exceptions of ^{Holsti (1982)} and ^{Welch (2005)}, whose comparative work used specific criteria to look for patterns, and represents a significant advancement in the study of FOC by relating specific conditions with outcomes. Case studies have used the concept of a “policy window” – developed by ^{Kingdon (1984)} and brought to FPC studies by ^{Gustavsson (1998)} – to study FPC processes (^{Checkel 1993}; ^{Gorjão 2005}; ^{Ziv 2011}; ^{Doeser 2011}; ^{Doeser and Eidenfalk 2013}). The literature identifies several relevant contextual conditions as being adequate to initiate processes of FPC, such as: economic or political crisis (^{Gustavsson 1998}; ^{Hay 1999}); political failure (^{Walsh 2006}, ^{Welch 2005}); critical junctures (^{Hogan 2006}; ^{Soifer 2012}); and regime and leader change (^{Hagan 1989}; ^{Hermann 1990}; ^{Gustavsson 1998}).

Seizing these opportunities, policy entrepreneurs¹ act to place new ideas on the agenda and push towards an FPS. On the one side, ^{Gustavsson (1999)} states that, during policy windows, the domestic and international environments lead decision-makers to reevaluate policy guidelines. On the other, ^{Doeser and Eidenfalk (2013}: 393) argue that ^{Gustavsson (1999)} wrongly interpreted the concept of a policy window, and that policy entrepreneurs already have reformist policy ideas in mind before the window ever opens.

According to the literature, the exceptional character of policy windows guarantees the relative urgency in FPS processes, by lowering political costs and providing legitimacy. During these periods, the effects of foreign policy stabilizers² are softened, facilitating the implementation of policy changes by political actors. This also has the impact of concentrating decision making, and loosening the restraining effects of standard bureaucratic operational procedures.

Therefore, while qualitative case study analyses lack substantiated and replicable selection criteria (^{Rynhold 2007}; ^{Yang 2010}; ^{Eidenfalk 2009}; ^{Doeser 2011}; ^{Ziv 2011}; ^{Doeser 2013}; ^{Doeser and Eidenfalk 2013}), quantitative frameworks have failed to create a specific concept-related measurement of FPS (^{Huxsoll 2003}; ^{Mattes, Leeds and Carroll 2015}). Despite presenting a convincing measurement metric for FPC, ^{Volgy and Kenski's (1976}; ¹⁹⁸²) concept of “distance change” is too complex and diverse to be widely used. The present research fills this gap by using a concept definition that follows established case selection criteria that are replicable across a wide range of countries and time periods. A comparative framework helps to characterize FPS processes, permitting a further understanding of how crises, critical junctures, or regime/leader changes contribute to FPS. Finally, it also contributes to two additional areas: the use of UNGA voting record as a tool to observe FPS, discussed in the following section; and the theoretical discussion on Latin American foreign policy, which will be discussed in the second section.

Using this research design, it is possible to identify two causal paths that led to FPS. By one account, growing discontent with longstanding dictatorships leads to political polarization and subsequent succession crises, including civil wars and social unrest, from which new regimes/leaders emerged, reversing existing policies. By another, international isolation worsens economic conditions, causing leaders to implement administrative reforms and alter their foreign policy by conceding to pressures from major powers.

Patterns in the outcomes are also identified. On the one hand, the failure of economic policy programs led to a strategy of accommodation after leaders, bending to international pressure which had severely impacted their countries' economies, implemented administrative reforms and changed foreign policy guidelines. On the other, succession crises – whereby international and domestic actors did not successfully mediate the emergence of moderate political groups – led to two different outcomes: a) revolutionary political groups, which substituted former political adversary after civil wars, adopted independent foreign policies; b) international-friendly regimes established by international military interventions adopted pro-Western foreign policies. Previous political polarization seems to have had an important role in determining the extremity in FPS.

The article is structured as follows: in the first section, methodological matters are addressed, the concept definition and criteria for selection are presented, and the dependent variable is operationalized. The following section presents the cases and discusses the relevant Latin American foreign policy literature. Next, the environmental conditions under which FPS occurs are described and patterns identified; conjectural conditions are then related to typologies of FPS. In the fourth section, each of the FPS typologies are illustrated using the case studies. Finally, the concluding remarks discuss contributions to the literature.

II. DEPENDENT VARIABLE AND CASE SELECTION TECHNIQUE

The focus of this research is on cases with an extreme value in the dependent variable, conceptualized as an FPS, understood as an FPC transition from one pole of the political spectrum to the other. For these cases, the principle of “conceptual opposites” is applied, establishing that the meaning and measurement of the concept and its opposite are not symmetrical (^{Goertz and Mahoney 2012}: 164). These cases were considered to be qualitatively different from the others, which, according to ^{Ragin (1987}: 22), justifies special attention with regard to these data points, despite their relative infrequence.

Even though in quantitative studies bias occurs when the selected cases do not represent the total population, resulting in inferences that are invalid and not generalizable, in qualitative research designs, cases may be deliberatively selected by the researcher based on a common outcome aiming to identify which variables are necessary and/or sufficient for the occurrence of the outcome. As discussed by ^{George and Bennett (2005}: 24-29), the main problem of case selection in qualitative studies occurs when the selects specific cases to sustain a particular hypothesis.

To avoid this problem, objective selection criteria were defined, allowing a “blind” selection of cases. This prevents the use of questionable justifications aiming to use cases to prove a theory. The main criteria for the inclusion of an observation as an occurrence of an FPS are: a) speed; b) thematic comprehensiveness; c) radicalism. To be considered a shift, a case of FPC needs to be: abrupt – that is, to happen suddenly with implementation beginning in a short period of time; widespread – involving the most thematic areas in foreign policy; and radical – the largest displacement in the political spectrum. All these characteristics need to be present for a case to be considered as an FPS.

Therefore, a numerical threshold was defined to select cases, with two objectives in mind: to avoid deliberately selecting certain cases; and to select the most extreme cases of FPS using a common index applicable to a large sample. To define the threshold, first, an FPC variable index was calculated as follows: a country's UNGA ideal point location for a session in year_t was subtracted from the same country's UNGA ideal point location in session in year_t-1. From this variable, the standard deviation was calculated.³ Whenever an observation presented an FPC index that surpassed four standard deviations, the case was considered as an FPS. The threshold of four standard deviations has the objective of selecting cases that most closely represent processes of FPS (and its extremity): the largest variation on a country's UNGA ideal point should represent the criteria of quickness (from year to the next), radicalism (largest displacement in political spectrum) and thematic comprehensiveness (as the UNGA sessions deal a wide range of issues if the international agenda).

Unfortunately, this technique also hampers a comparison with negative outcome cases, as corresponding criteria for selecting negative cases are not available; since the universe of negative cases encompasses 1676 cases, with a wide numerical (and theoretical) variation among them, it was not possible to select negative cases. ^{Mahoney and Goertz (2006}: 240) argue that, as in qualitative research designs, cases are selected based on a phenomenon of interest, positive outcomes rarely occur (e.g. wars, revolutions), while negative cases are almost infinite (e.g. nonwars, nonrevolutions). The only precaution needed is to avoid inferring about standard behavior in foreign policy (continuity, or negative outcome), focusing on developing theoretical claims exclusively on deviant behavior (FPS, or positive outcome).

The FPC variable has a mean of −0.025 and the standard deviation is 0.191; among 1688 observations, only 12 presented such an extreme value, around 1%, while the remaining 99% represent the remaining variance along the foreign policy spectrum. ^{George and Bennett (2005}: 24-29), debating the selection of cases based on extreme values of independent or dependent variables, suggest that the extremity of the case may be defined based on the mean and standard deviation of the variable in question.

Graph 1 presents the distribution of FPC indices using a histogram, showing the relative infrequence of cases that exceed four standard deviations from the variable mean. The thresholds for a case to be considered FPS were −0.789 and 0.739, and the FPC indices vary from −1.507 to 1.396.

Graph 1 Histogram of FPC indexes 

UNGA voting records have been widely used by scholars working in the IR literature and attend to the theoretical criteria previously explained. It is assumed here that UNGA voting records represent a complete foreign policy spectrum, and that countries' international behavior is generally represented by their political positioning in that arena, as is broadly used in the International Relations literature.

^{Bailey, Strezhnev, and Voeten (2017)} point out that 75 articles have used this indicator as a variable to represent states' international behavior. It has traditionally been used as a proxy to identify a country's foreign policy behavior change (^{Vengroff 1976}; ^{Hagan 1989}; ^{Huxsoll 2003}; ^{Ratner 2009}; ^{Mattes, Leeds and Carroll 2015}). Recently, a growing body of work has applied it to analyze Latin American countries' foreign policies, advancing that field of study in the region (^{Amorim Neto 2012}; ^{Mourón and Urdinez 2014}; ^{Amorim Neto and Malamud 2015}); this article contributes to this discussion by speaking to the causes of changes in patters of foreign policy behavior.

The ideal points measure was chosen as it uses the complete voting record,⁴ and ^{Bailey, Strezhnev, and Voeten (2017}: 13-23) confirm its validity in three ways: (1) by comparing the scores with historical knowledge; (2) by relating to agenda changes in the UNGA; and (3), by associating the scores with domestic events. Originally developed by ^{Poole and Rosenthal (1991}, ²⁰⁰¹), ^{Bailey, Strezhnev, and Voeten (2017}: 1) estimated these points according to a country's position in roll-call voting for each UNGA annual session. Ideal points estimate states' positions regarding the liberal order, representing a measure of political proximity regarding to most important representative and leader, the US.

III. THE “INVESTIGATION DOMAIN”

The sample was restricted to Latin American and Caribbean countries, from 1945 to 2008, totaling 33 countries and 1688 observations. This boundary establishment to case selection is called by ^{Berg-Schlosser and De Meur (2009}: 20-21) as the “investigation domain,” which determines that cases need to be comparable at least along specific dimensions, enabling us to consider these basic variables as constants.

Table 1 presents cases selected using the aforementioned threshold. The first column presents countries, while the second shows the years that account for the change in the ideal point of states from one UNGA session to the next; the third column represents the difference in these indexes: positive values represent political rapprochement towards the US, while negatives values indicate distancing. Three of the analyzed cases – Cuba (1958-1959; 1959-1960), Cuba (1968-1969; 1969-1970) and Grenada (1983-1984; 1984-1985) – encompass the identification of two extreme scores of FPC in subsequent years. Consequently, historical analysis was done by treating them as one, whereas they refer to processes triggered by the same factors.

The universe of this sample has internal homogeneity as the foreign policy analysis literature has traditionally used the alternation between autonomy/distancing and dependence/proximity regarding the US as an analytic framework (^{Giacalone 2012}); throughout that period, the US maintained hegemony over the region, materialized after the end of World War II and institutionalized after the creation of the Organization of American States (OAS) and the Inter-American Treaty of Reciprocal Assistance (TIAR). ^{Hey (1997}: 632) argues that the region has sufficient common characteristics to accommodate a unique approach, as its states have similar colonial and neocolonial history, highly export-dependent economies, close social and cultural backgrounds, similar foreign policy behavior, and are located in a hemisphere dominated by a superpower.

In discussing Latin American foreign policies, some analytical dimensions need to be considered. ^{Gardini (2011)} uses pragmatism and ideology as the primary explanatory factors, with the alternation and combination among these components impacting on foreign policy behavior. Pragmatism is defined as “the priority of action over doctrine, of experience over fixed principles […] characterized by medium-term planning and state, rather than government, policy,” while “ideology prioritizes preconceived positions and remedies over their actual viability and usefulness” and is associated with “relatively short-term planning and a personalized vision of international relations related to a specific leader” (^{Gardini 2011}:17). On the one side, pragmatism aims at achieving a greater role in the international arena and is conditioned by the acceptance of new responsibilities, and is there associated with resource availability. On the other, ideology aims at a short path to international visibility by contesting status quo and is consequently related to the lack of resources. The author also relates ideology-based foreign policies with restricted political systems dominated by isolated elites, and pragmatic foreign policies with professionalized bureaucracies and broader societal participation.

^{Hey (1997)} argues that authors working in this area have so far “failed to build on each other's work in a way that would move them towards a general theory” (632). To assist in that process, she proposes three different dimensions of analysis: 1) pro-core vs. anti-core; 2) autonomous vs. dependent; and 3) economic vs. political-diplomatic. The above-mentioned variables are used to explain states' susceptibility towards one side of each dimension.

The first dimension is divided between behaviors of support for the international order and major powers, or participation in anti-status quo organizations, such as the Non-Aligned Movement and the United Nations Conference on Trade and Development. The second is related to governments' ability to make decisions according to perceived national, political, or personal interests. Finally, the economic dimension would explain the foreign policy predominance of nationalist or liberal economic policies, while the political dimension relates to cooperation on specific topics, such as collaboration with the US in disrupting narcotrafficking.

The majority of selected cases come from Central America and the Caribbean, with Argentina as the exception. Despite differing from the others, Argentina was not withdrawn from the sample, as doing so would hamper the use of common and objective criteria for case selection. ^{Malamud (2011)} considers Argentina's foreign policy as subordinated to short-term domestic concerns, with variations in foreign policy following alternations in the presidency, which gives support including this case. A later study by ^{Amorim Neto and Malamud (2015)} using UNGA data also supports this perception, as Argentine foreign policy was considered more erratic than either Brazilian and Mexican policy.

Regarding the remaining cases, according to the literature discussed, the predominance FPSs in Central American and Caribbean countries can be explained by both international and domestic variables. With respect to international sources, geographical closeness to the US (inside its security perimeter) and its historical policy of interventions in the hemisphere serve to explain the foreign policy instability in these countries, as financial support, economic embargos, and military interventions played a pivotal role in overthrowing adversary administrations and/or placing friendly regimes, all of which influenced leaders' decision making. Complementarily, their domestic characteristics may also be related to the occurrence of FPS processes, as according to ^{Gardini (2011)}, the lack of resources leads to ideology-dominated foreign policies, aiming for short-term results and a short path to international visibility. Additionally, they are characterized by restrictive political systems, where leaders' interests may dominate the political agenda, enhancing the chances of FPS, leaders they do not face institutional constraints and use foreign policy as a tool to attain domestic political goals.

This supports ^{Hey (1997)} when she affirms that the most frequent independent variables used by Latin American foreign policy literature to explain its states behavior are: US influence, lack of economic resources, leader or regime ideology, and global distribution of power and wealth.

As FPS processes tend to occur during political and/or economic crises or after changes of administration/regime – periods typified by power concentration, when standard operational procedures in policy decision-making are disrupted – the role of the leader is primordial implementing processes of FPS. Additionally, bureaucratic variables do not have such a determinant impact, as most Latin American and Caribbean countries' institutions are poorly structured and easily influenced by leaders, with a few exceptional cases, such as Brazil and Chile.

IV. UNDERSTANDING FOREIGN POLICY SHIFTS

After analyzing the case studies and reviewing historical processes that led to FPS, two different pairs of conditions can be identified. Figure 1 presents an organogram illustrating conditions present in each case, the crisis output, and how conditions combined leading to the outcome.

Figure 1 Organogram of FPS processes 

First, labeled as pre-conditions, a couple of conjunctural conditions represent the initial scenarios: (i) political polarization; (ii) international isolation. On the one hand, political polarization was a result of political discontentment with longstanding dictatorships and/or institutional rupture. On the other hand, isolation was a result of nationalist economic policies and/or expansionist foreign policies, considered defiant behavior by hegemonic powers.

These adverse economic and/or political conditions led to another set of conjunctural conditions: (a) civil war or social chaos; (b) economic downturn. These conditions represent critical junctures, short periods during which drastic changes occur, shaping policies that are later institutionalized during longer periods of path dependence (^{Capoccia and Kelemen 2007}: 341-343). Despite the diverse sources of these determinants (domestic or international environments), they created adverse political scenarios and rising political pressure against incumbents, which functioned as policy windows for FPS. While accommodation was only achieved after leaders implemented administrative reforms with a significant impact on foreign policy guidelines, political crises that resulted in regime and leader change were characterized by the alternation between polarized opponent political groups, also resulting in FPS.

A deadlock in negotiations emerged either due to the inability of political groups to reach any kind of agreement, or to the ineffectiveness of embargos and/or political mediations by major powers. Argentina (both cases) and Cuba (1968-1970) presented similar paths from common pre-conditions and conjunctural conditions, mainly led by economic problems and major power sanctions/coercion to nationalist/independent foreign policy behavior; the response was an (I) accommodation strategy. The other cases presented diverse combinations of conjunctural conditions and outcomes, while presenting the same preconditions. Discontentment with political incumbents resulted either in civil wars or in economic crisis and international isolation; worsening political and social conditions caused the participation of international actors (major powers and international organization), leading to three different outcomes. First, when political forces failed to promote political transition to moderate groups, revolutionary regimes rose after civil wars, leading to (II) independent foreign policy – such as in Costa Rica (1947-1948), Cuba (1958-1960), and Nicaragua (1978-1978). Second, when the crisis was not managed by political mediation, a military intervention was undertaken, and friendly regimes took office resulting in (III) the restoration of pro-Western foreign policies – as in Grenada (1983-1985) and Haiti (2004-2005). Finally, when a political transition was not achieved, (IV) continuity (negative outcome) was maintained.⁵ Cases included in categories (II) and (III) represent processes of what ^{Pastor (2001}: 154) called succession crisis: when a declining dictator “does not allow for peaceful change by free elections,” leading the opposition to resort to violent change, either taking power quickly by a coup d'état or through an extended succession crisis. The crises outputs/solutions represented the periods of policy windows. Table 2 contains dichotomous variables indicating in which cases each of these (pre) conditions were present.

V. FOREIGN POLICY OUTCOMES CATEGORIZATION AND CASE ANALYSES

VI. CONCLUDING REMARKS

The conjunctural conditions discussed in this article have been utilized to depict how processes of FPS take place. Several causal paths, with distinct combinations of (pre)conditions, led to similar outcomes. The historical analysis allowed the identification of conditions that led to FPS.

In FPA it is important to highlight the importance of both domestic and international variables, and again this has been proven to be essential. While the success or failure of the mediation of international actors in political crises played a determinant role in the outcome, domestic variables were vital in leading to the most extreme processes of FPS, as pre-conditions of social and political chaos first created the scenario for instability and change and domestic political polarization was essential for radical outcomes. Thus, the conjunctural conditions identified – political polarization or social chaos, and economic downturn – represented the policy window for leaders to legitimize and implement FPS. Complementarily, pre-conditions were essential determining the extremity in processes of FPS.

Both independent and restoration FPS typologies were the most extreme, as they represented processes of complete transition on a theoretical foreign policy spectrum, despite going in opposite directions: towards or away from the US. Civil war scenarios and chaotic social conditions led to what ^{Pastor (2001)} conceptualized as succession crises, resulting in different outcomes, depending on leaders' responses to domestic and international political pressures. Political polarization played an important role in triggering these processes and defining the distancing among foreign policies of following administrations.

Accommodation-type outcome still represented a radical FPS process, but not a complete transition across the foreign policy spectrum, as policy change took place with leaders adjusting their course by implementing internal administration reforms and substituting key political posts within the current institutional order and reforms.⁸ Economic crisis presented a path for the adoption of accommodation strategies, aiming to remove international restraints. Incumbents acted as political entrepreneurs leading the processes of FPS by promoting reforms and changing key political posts.

As the extremity of FPS processes were determined by the antagonism among succeeding political groups, leaders that ascended after these succession crises – being previously excluded from the political system – acted as policy entrepreneurs and implemented foreign policy guidelines that they defended before stepping in office, and profited the policy window opening after revolutionary processes or international military intervention. Even during the same administration, as exemplified by the accommodation typology, leaders reviewed current policies and changes in key political offices were used to implement new foreign policies. Power concentration in the Executive branch, traditional in Latin American and Caribbean countries, the exceptional characteristics of these periods of policy windows, and the leaders' advocacy for FPS, guaranteed conditions for implementing new foreign policy guidelines without facing significant domestic restraints. These evidences support ^{Doeser and Eidenfalk's (2013)} arguments that policy entrepreneurs already have reformist ideas in mind before the window opening.

Previous theoretical discussion on Latin American foreign policy serves to contextualize the cases analyzed here. Despite having internal differences within the typologies, the findings are consent with the previous literature in the sense that all cases represent the alternation among either pro- or anti-core, or dependent and autonomous foreign policies, as advocated by ^{Hey (1997)}; or the struggle among ideology and pragmatic-founded foreign policies, as proposed by ^{Gardini (2011)}.

Another important feature in differentiating cases and typologies is related to the international system and domestic variation among cases. The sample included countries with different strategic positions in the American system, and with varying state structures and sizes. More extreme FPSs occured more frequently in small/weak and vulnerable states, while accomodation was more common in larger or more politically stable states, although further comparison would be need to confirm this finding. This gives support to ^{Gardini's (2011)} theoretical propositions, as FPSs took place mostly in Central American and Caribbean countries, where political systems are dominated by isolated elites, with ideology being used as a tool to contest the status quo, seeking a short path to international visibility. Additionally, geographical closeness to the US also appeared as an important variable to explain foreign policy instability in these weak countries, since hegemonic activism – through military interventions and embargos – resulted in foreign policy instability. This lends support to ^{Rosenau's (1966)} argument that idiosyncratic and systemic variables are the strongest determinants of foreign policy of small, developing countries. Furthermore, it corroborates ^{Hey's (1997)} arguments that the most frequent independent variables used to explain Latin American foreign policies are US influence, lack of economic resources, leader or regime ideology and global distribution of power and wealth. Finally, it endorses the consensus in the literature identified by ^{Below (2010)}, that the two most important factors affecting Latin American foreign policy are political leadership and US hegemony.

With respect to methodological matters, the measure used to select the most extreme processes of FPC in Latin America was mostly adequate, as a careful conceptual definition and operationalization as an FPC index served to select case studies that represented FPS processes, with only one exception. This avoided deliberate case selection and the use of isolate cases to backup specific arguments. The use of ideal point estimates based on UNGA voting record, which controls for agenda changes, guaranteed close association among historical events and change of states' behavior; that enhances our confidence in its appropriateness for future studies. Despite finding differences in the degree of FPSs across cases, it worked well to identify similar cases using common criteria. Additionally, the article sums efforts in analyzing Latin American and Caribbean countries foreign policies using UGNA, embodying a growing local literature. The technique advanced here might be used for further study case selection and quantitative research designs, as data is publicly available for all UNGA participants since their admission.