Foreign Policy Change in Latin America: Exploring a Middle Range Concept

In this article we examine patterns of change and continuity in Latin American foreign policies. We do so by asking two interrelated questions: How can we conceptually and empirically account for foreign policy change? And why do states change their foreign policies in Latin America? To answer these questions, we used the results of a new expert survey on foreign policy preferences in the region between 1980 and 2014. The results we obtained using both linear and non-parametric specifications are very clear and consistent: presidential institutions and ideology are what matters the most.


Introduction
Foreign policy change (FPCh) has been, and continues to be, a common concern among Latin American scholars. Pundits and practitioners from different ideological persuasions often lament the abrupt changes in foreign policy and the lack of what they call a política exterior de estadothe Spanish phrase for a stable, long-term, and consensual foreign policy. Simply put, foreign policy in Latin American countries has been alternatively depicted as 'erratic' (Gomez Mera 2009, 747), 'inconsistent' (Margheritis 2010, 1) or merely 'passive' (Tulchin 2016, 14). And yet, Latin America is too complex a region to be captured by a single narrative. Take Mexico's foreign policy, for instance, and compare it with Argentina's. While continuity and responsiveness to the international environment appear to be the dominant features of the former, fickleness and sensitivity to domestic whims has been a central trait of the latter. Also, countries like Brazil might be placed somewhere in the middle of the continuum, between Mexico and Argentina (Amorim Neto and Malamud 2015, 21).
Although FPCh is often depicted as a handicap, it can, under certain conditions, be an advantage. Foreign policies that adapt swiftly to changes in the international environment can benefit from this flexibility. It, hence, follows that in order to tell better-informed stories on Latin American foreign policy we need to be agnostic about the phenomenon and empirically characterize it, before understanding the drivers leading different countries to a diverse pattern of change and continuity. In the discussion that follows, we take a first step in that direction. We do so by asking two interrelated questions: How can we conceptually and empirically account for FPCh? And why do states change their foreign policies in Latin America?
The first of these questions is particularly relevant since the conceptual tools to address FPCh have so far been elusive. On the one hand, much of the Foreign Policy Analysis (FPA) scholarship has focused on the set of different strategies by which a country relates to the world. This scholarship has examined the foreign policies of specific countries and produced overarching typologies. The result has been the proliferation of thickly descriptive, case-oriented studies that, given their level of complexity, seldom render generalizable explanations (Hey 1997;Mora and Hey 2003). On the other hand, a more recent scholarship has introduced variable-oriented studies more amenable to large-n statistical methods, aiming for greater generalization. Yet, despite all its findings on a variety of issues, this approach has focused on single foreign policy issues or events (e.g. the signature of trade agreements, alliance formation or voting patterns at the UNGA), thus thwarting the evolution of FPA towards higher levels of aggregation, i.e. comparing foreign policies per se. In sum, FPA scholarship has been trapped between the qualitative analysis of broad foreign policies and quantitative explanatory designs addressing thin foreign policy issues. This deadlock calls for conceptual innovation and integrated methodological approaches (Goertz 2017).
It is in this intellectual background that a growing interest has developed within FPA scholarship to advance a middle ground approach (Alden and Aran 2017, 159) that can look at the evolution of foreign policy more broadly and, at the same time, do it in simplified terms that allow for meaningful comparisons in relatively large samples. In Latin America, this movement is still very incipient (cf. Belem Lopes, Pimenta de Faria, and Santos 2016;Amorim Neto and Malamud 2016;Jenne, Schenoni, and Urdinez 2017).
Most of the literature examining FPCh in Latin America focuses on few cases and emphasizes normative concerns. Scant academic effort has been put into developing rigorous comparisons in a relatively large-n context. Even more important, the existing literature has shown little interest, if any, in the development of middle range concepts and theories. This kind of research, we argue, would be fitter for statistical analysis and potentially lead to illuminating generalizations to the larger population of Latin American countries and beyond.
In this article, we examine patterns of change and continuity in Latin American foreign policies by defining a new middle-rage concept: FPCh. We believe that this concept retains our interest in the gross foreign policies of individual countriesi.e. the aggregate of several foreign policieswhile avoiding thick descriptions and identifying the relevant characteristics of that aggregate that we want to explain. Said otherwise, FPCh does not fragment foreign policies in several parts, but adds adjective to the whole, e.g. stable/unstable (Collier and Levitsky 1997;Goertz 2006). This approach, we argue, can be applied to many regions and epochs, but can be especially suitable to create a more fruitful dialogue between FPA and Area Studies, namely studying long-established Latin American foreign policy themes (e.g. trade policies, North/South interactions or the relation with the U.S) from a FPA/IR perspective. The outcome, therefore, will not be a unified theory of foreign policy but a partial explanation to 'provide insight into limited pieces of the overall puzzle of world politics' (Lake 2011, 467). In this case our focus will be on FPCh in Latin America.
To examine change and continuity in Latin American foreign policy, we used the results of a new expert survey on foreign policy preferences in the region between 1980 and 2014 (Merke and Reynoso 2016). This survey informed a database of 142 administrations and their respective orientations on three policy domains, namely 1) the pro-anti US positioning; 2) the North-South orientation of political partnerships; and 3) the place in the liberal-protectionist continuum. We do not look at FPCh as the binary opposite of continuity. We rather aim to examine varying degrees of change and continuity across time and space.
The next section of this article sketches our theory of FPCh, which follows a long tradition in focusing on the characteristics of the executive power and its capacity to mold this policy. Then we explain the way in which we conceptualize, operationalize and measure the general orientation of Latin American foreign policies, to estimate their change over time. Our goal is to develop a variable capturing a value of change for each Latin American president roughly from 1980 to 2014. Subsequently, the article examines the literature on FPCh and puts our hypotheses under test together with a number of alternative hypotheses that flow from it. The third section outlines the empirical approach to evaluating these hypotheses. This is followed by a discussion of the results from the statistical analysis. The article concludes by highlighting the implications of these findings for the study of Latin American international relations and FPCh more broadly.

Explaining FPCh in Latin America: An Institutionalist Approach
Why does foreign policy change? One of the striking lacunae in the study of foreign policy is the paucity of IR and FPA empirical analyses on the causes of FPCh. In a thorough review of the literature, Gustavsson (1999, 87) observed that 'there are relatively few' empirical studies on FPCh, and Garrison (2006, 291) then agreed to say that change received 'only sporadic attention' in the FPA literature. More recently, Blavoukus and Bourantonis (2014, 483) stated that FPA tended to focus 'on continuity and stability' with only 'sporadic attempts to account for change', and Alden and Aran (2017, 12) concluded that change 'is a neglected aspect of the study of foreign policy'.
Although a comprehensive review on FPCh is beyond the scope of this article, we observe four salient features on the study of change. Firstly, research on foreign policy change is inherently interdisciplinary. As such, FPCh can be a site of intersection between different related fields, e.g. political science, international relations, comparative politics or political psychology (Alden and Aran 2017, 134). In this article, we are particularly interested in observing how domestic political changes interact with global shifts thus yielding different foreign policy outcomes along the stability-instability spectrum.
Secondly, and related, the study of change has been, and continues to be, a multi-causal enterprise. Why do states behave as they do in world politics? From a FPA scholarship there is no simple answer to this question and the literature has relied on international and domestic parameters, on material and ideational drivers and on human agents and structural constraints. As a result, there is no single, unified theory of change, nor even a limited yet consistent set of variables to explain shifting behaviors. Explanations go from the interplay of structure and agency (Skidmore 1994, Goldman 1988 to cyclical models (Carlsnaes 1992, Rosati 1994 to cognitive biases (Welch 2005).
Thirdly, FPA scholarship in Latin America has been no exception to these lacunae.
Undoubtedly, the literature exhibits a large scholarly production that examines, for example, the change of foreign policy in Argentina, Brazil or Venezuela. But efforts to theorize these changes beyond a particular country have been rather scarce. As a result, we now have good stories to tell about individual foreign policies but we are still a long way to establish generalizable observations. And fourthly the literature on change has notably overlooked the impact of domestic institutional variables, particularly the structure and role of the executives. As Alden and Aran note 'FPA says little about the sources and conditions giving rise to significant alterations in a state's foreign policy ' (2017, 13). 'Understanding and integrating 'change' into analyses of foreign policy', they affirm, 'requires accounting for its impact in relation to individual decision makers, institutions and structures of decision making as well as the wider socio-political and external context within which such change occurs ' (14). In this article, we take this observation seriously. And we do so by studying the executive and its institutional context while controlling for other domestic and international variables.

The President and its Institutional Context
Alden and Aran claim that the study of FPCh 'through the three prisms of the individual, state institutions and political regimes, provides some of the foundational pillars required for the development of a theory of foreign policy change ' (2017, 138). Given that foreign policy is preeminently in hands of the executive power, we follow these authors' advice and focus on how presidential preferences are constrained by the institutions of the executive power in democracies.
A large body of literature has argued that the executive is the main actor in foreign policy decision-making and therefore should be the central focus of any theory of FPCh.
According to Milner and Tingley (2016, 263) this literature highlights that the executive 'negotiates and interacts with foreign leaders'; is the main point of 'governmental contact with foreign governments'; 'transmits the international environment into the domestic process of policy making' and 'his perceptions and views on what the international environment is like tend to dominate the domestic policy process'. Thus, a narrow focus on the executive seems to be the natural starting point for any theory of FPCh.
This observation particularly holds for Latin America. As Mora and Hey (2003, 4) observe, Latin American foreign policy 'has traditionally been the preserve of the executive and a narrow elite.' Typically, presidents have enjoyed a wide margin of maneuver to conduct foreign affairs and thus much of Latin American foreign policy can be explained by studying presidential preferences (Malamud 2014;Jenne, Schenoni, and Urdinez 2017; see also Burges and Chagas Bastos 2017).
Recent research suggests that this trend has only grown over the last twenty years. Malamud (2005, 158) examined presidential intervention in the management of Mercosur to conclude that presidents exhibited a significant degree of autonomy, 'acting not only as decision makers but also as dispute settlers and guarantors of commitments' (158). For Malamud (2015) president-to-president interactions in Latin America can be understood as a dominant type of diplomacy, namely 'interpresidentialism'. Emerson (2015, 488) explored the growing importance of the executive to the foreign policies of Brazil and Venezuela and concluded that the 'presidentialization' of foreign affairs has led to a 'short-term focus aligned with the electoral cycle, and a greater degree of personalism'. In sum, research on Latin America foreign policy suggests that the study of presidential preferences can illuminate important aspects of foreign policy change. In what follows, we introduce alternative hypotheses that flow from this reasoning. This is not to say, however, that presidential preferences have been uncontested domestically (by the military, the opposition or the private sector) or that presidents have been immune to international constraints. What we argue instead is that, typically, presidents have enjoyed a wide margin of maneuver to conduct foreign affairs and thus much of Latin American foreign policy can be explained by studying presidential preferences and constrains. If presidents are driving foreign policy, then a first observational implication of our theory is that a change in the president's preferences will be associated to FPCh.
And yet, presidential democracy in Latin America exhibits various political models.
Comparative political scientists have examined how strong presidents are relatively to each other, finding variation within the region. Following insights from the rational choice literature on veto players (Tsebelis 2002), some (Lima and Santos 2001;Schenoni and Aztiria 2014) have already interpreted FPCh as caused by the concentration of power in the president, and thus we have reasons to believe that presidential systems that show higher levels of power concentration in the Chief of State are intrinsically more prone to policy change. In other words, the more presidential power, the more FPCh.
Although institutional characteristics of presidential power are relatively constant, presidential independence from political actors and institutional constraints can vary through time in the same political system leading to different degrees of FPCh. A radical form of this phenomenon occurs when presidents are in overt conflict with other powers and de facto insulated from those political influences during presidential crises (Pérez-Liñán 2007;Llanos and Marsteintredet 2010). Previous research has shown that the attention to international incentives is reduced in contexts when presidential mandates are at stake (Schenoni 2017) and it is equally plausible that foreign policies will be disturbed through similar mechanisms. We thus expect that presidential crises will be associated to FPCh.
Finally, presidents can face very different bureaucracies constraining them. In the Latin American context the Brazilian Ministry of Foreign Affairs, for instance, is widely known for its bureaucratic powers, espirit de corps, and outstanding training (Da Faria, Lopes, and Casarōes 2013). Brazilian diplomats are highly influential in the foreign policy circle and Foreign Ministers themselves often come from their ranks. Other bureaucracies are not as professional or as influential. Some countries do not even have a particular institute to form their future diplomats. In many such countries ambassadorships are granted to politicians with no credentials. More generally, executive bureaucracies beyond the diplomatic corps can have a strongly constraining effect over presidents in key foreign policy agendas such as trade, defence, finance, environment or migration. This can be either due to the influence of their standard operative procedures, bureaucratic cultures, and even their political power (Allison and Zelikow 1999). In any case, they should act as stabilizers (Goldman 1988;Skidmore1994). This leads to a clear prediction: The weaker executive bureaucracies, the higher FPCh. 1 In sum, our model predicts that a change in the incumbent's preferencese.g. a change in presidential preferenceswill ceteris paribus result in FPCh. They also predict that institutions that strengthen the president or critical junctures where the president is particularly isolated from other powers will increase FPCh. These were the conditions, for instance, under which the government of Carlos Menem brought about a radical turn in Argentina's foreign policy towards the United States or Fernando Collor de Melo abandoned a third-world perspective in Brazil's foreign policy. Conversely, we expect that institutions constraining the power of the president, such as strong executive bureaucracies, will have a stabilizing effect and thus affect FPCh negatively. This has been the case in Chile since the transition to democracy in 1989, where socialists, Christian democrats and rightwing presidents, albeit with differences, have converged on the fundamentals of Chile's foreign policy.

Measuring FPCh in Latin America: Themes, Data, and Patterns
To unpack the range of variability in FPCh it is helpful to focus first on the key dimensions of Latin American foreign policy. Given space consideration, the task of explaining Latin American foreign policy goes beyond the scope of this article. This section therefore aims to flesh out the essential contours of Latin American foreign policy and then estimate variation within and between countries. Although Latin America is too extensive a region to generalize, some themes emerge throughout the literature (Hey 1997, Giacalone 2012. We present here three dimensions of foreign policy, which, we believe, capture some important and enduring dilemmas in the region. Undoubtedly, the three dimensions are empirically connected yet they are analytically distinguishable. Firstly, the study of Latin American foreign policy has been particularly concerned 'with the United States and the degree to which its activities and policies constrain Latin America's foreign policy aspirations and behavior' (Mora and Hey 2003, 4). Indeed, the literature on U.S.-Latin America relations is comparatively the largest (Giacalone 2012).
The dynamics between Latin America and the U.S. are deep and broad, spanning from security issues to finance and trade, and involving myriad actors (Carothers 1993;Kryzanek 1990;Schoultz 1998;Atkins 1999;Pastor 2001;Rivarola Puntigliano 2008;Smith 2008;Weeks 2009;Pastor and Long 2010;Williams 2011;Bertucci 2013) with diversified foreign policy standpoints (Russell y Tokatlian 2008). Both dependency theorists and realists highlighted the asymmetrical power relations that pervade the Western Hemisphere and asked to what extent Latin America dependence was reflected or not in its 'foreign policy under asymmetry' (Long 2015).
Secondly, another recurrent theme in Latin America has been its place in the North-South divide and the discussion over the benefits that can be obtained from an alliance with industrialized/developed countries vis-à-vis the developing world (Puig 1980;Jaguaribe 1985;Escudé 1992;Hey 1997 And thirdly, there is the political economy of foreign policy. The literature examining this dimension focuses on several issues including the political and commercial relationship with the United States (Bergsten 2002;Feinberg 2002); the role given to transnational corporations; the relationship with the global flow of capital (Robinson 2002) and the positions adopted vis-a-vis global and regional trade (Bouzas 2000, Phillips 2003and Tussie 2009). In these debates, trade policy has been a fundamental instrument to address Latin America's regional and global economic integration. Yet the region exhibits a recurrent discussion on the benefits and desirability of international trade integration. We want to know how foreign policy changes along these three dimensions from one administration to another and the degrees of change and continuity between countries over the same period of time. Take Menem in Argentina (1989Argentina ( -1999  Though by no means a comprehensive analysis, we believe these dimensions do serve to capture fundamental patterns of change and continuity in Latin America's foreign policy.
In order to examine preferences regarding foreign policy, we asked a total of 134 experts in 19 Latin American countries to locate administrations in their own countries from 1980 to 2014 in the three dimensions outlined above on a 7-point scale. Expert survey is an extensively used method to locate agents (e.g. parties, governments or presidents) in a given policy space (Budge 2000). And yet it has often been disqualified for reflecting subjective biases and conceptual inconsistencies among experts (Bollen and Paxton 2000). When it is done following certain rules, however, they can help to provide more accurate estimates. To avoid pitfalls and enhance the production of genuine expertise, Schedler (2012) recommends, among other things, working with a consistent selection of experts; an elaborate conceptual disaggregation; and an additive procedure (e.g. arithmetic means). We took his recommendations seriously by identifying foreign policy experts and not country experts. We also asked them not to estimate each president on a single foreign policy dimension but on several (e.g. ideology, the U.S., geopolitics and Once we have captured fundamental foreign policy orientations -our middle-range obsevrations of interest -the next logical step is to measure FPCh, which we understand as the level of variation in foreign policy from one presidential term to the next. The data transformation that we propose is relatively straightforward: As defined above, for every country, FPCh results from a change ∆ which we then square and take the square root of because we are not interested in the relative direction of such change but in the absolute value of change. In other words, we do not care whether the Latin American president had a more or less 'liberal' foreign policy. Instead, we focus on whether FPChno matter the directionwas more or less abrupt. Online appendix 1 shows the figures of FPCh for the 19 Latin American countries in our sample.  As one would expect, Cuba is the country that altered its foreign policy the least between administrations. This may be understood in relation to the stability of its political regime. Since 1959, Fidel Castro's leadership style and ideology 'have defined the nature of the regime and its foreign policy' (Fernández 2003, 98). For all its continuity, however, Cuba also faced the need to make important adjustments as seen in Figure

So far, we have introduced an empirical operationalization of foreign policy in Latin
America and defined what we mean by change. The result is a first estimate of FPCh in the region of which we offer brief descriptive statistics in figures 1 and 2. This description allows us to conclude a few things about FPCh in Latin America. In particular, they unveil a clear hierarchy between countries: While Panama, Dominican Republic, Chile, Cuba, and Costa Rica exhibit significant levels of continuity between administrations, Honduras, Ecuador, Bolivia or Argentina tend to show higher levels of change. Hence the labels of 'inconsistent' or 'erratic' foreign policies that are uniformly granted to these countries by the press, when studied from an academic perspective, should consider this relative order.
In the following section, we will briefly examine the performance of our theoretical expectations based on the characteristics of the executive power, together with other domestic and international factors that may possibly account for these diverse patterns.
Although the stable hierarchies depicted in this section suggest long-term factors might be determining the stability of foreign policy (and perhaps the stability of policy, more broadly) in a path dependent manner (Mahoney 2010, Soifer 2015 we will focus mostly on factors that have varied in the 1980-2014 period and use fixed effects in our models to control for those time-varying features. Our quantitative approach, however, does not preclude the possibility that historical events previous to 1980 might have set these countries in their current course of relative foreign policy stabilitya possibility that further research could scrutinize.

Hypotheses Definition: Presidential and Alternative Explanations
Recapitulating, our model of FPCh, which we introduced in section two, bases its predictions on the characteristics of the executive. More precisely, it predicts that changes in executivein the Latin American context, presidentialpreferences, power, and institutional environment would be the best predictors of FPCh. We use a set of four indicators to capture our president-related hypotheses of interest.

Presidential Variables
The first and most clear prediction of our theory is straightforward: H1: A change in the president's preferences will be associated to FPCh.
To estimate the aggregated preferences of each president, we rely on data measuring presidential ideology from Merke and Reynoso (2016). Ideology is measured in a 7-point scale, where 1 is left and 7 is right, and change is measured in the same way we described above. Yet, presidents also vary in their power so we expect: H2: The more presidential power, the more FPCh.
We use Shugart and Carey index of presidential powers  retrieved from Mainwaring and Perez-Liñan (2013). This index reflects the sum of two point-score measures: an index of legislative powers, reflecting proactive and reactive powers vis-àvis the legislature, and an index of non-legislative powers, reflecting presidential control over appointments and over the dissolution of congress. These institutions, however, capture presidential power de jure but fail to capture the ebb and flow of de facto presidential influence. To compensate for that we also predict: H3: Presidential crises will be associated to FPCh.
We use here a simple dichotomous indicator adopting the value of one (1) if the administration was preceded by an episode of presidential crisis (Perez-Liñan 2007).
Finally, very strong presidents, both institutionally and factually, could be relatively weaken by the presence of a strong executive bureaucracies, and thus we propose: H4: Weak bureaucracies will be associated to FPCh.
Since there are no available indicators of the relative strength of the diplomatic bureaucracy, we proxy this by using the Varieties of Democracy (V-Dem) public sector corruption index. The indicator captures to what extent public sectors officials act in a meritocratic fashion and according to bureaucratic rules or grant positions and benefits in exchange for personal economic or political gain (Coppedge et al 2017). This indicator has significant face validity, with countries like Brazil, Chile, Costa Rica, and Uruguay, known for their strong state capacity and diplomatic corps, at the top of the rank.

Partisan and Regime Variables
In FPCh is due to the change in presidential preferences, or that of her coalition, we consider the possibility that a change in the ideology of the party of the president is explaining FPCh.
H5: Changes in the incumbent party's ideology will be associated to FPCh.
To measure incumbent party ideology we use data provided by Murillo, Oliveros, and Vaishnav (2011)  H6: Higher electoral volatility will be associated with higher FPCh.
We use scores of electoral volatility in the elections for the lower chamber during the year that the president was elected (if presidential and legislative elections were concurrent) or those of the first legislative election after the president was elected. Data was retrieved from Mainwaring (2017).
Hypotheses H2 and H3 can only take place if the countries are democracies or very sophisticatedly institutionalized authoritarianisms. Also, by eyeballing Figures 1 and 2 above, one could surmise that autocracies are less prone to change than democracies due to the relative stability of those in power. Cuba, with the lowest average FPCh in the region, is a case in point. Thus, a control for regime type is required.

H7:
The greater the level of democracy, the higher FPCh.
Drawing on Freedom Index, we follow the recodification used in Mainwaring and Perez-Liñan (2013) in which scores are recoded to range between 0 and 12, where 12 is most democratic. The observations represent the mean Freedom Index score during a given presidency.
As a final domestic variable of interest we consider the possibility that extreme polarization or radicalization might be leading to FPCh, as different political actors might be stuck in centrifugal dynamics that lead them to support extremely different foreign policies.
H8: The greater the levels of radicalization, the higher FPCh.
We use Mainwaring and Perez-Liñan's (2013) measures of radicalism in all government and opposition political actors. This indicator varies between low (0) and high (1) levels of radicalization.

System-Level Variables
As argued above, the international factors that can provoke change in a given foreign policy are many, ranging from the coercion of a hostile third state to smooth adaptation to evolving international norms or regimes. In Latin America, however, the literature seems to agree that changes in foreign policy are attributable to a narrow set of international issues, namely the international economic environment (e.g. the so-called 'commodity boom'), the varying influence of Washington (from Monroe Doctrine to Roosevelt's Big Stick to the 'Washington Consensus') and the shock of crucial global changes (e.g. the end of the Cold War or the beginning of the Global War on Terror).
International factors can take the form of single events or critical junctures affecting the foreign orientation of several countries at the same time. In this light, almost nobody disputes that the end of the Cold War had a profound impact on the pattern of international relations. One could expect, for instance, that the end of the Cold War resulted in incentives for Latin American countries to realign with the U.S. due to a new distribution of power in which the U.S. remained as the sole superpower. It, hence, follows that:

H9: The end of the Cold War will lead to a change in foreign policy
We use here a simple dichotomous indicator, which adopts the value of one (1) if the dissolution of the Soviet Union took place during the presidency.
In the same way, the 9/11 attacks to the World Trade Center had enormous strategic consequences for the U.S. and the international society. They changed conventional attitudes towards terrorism dramatically and irrevocably. And they changed U.S. foreign policy by declaring a global war on terror. Did 9/11 alter Latin American foreign policies? Gardini and Lambert (2011, 5) argue that 'Latin America is a region where the impact of 9/11 and the so-called War on Terror has been less strongly felt.' Yet there is no empirical analysis of this observation covering the whole region or at least the most significant states thereof. The hypothesis, therefore, is as follows: H10: 9/11 will provoke a change in foreign policy As with H9, we use here a simple dichotomous indicator, which adopts the value of one (1) if 9/11 occurred during the presidency.
Economic growth is another variable of interest for our analysis. Both cyclical (Carlsnaes 1992;Rosati 1994) and loss-aversion (Welch 2005) models expect FPCh to take place when the economy is doing badly. An important amount of literature has been devoted to explore how economic constraints provoked the realignment of Latin American countries' foreign policies as a way to access credit, investment and even aid (Feinberg 2002;Phillips 2004;Grugel and Riggirozzi 2012). The implicit understanding of this literature is that countries in dire straits have to make substantial changes in their foreign policies if they want to get help from the markets or a more powerful country. For Latin Americans, this may evoke recurrent debates about how to 'return' to world politics -'la reinsersión del país en el mundo'that use to follow long periods of retraction or isolation. In this light, Latin American foreign policy has its own flavor, say Mora and Hey (2003, 9), and it is 'peppered by the role of core actors and domestic economic crisis'.
Conversely, countries that do well economically would not have a reason for changing their current foreign policy course. This literature has specific observational implications that we try to test in this article: H11: Lower economic growth will lead to higher FPCh.
To examine economic growth, we will look at growth in per capita GDP. We use The World Bank's World Development Indicators for the first year of the presidency.
Alternatively, some have argued that FPCh is a function of U.S. influence over Latin American countries and the extent to which these countries have enough leverage to resist the mandates of Washington. Applying this argument to the democratization debate, Levitsky and Way (2010) have recently proposed that the probability that a given country will evolve towards a competitive authoritarian regime is a function of the linkages and leverage the U.S. has vis-à-vis with that particular country. In general, one could argue that the regional hegemony of the U.S. might be a potent stabilizing factor (Mearsheimer 2001). The analogy with FPCh is thus all too straightforward:

H12: The higher U.S. influence the lower FPCh
To estimate U.S. influence in the region we use trade data provided by Barbieri, Keshk and Pollins (2009). We compute total trade as the sum of exports and imports to the US and then divide it by population to get to a per capita specification that controls for country size.
Lastly, Urdinez et al (2016) find that Chinese influence has acted as a countervailing force to U.S. influence in Latin America, at least since 2003. Their article was followed by a literature confirming that China might have stirred some foreign policies away from their normal course, which leads to our last alternative hypothesis:

H13: The higher China's influence the lower FPCh
We measure the influence of Beijing in the same way we did with Washington's.

Statistical Model
Given the structure of our data and the distribution of our dependent variable (FPCh) we use two main statistical frameworks to test the aforementioned hypotheses. We start with a standard linear approach, i.e. a simple ordinary least squares (OLS) model with autocorrelation and cluster correction, robust standard errors, and country fixed effects.
Since we are exploring a measure of change, autocorrelation in these models is theoretically and empirically of little relevance, which we corroborate by including a lagged specification of FPCh in robustness checks. This linear specification provides a correct (Beck and Katz 1995), yet probably too parsimonious heuristic. This is so particularly because of the characteristics of our dependent variable.
In a second approach, we take two characteristics of FPCh more seriously: That our index is bounded between 0 and 1, and that it is continuous between the two, with a distribution that is skewed towards zero and with almost no observations beyond a .6 threshold. This distribution suggests that a specific type of generalized least square (GLS) model is necessary and a logit or probit specification might not be particularly suitable. The type of model that better fits this distribution is a beta regression (Ferrari and Cribari-Neto 2010), which performs maximum likelihood estimation for the parameters of each variable. This second model should capture any non-linearity better.

Findings
Looking at the R 2 statistics in the linear models reveals that most of FPCh is explained by domestic variables alone. Although economic growth seems to have had a significant and non-trivial effect on foreign policy realignment, it is not clear that this is per se a systemic variable, and international variables account for only a small part of the overall variation in FPCh. However, it is interesting that the effect of economic growth is in line with our expectations, with more growth affecting FPCh negatively.
Among the domestic variables the effect of ideological change at the level of the incumbent president is no doubt the most impressive one, both statistically and substantively. As we expect, presidential instability and weak bureaucracies also seem to be relatively powerful predictors of FPCh, but are outshined by the effects of a change in the president's ideological persuasion.
Following the results of the linear specification of Model 3, the maximum change in presidential ideology observable in our samplea value of roughly 4 in the 7-point scale leads to the maximum observable change in FPCha value of .6 in the 0-1 continuum. Figure 4 shows that the non-parametric models (in this case the beta specification of Model 3) almost exactly replicate these results.

Figure 3. Maximum likelihood prediction with 95% confidence intervals
Looking retrospectively at our data we find this relation is not surprising. Our composite index of FPCh and change in presidential ideology correlate very highly (.8) in our sample. There is at least one interpretation of these results that is clear: presidents are very important in defining foreign policy in Latin America. Although we already knew this was the case, the finding serves as corroboration and a reminder that presidents often follow their preferences in foreign policy across the board, despite variation in presidential power from country to country. Only their bureaucracies and economic growth seem to have a stabilizing effect, but the whims of presidents can also be exacerbated in contexts of presidential crises.
In another alternative interpretation, radical ideological change and the positive effect of presidential crises could be capturing the effect of intra-elite struggle (Schweller 2006;Schenoni 2017). It might be that big changes in presidential ideology happen in countries where this phenomenon is actually present. Take Argentina. Malamud (2011, 88) observes that in the last decades the foreign policies 'flowing out of the Casa Rosada have been at least as many as the presidents themselves.' Foreign-policy reorientation, he continues, 'have occurred in the wake of both regime change and administration change.' Upon these observations, Malamud (2011, 87)  were also systematically unstable throughout the period. In sum, change in presidential ideology could be reflecting historical, even country-specific trends, despite the statistical tools we have applied to control for those.

Conclusions
In this article we have provided a first approximation to the phenomenon of FPCh in Latin America. To do so, we identified the three dimensions that largely define foreign policy orientation in the region. We then aggregated them into a single index, and calculated the change between different administrations. We did that for 19 Latin American countries between 1980 and 2014, but our procedure could be replicated to analyze FPCh in other time periods and regions as well.
After providing several descriptive statistics of FPCh in Latin America, we switched to an explanatory focus. Having answered what FPCh is, we asked how one could predict it. The results we obtained using both linear and non-parametric specifications were very clear and consistent: presidential preferences matter the most. Put it simple, a change in the preferences of the president produces a change in foreign policy that is almost equivalent in magnitude, all other theoretically relevant factors set to their means. Three answers arise from these findings. First, presidents virtually define FPCh in Latin America. Second, it is the change in their ideologynot in their powerthat afects FPCh.
Third, presidents can be limited by strong bureaucracies but can also be whimsical when their power is at stake in the event of presidential crises.
The argument pursued here is not meant to exclude the international system. We do not take the strong assertion that transnational pressures have no influence on Latin America's foreign policy twists and turns; instead our empirical strategy has been to take a tighter focus on the ways domestic politics do. Similarly, our findings do not preclude the possibility that other domestic variables might affect FPCh, in fact it is our understanding that most of the mechanisms we try to uncover still remain obscure. The main contribution of this work, if any, is to conceptualize FPCh and set the stage for a potentially fruitful debate. Future research could and should revise our findings by expanding the set of hypotheses we have considered and delving into the nuances of our key findings.